**Contents**




## **About the Editor**

**Sunkuk Kim** studied Construction Engineering and Management at the department of architectural engineering, Seoul National University. He had joined three Korean construction firms, Dealim Industrial Co., Ltd., Deadong Coporation Co., Ltd. and Seoktop Construction Co., Ltd., for 12 years. As a visiting scholar, he researched about the construction management & organization at the department of civil engineering, Stanford University from 1994 to 1995. Since September, 1995, he has served at Kyung Hee University as a professor.

Kim served as a dean of the Graduate School of Technology Management from 2015 to 2018. In addition, he was an organization chair in three international conferences including ICCEPM (2009) and MOC (2015). He was also a vice president of Korea Institute of Ecological Architecture and Environment and Korean Council on Tall Buildings and Urban Habitat (K-CTBUH). He also served as a guest editor for two special issues of Modular & Offsite Construction (2017) and Global Convergence in Construction (2011), Automation in Construction and special issue of Global Convergence in Construction (2010), International Journal of Strategic Property Management. Currently he is the guest editor for special issue of Technology and Management for Sustainable Buildings and Infrastructures, Sustainability.

Kim has concentrated on the research such as health performance evaluation of buildings, development of sustainable construction technology and management, simulation, optimization and risk management, construction information technology. Especially, for about a decade, he has participated in the development of SMART frame, a sustainable structural system, and production technology of free-form concrete panels.

## *Editorial* **Technology and Management for Sustainable Buildings and Infrastructures**

**Sunkuk Kim**

Department of Architectural Engineering, Kyung Hee University, Yongin-si 17104, Gyeonggi-do, Korea; kimskuk@khu.ac.kr; Tel.: +82-31-201-2922

According to a report published in 2019 by the United Nations Environment Program (UNEP), the building sector accounts for 38% of all energy-related CO2 emissions when adding building construction industry emissions [1]. Yudelson (2008) argued that the building sector accounted for 45% to 65% of landfill waste [2].

Given this fact, the building sector must be one of the major causes of global warming and the resulting climate catastrophe. Therefore, research on the technology and management of the entire process including design, construction, O&M, and decommissioning is urgently needed for sustainable buildings and infrastructure that minimize energy use throughout their life cycle. At this point, it is judged that it was timely to hold a Special Issue under the topic of "Technology and Management for Sustainable Buildings and Infrastructures".

At the time that the world is struggling with the COVID-19 pandemic, this special issue has been published in 27 research papers [3–29], 1 review paper [30], and 2 technical notes [31,32], and with the help of many research colleagues and reviewers. A total of 30 papers were published. A total of 104 authors from 9 countries including Korea [3,5,6,8,12–15,17,19,20,22,23,25–28,30–32], Spain [11,18,21], Taiwan [4,24], USA [16,17,25], Finland [10], China [29], Slovenia [9], the Netherlands [7], and Germany [21] participated in writing and submitting very excellent papers that were finally published after the review process had been conducted according to very strict standards.

Among the published papers, 13 papers directly addressed words such as sustainable, life cycle assessment (LCA) and CO2 [5–7,11,12,14,16,19,20,22,25,27,28], and 17 papers indirectly dealt with energy and CO2 reduction effects [3,4,8–10,13,15,17,18,21,23,24,26,29–32]. Sustainability research related to CO2 and the resulting climate change started in the construction field more than 20 years ago. Although life cycle cost analysis (LCCA) has dealt with the energy use of buildings for more than 40 years, it focuses on cost rather than CO2 reduction. In the 21st century, research on net zero or near zero energy use of buildings has been conducted, but research on embodied CO2 resulting from the design and construction stage has been excluded because it is limited to the operation and maintenance stage. Until recently, many design and construction studies focused on maximizing economic benefits, and rarely focused on carbon neutrality or CO2 emission minimization. As a result, there are not yet many papers directly dealing with energy and CO2 reduction throughout the construction project life cycle.

Among the published papers, there are 6 papers [4,6,9,18,29,32] dealing with construction technology, but a majority, 24 papers [3,5,7,8,10–17,19–28,30,31] deal with management techniques. The reason is that construction management can be approached more easily than construction technology when considering research cost, time, and effort. Among all the papers, 15 studies focused on buildings [7,8,10,12,14–16,18,19,22,25–27,31,32], 9 studies on infrastructures [6,9,11,13,17,20,24,28,29], and 6 papers could apply to both [3–5,21,23,30]. With the development of science and technology, there is a tendency for buildings to become taller, larger, and more luxurious, and the energy use tends to increase rapidly. In particular, this trend is conspicuous in the developed countries where most of the papers

**Citation:** Kim, S. Technology and Management for Sustainable Buildings and Infrastructures. *Sustainability* **2021**, *13*, 9380. https:// doi.org/10.3390/su13169380

Received: 13 August 2021 Accepted: 19 August 2021 Published: 20 August 2021


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

have been submitted. In addition, in the developed countries, infrastructures such as roads, bridges, seaports, airports, and power plants are sufficiently established. Therefore, it is presumed that there are more papers on sustainable buildings than on sustainable infrastructures.

The authors of the published papers used various analysis techniques to obtain the suggested solutions for each topic. Listed by key techniques, various techniques such as Analytic Hierarchy Process (AHP) [3,12], the Taguchi method [4], machine learning including Artificial Neural Networks (ANNs) [5,28], Life Cycle Assessment (LCA) [6,7], regression analysis [13,17,19,25,28], Strength–Weakness–Opportunity–Threat (SWOT) [11], system dynamics [16,26], simulation and modeling [10,19,22–24,29,31,32], Building Information Model (BIM) with schedule [21,24,27], and graph and data analysis after experiments and observations [8,9,14,15,18,20,27,29–32] are identified.

As mentioned above, although the construction sector is a key influencer that harms the global environment, many studies have been focused on cost, time, quality, and safety. However, future research should be conducted on the basis of carbon neutrality or CO2 emission reduction. For example, previous cost minimization studies should be conducted as cost optimization studies based on CO2 emission reduction or minimization. As such, if all research is conducted in the direction of pursuing sustainable buildings and infrastructures, the global environment will be gradually improved.

Finally, I would like to thank Maggie Sun of MDPI and others for their active cooperation in making this Special Issue successful, research colleagues who submitted excellent papers, and reviewers who have been active in the review process.

**Author Contributions:** Conceptualization, S.K.; methodology, S.K.; validation, S.K.; formal analysis, S.K.; writing—original draft preparation, S.K.; writing—review and editing, S.K.; supervision S.K. The author has read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The author declares no conflict of interest.

## **References**


## *Article* **Development of Key Performance Indicators for Measuring the Management Performance of Small Construction Firms in Korea**

**Doyeong Kim 1, Wongyun Oh 2, Jiyeong Yun 2, Jongyoung Youn 2, Sunglok Do <sup>2</sup> and Donghoon Lee 2,\***


**Abstract:** Large construction firms execute management control in preparation for a fast-paced business environment, but small ones are unable to do so. This is because there is no management control model tailored to them. The current study derived Management Performance Evaluation Indicators (MAPEIs) for small construction firms for measuring the management performance of construction firms with 10 or fewer employees, considering the characteristics of small construction firms. MAPEIs consist of BSC (Balanced Scorecard), performance, and the hierarchy and weighted value of KPIs (Key Performance Indicators). After an interview with an expert, based on the management performance indicators of large construction firms, a final hierarchy of small construction firms was constructed through modification and supplementation. The KPIs of the hierarchy were analyzed through a survey using the AHP (Analytic Hierarchy Process) method to finalize MAPEIs for small construction firms in Korea. The final MAPEIs underwent a feasibility evaluation to apply them to real life. It is expected that they can be used as fundamental resources for system development for small construction firm management performance and control. In addition, further studies to resolve the limitations would improve the competitiveness of small construction firms.

**Keywords:** management performance evaluation indicators (MAPEIs) for small construction firms; AHP; key performance indicators (KPIs); corporation management; small construction firms

## **1. Introduction**

Recently, the Construction and Economy Research Institute of Korea concluded that the construction industry of Korea has officially been in a depression since the second half of 2018 and anticipated that it would likely continue until the early to mid-2020s. They mentioned that it would be necessary to develop management strategies suitable for the period. The downturn's impact on the construction industry is greater for smaller firms compared to larger firms. Despite radical changes, the number of construction firms registered in Korea increased by about 120% from 10,921 in 2013 to 13,168 in 2020 [1].

Statistics Korea classified the construction firms in Korea into scales based on the number of full-time employees. The number of firms with fewer than 50 employees was 97,314 out of 100,654. This means that small construction firms account for 96.7% of Korea's construction market according to the criteria of the Construction Association of Korea, and most of the construction firms being added to the list are small ones [1,2].

These days, there is not much call for construction work, and the number of projects to bid for is very limited. An increase in the number of small construction firms increases competition and makes the probability of winning a bid very slim. The Construction and Economy Research Institute of Korea released a BSI Report in 2019. The average BSI (Business Survey Index) of small construction firms was 79.5 and did not exceed 100 in 2019, reflecting the overall worsening of the industry and greater burdens on the management of firms [2].

**Citation:** Kim, D.; Oh, W.; Yun, J.; Youn, J.; Do, S.; Lee, D. Development of Key Performance Indicators for Measuring the Management Performance of Small Construction Firms in Korea. *Sustainability* **2021**, *13*, 6166. https://doi.org/10.3390/ su13116166

Academic Editor: Ali Bahadori-Jahromi


Received: 12 March 2021 Accepted: 24 May 2021 Published: 30 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The management difficulties experienced by small construction firms are caused by both external and internal factors. First, there is not a sufficient management control system. Large construction firms analyze management characteristics, along with the external environment and internal capabilities, using an adequate management control system to establish management strategies and plans according to the management characteristics. Management performance is measured to verify whether the objectives have been met. However, small construction firms are incapable of identifying the causes of difficulties due to a lack of management control systems and difficulties in measuring the management performance. Second, it is difficult to respond to and prepare for changes in the external environment. Due to the difficulties in management control, it is difficult to respond to the fast-paced environment of the construction industry, which involves a high rate of unpredictability, and impossible to prepare for the changes that they may encounter. Third, the unorganized structure of firms is a challenge. Unlike large construction firm, small construction firms find it hard to organize because there are only a few members and they lack management expertise. As mentioned above, it is impossible for small construction firms to execute management control due to difficulties in management. Therefore, the first step to take would be to understand the current circumstances of small construction firms. The position and status of each firm shall be identified by analyzing the problems and measuring the management performance. Then, sustainable management control shall be executed. For that purpose, this study was conducted to derive the MAPEIs (Management Performance Evaluation Indicators for Small Construction Firms) to analyze the management characteristics of small construction firms in Korea and measure their management performance.

In order to obtain the performance indicators of small construction firms, experts were interviewed based on the management performance indicators of large construction firms derived from preceding studies. The first expert survey was conducted with the top managers of five small construction firms, and the second survey was conducted with 33 top managers and engineers. The indicators were supplemented and modified to create a final hierarchy suitable for the scope of this study. KPIs (Key Performance Indicators) of the hierarchy were analyzed through a survey using the AHP method to finally derive the MAPEIs of management performance of small construction firms. Finally, the MAPEIs were tested in a real-life environment.

## **2. Preliminary Study**

The balanced scorecard (BSC) of Kaplan and Norton is a strategic management system developed to measure the management performance of companies. BSC measures and controls performance in four balanced perspectives of finances, customers, internal processes, and learning/growth. Various studies have been conducted by companies, organizations within companies, and other areas that require individual competence and strategic systemization since the development of BSC (Balanced Scorecard) [3,4]. Small construction firms also need to manage intangible assets as well as tangible assets. This study applies BSC to measure the management performance of small construction companies.

Kim (2010) interviewed the management officers of companies to develop a model for analyzing the management performance of large Korean construction firms. He adopted a program to feed back on goals, management strategies, management plans, and management performance evaluation to suggest the importance of effective management strategies and efficient management control. Additionally, he identified errors and misses when applied to real life, unlike the preceding studies, and analyzed the findings to suggest the general process of management performance for construction firms and the errors and solutions to consider when measuring performance [5]. Jung (2005) comparatively analyzed the weighted value according to the scale of companies to measure the management performance of construction firms. An AHP analysis was applied to calculate the weighted value of performance indicators and analyze their importance for small/medium versus large companies. The study of Jung is different from other studies in terms of the subject of

analysis, survey method, and findings [6]. Yu (2004) analyzed past cases of other countries to suggest that it is necessary to develop key performance indicators for the construction firms in Korea along with a PMS framework to control them [7]. In other countries, KPIs were applied to the PMS Islam Bank based on BSC and AHP [8], and BSC was used to improve the efficiency of company operations for Luka Koper and d.d. Company [9]. Among the top 1000 companies named by *Fortune*, a U.S. magazine on economics, about 60% are assumed to have adopted the concept of BSC [10–12].

Management diagnosis refers to hiring an outside management expert to address management issues that cannot be resolved internally or to identify directions for future development. Management diagnosis models are mostly used by consulting firms or individual companies [13]. The management diagnosis model for small construction firms in Korea is still in the theoretical development stage and the only available models are modified forms of generalized models.

The Korea Small Business Institute has suggested a corporate diagnosis model to select businesses for a small business support project. The model suggested indicators of diagnosis for categories including attractiveness, competence, systems, and CEO.

The government of Korea is also developing various evaluation indicators, such as "the Small/Medium Business Healthcare System" and the "INNO-BIZ Evaluation Model," with continued efforts to enhance the management control capabilities of small construction firms. Management performance was analyzed according to the scale of construction firms and characteristics of organization, and a model for management performance evaluation has been developed. Additionally, there have been continued efforts to develop management diagnosis models to enhance the management control capacities of small construction firms to enhance their competitiveness. However, the study of performance evaluation models for small construction firm management control has not been sufficient, as there are many limitations when applying the management diagnosis models developed for small construction firms.

Research has been conducted into the management of small and medium-sized companies in Korea, with differences in the target companies and objectives from this study [14]. We conducted a study to evaluate the management performance of small construction companies with fewer than 10 employees. There are differences between management diagnosis strategies and management performance evaluation models. Management diagnosis is the process of identifying problems with a company's management, identifying the cause of the problem, and deriving improvements to these problems. It is a good tool for improving current management problems and providing future management directions in corporate management. However, this differs in purpose and process from the assessment of management performance. In addition, BSC has been used to develop a framework for small/medium businesses and a performance control system has been constructed by a small nonprofit organization using BSC. As a result, BSC made it possible to search for and correct problems, but there is a limitation in that it cannot be used in many areas [15,16]. It is necessary to continue studying various models suitable for small companies [17,18]. Therefore, the current study suggested KPIs to measure the management performance of small construction firms in order to improve their competitiveness and pursue gradual corporate growth.

Kim proposed MAPEIs to evaluate the management performance of large construction firms. The MAPEIs are composed of the hierarchy and weighted value of BSC, performance, and KPIs to derive the management performance evaluation indicators of small construction firms. Figure 1 shows the basic structure of MAPEIs [19]. In addition, MAPEIs were established as a practical evaluation management system. This study applied the concept of measuring the management performance of construction firms.

**Figure 1.** Basic structure of MAPEIs [19].

#### **3. Methodology**

Management Performance Evaluation Indicators (MAPEIs) for small firms consist of evaluation indicators with various hierarchies and weighted values for each KPI. The management performance evaluation models for construction firms vary according to each firm's business and scale, knowledge informatization level, brand value in Korea or abroad, and soundness of management control. Therefore, it is necessary to provide appropriate indicators. The current study's MAPEIs, as mentioned above, may be applied to small construction firms in Korea. The study was limited to small construction firms in Korea with no more than 10 full-time employees, no construction projects abroad, and businesses not including civil works and plants.

Construction firms have many factors to consider when measuring performance due to the uncertainties in the market environment. Therefore, the current study applied a Balance Scorecard (BSC). A BSC consists of four areas—finances, customers, internal processes, and learning/growth—and is applied to the management control of firms in good standing in Korea and abroad [19]. The BSC of construction firms is the same as that of other companies, but performance and KPIs differ due to corporate characteristics. Management characteristics also vary according to the size of corporations, even if they are in the same industry. Therefore, the current study considered the characteristics of small construction firms to develop the hierarchy of MAPEIs. KPIs were derived by analyzing the characteristics of small construction firms, and all items for performance evaluation had weighted values. MAPEIs serve as KPIs to measure the management performance of small construction firms.

We selected the items used to evaluate performance (Figure 2) and deleted and supplemented items through expert interviews to configure the hierarchy. The survey was performed based on the hierarchy and the weighted values were tabulated by analyzing the importance of each item to derive the MAPEIs of small construction firms.

**Figure 2.** How to derive MAPEIs.

### **4. MAPEIs of Small Construction Firms**

#### *4.1. Selection of Management Performance Evaluation Indicators for Small Construction Firms*

The current study was conducted to derive KPIs for the management performance evaluation of small construction firms. Preceding studies analyzed the strategies, plans, and goals of construction companies to configure the performance areas of a subject to identify KPIs. Kim (2010) conducted a survey on managers of large construction companies and derived the following evaluation items and weights by an AHP analysis. This indicated the evaluation items of large construction firms that allow for systemized management control but cannot be KPIs of small construction firms. As shown in Table 1, we used the hierarchy of Kim (2010) as the preliminary indicators to derive MAPEIs [19].


**Table 1.** Weighted value of factors of MAPEIs [19].

The primary tier of the hierarchy consists of 14 performance areas and 31 KPIs. These are MAPEIs for large construction firms and cannot be indicators for the management performance evaluation of small construction firms. Therefore, there must be KPIs suitable for small construction firms. In order to configure a hierarchy for the purpose, the top management of five small Korean construction firms were interviewed and the findings are shown in Table 2. The respondents to this survey were top management who had managed construction companies for a long time.

**Table 2.** Details of survey subjects.


This study removed unnecessary KPIs following interviews. The interviews surveyed the items that realistically reflect the management performance of current companies among the items in the primary hierarchy. Tables 3–6 show the performance evaluation of small construction firms. The respondents selected items necessary for management evaluation. The score is the sum of the choices. For a maximum score of 5, all the respondents of the five companies analyzed deem that KPI is relevant. In fact, it is not easy for small construction firms to analyze management performance in various areas. Therefore, this study deleted items selected by fewer than half of the companies.

**Table 3.** Reflection of management performance evaluation on finance.


**Table 4.** Reflection of management performance evaluation on customers.



**Table 5.** Reflection of management performance evaluation on internal process.

**Table 6.** Reflection of management performance evaluation on learning and growth.


In the KPIs of preceding studies, finance consists of five areas, including profitability, growth, stability, activity, and order, as in Table 3. Profitability areas consist of ROIC (Return on Invested Capital), cost of sale ratio, and ordinary profit. Growth consists of an increase in revenue in Korea and increase in revenue abroad. Stability includes the debt ratio and achievement of collection goals, while activity includes the turnover ratio of total liabilities and net worth. Orders consist of amounts of new orders. In each area of performance, the number of new orders was selected as a major KPI by all five companies. Cost of sale ratio and achievement of collection goals were also representative. Achievement of collection goals was widely reflected, as poor collection is likely to lead to poor performance, inactivity, or unprofitability for small companies. On the other hand, ROIC and increase in revenues abroad are rarely representative. ROIC is a return on invested capital and may be evaluated based on the cost of sale ratio or ordinary profit as it is the actual assets invested in projects. This is mostly applied to companies where responsible management is possible, so it is difficult to use with small companies that lack systemized management control. The increase in revenues abroad is unrealistic for small construction firms that receive few orders from abroad.

Customers, as shown in Table 4, account for three performance areas, including satisfaction of external customers, satisfaction of internal customers, and market share. Satisfaction of external customers consists of awards won in competitions, customer satisfaction, corporate image, and social contribution, while satisfaction of internal customers is composed of employee transfer rate, work environment, and corporate culture. In the customer area, corporate image, customer satisfaction, employee transfer rate, work environment, and corporate culture are widely reflected. However, awards won in competitions and social contributions that have an additional impact on corporate image are not reflected

as much and the market share of orders abroad is also rarely reflected as small construction firms receive few orders from abroad, as seen in Table 3.

The internal process consists of three performance areas—investment in R&D, technology capacities, and work efficiency—as shown in Table 5. Investment in R&D consists of the cost of R&D to revenue, and the effect of new technology on the cost of development. Technology capacity consists of the application of internally developed technology and intellectual property rights, while work efficiency consists of selling and administrative expenses to revenue, compliance with guidelines, accident rates, and reuse/recycling of waste. The KPIs of the internal process were generally reflected less frequently than other areas were. On the other hand, the accident rate of efficiency area was widely reflected. This is because construction projects are generally large in scale and the losses related to accidents may be massive. Therefore, the accident rate is frequently applied to small construction firms.

Learning and growth, as shown in Table 6, consist of three performance areas: training, organizational capacity, and informatization. Manpower training includes index of excellent workforce, cost of training per employee, and satisfaction of trainees. Organizational capacity includes the knowledge sharing and productivity of employees, while informatization includes the informatization capacity index. In learning and growth, the productivity of employees was widely reflected. The number of employees is smaller than it is for large construction firms. Therefore, each member has a great impact on the organization, and the productivity of employees is significant. The index of excellent workforce is also frequently reflected because the competence of each individual employee is significant due to the smaller scale of firms. In a fast-paced business environment, informatization knowledge of construction is used as a strategic resource for the construction market and plays a major function. Therefore, the informatization capacity index is widely used for the evaluation of firms.

Based on the preliminary hierarchy, the top managers of firms were interviewed to survey the reflection of KPIs. In order to configure the evaluation indicators suitable for small construction firms based on the surveyed resources, the items that could not be assigned 3 points or more were deleted to configure the hierarchy. The secondary hierarchy of MAPEIs, configured based on the aforementioned standards, consisted of 13 performance areas and 18 KPIs.

However, there are many differences in management methods between large companies and small companies, and different sets of evaluation items apply for appropriate management control. In order to bridge the differences, the items that are considered most important by small construction firms for performance evaluation were assessed in addition to the evaluation indicators of large construction firms. Major MAPEIs of small construction firms included 10 indicators: net profit of construction projects, accident rate, complaint handling capacity, possibility of open bidding, construction performance rate, cost of construction, employees' task-processing capacity, revenue, gain, and accidentfree rate.

The items' similarity to the pre-existing evaluation indicators was analyzed through interviews with experts. Net profit of construction projects, cost of construction, revenue, and gain refer to the profitability of companies and overlap with the cost of sale ratio and the ordinary profit of profitability area under finance heading. The possibility of open bidding and construction performance rate are items that evaluate the profitability, growth, and number of orders of companies and are similar to the detailed items of finance. The accident rate was similar to the accident rate of the internal process area, while employees' task-processing capacity was similar to the index of excellent workforce in learning and growth. However, the complaint handling capacity, although it may be considered part of corporate image, was judged to be a new item for evaluating the management performance of small construction firms based on corporate characteristics.

The corporate image of large construction firms includes quality, brand, customer service, market reputation, stock prices, corporate value, and defects, as in Figure 3. These are auxiliary factors of corporate image for management performance evaluation and do not have a significant impact on performance evaluation. However, they may have a significant impact on small construction firms. In other words, the factors of corporate image can be a significant indicator for small construction firms. Therefore, complaintprocessing capacity was included in the work efficiency area of internal process as an indicator of performance evaluation.

**Figure 3.** Corporate image of large companies.

As mentioned above, the performance evaluation indicators were analyzed for deletion, modification, and supplementation. Then, the findings were used to derive the final hierarchy. The final hierarchy became the hierarchy of MAPEIs and consisted of 13 performance areas and 19 KPIs, as in Figure 4.

**Figure 4.** Hierarchy of MAPEIs.

## *4.2. Tabulation of Weighted Values*

Each item comprising the hierarchy of MAPEIs becomes an indicator for the management performance evaluation. However, not all performance areas and evaluation indicators have equal weight. Therefore, each item shall be assigned a weighted value to evaluate management performance by considering the weight of each indicator.

An AHP survey was performed to assign a weighted value to each item. The survey was comprised of an importance analysis of each BSC area, an importance analysis of the performance of each BSC area, and an importance analysis of KPIs of each performance area, and the overview of the survey is as shown in Table 7.


**Table 7.** Details of subjects for tabulation of weighted values.

The survey took about one month and the subjects were 33 members of top management or engineers of small construction firms. As the AHP survey was conducted, the consistency of responses was verified. The validity range of the consistency index was limited to 0.1 and the number of questions satisfying the consistency index was identified. The survey results satisfying the consistency index were analyzed for relative importance through a paired comparison analysis.

Figure 5 shows the weighted value of performance areas of BSC. The weighted value of the finance area was the highest at 0.379 and for the customer area it was 0.217. A weighted value of 0.198 was assigned to internal process and 0.206 to learning and growth. The importance of BSC of small construction firms was in the following order: finance, customers, learning and growth, and internal process. The highest weighted value of performance in finance was 0.115, assigned to profitability, followed by stability, orders, growth, and activity. Performance in the customers area assigned the highest value of 0.089 to satisfaction of external customers, followed by satisfaction of external customers, market share, and satisfaction of internal customers. Performance in the internal process area assigned 0.100 to work efficiency, which was a weighted value greater than that of technological capacity. The highest weighted value of 0.078 was assigned to the organizational capacity area in terms of performance on learning and growth, followed by manpower training and informatization. Table 8 lists the weighted values of all items tabulated through an importance analysis with AHP.

As mentioned above, the weighted value of finance was highest in BSC. In detail, profitability was assigned to the highest weighted value in finance, satisfaction of external customers in customers, work efficiency in internal process, and organizational capacity in learning and growth.

MAPEIs were compared between Tables 1 and 8. The importance of items for large construction firms is different from that for small construction firms. For large construction firms, the weighted value of customers was 0.34 and highest in BSC, followed by finance and learning and growth. However, the highest weighted value was 0.379 for finance, followed by customers, learning and growth, and internal process, for small construction firms. In the performance area of finance, the importance of orders and profitability was high for large construction firms, whereas the importance of profitability and stability was high for small construction firms. Unlike large construction firms, where orders are considered important with a weighted value of 0.32, small construction firms assigned greater importance to stability with a weighted value of 0.246 when the value of orders is

0.184. This is because stability is considered a very important indicator due to the constantly decreasing orders for small construction firms. This shows that the priority of performance evaluation indicators varies even for companies within the construction industry, according to their scale, management environment, and management characteristics.

**Figure 5.** Performance areas.



## *4.3. Derivation of MAPEIs for Small Construction Firms*

MAPEIs are the KPIs for management performance evaluation of small construction firms and consist of the hierarchy and weighted value of BSC, performance areas, and KPIs. There are four BSC areas, 13 performance areas, and 19 KPIs, and the weighted value of each item is as follows.

BSC-W refers to the weighted value of BSC, and the sum of BSC-W assigned to finance, customers, internal process, and learning and growth is 1. Performance-W refers to the weighted value of performance and is the product of BSC-W and the weighted value of performance, as in Equation (1). The sum of all weighted values of Performance-W is 1. KPI-W refers to the weighted values of KPI and is the product of Performance-W and the weighted value of KPIs, as in Equation (2). The sum of all weighted values of 'KPI-W' is 1.

$$\text{Performance} - W = \text{Weighted Value of Performance of BSC} - W \tag{1}$$

$$\text{KPI} - \text{W} = \text{Weighted Value of KPI of Performance} - \text{W} \tag{2}$$

MAPEIs are the most detailed items and the key indicators of management performance. Generally, BSC-W and Performance-W were highest in finance and profitability, so KPI-W would be highest for the items of finance. However, KPI-W was highest for possession of intellectual property rights in technology capacity at 0.098, as in Table 9. As KPI is applied to the hierarchy of performance, however, the items evaluating profitability were further categorized to reduce the weight of each item. The KPI of technology capacity applies to the possession of intellectual property rights only, whereas profitability was divided into two items of cost of sale ratio and ordinary profit. This implies that finance is important for evaluating the management performance of companies and the many evaluation indicators allow for accurate evaluation.



## *4.4. Evaluation of Applicability of MAPEIs (Small Construction Firms)*

The current study evaluated management performance to verify the applicability and necessity of MAPEIs. The subject applying MAPEIs was one small construction firm within the scope of study and three years' management performance was evaluated using a five-point scale. Figure 6 gives the MAPEI scores applying the weighted values and the MAPEI scores not applying the weighted values based on the evaluation results of the firm.

**Figure 6.** Results of management performance evaluation with/without weighted values of MAPEIs. (A) Results when not applying the weighted values to MAPEIs and assigning 60 points to 2017, 63.16 points to 2018, and 58.78 points to 2019; (B) results applying the weighted values to MAPEIs and assigning 61.34 points to 2017, 55.18 points to 2018, and 54.74 points to 2019. In (A), the management performance evaluation score of 2018 was 6.16%P lower than the previous year and showed a worsening of performance. The evaluation in (A) could not reflect the decline in management that was identified when analyzed by KPIs and importance (B). The management performance evaluation score of 2019 decreased by 8.42%P compared to the previous year in (A) but increased by 3.60%P with (B).

> The management performance of companies varied greatly according to the application of weighted values to MAPEIs. This is because the results are distorted by applying the same weight value to all items affecting the management of firms. When the same weight is applied to all items, the performance of items with minimal impact is exaggerated and the performance of items with greater impact is lessened, which can cause errors. In other words, critical situations that may have a negative impact on management may be misinterpreted as an improvement in management. Therefore, it is important to apply weighted values to the evaluation items for the accurate evaluation of management performance.

#### **5. Conclusions**

The construction market in Korea is constantly being depressed due to the poor management of construction firms in Korea, and this has a significant impact on management performance. However, most firms in the construction market are small and the impact on management performance is tremendous. Additionally, small construction firms lack sufficient management control systems, response to and preparation for changes in the management environment, and structure of organization to improve the management. Therefore, the current study derived the MAPEIs (Management Performance Evaluation Indicators) for small construction firms for management performance evaluation.

The current study applied the management performance evaluation indicators of large construction firms from preceding studies as preliminary indicators to derive MAPEIs. Five small construction firms in Korea were selected, and we interviewed the top management about the items that are realistically adopted by companies for management performance evaluation. A secondary hierarchy was created by analyzing the items surveyed, and items besides preliminary indicators were surveyed to finalize the hierarchy through deletion and supplementation. Complaint-handling capacity was added to the final hierarchy. This was derived from the corporate characteristics of small construction firms. The final hierarchy of MAPEIs consisted of 13 performance areas and 19 KPIs.

Not all performance areas and evaluation items of the final hierarchy have equal weight values. When the same weight value is applied to all items, the management performance of companies may be distorted. Therefore, the weight values of items shall be tabulated for accurate evaluation. An AHP survey was conducted for top management and engineers and the weight values of items were tabulated through a paired comparison. The survey involved analysis of BSC and performance areas of BSC and analysis of importance of KPIs of each performance area. As a result of the importance analysis, the highest values were applied to finance of BSC, profitability of performance, and possession of intellectual property rights of KPIs. This shows that a performance evaluation based on financial factors is more important than customer-centered performance for small construction firms.

MAPEIs are KPIs for the management performance evaluation of small construction firms and consist of the hierarchy and weighted values of BSC, performance areas, and KPIs. In order to verify the feasibility of MAPEIs, one small construction firm in Korea was selected for the applicability evaluation. The evaluation results varied according to the application of weighted values to MAPEIs and the need to apply weighted values to MAPEIs was confirmed as the management performance evaluation scores were distorted when the same weighted value was applied to all indicators.

The current study analyzed the characteristics of small construction firms and selected the evaluation items through an actual corporate survey to derive a weighted value for each item. Additionally, applicability was evaluated to verify the feasibility and applicability of MAPEIs. MAPEIs are fundamental to the study of management control in small construction firms; KPIs can be applied to construction companies with no more than 10 full-time employees. However, the items can be modified and supplemented to fit the characteristics of each company, and further studies and the development of performance evaluation systems for the performance evaluation of small construction firms to resolve the limitations would improve their management evaluation and achieve competitive management control. Therefore, the results of this study can be used as basic data not only for measuring management performance, but also for developing a system for the management of small construction firms. In addition, this study has limitations because it was conducted for construction companies in Korea. This study used the MEPAI model, which is the result of existing research on the management performance of construction companies. Not applying various models can be another limitation of this study.

**Author Contributions:** Conceptualization, D.K. and D.L.; methodology, D.K., and D.L.; software, D.L. and W.O.; validation, D.K., W.O., J.Y. (Jiyeong Yun), J.Y. (Jongyoung Youn), and S.D.; formal analysis, W.O.; investigation, J.Y. (Jiyeong Yun); resources, D.K.; data curation, J.Y. (Jiyeong Yun); writing—original draft preparation, D.K., W.O., and D.L.; writing—review and editing, D.K., J.Y. (Jongyoung Youn), W.O., J.Y. (Jiyeong Yun), S.D., and D.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2020R1C1C1012600).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Application of the Taguchi Method for Optimizing the Process Parameters of Producing Controlled Low-Strength Materials by Using Dimension Stone Sludge and Lightweight Aggregates**

**How-Ji Chen 1, Hsuan-Chung Lin <sup>1</sup> and Chao-Wei Tang 2,3,4,\***


**Abstract:** In view of the increasing concerns over non-renewable resource depletion and waste management, this paper studied the development of low-density controlled low-strength material (CLSM) by using stone sludge and lightweight aggregates. First, the investigation was performed at a laboratory scale to assess the effects of the composition on the properties of the resulting low-density CLSM. The Taguchi method with an *L*9(34) orthogonal array and four controllable three-level factors (i.e., the stone sludge dosage, water to binder ratio, accelerator dosage and lightweight aggregate dosage) was adopted. Then, to optimize the selected parameters, the analysis of variance method was used to explore the effects of the experimental factors on the performance (fresh and hardened properties) of the produced low-density CLSM. The test results show that when the percentage of stone sludge usage was increased from 30% to 60%, the initial setting time approximately doubled on average. Moreover, at the age of 28 days, the compressive strength of most specimens did not exceed the upper limit of 8.83 MPa stipulated by Taiwan's Public Construction Commission. Further, the material cost per cubic meter of the produced CLSM was about NT\$ 720.9 lower than that of the ordinary CLSM, which could reduce the cost by 40.6%. These results indicate that the use of stone sludge as a raw material to produce CLSM could achieve environmental sustainability. In other words, the use of stone sludge and lightweight aggregates to produce low-density CLSM was extremely feasible.

**Keywords:** stone sludge; lightweight aggregates; controlled low-strength materials; Taguchi method

## **1. Introduction**

The continuous progress of science and technology has improved social productivity and material standard of living, and has achieved unprecedented prosperity and development of human society. However, the widespread application of science and technology in various fields has also caused a certain degree of environmental pollution, ecological destruction, and resource scarcity, which may cause devastating potential threats to the entire earth of human life. According to the second edition of NACE [1], the total amount of waste generated by the economic activities of households and businesses in the EU in 2018 was 2.609 billion tons, and in 2008, the amount of recyclable waste was 202 million tons. In view of this, a key element of the EU's environmental policy is to manage waste in an environmentally sound manner and make full use of the auxiliary materials contained therein. Similarly, with the rise of interest in the concept of sustainable development and the awareness of environmental protection, it has become increasingly difficult to obtain

**Citation:** Chen, H.-J.; Lin, H.-C.; Tang, C.-W. Application of the Taguchi Method for Optimizing the Process Parameters of Producing Controlled Low-Strength Materials by Using Dimension Stone Sludge and Lightweight Aggregates. *Sustainability* **2021**, *13*, 5576. https://doi.org/ 10.3390/su13105576


Academic Editor: Sunkuk Kim

Received: 1 April 2021 Accepted: 15 May 2021 Published: 17 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

raw materials for concrete production in Taiwan. For these reasons, Taiwan promulgated (amended on 21 January 2009) the Resource Recycling Act [2] on 3 July 2002, to reduce the use of natural resources, reduce waste generation, promote the recycling and reuse of materials, reduce the environmental load and build a society in which resources can be used continuously. Currently, the Ministry of Economic Affairs of the Republic of China in Taiwan has announced 57 types of industrial waste for reuse, which can be divided into engineering, agricultural and other uses. Among them, there are 22 types of recycling for engineering purposes (such as waste casting sand, stone sludge, stone waste, coal ash, waste wood, waste glass, waste pottery, waste porcelain, waste bricks, waste tiles, etc.). These can be recycled for the production of civil engineering and construction materials.

According to the definition of the American Concrete Institute (ACI), controlled lowstrength materials (CLSM) is a self-filling, self-leveling and cementitious material that is mainly used to replace traditional backfill soil and structural fillers. Generally, the compressive strength of CLSM at 28 days is 8.3 MPa or lower [3]. Although CLSM's requirements for its constituent materials are not as strict as those of ordinary concrete materials, it provides suitable engineering characteristics at a lower cost for most needs and is in line with energy-saving and ecological benefits [4]. Essentially, CLSM has excellent rheological properties and anti-segregation ability, and it can be easily filled in a narrow excavation surface without any tamping. Moreover, CLSM has a low and sufficient loadbearing strength, which can facilitate future excavation [5–7]. Therefore, the application of CLSM has become more and more common in countries around the world and in Taiwan.

The stone processing industry is an important industry in Taiwan (with an annual output value of nearly 1.42 billion U.S. dollars), but the impact of its wastewater, solid waste and dust in the environment cannot be underestimated. In the cutting process of natural stone, a large amount of stone chips and waste will be produced. In order to cool the cutting saw blade, a large amount of cooling water must be used, resulting in a considerable amount of mud-like waste (a mixture wastewater and stone chips), which is called dimension stone waste mud. The dimension stone waste mud is discharged to the wastewater treatment plant, and the waste produced after sedimentation, separation and dehydration is called stone sludge. In Europe, the amount of annual stone sludge is estimated to be 5 million tons [8]. The annual output of Taiwan's stone sludge and stone waste exceeds 1.1 million tons [9], which in general comprises industrial wastes. However, due to the relatively low cost of the burial treatment of stone sludge or the poor marketing of resource-recycling products, the amount of resource treatment is quite limited, and the benefits of its resource application cannot be brought into full play.

Because concrete has the advantages of durability, fire resistance, high compression resistance, good shape ability, economy, etc., it is currently the most widely used man-made material. However, the production of cement is the main source of greenhouse gas emissions, which has severely damaged the earth's climate and environment and threatened the sustainable survival of mankind. From the perspective of maintaining the ecological sustainability of the natural environment and the overall economic benefits of society, actively developing alternative sources of construction materials to replace some of the traditional main sources is an important issue that cannot be delayed. In terms of the sustainable development of concrete materials, we can start with the design of materials and mix proportions. For example, the use of renewable resources to replace part of cement or aggregates will not only greatly contribute to energy saving and carbon reduction, but also improve the fresh property and durability of concrete materials. Martínez-García et al. [10] evaluated the viability of incorporating fine recycled concrete aggregates (FRCA) from urban demolition and construction waste for the manufacture of cement-based mortars. The results showed that the optimal percentage of substitution of fine natural aggregates for FRCA was 25% with respect to compressive and flexural strength tests. López Boadella et al. [11] analyzed the feasibility of using waste from a granite quarry to replace the micronized quartz in ultra-high-performance concrete (UHPC). The results showed that when the substitution rate was 35%, the flexural strength and tensile strength increased, and the values obtained

even for 100% substitution was acceptable. This confirmed that granite cutting waste instead of commonly used micronized quartz powder was a viable alternative to the expected more sustainable UHPC. Zamora-Castro et al. [12] reviewed the latest research on the performance of different sustainable concrete types. They recommend the use of standardized testing to ensure reliable results of the impact of sustainable materials on the physical and mechanical properties of concrete specimens. In addition, because recycled materials mixed into concrete showed high absorption capacity, it could cause workability problems of fresh concrete, thereby affecting its mechanical strength. Therefore, they recommend finding the best combination of materials from different sources to improve these properties of sustainable concrete. In addition, Chen et al. [13–15] used reservoir sediments, paper sludge and tile grinding sludge to produce lightweight aggregates, turning waste into renewable resources.

In order to achieve the purpose of waste reduction and resource reuse, renewable resources, and industrial wastes from all over the world have been used in large quantities in the production of CLSM, such as coal ash and new pozzolanic materials [16], cement kiln dust [17], stockpiled circulating fluidized bed combustion ashes [18], circulating fluidized bed combustion ash and recycled aggregates [19], bottom ash of municipal solid waste incinerator and water filter silt [20], solid wastes/byproducts from paper mills [21], circulating fluidized bed combustion ash [22,23], waste oyster shells [24], treated oil sand waste [25], alum sludge and green materials [26], waterworks sludge [27], water purification sludge [28], etc. Hung et al. [29] established a prediction model for the compressive strength and surface resistivity of controlled low-strength desulfurization slag. They suggested that expanding the use of unqualified raw materials and man-made waste as secondary raw materials could be one of the most important directions for creating a waste-free process to ensure the most reasonable use of natural resources and reduce the negative impact on environmental conditions. Park and Hong [30] analyzed the influence of the mixing conditions of wastepaper sludge ash (WPSA) on the strength and bearing capacity of controlled low-strength materials (CLSM). The results showed that CLSM and WPSA could be used as backfill materials for sewage pipes, which could ensure higher stability compared with soil backfill.

In the past, the disposal cost of stone sludge in Taiwan was relatively low, and the related reuse or volume reduction technology has not received much attention from the industry. However, due to the lack of natural resources and the increasing difficulty of finding waste disposal sites, stone sludge should be prioritized for resource utilization. On the other hand, the composition of stone sludge and stone waste is not significantly different from the parent stone processed [31], and its output is huge, representing a large resource that can be recycled and reused with great resource potential and economic value. If stone sludge replaces the fine particles in CLSM as an alternative material source, the cost of raw materials can be reduced. In addition, replacing ordinary aggregates with lightweight aggregates can produce low-density CLSM, which can reduce the load on the underground structure and the soil or filler below it.

In view of the above, the development of low-density CLSM by using stone sludge and lightweight aggregates was explored in this study. First, the investigation was performed on a laboratory scale to assess the effects of the composition on the properties of the resulting low-density CLSM. The Taguchi method with an *L*9(34) orthogonal array and four controllable three-level factors (i.e., stone sludge, water/binder ratio, accelerating agent and lightweight aggregate content) was adopted. Then, in order to optimize the selected parameters, the analysis of variance method was used to explore the effects of the experimental factors on the performances (fresh and hardened properties) of the produced low-density CLSM. This study confirmed that the use of stone sludge and lightweight aggregates to produce low-density CLSM was extremely feasible. Especially, in view of the various engineering requirements of CLSM, the Taguchi method could be used to optimize the process parameters of using size stone sludge and lightweight aggregates to produce controlled low-strength materials.

## **2. Materials and Methods**

## *2.1. Materials*

A large amount of stone sludge in Taiwan has created an environmental burden. Therefore, this study uses stone sludge in the production of CLSM, and its purpose is to treat stone sludge through recycling and reuse to avoid secondary pollution. The materials used in this study included cement, ground-granulated blast-furnace slag, water, ordinary fine aggregates, lightweight aggregates, stone sludge, accelerating agent and air-entraining agent. The cement used was Type I Portland cement with a specific gravity of 3.15 produced by Taiwan Cement Corporation. The ground-granulated blast-furnace slag was produced by CHC Resources Corporation and its specific gravity was 2.9. Ordinary fine aggregates were taken from the nearby sand and gravel plant, with a specific gravity of 2.6 and a water absorption rate of 1.1%. Lightweight aggregates were purchased from Ming Chun Ceramic Corporation, and their specific gravity was 1.6. Two kinds of stone sludge were taken from Stone and Resource Industry Research and Development Center in eastern Taiwan, and their chemical composition is shown in Table 1. These stone sludge is characterized by a very fine size distribution, which is mainly made up of the same compounds as the processed stones. Since the particle size of marble stone sludge was larger than that of granite stone sludge, its particle size distribution is shown in Figure 1. Therefore, this study used marble stone sludge with a specific gravity of 2.6 to replace part of the ordinary fine aggregates. The accelerating agent and air-entraining agent were purchased from Guanghui Building Materials Company, in line with the Chinese national standards or the American Society for Testing and Materials specifications, and their specific gravity was 1.05.


**Table 1.** The composition of the stone sludge initially examined for research purpose.

**Figure 1.** Grain size distribution of the used marble stone sludge.

## *2.2. Experimental Design*

This study aimed to produce low-density CLSM using stone sludge and lightweight aggregates. The main component of stone sludge was suspended solids in the raw water—that is, fine sand particles with a specific gravity between 2.6 and 2.8. Because the particle size of stone sludge was relatively fine, this study used it as a filler to replace part of the fine aggregates.

Depending on the application field, such as backfill, utility bedding, void fill, and bridge approach, the important characteristics a CLSM must have are different. The mix proportions of the ingredients in a CLSM mixture depends on the required properties of the CLSM in two states, namely the plastic state and the hardened state. Basically, a CLSM mixture is made by mixing cementitious materials, aggregates, and water in a designed ratio. If necessary, chemical admixtures or mineral admixtures can be used to ensure that the CLSM meets the requirements of fluidity, setting time, and low strength. According to the setting time of CLSM, its mix design is divided into early-strength type and general type. The early-strength CLSM is developed for pipeline projects in urban areas and traffic hubs. It can solve urban construction troubles and keep traffic flow; in addition, it can effectively ensure project quality and construction safety. For conventional CLSM, it is developed for projects that do not require early-strength, non-emergency, and more cost considerations. It is more suitable for pipeline projects in suburbs and industrial areas. Compared with traditional backfilling methods, it can not only improve the quality and safety of general backfilling projects, but also is very economical.

The use of different types and sources of renewable resources is the most important factor that affects the water demand of CLSM. In addition, the water-cement ratio and the type of renewable resources are variables that affect the compressive strength of CLSM. Because CLSM contains a large number of materials that exceed traditional specifications, there is currently no unanimously accepted mix design method. Therefore, based on previous experience and trial and error, we screened out four test variables that needed to be investigated; namely, the percentage of the stone sludge to replace the fine aggregates, the water–binder ratio, the percentage of the accelerating agent and the dosage of the lightweight aggregates. After the experimental control factors were selected, the level of each factor was set to keep the level value within a reasonable range as far as possible. The evaluation indicators of the test included the initial setting time, final setting time, slump, slump flow, unit weight, air content and compressive strength of the CLSM produced.

The main performance characteristics of CLSM include high fluidity and a controllable low strength, and the characteristics of the constituent materials and their proportion in the mixture are the main parameters that affect the performance of CLSM. So far, there is no standard method for CLSM ratio design. The different control levels of each control factor are shown in Table 2 to explore the performance of each control factor level combination

on the characteristics of CLSM and evaluate the best treatment. The literature showed that renewable resources with fine particle size could be used to replace fine aggregates, and the replacement percentage was usually between 0–90% [14,17,21]. Due to the high-water absorption characteristics of stone sludge, this study believes that the percentage of stone sludge to replace fine aggregates should not be too high. Under the condition of a fixed amount of sand (1450 kg per cubic meter), we planned a total of three stone sludge dosages, and the weight percentages of the stone sludge to replace the fine aggregates were 0%, 30% and 60%. Of these, the proportion of 0% was the control group, and the other proportions made up the experimental group. In addition, for economic cost considerations, the amount of cement was fixed at 125 kg per cubic meter, the amount of ground-granulated blastfurnace slag was fixed at 50 kg per cubic meter, the amount of accelerating agent was between 2% to 4% of the binder content and the amount of air-entraining agent was fixed at 1% of the binder content. It is worth mentioning that the use of accelerating agent was to enhance the early strength of CLSM. However, among various raw materials, the unit price of accelerating agent and air-entraining agent was relatively expensive, so their usage was less.


**Table 2.** Factors and design levels for test mixtures.

Under the condition of four factors and three levels for each factor, if the full factor experiment was carried out, the scale of the experiment would be very large (with 34 experimental combinations). In this study, an experimental orthogonal array *L*9(34) was selected to arrange the test plan, as shown in Table 3. Then, through the use of range analysis and analysis of variance, it was possible to quickly analyze the effect factors that had a significant impact on the experimental characteristic indexes among many factors to determine the factor combination that could obtain the best characteristic indexes.

**Table 3.** Orthogonal array (*L*9(34)) for test mixtures.


Notes: \* The numbers in parentheses indicate the level of the factor. AA = accelerating agent; LWA = lightweight aggregate.

#### *2.3. Mix Proportions Design*

According to the experimental combination of CLSM in the orthogonal design table, the amount of each constituent material was calculated as shown in Table 4. According to the design of the aforementioned various combinations, a horizontal twin-shaft mixer was used to mix CLSM. After the mixing of each group of CLSM was completed, the fresh properties (slump, slump flow, setting time, unit weight and air content) were measured

and recorded; after that, ninety-one cylindrical test body (100 mm in diameter and 200 mm in height) was cast. The specimens were disassembled according to the planned age and then placed in a water bath for curing. They were not taken out until the day before the test age for the compressive strength test.


**Table 4.** Mix proportions of test mixtures.

Notes: LWA = lightweight aggregate; FA = fine aggregate; AA = accelerating agent; AE = air-entraining agent.

#### *2.4. Test Methods and Data Analysis*

All the mixtures were evaluated in terms of their initial setting time, final setting time, slump, slump flow, unit weight, air content and compressive strength. The slump of the mixture was measured using Chinese National Standard (CNS) 1176 [32], while the slump flow of the mixture was measured using CNS 14842 [33]. The time of setting the mixture was measured using American Society for Testing and Materials (ASTM) C403/C403M-16 [34]. The unit weight and air content of the fresh mixture were measured using ASTM D 6023 [35]. The compressive strengths of the hardened mixture were measured at curing ages of 12 h, 24 h, and 28 days, respectively. Three specimens from each mixture were tested for compressive strength. Only one specimen was taken for other test items. The preparation and testing of the mixture specimens were in accordance with CNS 1232 [36].

In the process of discussing optimization, the most important factor is to find the objective function that can best express the quality characteristics. For example, maintaining the overall average value of the product close to the set value or reducing the variation between products can be used as an objective function to improve quality. During the product life cycle, the total price paid by the entire society is called quality loss. The less the quality loss, the higher the quality. In Taguchi's quality concept, parameter design is the most important step to achieve high-quality and low-cost goals. Taguchi believes that the quality characteristics should be concentrated around the target value, and the further away from the target value, the greater the loss. Dr. Taguchi used the loss function of the quadratic curve to measure quality characteristics [37]. When the quality characteristic completely met the target value, the quality loss was zero. When the quality characteristic deviated from the target value, the quality loss increased at the speed of a quadratic curve. The quality loss function (*L*) is a second-order function, which is defined as: "The quality loss is equal to the square of the difference between the actual value and the target value, multiplied by a quality loss coefficient." It is a criterion for evaluating the quality of a product. Its mathematical function can be expressed as follows [37]:

$$L(y) = k(y - m)^2 \tag{1}$$

where *L* = quality loss function; *y* = quality characteristic; *m* = target value; *k* = quality loss coefficient. The total quality loss can be calculated as follows:

$$Total\ quality\ loss = \sum\_{i=1}^{n} k(y\_i - m)^2\tag{2}$$

where *y*<sup>i</sup> = test value; *n* = measurement or observation times; *m* = target value.

There are three standard types of calculation loss function: smaller-the-better, largerthe-better, and nominal-the-better. The signal-to-noise ratio (*S*/*N*) is an important evaluation index in Taguchi's quality engineering approach. According to different quality characteristics, the calculation of the *S*/*N* ratio (*η*) can be divided into three types [37,38]:

$$\text{Smaller} - \text{the} - \text{better} : \eta = -10 \times \log \left( \frac{1}{n} \sum\_{i=1}^{n} y\_i^2 \right) \tag{3}$$

$$\text{Larger} - \text{the} - \text{better} : \eta = -10 \times \log \left( \frac{1}{n} \sum\_{i=1}^{n} \frac{1}{y\_i^2} \right) \tag{4}$$

$$\text{Nominal} - \text{the} - \text{better} : \eta = -10 \times \log \left( \frac{1}{n} \sum\_{i=1}^{n} (y\_i - y\_0)^2 \right) \tag{5}$$

where *n* is the number of repetitions or observations, *yi* is the observed data, and *y*<sup>0</sup> is the nominal value desired.

In this study, the objective functions of the mixture specimens were set separately according to engineering requirements. The objective function of the initial setting time, final setting time and unit weight was a smaller-the-better type, while the objective function of the slump, slump flow, air content and compressive strength was a larger-the-better type. On the other hand, the statistical range analysis method was used to explore the relationship between CLSM properties and various control factors. The range was used to represent the measures of variation in statistical data. The difference between the maximum value and the minimum value was taken, which was the data equal to the maximum value minus the minimum value. The range analysis benefits from a simple calculation and intuitiveness and was simple and easy to understand; thus, it is the most used method for the analysis of orthogonal test results. The range analysis process for the test results included the following steps: calculating the range of each factor, determining the order of importance of the factors, drawing a trend graph of the factors and indicators, and determining the optimal level and the optimal level combination of the test factors.

Moreover, an analysis of variance (ANOVA) was used to detect the optimization of the observed values. This was accomplished by separating the total variation of the *S*/*N* ratios into contributions by each of the process parameters and the error [39]. In other words, the total variation can be decomposed into two parts: variation due to changes in various factors, and variation due to experimental errors. The different experimental values measured under exactly the same process conditions are all attributed to "experimental error", or simply "error". Analytically, the total sum of square deviation (*SST*) of the *S*/*N* ratio can be calculated as [37]:

$$SS\_T = \sum\_{i=1}^{n} (\eta\_i - \eta\_{\text{m}})^2 \tag{6}$$

where *n* is the number of experiments in the orthogonal array; *η<sup>i</sup>* is the mean *S/N* ratio for the *i* th experiment; and *η<sup>m</sup>* is the grand mean of the *S/N* ratio. Then, the sum of squares of the measured parameter *Z* (*SSZ*) can be calculated as [37]:

$$SS\_Z = \sum\_{j=1}^{r} \frac{Z\_j^2}{t} - \frac{1}{n} \left(\sum\_{i=1}^{n} \eta\_i\right)^2 \tag{7}$$

where *Z* represents one of the tested parameters; *j* is the level number of parameter *Z*; *r* is the number of levels of parameter *Z*; *t* is the number of repetitions of each level of parameter *Z*; and *Zj* is the sum of the *S*/*N* ratio involving parameter *Z* and level *j*. The sum of squares of the error parameter (*SSe*) can be calculated as follows [37]:

$$SS\_t = SS\_T - SS\_F \tag{8}$$

where *SSF* represents the sum of squared deviations due to each parameter.

On the other hand, the application of statistical *F*-test can determine which process parameters have a significant impact on performance characteristics. In order to perform the *F* test, it is necessary to calculate the average of the squared deviation (variation) due to each process parameter and error term, as shown below [37]:

$$\text{MS}\_{\mathbb{Z}} = \text{SS}\_{\mathbb{Z}} / df\_{\mathbb{Z}} \tag{9}$$

$$MS\_{\varepsilon} = SS\_{\varepsilon} / df\_{\varepsilon} \tag{10}$$

Among them, *MSZ* is the average value of the square deviation attributed to the parameter *Z*; *dfZ* is the degree of freedom of the parameter *Z*; *MSe* is the average value of the square deviation due to the error term; and *dfe* is the degree of freedom of the error term. Then, the *F* value of the parameter *Z* (*FZ*) can be calculated according to the following formula [37]:

$$F\_{\overline{Z}} = MS\_{\overline{Z}} / MS\_{\mathfrak{e}} \tag{11}$$

In an orthogonal array experiment, when the *F* value of a control factor is large, it means that the control factor is influential (important). The corrected sum of squares (*SS*∗ *Z*) can be calculated as follows [37]

$$SS\_{\overline{Z}}^{\*} = SS\_{Z} - MS\_{\varepsilon} \times df\_{Z} \tag{12}$$

Finally, the percentage contribution of parameter *Z* (*PZ*) can be calculated as follows [37]:

$$P\_{\mathbb{Z}} = S S\_{\mathbb{Z}}^{\*} / S S\_{\mathbb{T}} \tag{13}$$

*PZ* can be used as a simple indicator to represent the influence of a factor's change, so it can be used as an indicator of the "importance" of a factor.

## **3. Results and Discussion**

#### *3.1. Fresh Properties of Tested Mixtures*

The fresh properties (slump, slump flow, setting time, air content and unit weight) of mixture specimens are shown in Table 5. As with ordinary concrete, the fresh properties of CLSM were also affected by its composition. It can be seen from Table 5 that the fresh properties of the prepared CLSM were a function of the amount of stone sludge, the water–binder ratio, the accelerating agent, and the lightweight aggregate. The slump of tested mixtures was between 2.0 and 22.5 cm, as shown in Figure 2. Of these, the slump of the M4 mixture was the smallest, and the slump of the M9 mixture was the largest. On the other hand, the slump flow of tested mixtures was between 20 and 38 cm, as shown in Figure 2. Of these, the slump flow of the M2 and M4 mixtures was the smallest, and the slump flow of the M9 mixture was the largest. Regarding the setting time of tested mixtures, the initial setting time was between 159 and 600 min, and the final setting time was between 396 and 1855 min, as shown in Figure 3. Of these, the initial setting time of the M4 mixture was the shortest, the initial setting time of the M3 mixture was the longest, the final setting time of the M6 mixture was the shortest, and the final setting time of the M3 mixture was the longest. In terms of the air content of tested mixtures, its value was between 2.4% and 5%. The air content of the M1 and M2 mixtures was the smallest, and the air content of the M5 mixture was the largest, as shown in Figure 4. In terms of the unit weight of tested mixtures, its value was between 1961–2346 kg/m3. The unit weight of the M3 mixture was the smallest, and the unit weight of the M4 mixture was the largest, as shown in Figure 4. From this point of view, the M1–M3 and M9 mixtures can be seen to produce low-density CLSM.


**Table 5.** Fresh properties and the corresponding signal-to-noise ratios of tested mixtures.

**Figure 2.** Slump and slump flow test results of tested mixtures.

**Figure 3.** Setting time test results of tested mixtures.

**Figure 4.** Air content and unit weight test results of tested mixtures.

On the other hand, Kaliyavaradhan et al. [40] pointed out that researchers who used different types of wastes with high water absorption rates as fine aggregate substitutes

in CLSM often observed negative effects on the fresh properties. For example, the test results showed that when the percentage of stone sludge usage was increased from 30% to 60%, the initial setting time approximately doubled on average. In other words, the setting time increased with the increase of stone sludge dosage in the CLSM mixture. In order to further understand the appropriateness of the properties of the prepared CLSM specimens, a comparison was made with the test results of other researchers. Although different researchers used different composition materials, the test results were still available for reference, as shown in Table 6. On the whole, the results of the fresh properties of this study were included in the scope of the existing literature. Taking the slump flow as an example, the results of this study ranged from 20 to 38 cm, while the literature range was 0 to 65 cm. This means that the various properties of the produced CLSM were within a reasonable range. From this point of view, it is feasible to produce CLSM from stone sludge.


**Table 6.** Comparison of test results and literature.

In order to analyze the effect of each factor, we calculated the average value of the *S*/*N* ratio of each factor at the same level, and then calculated the main effect value (delta) of the factor level. In this way, these data could be made into an auxiliary table as shown in Table 7. If the main effect value of one of the factors was larger, it meant that the influence of the factor on the whole system was greater, and the quality of improvement was also greater. It can be seen from Table 7 that the influence of A1 (i.e., the level of A factor was under the condition of the first level) was reflected in the experimental combination of M1–M3, the influence of A2 was reflected in the experimental combination of M4–M6, and the influence of A3 was reflected in the experimental combination of M7–M9. In Table 7, delta represents the calculated range; that is, the difference between the maximum value and the minimum value in levels 1–3. In the same way, the influence of the water–binder ratio, the accelerator dosage and the lightweight aggregate dosage on the performance parameters can be analyzed. In principle, the greater the delta value, the greater the influence of the level change of the factor on the performance parameters; that is, the more important the factor. On the other hand, a statistical method, ANOVA, was used to further explore the test results. The results of the ANOVA for the fresh properties of tested mixtures are shown in Table 8. The total variation represents the possible total quality loss. In addition, the *F* value and *P* value obtained under a confidence level of 95% and the contribution percentage of each parameter are also listed in the table, which represents the proportion of the variation of the factor to the total quality loss.


**Table 7.** Range analysis for fresh properties of tested mixtures.

Note: *S/N* = signal-to-noise ratio; LWA = lightweight aggregate.

**Table 8.** Analysis of variance and *F* test for fresh properties of tested mixtures.



**Table 8.** *Cont.*

Note: LWA = lightweight aggregate.

## 3.1.1. Setting Time

For the setting time of tested mixtures, the objective function was a smaller-the-better type. From the range analysis results in Table 7, in order to shorten the initial setting time of tested mixtures, the order of importance of the control factors was the stone sludge dosage (factor A), the lightweight aggregate dosage (factor D), the accelerator dosage (C factor) and the water–binder ratio (B factor); the corresponding delta values were 8.707, 2.238, 1.419 and 0.323, respectively. Moreover, the *S*/*N* response graph for the initial setting time of tested mixtures is shown in Figure 5. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant increase in the initial setting time. It can be seen from the test results that the initial setting time was increased by approximately 2.7 times on average. In addition, when the percentage of stone sludge usage was increased from 30% to 60%, the *S*/*N* ratio decreased significantly, which reflected a significant increase in the initial setting time. It can be seen from the test results that the initial setting time approximately doubled on average. Therefore, the stone sludge dosage had the greatest impact and was the main factor. On the other hand, from the ANOVA results in Table 8, the most significant factor affecting the initial setting time was the stone sludge dosage, and its contribution percentage (*PZ*) was 90.95%. The purpose of the prepared CLSM is to emphasize the need to shorten the initial setting time; that is, the shorter the initial setting time, the better. According to the experimental results, the optimal combination is A2B3C1D2, and the shortest initial setting time was 159 min. However, the best combination estimated by the range analysis and ANOVA was A2B1C2D2, i.e., stone sludge dosage at level 2, water–binder ratio at level 1, accelerator dosage at level 2, and lightweight aggregate dosage at level 2.

In terms of shortening the final setting time of the tested mixtures, the range analysis results in Table 7 show that the order of importance of the control factors was the stone sludge dosage, the accelerator dosage the accelerator dosage, the lightweight aggregate dosage, and the water–binder ratio; the corresponding delta values were 11.267, 1.694, 1.686 and 1.153, respectively. In addition, the *S*/*N* response graph for the final setting time of tested mixtures is shown in Figure 6. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant increase in the final setting time. It can be seen from the test results that the final setting time was

increased by approximately 2.7 times on average. In addition, when the percentage of stone sludge usage was increased from 30% to 60%, the *S*/*N* ratio decreased significantly, which reflected a significant increase in final setting time. It can be seen from the test results that the final setting time increased by approximately 2.7 times on average. From this point of view, the stone sludge dosage (factor A) had the greatest impact and was the main factor, while the accelerator dosage was the secondary factor. On the other hand, from the ANOVA results in Table 8, the most significant factor affecting the final setting time was the stone sludge dosage, and its contribution percentage (*PZ*) was 93.55%. The purpose of the prepared CLSM is to emphasize the need to shorten the final setting time. That is to say, the shorter the final setting time, the better. According to the experimental results, the optimal combination was A2B3C1D2, and the shortest final setting time was 396 min. In contrast, the best combination estimated by range analysis and ANOVA was A2B2C2D2, i.e., stone sludge dosage at level 2, water–binder ratio at level 2, accelerator dosage at level 2, and lightweight aggregate dosage at level 2.

**Figure 5.** Signal-to-noise (*S*/*N*) response graph for the initial setting time of tested mixtures.

**Figure 6.** Signal-to-noise *(S*/*N*) response graph for final setting time of tested mixtures.

### 3.1.2. Slump and Slump Flow

For the slump and slump flow of tested mixtures, the objective function was a largerthe-better type. From the results of the range analysis in Table 7 when increasing the slump of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the water–binder ratio (factor B), the lightweight aggregate dosage (factor D)

and the accelerator dosage (factor C); the corresponding delta values were 13.835, 5.233, 4.614 and 3.447, respectively. In addition, the *S*/*N* response graph for the slump of tested mixtures is shown in Figure 7. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio increased, which reflected an increase in the slump. It can be seen from the test results that the slump was increased by approximately 1.6 times on average. In addition, When the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio increased significantly, which reflected a significant increase in the slump. In addition, as the water–binder ratio increased, the *S*/*N* ratio increased significantly, which reflected a significant increase in the slump. Therefore, the stone sludge dosage (factor A) had the greatest impact and was the main factor, while the water–binder ratio (factor B) was the secondary factor. Furthermore, from the ANOVA results in Table 8, the most significant factor affecting the slump was the stone sludge dosage, and its contribution percentage (*PZ*) was 71.57%. The purpose of the prepared CLSM is to emphasize the need for increasing slump. The larger the slump, the better. According to the experimental results, the optimal combination was A3B2C2D1, and the maximum slump was 22.5 cm. However, the best combination estimated by range analysis and ANOVA was A3B3C1D1, i.e., stone sludge dosage at level 3, water–binder ratio at level 3, accelerator dosage at level 1, and lightweight aggregate dosage at level 1.

**Figure 7.** Signal-to-noise (*S*/*N*) response graph for slump of tested mixtures.

It can be seen from Table 7 that when increasing the slump flow of tested mixtures, the order of importance of the control factors was the stone sludge dosage (factor A), the water–binder ratio (factor B), the lightweight aggregate dosage (factor D) and the accelerator dosage (factor C). Moreover, the *S*/*N* response graph for the slump flow of tested mixtures is shown in Figure 8. When the use percentage of stone sludge was reduced from 30% to 0%, the signal-to-noise ratio did not change, which reflected that there was basically no change in the slump flow. In addition, when the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio increased significantly, which reflected a significant increase in the slump flow. Moreover, as the water–binder ratio increased, the *S*/*N* ratio increased significantly, which reflected a significant increase in the slump flow. Therefore, the stone sludge dosage (factor A) had the greatest impact and was the main factor, while the water–binder ratio (factor B) was the secondary factor. Furthermore, from the ANOVA results in Table 8, the most significant factor affecting the slump was the stone sludge dosage, and its contribution percentage (*PZ*) was 71.57%. The purpose of the prepared CLSM is to emphasize the need for increasing slump. The larger the slump, the better. According to the experimental results, the optimal combination was A3B3C2D1, and the maximum slump flow was 38 cm. Coincidentally, the best combination estimated by range analysis and ANOVA was also A3B3C2D1.

**Figure 8.** Signal-to-noise (*S*/*N*) response graph for slump flow of tested mixtures.

#### 3.1.3. Air Content and Unit Weight

For the air content of tested mixtures, the objective function was a larger-the-better type. From the range analysis results in Table 7, in order to increase the air content of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the accelerator dosage (factor C), the water to binder ratio (factor B) and the lightweight aggregate dosage (factor D); the corresponding delta values were 3.782, 3.302, 1.548 and 0.774, respectively. Moreover, the *S*/*N* response graph for the air content of tested mixtures is shown in Figure 9. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the air content. Moreover, when the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the air content. In addition, as the accelerator dosage increased, the *S*/*N* ratio increased significantly, which reflected a significant increase in the air content. Therefore, the stone sludge dosage had the greatest impact and was the main factor, while the accelerator dosage was the secondary factor. On the other hand, from the ANOVA results in Table 8, the most significant factor affecting the air content was the stone sludge dosage, and its contribution percentage (*PZ*) was 49.74%. The purpose of the prepared CLSM is to emphasize the need to increase the air content; that is, the larger the air content, the better. According to the experimental results, the optimal combination was A2B2C3D1, and the largest air content was 5%. However, the best combination estimated by range analysis and ANOVA was A2B1C3D3, i.e., stone sludge dosage at level 2, water–binder ratio at level 1, accelerator dosage at level 3, and lightweight aggregate dosage at level 3.

For the unit weight of tested mixtures, the objective function was a smaller-the-better type. From the range analysis results in Table 7, in order to reduce the unit weight of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the lightweight aggregate dosage (factor D), the water–binder ratio (factor B), and the accelerator dosage (factor C); the corresponding delta values were 1.342, 0.286, 0.175 and 0.168, respectively. Moreover, the *S*/*N* response graph for the unit weight of tested mixtures is shown in Figure 10. When the use percentage of stone sludge decreased from 30% to 0%, the S/N ratio increased significantly, which reflected a significant decrease in the unit weight. In addition, when the percentage of stone sludge usage increased from 30% to 60%, the S/N ratio increased significantly, which reflected a significant decrease in the unit weight. Therefore, the stone sludge dosage had the greatest impact and was the main factor, while the lightweight aggregate dosage was the secondary factor. On the other hand, from the ANOVA results in Table 8, the most significant factor affecting the unit weight was the stone sludge dosage, and its contribution percentage (*PZ*) was 90.95%. The purpose of the prepared CLSM is to emphasize the need to reduce the unit weight; that

is, the lighter the unit weight, the better. According to the experimental results, the optimal combination was A1B3C3D3, and the lightest unit weight was 1961.4 kg/m3. However, the best combination estimated by range analysis and ANOVA was A1B3C1D1, i.e., stone sludge dosage at level 1, water–binder ratio at level 3, accelerator dosage at level 1, and lightweight aggregate dosage at level 1.

**Figure 9.** Signal-to-noise (*S*/*N*) response graph for air content of tested mixtures.

**Figure 10.** Signal-to-noise (*S*/*N*) response graph for unit weight of tested mixtures.

## *3.2. Compressive Strength of Tested Mixtures*

The test results of the compressive strength of mixture specimens are shown in Table 9. It can be observed from the table that the compressive strengths of tested mixtures at 12 h, 1 day and 28 days of age were between 0.02–0.62, 0.2–1.49, and 1.37–9.86 MPa, respectively. In addition, it can be seen from Figure 11 that the compressive strength of the M1–M3 mixtures without stone sludge at each age was generally lower, and the 28-day compressive strength of the age was between 1.37–2.13 MPa. The weight percentage of the stone sludge to replace the fine aggregates in the M4–M6 mixtures was 30%. These mixtures had a higher compressive strength at all ages, and the 28-day compressive strength was between 8.56–9.86 MPa. The weight percentage of the stone sludge to replace the fine aggregates in the M7–M9 mixtures was 60%. Their compressive strength at each age was roughly between the first two groups, and the compressive strength at 28 days was between 3.88 and 6.61 MPa.


**Table 9.** Compressive strengths and the corresponding signal-to-noise ratio of tested mixtures.

Note: Data in parentheses indicate standard deviation.

**Figure 11.** Compressive strength test results of tested mixtures.

On the other hand, in many cases, CLSM must be designed to have a strength equivalent to the surrounding soil after hardening, so that it can be used for future maintenance and excavation operations. Therefore, the excavability of hardened CLSM in the late age is an important consideration for many projects. In order to further understand the appropriateness of the harden properties of the prepared CLSM specimens, a comparison was made with the research results of other researchers, as shown in Table 6. It can be seen from Table 6 that the compressive strength results of this study were included in the scope of the existing literature. Taking the one-day compressive strength as an example, the results of this study ranged from 0.20 to 1.49 MPa, while the literature range was 0.04 to 1.67 MPa. In order to facilitate the subsequent construction of the paving layer, most of Taiwan's CLSM regulations require that the early strength of CLSM used for road trench backfilling be greater than 0.69 MPa. The results of the M4–M8 specimens in this study all met this requirement. As for the 28-day compressive strength, except for the M4 specimen, the 28-day compressive strength of the remaining specimens did not exceed 8.83 MPa, which was consistent with most of Taiwan's CLSM regulations. From this point of view, the compressive strength range of CLSM produced by using stone sludge could meet the needs of engineering operations.

The corresponding *S*/*N* ratios of the compressive strength of mixture specimens is shown in Table 9. Then, the *S*/*N* ratios data in Table 9 are compiled into Table 10 to analyze the impact of each level of various experimental control factors on the compressive strengths of tested mixtures. On the other hand, the results of the ANOVA of the compressive strengths of tested mixtures are shown in Table 11.


**Table 10.** Range analysis for compressive strengths of tested mixtures.

Note: LWA = lightweight aggregate.

**Table 11.** Analysis of variance and *F* test for compressive strengths of tested mixtures.


Note: LWA = lightweight aggregate.3.2.1. Twelve-Hour Compressive Strength.

For the compressive strength of tested mixtures, the objective function was a largerthe-better type. From the range analysis results in Table 10, in order to increase the 12-h compressive strength of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the water–binder ratio (factor B), the accelerator dosage (factor C) and the lightweight aggregate dosage (factor D); the corresponding delta values were 17.384, 8.908, 7.744 and 5.521, respectively. Moreover, the *S*/*N* response graph for the compressive strength of tested mixtures is shown in Figure 12. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the 12-h compressive strength. Moreover, when the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio decreased, which reflected a decrease in the 12-h compressive strength. In addition, as the water– binder ratio increased, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the 12-h compressive strength. Therefore, the stone sludge dosage had the greatest impact and was the main factor, while the water–binder ratio was the secondary

factor. On the other hand, from the ANOVA results in Table 11, the most significant factor affecting the 12-h compressive strength was the stone sludge dosage, and its contribution percentage (*PZ*) was 60.22%. The purpose of the prepared CLSM is to emphasize the need to increase the 12-h compressive strength; that is, the larger the compressive strength, the better. According to the experimental results, the optimal combination was A2B3C1D2, and the 12-h compressive strength was 0.62 MPa. However, the best combination estimated by range analysis and ANOVA was A2B1C1D1, i.e., the stone sludge dosage at level 2, water–binder ratio at level 1, accelerator dosage at level 1, and lightweight aggregate dosage at level 1.

**Figure 12.** Signal-to-noise (*S*/*N*) response graph for the 12-h compressive strength of tested mixtures.

## 3.2.1. One-Day Compressive Strength

From the range analysis results in Table 10, in order to increase the one-day compressive strength of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the water–binder ratio (factor B), the accelerator dosage (factor C) and the lightweight aggregate dosage (factor D). The corresponding delta values were 12.905, 4.466, 2.518, and 1.842, respectively. This result is consistent with the result of the 12-h compressive strength. Moreover, the *S*/*N* response graph for the one-day compressive strength of tested mixtures is shown in Figure 13. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the one-day compressive strength. Moreover, when the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in one-day compressive strength. In addition, as the water–binder ratio increased, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in one-day compressive strength. Therefore, the stone sludge dosage had the greatest impact and was the main factor, while the water–binder ratio was the secondary factor. On the other hand, from the ANOVA results in Table 11, the most significant factor affecting the one-day compressive strength was the stone sludge dosage, and its contribution percentage (*PZ*) was 82.97%. The purpose of the prepared CLSM is to emphasize the need to increase the one-day compressive strength. That is to say, the larger the one-day compressive strength, the better. According to the experimental results, the optimal combination was A2B1C2D3, and the compressive strength was 1.49 MPa. However, the best combination estimated by range analysis and ANOVA was A2B1C1D2, i.e., stone sludge dosage at level 2, water–binder ratio at level 1, accelerator dosage at level 1 and lightweight aggregate dosage at level 2.

**Figure 13.** Signal-to-noise (*S*/*N*) response graph for one-day compressive strength of tested mixtures.

#### 3.2.2. Twenty-Eight-Day Compressive Strength

From the range analysis results in Table 10, in order to increase the 28-day compressive strength of CLSM, the order of importance of the control factors was the stone sludge dosage (factor A), the water–binder ratio (factor B), the lightweight aggregate dosage (factor D) and the accelerator dosage (factor C); the corresponding delta values were 14.347, 2.702, 2.020 and 0.345, respectively. Moreover, the *S*/*N* response graph for the 28-day compressive strength of tested mixtures is shown in Figure 14. When the use percentage of stone sludge decreased from 30% to 0%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the 28-day compressive strength. Moreover, when the percentage of stone sludge usage increased from 30% to 60%, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the 28-day compressive strength. In addition, as the water–binder ratio increased, the *S*/*N* ratio decreased significantly, which reflected a significant decrease in the 28-day compressive strength. Therefore, the stone sludge dosage had the greatest impact and was the main factor, while the water–binder ratio was the secondary factor. On the other hand, from the ANOVA results in Table 11, the most significant factor affecting the 28-day compressive strength was the stone sludge dosage, and its contribution percentage (*PZ*) was 94.1%. The purpose of the prepared CLSM is to emphasize the need to increase the 28-day compressive strength. That is to say, the larger the 28-day compressive strength, the better. According to the experimental results, the optimal combination was A2B1C2D3, and the compressive strength was 9.86 MPa. However, the best combination estimated by range analysis and ANOVA was A2B1C1D2, i.e., stone sludge dosage at level 2, water–binder ratio at level 1, accelerator dosage at level 1, and lightweight aggregate dosage at level 2.

## *3.3. Confirmation Test*

The results of this study show that in order to meet the requirements of most CLSM regulations in Taiwan that the early strength of CLSM used for road trench backfilling needs to be greater than 0.69 MPa, the percentage of stone sludge to replace fine aggregates can be 30%. However, in order to facilitate future maintenance and excavation, the late strength development of CLSM should not be too great. As the percentage of stone sludge substituted for fine aggregates increased from 30% to 60%, the compressive strength of CLSM at all ages decreased, and the 28-day compressive strength did not exceed 9 MPa. This indicates that stone sludge is a viable material for CLSM.

For various performance parameters, the experimental optimal combination and estimated optimal combination of CLSM prepared by this research are shown in Table 12. Of these, the experimental optimal combination of slump fluidity was consistent with the estimated optimal combination. The experimental optimal combination of other perfor-

mance parameters was inconsistent with the estimated optimal combination. In order to verify that the best combination of experimental control factors can be obtained using the Taguchi method, four sets of confirmation test combinations were planned as shown in Table 12. It can be seen from Table 13 that the initial setting time of the confirmation test combination A2B1C2D2 was 143 min, which was shorter than the optimal combination of the experiment. The unit weight of the confirmation test combination A1B3C1D1 was 1950.2 kg/m3, which was less than the optimal combination of the experiment. The 12-h compressive strength of the confirmation test combination A2B1C1D1 was 0.78 MPa, which was higher than the experimental best combination. The one-day compressive strength of the confirmation test combination A2B1C1D2 was 1.52 MPa, which was higher than the experimental best combination. The confirmation test results showed that the optimal combination of experimental control factors proposed by the Taguchi method could obtain the best results for the performance parameters.

**Figure 14.** Signal-to-noise (*S*/*N*) response graph for 28-day compressive strength of tested mixtures.




## *3.4. Cost Analysis of Production CLSM with Stone Sludge and Lightweight Aggregates*

This research aimed to produce and characterize low-density CLSM using stone sludge. There were four parameters selected in this study and each had three levels, which were the amount of stone sludge (0%, 30%, and 60%), the water–binder ratio (0.9, 1.0 and 1.1), the amount of accelerator (2%, 3%, and 4%) and the amount of lightweight aggregate (250, 300 and 350 kg per cubic meter). Considering the performance of the fresh properties, compressive strength, utilization rate of renewable resources and economy, the following mixture design of the large-scale production CLSM containing stone sludge was selected: a water–binder ratio of 1.1, an amount of accelerator agent of 3%, replacement of 60% fine aggregates with stone sludge and a lightweight aggregate content of 250 kg per cubic meter. These mixture proportions had a higher amount of stone sludge, and the mechanical performance could also meet the requirements of CLSM.

Table 14 shows the unit price of various materials required for the production of CLSM. The processing fee of stone sludge is currently NT\$1000–1500 per ton because it is a general industrial waste. Therefore, the unit price of stone sludge in the table is NT\$-1.0/kg, which means that the cost can be reduced. The unit price of general commercial CLSM in Taiwan varies from region to region, and the price per cubic meter applied to public works ranges from NT\$1230 to NT\$2200. Table 14 shows the mix proportions and material cost analysis of ordinary CLSM and stone sludge CLSM. The material cost per cubic meter of ordinary CLSM is about NT\$1777.3. The material cost of stone sludge CLSM is NT\$697.8 per cubic meter. The material cost per cubic meter of stone sludge CLSM is about NT\$1079.5 lower than that of ordinary public works CLSM, which can reduce the cost by 60.7%. In terms of economy, the production of CLSM from stone sludge is indeed quite competitive.

## *3.5. The Sustainability Effect of CLSM with Stone Sludge and Lightweight Aggregates*

In Taiwan, it is quite common to use slices of stone such as marble or granite as decoration materials for building structures. However, the stone sludge produced by stone slicing in Taiwan exceeds one million metric tons every year. The sludge produced by cutting is extremely small and has a high-water content, which causes great trouble in the terminal disposal. At present, only a small part of it was recycled into raw cement, and most of it became earthwork. The amount of stone sludge that has not been reused and piled-up is already huge, which has caused an environmental burden in Taiwan. On the other hand, Taiwan's sludge treatment regulations are becoming stricter, and the cost of sludge disposal is gradually increasing. Therefore, stone sludge must be managed from the source, recycled and reused, and strengthened in the final treatment and management, so that the stone sludge can be reduced, stabilized, harmless, and recycled, thereby avoiding secondary pollution.

Although the definition of sustainability has not yet been consistent, it is usually defined as the process and action by which humans avoid the depletion of natural resources to maintain ecological balance without reducing the quality of life in modern society. Especially, it is composed of economy, society, and environment. Taiwan is small in area, densely populated, and limited land resources. In the face of the difficulty of finding disposal sites, reducing waste production and recycling resources utilization has become an important direction for current disposal. For this, in order to achieve the purpose of sustainable environmental management, the disposal and reuse methods of stone sludge need to be developed urgently.

Since CLSM is mostly used in backfill sites that require re-excavation, the required strength is not high for the purpose of continuous re-excavation. In view of this, the Taiwanese government encourages the use of recycled materials to manufacture CLSM. In this study, the back-end waste (i.e., stone sludge) produced by stone cutting was reused as a renewable material in CLSM. Its appearance is not much different from the general CLSM, as shown in Figure 15. From the test results and economic cost analysis, the use of stone sludge to produce CLSM could not only achieve waste recycling, but also had economic feasibility. These results also showed that waste reduction and resource disposal could be achieved, and the number of fine aggregates used could also be reduced. In other words, the production of CLSM with stone sludge as a raw material could reduce the impact of carbon emissions on the environment, thereby achieving the goal of environmental sustainability.


**Table 14.** Unit price of controlled low-strength material (CLSM) materials and cost analysis of CLSM.

Notes: AA = accelerator agent; LWA = lightweight aggregate; CA = coarse aggregate; FA = fine aggregate; WRA =

water-reducing

 agent; AE = air-entraining agent.

**Figure 15.** Appearance of CLSM samples: (**a**) CLSM produced by stone sludge and (**b**) CLSM produced by reservoir sediments.

#### **4. Conclusions**

In this study, dimension stone sludge was used to replace fine aggregates, and lightweight aggregates were used to replace ordinary coarse aggregates to explore the feasibility of producing low-density CLSM. The test results showed that the use of stone sludge and lightweight aggregates to produce low-density CLSM was extremely feasible. Based on the aforementioned test results and analysis, the following conclusions can be drawn.

The increase in the percentage of stone sludge to replace fine aggregates prolonged the setting time. Moreover, at the age of 28 days, the compressive strength of most specimens did not exceed the upper limit of 8.83 MPa stipulated by Taiwan's Public Construction Commission.

In view of the various engineering requirements of CLSM, the Taguchi method can be used for optimizing the process parameters of producing controlled low-strength materials by using dimension stone sludge and lightweight aggregates. When reducing the unit weight of CLSM, the order of importance of the control factors was the stone sludge dosage, the lightweight aggregate dosage, the water–binder ratio, and the accelerator dosage. Moreover, the ANOVA results showed that the most significant factor affecting the unit weight was the stone sludge dosage, and its contribution percentage was 90.95%.

For the improvement of the 12-h compressive strength of CLSM, the order of importance of the control factors was the stone sludge dosage, the water–binder ratio, the accelerator dosage and the lightweight aggregate dosage. In addition, the ANOVA results showed that the most significant factor affecting the compressive strength was the stone sludge dosage, and its contribution percentage was 60.22%.

Considering the performance of the fresh properties, compressive strength, utilization rate of renewable resources and economy, the following mixture design of the large-scale production CLSM containing stone sludge was selected: a water–binder ratio of 1.1, an amount of accelerator agent of 3%, replacement of 60% fine aggregates with stone sludge, and a lightweight aggregate content of 250 kg/m3. These mixture proportions had a higher amount of stone sludge, and the mechanical performance could also meet the requirements of CLSM. The material cost per cubic meter of stone sludge CLSM is about NT\$1079.5 lower than that of ordinary public works CLSM, which can reduce the cost by 60.7%. In economic terms, the production of CLSM from stone sludge is indeed quite competitive.

**Author Contributions:** Conceptualization, H.-J.C. and C.-W.T.; methodology, H.-J.C. and C.-W.T.; software, C.-W.T.; validation, H.-C.L. and C.-W.T.; formal analysis, C.-W.T.; investigation, H.-C.L. and C.-W.T.; resources, C.-W.T.; data curation, H.-J.C. and C.-W.T.; writing—original draft preparation, H.-J.C. and C.-W.T.; writing—review and editing, C.-W.T.; visualization, C.-W.T.; supervision, C.-W.T.; project administration, C.-W.T.; funding acquisition, C.-W.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Science and Technology of Taiwan grant number MOST 108-2622-E-230-003-CC3.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors are grateful to the Department of Civil Engineering of National Chung-Hsing University for providing experimental equipment and technical support.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Classifying the Level of Bid Price Volatility Based on Machine Learning with Parameters from Bid Documents as Risk Factors**

**YeEun Jang, JeongWook Son and June-Seong Yi \***

Department of Architectural & Urban Systems Engineering, Ewha Womans University, Seoul 03760, Korea; jyee@ewha.ac.kr (Y.J.); jwson@ewha.ac.kr (J.S.)

**\*** Correspondence: jsyi@ewha.ac.kr; Tel.: +82-2-3277-4454

**Abstract:** The purpose of this study is to classify the bid price volatility level with machine learning and parameters from bid documents as risk factors. To this end, we studied project-oriented risk factors affecting the bid price and pre-bid clarification document as the uncertainty of bid documents through preliminary research. The authors collected Caltrans's bid summary and pre-bid clarification document from 2011–2018 as data samples. To train the classification model, the data were preprocessed to create a final dataset of 269 projects consisting of input and output parameters. The projects in which the bid inquiries were not resolved in the pre-bid clarification had higher bid averages and bid ranges than the risk-resolved projects. Besides this, regarding the two classification models with neural network (NN) algorithms, Model 2, which included the uncertainty in the bid documents as a parameter, predicted the bid average risk and bid range risk more accurately (52.5% and 72.5%, respectively) than Model 1 (26.4% and 23.3%, respectively). The accuracy of Model 2 was verified with 40 verification test datasets.

**Keywords:** risk management; risk analysis; bid price volatility; uncertainty in bid documents; prebid clarification document; machine learning (ML), classification model; public project; sustainable project management

## **1. Introduction**

Sustainability refers to the whole life cycle from siting to design, construction, operation, maintenance, renovation, and deconstruction [1,2]. Traditional research focused on the design and construction stages to maximize profits has gradually expanded to the entire life cycle of construction projects to realize sustainable development. Accordingly, many researchers have conducted valuable studies to minimize the impact on the environment by improving the energy performance of buildings and reducing waste. As a result, many studies on sustainability have developed remarkably around maintenance and subsequent steps. Recently, this trend has been further expanded to realize the results of many studies conducted so far [3]. Therefore, many studies have refocused on project management, which corresponds to the preceding stages in terms of sustainability [4]. The success or failure of a construction project starts from the initial stage of the project. More precisely, the feasibility at the bid and contract phases, stipulating plans for the future, enables the completion of a sustainable project.

Contracts for construction projects are created based on competitive bids. In general, the bidder who offers the lowest price is selected as the final winner. Therefore, determining the final price is crucial for bidders [5]. It is also difficult because the bid price affects the likelihood of gaining a satisfactory profit and winning the project [6]. The client provides a bid document to the bidders, who then examine it to estimate the bid price. Thus, the bid document plays an essential role in determining the bid price. If the content of the bid document is uncertain, the intention of the construction object may be ambiguous and cause mistakes during the construction phase, which may lead to construction rework and

**Citation:** Jang, Y.; Son, J.; Yi, J.-S. Classifying the Level of Bid Price Volatility Based on Machine Learning with Parameters from Bid Documents as Risk Factors. *Sustainability* **2021**, *13*, 3886. https://doi.org/10.3390/ su13073886


Academic Editor: Sunkuk Kim

Received: 28 February 2021 Accepted: 25 March 2021 Published: 1 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

claims [7,8]. Thus, bidders include the cost in the bid price to cover those risks. In general, the bid price can be expressed as follows (Equation (1)):

$$B\_{\rm i} = \mathbb{C}\_{\rm i} (1 + M\_{\rm i}),\tag{1}$$

*Ci* represents the construction cost and *Mi* is the markup (i.e., contingency), which means the risk cost due to uncertainty in the bid document [9]. In other words, if the risk cost increases owing to uncertainty in the bid documents, the bid price increases [10].

The uncertainty factor in bid documents that causes the risk cost must be determined and investigated. However, because reviewing all bid documents in a limited time frame is difficult [11], businesses often rely on their experience rather than quantitative uncertainty measurements [12–15]. Because uncertainty in bid documents is affected by complex factors that are difficult to measure quantitatively, most qualitative research studies have been conducted in academia. Therefore, determining the risk cost remains a tough challenge [16–20]. Bid prices that are not adequately set negatively affect bidders, clients, and users alike. Bidders take the risk to a severe degree, which not only does not yield the expected return but can lead to more serious financial difficulties. Simultaneously, such a bid price may increase the cost of completion due to frequent design changes during project execution, increasing the burden on the client. Eventually, the project quality completed by this process could be worse, causing great inconvenience to users.

Although reviewing all documents may be challenging, pre-bid clarification documents contain much more uncertain information than other documents. This document type includes inquiries and answers from bidders and clients about the uncertainty factors in the bid documents; this information can be used as an input parameter for a machine learning-based model to construct a bid price. This study aims to examine whether the uncertainty measured in pre-bid clarification documents affects the bid price. This uncertainty may change the mean value or variance of the bid price. In this paper, these two changes are operatively defined as "bid price volatility."

In this study, a sample of data from the California Department of Transportation (Caltrans) in the US was used to see how uncertainty in the tender document changes bid price volatility. Analyzing the uncertainty of the bidding document is very difficult. In particular, the volume of bidding documents is enormous because of construction project size, making analysis difficult. Crucially, construction project data has an unstructured text format, making quantitative analysis even more difficult. However, the authors solved this problem using the pre-bid clarification document, which inquired about this uncertainty as a proxy, and used it with the bid summary. This study suggests that the uncertainty of the bid document affects the change in the bid price volatility. This allows bidders to execute the project at a reasonable price between earning a profit and winning the project. Further, this reasonable price can improve the project performance and realize the client's satisfaction. More ultimately, it can extend sustainability in terms of the life cycle of the project.

## **2. Preliminary Research: Project Risk in Bid Phase and Uncertainty in Bid Documents** *2.1. Definition of Project Risk in Bid Phase*

The definition of risk depends on the subject and purpose in a field. Because risk is a concept defined to quantify the uncertainty regarding danger, it differs from the latter; it is defined as the "possibility of loss or injury" in dictionaries. In academia, the risk is more clearly defined as a factor or condition that can cause loss or injury owing to uncertainty; this definition focuses more on the possibility of risk rather than the risk itself, typically expressed in the following equation [21]:

$$Risk\text{ }Magnituude = Risk\text{ }Severity \times \text{Risk}\text{ }Likelihood,\tag{2}$$

Risk magnitude is one of several attempts to measure risk [16]. It is a useful indicator for determining priorities among various risk factors using "risk" and "uncertainty" as variables. However, this also acts as a limitation because relative comparisons between risk factors are possible, but absolute comparisons are impossible. Therefore, the quantitative relationship with the bidding price resulting from the risk in the actual bidding process is blurred. This suggests that a new indicator that can reflect the risk of bidding price is necessary for at least construction projects, and there have been many studies related to this. Abotaleb and El-Adaway [9] attempted to measure bidding risk as a percentage of markups. Besides the total construction cost, bidders present the total construction cost plus a specific rate as the bid price for pursuing profit while preparing for risks. In addition, a study was conducted to determine whether the successful bid price was a price that had more risk than necessary by using the contrast between the successful bid price and the average bid price [22]. Lee et al. [23] attempted to measure the bid risk by using an equation similar to the equation of Williams [22], but in which the engineer's estimate replaced the successful bid price. However, the previously suggested equations have limitations in that they are challenging to use in this study in the following aspects. First, it is a matter of the possibility of utilizing the markups. It is correct that the contingency is included in the price, but a third party such as researchers other than bidders cannot check from the bid history. This is because the contingency is included in one or more of several bid items of the bill of quantities (BOQ). Second, the successful bid price and the average bid price are values determined after the bidding date. It is difficult to predict similar projects' risks during the bid phase using these values.

In this study, the risk is defined as the quantitative uncertainty regarding risk, and the risk factor represents a factor that causes uncertainty regarding time, cost, and quality risk. We use two metrics that match our definition of risk in Section 4.1.2. The scope of this study covers construction projects, and the project risk corresponds to the uncertainty regarding risks that arise from the characteristics of the construction project. The bid varies according to the project delivery method in actual construction projects. In this study, the bid phase is considered as the period in which the construction bid is made with the design–bid–build method. Uncertainty in a bid document is one of the many project risk factors in the bid phase.

#### *2.2. Project Risk Factors Affecting the Bid Price*

Several researchers have studied project risk factors that affect the bid price. Construction projects can be classified into several types depending on the case, and any project type can include risk. Therefore, researchers have analyzed the risk without considering the project type. In this study, risk factors that affect the bid price are extracted from 13 reviews in the field of transportation (Table 1).

The above studies are of great significance in that they have substantially advanced the critical risk identification stage in risk management. Many studies have extracted common factors as considerations when bidding for projects. Existing studies have facilitated more detailed risk management by deriving or breaking down the priorities of risks to be considered when performing projects based on surveys of most experts. On the other hand, some studies have analyzed how the number of bidders affects the bid price and predicts the bid price through simulation using multiple variables instead of one variable. However, there is a limitation in not considering how the project risk is integrated into the project's initial bid price.


**Table 1.** Project Risk Factors Affecting Bid Price According to Previous Studies.

## *2.3. Uncertainty in Bid Document as a Project Risk Factor*

Uncertainty in bid documents is one of the most crucial risk factors. The bid phase is the first stage of a project contract. The bidder submits the bid price after reviewing the extensive bid documents, which contain information on the following three aspects: (1) the bid procedure (e.g., the announcement, guide, participation application, participation notice, and bid), (2) contract (e.g., the general and special conditions), and (3) construction (e.g., the drawings, specification, and pre-bid clarification document). Each bid document has a different scope and form. For example, the specification document contains a set of documented requirements and the drawings that present the building requirements. The special conditions are contract clauses that apply only to the project subject to the contract; they are created by changing, adding, or deleting existing content in the General Conditions section. In other words, the bid documents present standards and procedures regarding the design, construction method, materials, and inspection for the completion of the construction object; thus, the bid documents constitute the basis for calculating the bid price. In addition, because bid documents are contract documents, they are the basis for judgment when legal problems arise in the future. Cost overrun can occur if the bidder fails to review the bid document's risk factors in advance [23]. Therefore, analyzing the uncertainty in bid documents is crucial.

Discrepancies, errors, and omissions cause uncertainty in bid documents; these are the leading causes of legal adjustment, arbitration, or litigation regarding the project costs. According to Tanaka [33], 74.4% of construction-related claims in the United States are due to uncertainty in bid documents; Erdis and Ozdemir [34] studied the dispute between a client and bidder, arguing that uncertain expressions in a bid document could lead to construction disputes.

Public projects in the US include a pre-bid clarification procedure that can resolve all uncertainty in bid documents before the bid. If bidders find uncertainty in a bid document during the quotation, they can contact the client, who must respond within the deadline. Relieving all uncertainty in bid documents through this approach helps bidders present the correct bid price [35]. New Work State in the US emphasized that the pre-bid clarification is a significant procedure for the client and bidder [36]. The former can calculate the project cost more accurately with less uncertainty. Pre-bid clarification is an institutional method that helps present accurate project costs and prevents possible future design changes, extensions, additional construction costs, and disputes [37].

## *2.4. Pre-Bid Clarification Document as a Proxy for Uncertainty in Bid Documents*

Uncertainty in bid documents includes (1) unclear communication caused by discrepancies, errors, or the omission of information or (2) unclear requirements regarding the project object. In general, the bid process for a construction project involves many bid documents [23]. Each document may independently contain risk factors; besides this, they may interact with each other, creating risk factors. Hence, all bid documents must be carefully reviewed to determine the uncertainty level. However, it is complicated for bidders to identify all hidden risks within a short bid preparation time [16–20,23]. Therefore, in the actual field and academia, the uncertainty of bid documents has been considered a complex problem to solve [11,38] and risk beyond control [28].

The uncertainty that arises in the pre-bid clarification procedure is caused by factors, which occur in all the bid documents that the bidders read. These documents are incorporated into the pre-bid clarification document, which can serve as a proxy variable that gauges the entire bid document's uncertainty. For example, Daoud and Allouche [39] analyzed pre-bid clarification documents to examine which uncertainty factors occur in the bid documents of construction projects.

## *2.5. Hypothesis Development*

From the literature review, there is a widely believed proposition: uncertain things during the bid phase affect bid price on the theoretical plane. However, the problem is that the factors classified as risks are mixed with what can be measured and what is not, what can be controlled and what is not possible, making quantitative analysis impossible. For this reason, when practitioners calculate prices, these uncertainties are guessed and reflected in prices without a factual basis. We made the following two assumptions to establish the hypothesis: (1) As the uncertainty increases, the bidders will reflect this in their prices, causing an increase in the overall average bid price. (2) The greater the uncertainty, the more significant the difference in prices offered by bidders will also increase, resulting in an increase in the range of bidding prices formed. Under these assumptions, we set up the following two hypotheses on the empirical plane.

**Hypothesis 1 (H1).** *Factors derived from bid summary and pre-bid clarification document affect F*1*(x), representing the volatility of bid price.*

**Hypothesis 2 (H2).** *Factors derived from bid summary and pre-bid clarification document affect F*2*(x), representing the volatility of bid price.*

In H1 and H2, the factors consist of seven independent parameters obtained from the bid summary and pre-bid clarification document. Then, *F*1(x) of H1 becomes Bid Average Risk (Equation (3)), and *F*2(x) of *H*<sup>2</sup> becomes Bid Range Risk (Equation (4)) discussed in Section 4.

## *2.6. Research Gaps and Research Questions*

According to the Project Management Body of Knowledge [40], risk management research is based on (1) risk identification, (2) risk assessment, and (3) risk plan and control. The risk assessment, which is a leading step in risk planning and control, quantifies the

potential impact of these uncertain factors [11]. However, many variables to be considered and interrelated make the analysis in the actual field and research studies difficult [14,24].

The fundamental reason is risk identification (which is a leading step); the general approach is to subdivide all project risk factors into controllable units based on specific criteria. Analyzing segmented risks can reduce uncertainty in the bid phase [37]. However, this approach has not been both quantitatively and qualitatively studied for risk factors in bid documents [23] because they contain vast amounts of information and differ in content depending on the project. Therefore, uncertainty in bid documents has been classified as an uncontrollable risk [28]. As mentioned in Section 2.2, researchers have only progressed to Level 1 by suggesting that uncertainty in bid documents is a risk factor; there are no sufficient specific studies on Level 2 [23]. In other words, published management studies have mainly focused on high-level risk factors and surveys with expert groups [11,38]. However, these data [41] only serve as references for determining the bid price in the bid phase.

The following factors must be investigated: first, regarding the social background, a reasonable bid price is crucial for establishing a reasonable project budget for the bidder and client [20]. This requires a decision support tool that can be used by practitioners who encounter difficulties in the bid price prediction. In research, a more quantitative study based on actual bid data is required to assess whether uncertainty in bid documents affects bid prices. This study aims to meet both academic and practical needs by analyzing whether uncertainty in bid documents affects the bid price.

Uncertainty in a bid document is expected to have the following effects on the bid price. First, each bidder will represent this risk factor in his/her bid price, thereby increasing the project's overall bid price (i.e., the bid price average). Besides this, the other bidders represent this risk factor in prices, which increases the range of the established bid price bands. In this study, these two X are defined as "bid average risk" and "bid range risk," respectively. Further, the bid price volatility comprises the two types of bid price fluctuations due to uncertainty in a bid document (i.e., the increase in the bid price average and range). This study aims to provide answers to the following two questions:


The results of this study are two types of bid price volatility level classification models:


In this study, the performance of Model 2 is evaluated to support decision-making about bid prices.

#### **3. Materials and Methods: Modeling Approach**

Regarding risk management, this study on assessment is different from risk plan and control, which supports decision-making on participation in the bid phase. This is because the decision-making process regarding bid prices of bidders who have already decided to participate is supported in this study. In general, risk assessment studies can be classified into studies of *Bi* and studies of *Mi* (Equation (1)). Because it is difficult to collect and analyze sufficient data, mainly *Mi* has been studied; in this study, *Bi* is empirically evaluated based on the actual bid results. In addition, in other published studies, the uncertainty factors of bid documents were analyzed with a proxy (i.e., a pre-bid clarification document).

Uncertainty factors in bid documents are natural phenomena because construction projects are typically one-off projects. Considering the toxin clause that partially exists in the special conditions, an uncertainty factor in bid documents is problematic because the artificial content clearly defines who should be responsible in certain circumstances. Therefore, the uncertainty factors in bid documents considered in this study are limited to those that occur naturally because of specific characteristics of the construction industry.

#### *3.1. Materials*

The data from the bid results regarding the project risk factors discussed in Section 2.2 are the variables of interest. To study their effects, the data in which the influence of other factors can be minimized should be analyzed. The public construction project of Caltrans meets this purpose because of the following reasons: first, the uncertainty in the bid documents can be analyzed. Because Caltrans includes a pre-bid clarification process in the bid phase, the pre-bid clarification document is publicly available. Second, the quantity and quality of available project data are sufficient. Caltrans invests approximately \$1.7 billion per year in approximately 450 projects, which is the largest of the 50 US states. The thousands of standardized project datasets of Caltrans have led to large amounts of high-quality data and excellent project management capabilities based on experience. Third, the absence of special conditions reduces influences from other than the variables of interest. Standard contracts used worldwide include the FIDIC (Fédération Internationale Des Ingénieurs-Conseils), JCT (Joint Contracts Tribunal), NEC (New Engineering Contract), and AIA (American Institute of Architects), mainly applied to private projects. By contrast, Caltrans uses federal-aid construction contracts (FHWA-1273) for public projects. In this case, only general conditions without special conditions (unlike private projects) are applied, which means that the projects are relatively standardized. The bid document of a standardized project reduces the influences of numerous external factors; it is considered suitable for observing the effect of uncertainty in bid documents on the bid price because of the absence of special conditions.

Caltrans has published all the bid results online since 2004 (they provide all bid documents, bid summaries, and important information). However, the online services for pre-bid clarification documents have been operated since 2011. The number of projects since the access date was 3584 during 2011–2018. In total, 3578 datasets were collected (six cases were excluded because they could not be accessed owing to system errors).

### *3.2. Methods*

*Pre-Data Analysis (Data Preprocessing Based on Bid Summary and Pre-Bid Clarification Document)*: in this step, information that can be obtained from the bid result is preprocessed into input parameters (IPs) and output parameters (OPs). Caltrans has published a bid summary containing the critical details of the bid results. In this study, the data related to project risk factors affecting the bid price (which was discussed in Section 2.2) are extracted from the bid summary (B. S.) and pre-bid clarification document (P. C. D.). Subsequently, the final dataset is constructed from the raw data.

*Data Analysis (Two Classification Models of Bid Price Volatility Based on Machine Learning)*: Methods of analyzing data can be classified into several categories depending on the purpose of the study and the characteristics of the data. When analyzing data that is large and composed of various factors, techniques such as data mining through machine learning (ML) are mainly used, and the data mining method is actively applied in recent risk analysis studies [42]. Such data mining can be classified mainly into a prediction technique that derives a regression equation, such as statistical analysis, and a classification technique that determines the category of data. Therefore, this study uses a machine learning-based data mining classification technique as a data analysis technique to classify the level of bidding risk with a large amount of data composed of various variables. In this study, machine learning can be used to classify the OPs in data consisting of multiple IPs. In this study, the class of the OP is designated such that the model algorithm learns to classify the levels of bid price volatility with MATLAB. To evaluate the model performance during training and validate it through validation tests in a post-data analysis, the pre-data analysis's final dataset is classified into training and validation data (for the training validation and validation test, respectively). As a result, Model 1 (which does not include the uncertainty

in the bid documents in the IPs) and Model 2 (which includes the uncertainty in the bid documents in the IPs) are generated.

*Post-Data Analysis (Validation)*: Models 1 and 2 are tested in a validation test to determine whether the models created in the data analysis step show similar performance characteristics for new data other than the data used for training. The test results are presented in a confusion matrix, which is analyzed and discussed.

## **4. Model for Classification of Level of Bid Price Volatility**

## *4.1. Pre-Data Analysis: Preprocessing of Data from Bid Summary and Pre-Bid Clarification Document* 4.1.1. Input Parameters

Because of the wide range of types and the number of bid documents, Caltrans's bid summary contains significant information about the bid. The pre-bid clarification document contains uncertain details of risk on the bid documents. Based on the factors discussed in Section 2.2, highly relevant parameters to the bid price are extracted from these two documents (Table 2).



<sup>1</sup> Numbers in "Relevant Risk Factors ... " match numbers in Table 1; <sup>2</sup> B. S.: bid summary; <sup>3</sup> P. C. D.: pre-bid clarification document; <sup>4</sup> Num.: numeric. <sup>5</sup> Categ.: categorical; <sup>6</sup> U. B. I.: unsolved bid inquiries.

Meanwhile, the information extracted from the bid summary requires preprocessing to be used as a model parameter. Further, text data must be standardized through nominalization, and numeric data must be filtered based on a chosen range such that the outliers do not affect the model performance. The pre-bid clarification document contains the inquiries and answers for the project (Table 3).

Most inquiries aim to accurately estimate the bid prices by resolving uncertainty; because some inquiries do not, they must be preprocessed to include only those related to uncertainty. If these inquiries can be resolved with appropriate answers, they are excluded. The following describes the seven input parameters presented in Table 2 in detail for each risk type.


**Table 3.** Examples of Inquiry and Answer in Pre-Bid Clarification Document.

## Time Risk

*IP-1 (Working Days)*: IP-1 represents the period of completion of the project required by the client. If IP-1 is relatively short considering the size of the project, the bid price may increase owing to required rush or night work. The projects considered in this study have IP-1 values between 47 and 1530.

*IP-2 (Project Location)*: the IP-2 of the raw data is an address close to the construction site. IP-2 is related to the local price index, affecting the bid price. In the US, the price index is generally determined at the state level; however, differences in prices can occur within a single state. Because California is a large state in the US, Caltrans divides its administration into 12 districts separate from their counties. Accordingly, in this study, IP-2 is coded as 1–12 according to the district.

#### Cost Risk

*IP-3 (Engineer's Estimate)*: the project cost to which bidders can refer for the bid price is IP-3 at the time of the announcement; because the raw data have a too wide IP-1 distribution, the range must be adjusted. In this study, projects in the range of \$10,000,000–\$280,000,000 are used in the model.

*IP-4 (Bid Preparation Days)*: IP-4 is when bidders have to review the bid document (including the uncertainties); thus, this time affects the accuracy of the bid price. IP-4 is calculated as the period from the bid announcement date to the bid opening date extracted from the raw data (values between 18 and 237).

*IP-5 (Number of Bidders)*: to use IP-5 as an input parameter, it must be checked whether the information is known before the bid opening. Researchers have argued that the variable IP-5 influences the bid price [24,25,32]. When it increases, the bidders deliberately lower the bid price to win the project [43,44]. Thus, the bidders are aware of the number of competitors in advance; Christodoulou [43] studied the optimal *Mi* (Equation (1)) based on this premise. Therefore, IP-5 is included as an input parameter with values from 2 to 12.

## Quality Risk

*IP-6 (Project Type)*: IP-6 can be mainly classified into roads (e.g., highways, freeways, or roadways) and bridges; the numbers "1" and "2" represent a road and bridge, respectively.

*IP-7 (Uncertainty of Bid Documents):* 3578 raw datasets are screened through Section 4.1.1, which result in 269 final datasets with 6682 bid inquiries. As mentioned in Section 2.4, there are two uncertainty factors in bid documents: unclear communication (BI. 1–3) and unclear requirements (BI. 4–5). Uncertain communication includes discrepancies, errors, and the omission of information in the bid document, each of which has overlapping meanings. For example, omission means that necessary information is missing owing to an error; thus, it can be interpreted as an error itself. Therefore, each term is clearly classified according to the mutually exclusive and collectively exhaustive principle. When certain identical information in various bidding documents causes conflicts, the case corresponds to case BI. 1: discrepancy. The case in which being inquired by an error of single information itself is categorized as case BI. 2: error. Uncertainty due to the lack of specific information

is classified as case BI. 3: omission. Furthermore, uncertain requirements are classified into two types that ask for insufficient but non-essential information (BI. 4: insufficient information) or accept alternatives to the existing guidelines (BI. 5: alternative information). That is, only inquiries corresponding to BI. 1-5 among the pre-bid clarification document content are regarded as uncertainty factors. Through this process, 52 bid inquiries are excluded. After excluding the questions, the uncertainty of which has been resolved with appropriate answers (4336), the number of unsolved bid inquiries is 1994 with values between 2 and 59 for each project (i.e., the IP-7).

All coded IPs are used to train the model in the normalized form.

#### 4.1.2. Output Parameters

As stated in Section 2.6, the output parameters of the models are the bid average risk (OP-1) and bid range risk (OP-2), which are based on the bid price of the raw data. The bid average risk is the ratio of the average price to the engineer's estimate (Equation (3)):

$$\text{Bid Average Risk} = \frac{\text{Average Bid Price}}{\text{Engineer's Estimate}} = F\_1(\mathbf{x}). \tag{3}$$

For example, if the engineer's estimates of projects A and B are \$10 billion and the respective average bid prices are \$10 billion and \$13 billion, the bid average risks are 1.0 and 1.3, respectively. Thus, it can be assumed that the bidders expect a higher risk for project B. Moreover, the bid range risk (OP-2) refers to the difference between the maximal and minimal bid price concerning the engineer's estimate (Equation (4)):

$$\text{Bid Range Risk} = \frac{\text{Max. Bid Price} - \text{Min. Bid Price}}{\text{Engineer's Estimate}} = F\_2(\mathbf{x}).\tag{4}$$

For projects A and B with estimates of \$10 billion and \$1 billion, respectively, the differences between the maximal and minimal bid prices would be identical (\$2 billion); however, the differences between the uncertainties of the two projects cannot be considered identical: project B is riskier than A. Finally, as mentioned in Section 2, the bid average risk and bid range risk defined in this section act as *F*1(*x*) of H1 and *F*2(*x*) of H2.

## 4.1.3. Impact of IP-7 on Bid Price Volatility

The raw data of 3578 projects are preprocessed to create the final dataset of 269 projects. To answer research question 1 in Section 2.6, the final dataset should be classified into groups with and without uncertainty factors. In this study, projects with IP-7 values between 2 and 59 are considered a group with uncertainty factors, and projects with IP-7 values between 0 and 1 are considered a group without uncertainty factors. The OP-1 and OP-2 in each group are presented in Figure 1 and Table 4.

**Figure 1.** Comparison between OP-1 and OP-2 of Projects with IP-7 Values of (1) 2–59 and (2) 0–1.


**Table 4.** Comparison between Data Statistics of OP-1 and OP-2 of Projects 1 and 2.

As shown in Figure 1 and Table 4, projects with uncertainty (Project 1) score higher values for the bid average risk and bid range risk than projects without uncertainty (Project 2). In other words, uncertainty in bid documents increases the bid price volatility.

## *4.2. Data Analysis: Two Classification Models Based on Machine Learning for Bid Price Volatility* 4.2.1. Design

In the data analysis stage, the final data in Section 4.1 is used to implement a model for classifying the level of volatility in bid prices. This study presupposes that the uncertainty of the bid document is related to the bid price and ultimately tries to improve the accuracy of the bidding price volatility level classification model by adding the variable of the uncertainty of the bid document to the existing bid-related variables.

SPSS, MATLAB, R, and Python are mainly used for machine learning-based data mining classification. In this study, data analysis and model development were performed using Mathworks' MATLAB software as a data analysis tool.

Because this model predicts the classes of OP-1 and 2 with input parameters, the models are trained through supervised learning after the class designation. A suitable number of classes is significant because too many classes decrease the reliability of the prediction results; too few classes make the interpretation of the results difficult. In this study, four levels according to the bid price volatility distribution are considered.

When the boundary between the classes is set, a natural breakpoint is preferable; the breakpoint is the point at which the distribution of the OP values suddenly breaks. If there is no natural breakpoint, a boundary should be set such that the data of each class is evenly distributed to ensure reliability. Figure 2 shows the distributions of OP-1 and OP-2 of the 269 final projects; the parameters do not have natural breakpoints. The set boundaries between the classes of the models are presented in Table 5. The OP-1 and OP-2 have a total of 4 classes: OP-1 is a "–" class that is much smaller than 0, a slightly smaller class is "-," a slightly larger class is "+," and a much larger class is "++." On the other hand, classes of OP-2 were named with the symbols "+," "++," "+++," and "++++" in the order of close to 0.

**Figure 2.** Distribution of Bid Price Volatility of 269 Projects.


**Table 5.** Class Designation for Output Parameters.

Not all final datasets are used to train the model. The remaining datasets are used to check whether the classification model performs consistently for new data (typically 30% of the total data). Half these data is used for the validation during training, and the rest is applied in the validation tests (Section 4.3). Accordingly, 269 datasets are allocated to 40 for the training validation, 40 for the validation test, and 189 for the model training with machine learning. There are several machine learning algorithms for training classification models; in this study, neural net (NN), which shows good performance, is used. In addition, three algorithms are combined to evaluate the results derived with neural net. In total, four classification algorithms are used for the models: NN, decision tree (Tree), support vector machine (SVM), and K-nearest neighbor (KNN).

## 4.2.2. Results

In the bid price volatility level classification based on the model design in Section 4.2.1, Model 1 is trained without IP-7, whereas Model 2 is trained with IP-7. The performance of the model is expressed as the accuracy (i.e., the number of correct classifications compared to the total number of classifications):

$$Accuracy\left(\%\right) = \frac{Number\ of\ Correct\ Class\ fibrations}{Total\ Number\ of\ Class\ fibrations}.\tag{5}$$

In this section, the accuracies of Model 1 trained without IP-7 and Model 2 trained with IP-7 are compared to answer research question 2 in Section 2.6 (Table 6).


**Table 6.** Accuracies of Classification Models.

According to Table 6, Model 2 performs better than Model 1 for all algorithms, including the NN. Second, the classification Models 1 and 2 result in higher accuracies for OP-2 than for OP-1 in most cases. Third, the NN exhibits the best performance in Models 1 and 2.

Model 1 results in accuracies with the order of SVM > KNN > Tree, and Model 2 results in accuracies with the order of Tree > SVM > KNN. Thus, IP-7 (a parameter of the uncertainty in bid documents) improves the performance of the model that classifies the level of the bid price volatility. Regarding the NN algorithm classifier, the performance scores of Model 1 (without IP-7) are 37.5% (OP-1) and 42.5% (OP-2), whereas those of Model 2 (with IP-7) are 63.9% (OP-1) and 65.8% (OP-2); these results are 26.4% and 23.3% higher than those of Model 1, respectively. This trend is identical for the averages of all

algorithms: the accuracies of Model 1 are 34.1% (OP-1) and 38.4% (OP-2), while those of Model 2 are 60.9% (OP-1) and 60.8 % (OP-2).

#### *4.3. Post-Data Analysis: Validation*

In the validation test, the accuracy of a classification model is evaluated with data that have not been used for training. In this section, the training performance of Model 2 with IP-7 is validated based on the NN. There are various approaches for validating a machine learning model (e.g., the N-fold cross-validation, bootstrap, and sliding window methods); N-fold cross-validation is the most used approach. The results of the validation test are presented in a confusion matrix, which enables the determination of the accuracy of the model and true positive rate (*TPR*, which represents the recall and sensitivity):

$$\text{True Positive Rate} \left(TPR\right) = \frac{TP \left(True \text{ Positive} \right)}{TP \left(True \text{ Positive} \right) + FN \left(False \text{ Negative} \right)} \tag{6}$$

where "*True Positive*" (*TP*) refers to samples in which the positive cases are correctly classified, whereas "*False Negative*" (*FN*) refers to samples in which the negative cases are incorrectly classified. In other words, the TPR is the ratio of correct samples to total samples classified by the model. Figure 3 presents the validation results of Model 2 trained with the NN for the classification of the bid price volatility.

**Figure 3.** Confusion Matrices of Model 2: Bid Average Risk and Bid Range Risk.

The validation accuracies of Model 2 for OP-1 and OP-2 are 52.5% and 72.5%, respectively; thus, the model correctly classifies the bid price volatility levels of 21 and 29 out of 40 projects. First, the TPR of Model 2 for OP-1 is highest in the "+" class (66.7%) and lowest in the "–"class (44.4%). However, according to the confusion matrix, incorrectly classified samples are classified into groups of similar rather than entirely different classes. This is because successive values are cut by the artificial boundaries of the class designation. Moreover, the accuracy of the model for the test data (52.5%) is slightly lower than the accuracy of the model for the training data (69.3%). Second, the TPR of Model 2 for OP-2 is highest in the "+++" class (76.9%) and lowest in the "++" class (66.7%). For OP-1, the model classifies most of the samples into same or similar classes. In addition, the performance of the model for the validation test data is 72.5%, which is better than the accuracy of the model for the training data (64.6%).

## **5. Discussion**

From a background in Introduction and Research Gaps, a decision support tool for bid price budgeting for the actual field and research studies was created based on quantitative data analyses with project-based bid results from other studies. Based on these requirements, two research questions were established, and Caltrans's bid results were used to select related parameters based on previous studies. Based on the 269 final project datasets obtained after preprocessing, the uncertainty level in bid documents was quantified as the number of inquiries corresponding to BI. 1-5 in the pre-bid clarification document that have not been resolved.

To answer research question 1, the 269 projects were classified into two groups: a group with IP-7 values of 0–1 and another group with IP-7 values of 2–59. The group with uncertainty in the bid documents generally had a higher bid average risk and bid range risk than the group without uncertainty. Thus, uncertainty in bid documents increases the bid price volatility. It is noteworthy that the project with an IP-7 value of 1 was also considered a project without uncertainty in this study because the considered uncertainty level depends on the stakeholders' opinion. For example, for projects with IP-7 values of 1, some bidders may believe that there is little uncertainty, whereas others may believe that the uncertainty level is relatively high. Therefore, the boundary of uncertainty can be expressed as follows: {project without uncertainty: IP-7 value of *x*|*x* = 0} ∪ {project with uncertainty: IP-7 value of *x*|*x* > 0} for risk-averse bidders and as {project without uncertainty: IP-7 value of *x*|0 ≤ *x* ≤ *n*, where *n* = 1, ... , 59} ∪ {project with uncertainty: IP-7 value of *x*|*x* > *n*} for risk-takers. In other words, for projects with IP-7 values of *n*, the level of perceived uncertainty depends on the stakeholder; the case *n* = 1 was presented as an example in this study.

Furthermore, Models 1 and 2 were trained with an NN to examine whether IP-7 affects the bid price volatility classification. The accuracies of Model 1 for the bid average risk and bid range risk were 37.5% and 42.5%, respectively; those of Model 2 were 63.9% and 65.8%, respectively.

The accuracies of Model 1 exceeded 25%, which corresponds to the mathematical probability when one of four classes is randomly selected; nevertheless, the accuracies were too low for decision making. This means that parameters that are not included in the model have a more substantial influence on the bid price. By contrast, the accuracies of Model 2 were better, and similar phenomena were observed for the three other algorithms and NN. In other words, the influence of IP-7 is relatively strong compared to those of the other parameters included in the model. However, the fact that the accuracy of the models has stopped at the current level proves that there are still other parameters that are not included in the models and that significantly affect the bid price. However, this effect can be attributed to the nature of IP-7 itself; IP-7 represents the number of unsolved bid inquiries; this surrogate endpoint is set because it is impossible to cost the situation mentioned in each inquiry. For example, two inquiries might be worth \$100 and \$10,000, respectively; however, they are treated equally as a value of 1 in the models. This may be why the models score higher accuracies for the bid range risk than for the bid average risk. From a superficial point of view, this implies that the model's application may be limited due to its accuracy. However, the more time it takes to consider accurately the cost of each inquiry, the less time it takes to identify more risks, which in turn increases the uncertainty of the bidding document. In this respect, obtaining such a relatively high accuracy only with the number of unsolved bid inquiries is a good sign.

Moreover, the results were validated with test data, and the answer to research question 2 was found. Bidders should determine the bid price by integrating projectrelated information. Before this, the bidders should add the cost of project uncertainty in the markups along with Ci (Equation (1)), which is calculated based on construction budgeting. Because "uncertainty" literally means "lack of knowledge," there is no raw data for calculating it; thus, Model 2 can be applied. For example, if the project's bid average risk is classified as "–" the average of the bid price is likely to be much lower than the engineers' estimate. If the currently calculated Ci is high, Mi must be lowered, or Ci must be adjusted to win the bid. If the results are below the lower bound of the expected profit, the bidder may stop participating in the project. Likewise, if the bid range risk is classified as "+++," many bidders may present prices that deviate much from the engineers' estimate; thus, the bidder may use this strategy to adjust Ci and Mi.

Since there are too many uncertainties affecting the bid price of a construction project and many of them are risks, researchers have performed many valuable studies, and practitioners have been through trial and error. However, when a specific point in time and space is determined, many risks are eventually settled. For example, a city's price index itself fluctuates over time, but finally, the city's price index is determined as one at the time and place of the project. So, for bidders, one dilemma between winning bids and project profits is, therefore, whether they should call more or less than expected. These finegrained adjustments are ultimately determined by project-intrinsic-risks that have not been finalized. The bidding document is a kind of contract document, and all the explicit contents contained in it are directly related to the price. However, there is a problem that uncertain content cannot be dealt with one-to-one with the price. Moreover, since the contents of the bidding document are all different for each project, it was impossible to solve this with the existing linear method. On the other hand, the authors measured the uncertainty of each bid document, not the content of the bid document, and analyzed the risk through the NN algorithm. This study provided an answer to how much the uncertainty in the contents of bid documents increases or decreases the already expected price.

Nevertheless, the method presented in this study has several limitations, which must be considered in future research studies. First, the data are limited. Caltrans's project data were used to minimize the influence of external factors; however, owing to the characteristics of standardized data, the impact of uncertainty in bid documents (which is one of the internal factors) may be relatively weak. In this respect, it might be difficult to generalize the results of this study. Another limitation is related to the preprocessing of IP-7. Because the pre-bid clarification document contains unstructured text data, which is challenging to be computationally processed, there is no automated method for quantifying it; thus, the authors manually analyzed 6682 datasets. Because this process can introduce human error, the authors mutually verified the results. Third, there are obvious limitations in that this paper does not address all of these factors and only deals with the risk of uncertainty in the bid document. The risk of fluctuations in the material and workforce market is a significant factor affecting the bid amount and must be considered in the bidding stage. The bidders scrutinize these through market research. Meanwhile, the risk of fluctuations in the material and labor market varies with time and location, which means they are variables. If they are determined, the risk of volatility could be determined, too. Since the bidders bid simultaneously for work performed at the same location, these costs will reach some agreed value. However, this cannot be confirmed as a single value, so it remains a risk of the bid price. Future studies have to further consider the remaining major fluctuation risks and use them as parameters. In that case, significant improvements are expected to improve model accuracy.

## **6. Conclusions**

In this study, a classification model for the bid price volatility level was developed by analyzing the relationship between the uncertainty in bid documents and bid price based on bid data. The model results in this study reveal that uncertainty in bid documents is causing bid prices to rise or fall more than necessary. Therefore, it is essential to conduct a thorough review of items causing uncertainty before bidding. We present these items as discrepancy, error, omission, insufficient information, and alternative information. Besides this, the inclusion of a pre-bid clarification process allows the price of the project to converge more appropriately if the remaining uncertainties are eliminated at the time of bidding. This study has the following contributions: the first step in quantifying the uncertainty in bid documents. The results of published qualitative studies of risk identification were evaluated with bid result data in this quantitative study. The model proposed in this paper enables risk management at a lower level: in the new approach, the uncertainty in bid documents considered to have an uncontrollable risk is analyzed with a pre-bid clarification document. Through this, the theoretical gaps are closed. The results of this study can help bidders to determine the bid price. According to the results, the pre-bid clarification in the bid phase is an essential process because the resolution of uncertainty in the bid documents can reduce the bid average risk and bid range risk. Accordingly, it is

expected that the bid price of the project, the risk of which has been resolved during the pre-bid clarification process, will converge to a more acceptable price. The construction objectives created with this price will improve bidders' profitability and meet the client's expectations, which ensures a successful construction project. The final contribution of this study is that the concept of sustainability has been further expanded in construction projects. In construction, the idea of sustainability is meaningful in that it extends the project management, which focused on the design and construction stages, to the entire life cycle. Until now, researchers have carried out valuable studies related to sustainable energy use, noting the importance of maintenance and operation phase after completion. However, the early stages of the project are also of great importance. This study tried to ensure the success and sustainability of the project through research on a reasonable project price. In the future, the authors will establish a method that comprehensively analyzes the uncertainty in unstructured text data from public projects of various institutions that provide pre-bid clarification documents for the automatic extraction of the IP-7 content. Further, the authors will combine this with the results of this study to establish a more general and highly accurate risk classification model.

**Author Contributions:** Conceptualization, Y.J. and J.-S.Y.; methodology, Y.J. and J.S.; software, Y.J.; validation, Y.J., J.-S.Y. and J.S.; formal analysis, Y.J.; investigation, Y.J.; resources, Y.J.; data curation, Y.J. and J.-S.Y.; writing—original draft preparation, Y.J., J.-S.Y. and J.S.; writing—review and editing, Y.J., J.-S.Y. and J.S.; visualization, Y.J.; supervision, J.-S.Y. and J.S.; project administration, J.-S.Y. and J.S; funding acquisition, J.-S.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is supported by the Korea Agency for Infrastructure Technology Advancement (KAIA) grant funded by the Ministry of Land, Infrastructure and Transport (Grant 21ORPS-B158109- 02). This study is also supported by the Ewha Womans University Scholarship of 2019.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Caltrans's raw data (2011–2018) used in this study can be accessed through the following link: http://ppmoe.dot.ca.gov/des/oe/project-bucket.php (accessed on 27 March 2021).

**Acknowledgments:** The authors would like to thank Lee, Assistant Professor of Department of Civil & Environmental Engineering and Construction of University of Nevada, Las Vegas, for helpful advice on various technical issues examined in this paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


## *Article* **Excavation Method Determination of Earth-Retaining Wall for Sustainable Environment and Economy: Life Cycle Assessment Based on Construction Cases in Korea**

**Youngman Seol 1, Seungjoo Lee 2, Jong-Young Lee 3, Jung-Geun Han 3,4,\* and Gigwon Hong 5,\***


**Abstract:** This study describes life cycle assessment (LCA) results of the excavation depth and ground condition of a medium-sized excavation ground in order to examine the effect of construction methods on environmental and economic feasibility for an earth-retaining wall. LCA is conducted in consideration of eight environmental impact categories according to the construction stage of the earth-retaining wall. In addition, the environmental cost of construction method for the earthretaining wall was calculated, and its selection criteria were analyzed based on the calculation results. The evaluation results of the environmental load of construction methods for the earth-retaining wall show that the H-Pile+Earth plate construction method has low economic efficiency because the construction method significantly increased the environmental load due to the increased ecological toxicity. The environmental load characteristics have a greater effect on the selection of construction methods in sandy soil than in composite soil when the excavation depth is the same. The evaluation result of the environmental cost of the construction methods for the earth-retaining wall shows that the environmental cost increased as the excavation depth increased, and the sandy soil conditions have higher environmental costs than complex soil conditions.

**Keywords:** LCA (life cycle assessment); earth-retaining wall; excavation; environment load; environment cost

## **1. Introduction**

LCA, called life cycle assessment or life cycle environmental load assessment, is defined as a technique that identifies life cycle flows, such as raw material and energy input, pollutant occurrence, and recycling in product production, and it assesses potential environmental impacts. That is, it is an evaluation of the environmental impact of the entire process of obtaining raw materials for products, production, application, and disuse, i.e., the entire process from the acquisition of raw materials to the final disposal of the product [1–3]. LCA, an environmental evaluation technique, is actively used as a technology evaluation method to secure source technologies to respond to climate change worldwide. [4–8]. LCA is not limited to assessing greenhouse gas emissions, but they are focused on in the literature review section of this study, because Korea is facing the considerable issue of greenhouse gas emissions in the field of construction.

Large-scale facilities are planned mainly in the construction industry. The application of LCA in this field can sufficiently consider the environmental impact, because there are

**Citation:** Seol, Y.; Lee, S.; Lee, J.-Y.; Han, J.-G.; Hong, G. Excavation Method Determination of Earth-Retaining Wall for Sustainable Environment and Economy: Life Cycle Assessment Based on Construction Cases in Korea. *Sustainability* **2021**, *13*, 2974. https://doi.org/10.3390/su13052974


Academic Editor: Sunkuk Kim

Received: 21 January 2021 Accepted: 5 March 2021 Published: 9 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

many types and quantities of materials, and high-energy facilities are applied. In particular, rapid decision support is possible for environmental issues if LCA is performed in the early stages of a project [9,10]. As a result of forecasting greenhouse gas emissions by the industry sector by 2030, Lee [11] predicted that emissions associated with the construction industry will increase by 2.2% by 2030. In 2015, the International Energy Agency (IEA) established a plan to induce and support greenhouse gas reduction activities with the aim of implementing greenhouse gas target management in the construction industry; in Korea, 8.34% and 2.07% reduction targets were established in the building and transportation sectors, respectively. As mentioned above, various studies on environmental impact assessment of greenhouse gases emitted from construction activities have been actively conducted in order to respond to the international situation [12–14].

The Korean construction market is expanding not only to the infrastructure sector but also to the energy and building sectors, mainly in the carbon emission rights market, renewable energy market, and green building market, so it is time to require a more aggressive response and greater investment. Overseas, it is reported that Europe classifies the construction industry as one of the seven major sectors that emit greenhouse gases, and the construction industry accounts for 36% of total industrial carbon emissions and 40% of total energy consumption. It was determined that the cause of these results is closely related to the fuel use of construction equipment and gas emissions due to various construction activities, and studies have been conducted to contribute to reducing the emission of greenhouse gases [15–17].

Research on LCA has actively been carried out abroad for more than two decades. Europe is a leader in the field of LCA research, and many studies have been conducted on methodology, life cycle inventory (LCI) DB (Database) construction, and program development in the field of the environment [18]. Japan is attempting a systematic approach to LCA, and Australia has constructed an LCI DB mainly of infrastructure facilities, such as buildings, raw materials, iron, minerals, and packaging materials. In addition, various case studies have been conducted to evaluate the environmental impact related to greenhouse gases on the foundation work of buildings and residential buildings [19–22]. Moreover, in many advanced countries, evaluation programs that take into account the life cycle of construction materials have been developed and put into use, and they have been set as sustainable development goals to reduce the environmental load in the construction industry [23–25]. Recently, research on LCA has been conducted in various environmental fields in Korea. It has been only 5 years since the study on the field of civil engineering took off in Korea, so the available data related to construction materials and construction are insufficient. Additionally, LCA is partly applied to SOC (social overhead capital) facilities, such as roads, bridges, and tunnels, in which the target facilities are standardized [26,27].

Therefore, this study aims to improve the process by which existing construction methods are selected by additionally applying the results of LCA analysis, such as constructability and economic feasibility, to the way a construction method is selected when considering various soil conditions. To this end, the earth-retaining wall, a representative soil structure, was selected as the target structure, and a case of securing stability through a series of design processes was established for various excavation conditions and construction methods after simplifying the excavation-related ground conditions. Afterward, the environmental loads for the eight major categories in the environmental product declaration (EPD), such as greenhouse gas emissions and energy consumption, which are the main management targets of the Greenhouse Gas and Energy Target Management System, were analyzed and applied to the established case. Based on this analysis, in order to minimize the environmental load when selecting a construction method for an earth-retaining wall, LCA analysis for an earth-retaining wall according to excavation depth and soil conditions was conducted to prepare improvement measures.

### **2. Theoretical Review of LCA Technique**

*2.1. Concept of LCA Technique*

LCA, also called "life cycle environmental load assessment," is a technique to identify the inputs of raw materials, energy, chemicals, etc. and outputs of wastes, pollutants, recycling, etc. in the life cycle of a product and to evaluate potential environmental impacts (Figure 1).

**Figure 1.** Overall process and input/output of life cycle assessment (LCA).

Raw materials, energy, and utilities are inputs, and air emissions, water system emissions, solid wastes, etc. in the manufacturing process, the use process, and the disposal process are outputs. Early stages of the construction process such as collection and transportation of raw materials are referred to as "upstream," whereas product use and disposal are "downstream."

General guidelines to LCA structures and procedures used to assess environmental performance in a series of processes can be found in ISO standards 14040 and 14044, international standards for environmental management (green management) established by the International Organization for Standardization (ISO) [1,2]. As shown in Figure 2, the LCA consists largely of objectives and scope definitions, inventory analysis (LCI), impact assessment (LCA), and interpretation of results.

**Figure 2.** Procedure of life cycle assessment (LCA).

LCA is used to provide a scientific basis for determining which of several processes has a significant environmental impact or which of several products is environmentally friendly. For instance, LCA can be performed to identify which construction method, A or B, has a smaller impact on the environment. This process makes quantitative numerical comparisons possible by collecting data on materials and equipment that are inputted during the construction and maintenance stages of a comparative construction method and by setting inputted material, energy, and resource usage units. LCA has recently been applied to the construction industry internationally to reflect various environmental impact assessments in the planning and design stages, making it possible to design alternatives by taking into account the environmental friendliness, such as the comparison of routes and construction methods. Therefore, as it is necessary to introduce and effectively apply decision-making methods for environmentally friendly development in the construction sector, the LCA, in which environmental performance in terms of construction environment and environmental value through the quantification of environmental load are evaluated, is a significant factor.

### *2.2. Application of LCA in the Construction Industry*

Although there have not yet been many cases in which an LCA evaluation was conducted in Korea, the results of analyzing various cases performed concerning roads/bridges, ports, and railways are as follows: First, the evaluation method was conducted by analyzing material and equipment inputs through information collection and analysis and then evaluating environmental and economic feasibility by calculating the environmental load through LCA evaluation by comparison. In addition, LCA analysis as a comparison method is performed in the application stage of a construction method, and LCA analysis of the basic plan and basic design is conducted after dividing it into the initial construction stage, maintenance stage, and dismantling and disposal stage. In other words, in life cycle cost (LCC) analysis as a comparison method, the environmental loads for eight environmental impact categories (abiotic resource depletion, global warming, ozone depletion, photochemical oxidant creation, acidification, eutrophication, ecotoxicity, and human toxicity) are calculated, and those for key contributors to global warming (carbon dioxide (CO2), sulfur dioxide (SO2), nitrogen dioxide (NO2), and carbon monoxide (CO)) out of all environmental impact categories are calculated and compared to alternatives. After analyzing the effect on the reduction of environmental load of the basic design, reflecting the final LCA evaluation results, with the reduced values of the environmental indices compared to the basic plan, the basic design was presented, which makes environmental economic feasibility or environmentally friendly design possible.

Meanwhile, more work is being done in foreign countries. The Netherlands has been developing LCA evaluation programs for the construction industry since 1994, with work being conducted by major construction-related organizations (e.g., the Ministry of Housing, Spatial Planning, and Environmental), with various types of data now being provided, such as the reliability of LCA. In Finland, LCA of the construction industry is conducted by the VTT Technical Research Centre of Finland; the scope of the LCA is set at each life cycle stage of an individual building, such as material production, transportation, construction, maintenance, and dismantling, and the environmental impact data obtained from these results are used in marketing, product display, system management, and product design [28]. Recently, Han et al. [29] developed a tool that considers cost and environmental impact together by utilizing building information modeling (BIM) based on information and communication technologies (ICTs) to link LCC throughout the life cycle of a building to LCA tools. In order to develop a database that can reflect greenhouse gas reduction, Japan developed an environmental load inventory for individual items by utilizing a method to correct the estimates with inter-industry relational tables based on the detailed DB calculated using the estimation method. By making use of these methods, the environmental load of new materials such as eco-cement to consider the environment can be updated from time to time through the DB, and a basis for conducting evaluations that reflect an environmental load of materials has been prepared. The American Society for Testing and Materials (ASTM) in the United States prepares guidelines for LCA of construction materials, design of green buildings, construction, and operation, and the National Institute of Standards and Technology (NIST) develops Building for Environmental and Economic Sustainability (BEES) to support the selection of economic and environmentally friendly construction materials.

As mentioned above, mainly overseas, evaluation software that considers all aspects of construction materials has been developed and utilized mainly in the construction sector, and various activities have been carried out to reduce the environmental pollution load in the construction sector with the goal of sustainable development. Table 1 shows these research activities by country.


**Table 1.** Life cycle assessment (LCA) application status by country (Kwon [30]).

## *2.3. Application of Similar Techniques for the Selection of Construction Methods on Civil Engineering Structures*

Bae [31] suggested a system for selection of construction method by classifying influential factors by applying the analytical hierarchy process (AHP) technique to the selection of construction methods for an underground retaining wall; Han and Lee [32] applied the AHP technique to work conducted by a group of experts in related fields when selecting the reinforcing method for a cut slope. Lee et al. [33] once presented a decision model for selecting soft ground improvement methods using AHP techniques, and Lee and Jeong [34] proposed a decision-making system using the AHP technique and preference function (PF) when selecting the basic construction method for structures.

In order to resolve the inaccuracies intrinsic to the subjective judging process and reduce the uncertainty and ambiguity of the AHP method in bridge construction projects, Pan [35] proposed the fuzzy AHP (FAHP) model by applying triangular and trapezoidal fuzzy numbers and the α-cut concept. Ebrahimian et al. [36] pointed out that application of the existing AHP technique has the drawback that the pairwise comparison required for hierarchy analysis is tedious and time-consuming in the planning phase of a construction project when complex interests are concerned, such as urban construction projects, and suggested a combined model of fuzzy AHP (FAHP) and compromise programming (CP).

Shen et al. [37] introduced text mining case-based reasoning (TM-CBR), which can extract the most similar case from a design by integrating the text mining technique into the CRB system in order to improve the efficiency of decision-making in environmentally friendly design. Lorenz and Jost [38] reported that the system dynamic model is an efficient way to select the best method for a given purpose; Tsai et al. [39] proposed the multiple criteria decision making (MCDM) approach to resolve the impact on the goal of the time, cost and environmental Impacts (TCEI) analysis, the selective issue on how decision-makers determine the most appropriate construction methods.

In order to rationalize selection of construction methods for a retaining wall, Kim et al. [40] used a neural network system to verify the rationality of the selection at approximately 160 sites and showed predictive results of 88% in the selection of a construction method and 90% in the selection of the wall retaining method. Furthermore, the selection of the construction method for a retaining wall has many factors to consider and is based on uncertain information, resulting in frequent design changes and consequent delays in construction and lots of economic loss. To overcome this issue, we highlight the limitation that artificial intelligence (AI) technology is limited to new projects even though it can be used to support complex decision-making processes [40,41]; when selecting tunnel construction methods, Park et al. [42] applied the AHP technique to the existing problems of value engineering (VE), and LCC and proposed the life cycle social cost (LCSC) evaluation method to convert social loss expenditures, which could not be applied in the LCC technique.

However, as mentioned above, in most previous research, several decision-making methods have been adopted to rationally select the construction method for an earthretaining wall, and most of them suggested only the applicability and rationality of appropriately applied construction methods based on the existing application cases.

That is, in order to select a rational construction method, an evaluation system that considers social loss expenses (environmental factors) and social factors has been used only with improvements. Therefore, there is a limit to using the mechanical relationship among construction methods, soil, material, and environment based on a stability-based design for various soil conditions when selecting construction methods for an existing earth-retaining wall.

Therefore, beyond the selection of a construction method that focuses on the given soil conditions and the usability and stability of the materials in each construction method, a study is needed that addresses how to select a construction method for earth-retaining walls that considers economic feasibility and environmental performance applied the conversion of environmental costs as well as LCA analysis considering environmental performance in the existing method.

## **3. Selection of Cases and Stability Review for LCA of Earth-Retaining Wall**

*3.1. Evaluation of Case Selection and Soil Characteristics*

In this section, we set the selection criteria with which the construction method for a retaining wall is applied and propose a resultant rational selection method. In this study, rational selection methods are classified by taking excavation size and excavation depth into account based on the "Special Law about Underground Safety Management (2018)" and "Review Guideline on Excavation for Safe Building Construction" of Seoul Metropolitan City, created for special safety management considering the stability of recent ground subsidence. The criterion for excavation depth under the Special Law about Underground Safety Management is 20 m, and the criterion for excavation depth of buildings under architectural design-review in urban areas is 10 m. Therefore, as shown in Table 2, the excavation depth at which the earth-retaining wall was installed was 15 m, the middle value of the two standards. This can be viewed as a criterion considering the fact that various construction methods use a 15 m excavation depth. Additionally, the characteristics of the soil to be installed are mostly distributed from the surface to the topsoil, weathered soil, weathered rock, soft rock, and hard rock, in that order, and the weathered soil is mostly composed of deposits. There is also a composition of the sandy soil layer and soft clay layer on the rock layer of a riverbank or shoreline, and, most commonly, it is to consist of composite stratum (typically weathered soil layers) on the rock layer. Thus, the new construction method can be applied if it is composed of only rock layers, so the general sediment layer consists of the sandy soil layer, soft clay layer (soft clay ground), and the mixed stratum of the sandy soil and soft clay. Therefore, we decided to conduct an analysis based on these soil compositions in this study (Table 3). The applied equipment and the construction management method are different according to the excavation scale and ground conditions in excavation construction. Thus, the excavation scale and the ground conditions were applied as comparative criteria in this study.


**Table 2.** Excavation conditions and soil conditions in each case.



The excavation area is medium-sized (50 × 50 m), and the deepest excavation point (excavation depth: 40 m) was determined to be 40 m, a depth which makes the application of the construction method for a retaining wall clearly distinguished, in consideration of the maximum possible construction depth (less than 50 m allowed).

## *3.2. Evaluation of Stability in Each Case*

The program used in the design case is Midas GeoX V.4.6.0. Earth pressure applied to the retaining wall causes stress and displacement of the structure. The deformation analysis of the retaining wall is generally performed by the elastoplastic analysis, because the stress and displacement of the retaining wall change depending on the excavation stage of ground. Midas GeoX V.4.6.0 allows the elastoplastic analysis considering the excavation stage.

All cases applied to the LCA analysis were assumed to have both internal and external stability at each excavation stage. The assessment of internal stability was conducted by a review of the cross-section of the structure (member), and the structural stability of H-Pile, C.I.P, Sheet Pile, S.C.W, Strut, Wale, etc., which form a wall, was evaluated by construction stage (excavation stage). External stability was evaluated by dividing it into the stability on the earth pressure acting on the retaining wall and the stability on the surrounding ground subsidence, etc. during the excavation stage and final excavation stages. Table 4 summarizes the application method of each item for the evaluation of stability performed for the earth-retaining wall in this study.


**Table 4.** Method for review of stability by construction method and item.

Figure 3 shows a schematic diagram of the numerical analysis carried out in this study. The underground water level is reflected in the analysis on the premise that it is lowered according to the stage of excavation and lowered to the excavation surface. The review of stability, such as the stability of the embedded unit, the stability of subsidence, and heaving in each case, considered only the impact on excavation depth because it was affected by the increase in stress depending on excavation depth and was independent of the excavation width.

**Figure 3.** Schematic diagram of the numerical analysis.

Table 5 shows the results of the stability review for each case based on analysis conditions. First, in the stability evaluation of the embedded depth (required safety factor: 1.2) based on the Earth-retaining Wall Design Standard of the Ministry of Land, Infrastructure and Transport in Korea [43] for a shallow excavation depth, the safety factors

are in the following orders from high to low: C.I.P, S.C.W, Sheet Pile, and H-Pile+Earth Plate construction method in the composite soil, and C.I.P, H-Pile, S.C.W, and Sheet Pile construction method in the sandy soil. For deep excavations in composite soil, the safety factors are in the following orders from high to low: S.C.W, C.I.P, Sheet Pile, and H-Pile+Earth Plate construction method. For deep excavations in sandy soil, they increase in the following order: C.I.P, S.C.W, Sheet Pile, and H-Pile+Earth Plate construction method. Furthermore, it was confirmed that the deeper the excavation depth, the greater the safety factor in soft clay ground.


**Table 5.** Stability review results by case.

Here, pre-final excavation stage—-<sup>1</sup> , final excavation stage—-2 .

Caspe [44] estimation of subsidence on the soil was based on a method redefined by Bowles [45], which is relatively consistent with actual data. However, this method has the premise that the displacement (subsidence) due to an increase in effective stress caused by a drop in groundwater level should be calculated separately. As input data for analysis, lateral displacement of the wall by depth, excavation depth, excavation width, and shear resistance angle are required, and for lateral displacement of the wall, computerized analysis data using the beam on elasto-plastic foundation analysis were used.

The deeper the excavation depth, the larger the maximum subsidence, and subsidence occurred more in sandy soil than in composite ground. In addition, in composite soil, when the excavation depth is shallow, the H-Pile+Earth Plate construction method produces the largest amount of subsidence, but the deeper the excavation depth, the greater the subsidence in the Sheet Pile construction method. When the excavation depth is shallow in sandy soil, the Sheet Pile and the H-Pile+Earth Plate construction methods have the largest subsidence, and the C.I.P construction method has the smallest subsidence. When the excavation depth is deep, the Sheet Pile construction method has the largest subsidence, and the S.C.W. construction method has the smallest one. Meanwhile, in soft clay ground, the deeper the excavation depth, the more rapidly the subsidence increases. This result is based on the design of the retaining wall structure with secured stability, so only a very small amount of subsidence occurs; only the tendency of the occurrence of subsidence was analyzed.

Boiling on the bottom of an excavation is generally assessed to increase the safety factor as excavation depth increases, and at this time, the safety factor applied to the boiling judgment was 2.0 [43]. When the excavation depth is shallow, in composite soil, the H-Pile+Earth Plate construction method has a smaller safety factor than do the other construction methods. In sandy soil, as the excavation depth increases, the safety factor increases rapidly, and the safety factor is high in the order of Sheet Pile, S.C.W, C.I.P, and H-Pile+Earth Plate construction method. On the other hand, if pile stiffness and penetration depth are met, a review of heaving is considered in the soft clay layer, so in Sheet Pile application, the deeper the excavation depth, the greater the calculated safety factor necessary to meet the safety factor requirements. The required safety factor was applied to 1.2 [43].

## **4. Analysis of LCA on Earth-Retaining Wall**

#### *4.1. Method and Scope of the Evaluation of Environmental Impact Assessment*

LCA analysis was performed on the applicable construction method of an earthretaining wall by each installation condition, and then the environmental impact characteristics were analyzed. In Korea, the environmental impact assessment of earth-retaining wall is considered as a temporary structure, which reflects only the production and consumption of input resources in the construction stage. Therefore, construction details of material and equipment usage, standards of construction estimates, and energy statistics data of Korea were used to perform inventory analysis on all items applied to the construction of the earth-retaining wall method in this study. In addition, the LCI DB of the Ministry of Environment (MOE) and Ministry of Trade, Industry and Energy (MPTIE) of Korea was used for inventory analysis of the surveyed resources that were required. LCA software (Tool for TypeIII Labeling and LCA, hereinafter referred to as TOTAL) suggested by the Ministry of Environment in Korea was used. The environmental impact assessment was performed on the temporary earth protection facility based on the results after inventory analysis for each case object was performed. Abiotic resource depletion (ARD), global warming (GW), ozone depletion (OD), photochemical oxidant creation (POC), and acidification (AC), eutrophication (EU), ecotoxicity (ET), and human toxicity (HT) were applied as impact categories in order to establish the evaluation comparison criterion. In the environmental load assessment, the construction cost considering the construction method and ground conditions of the earth-retaining wall was applied based on standard of construction estimates in Korea [46].

## *4.2. LCA Results of the Earth-Retaining Wall According to Excavation Depth*

4.2.1. Evaluation Results of Environment Load

Tables 6 and 7 and Figure 4 show the results of identifying and evaluating major environmental impacts through list analysis and impact assessment results for cases where the excavation area is medium scale (50 × 50 m) in shallow excavation (15 m) and deep excavation (40 m) depending on the ground conditions.

**Environmental Impact Factor Soil Condition Construction Method C.I.P S.C.W Sheet Pile H-Pile+Earth Plate** Abiotic Resource Depletion (ARD) Composite Soil 2.50 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 2.58 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 1.59 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 1.59 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Sandy Soil 2.56 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 2.67 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 1.79 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 1.74 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Soft Clay Soil 4.01 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Global Warming (GW) Composite Soil 5.37 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 5.49 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 2.70 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 5.94 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Sandy Soil 5.64 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 5.80 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 3.08 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 6.22 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Soft Clay Soil 6.78 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Ozone Depletion (OD) Composite Soil 1.40 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.41 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.37 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 2.25 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Sandy Soil 1.36 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.40 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.52 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 2.38 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Soft Clay Soil 3.48 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Photochemical Oxidant Creation (POC) Composite Soil 2.24 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 2.35 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.32 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 3.87 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Sandy Soil 2.35 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 2.48 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 1.50 <sup>×</sup> <sup>10</sup>−<sup>7</sup> 4.00 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Soft Clay Soil 3.32 <sup>×</sup> <sup>10</sup>−<sup>7</sup> Acidification (AC) Composite Soil 2.00 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 1.98 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 1.37 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 1.33 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Sandy Soil 2.13 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 2.13 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 1.57 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 1.47 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Soft Clay Soil 3.45 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Eutrophication (EU) Composite Soil 3.47 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 3.41 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 2.21 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 8.20 <sup>×</sup> <sup>10</sup>−<sup>9</sup> Sandy Soil 3.60 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 3.58 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 2.50 <sup>×</sup> <sup>10</sup>−<sup>9</sup> 8.42 <sup>×</sup> <sup>10</sup>−<sup>9</sup> Soft Clay Soil 5.58 <sup>×</sup> <sup>10</sup>−<sup>9</sup> Ecotoxicity (ET) Composite Soil 6.91 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 6.96 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 5.16 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 5.45 <sup>×</sup> <sup>10</sup>−<sup>4</sup> Sandy Soil 7.04 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 7.19 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 5.78 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 5.46 <sup>×</sup> <sup>10</sup>−<sup>4</sup> Soft Clay Soil 1.30 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Human Toxicity (HT) Composite Soil 4.53 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 4.63 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 2.49 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 3.79 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Sandy Soil 4.49 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 4.64 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 2.76 <sup>×</sup> <sup>10</sup>−<sup>6</sup> 4.03 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Soft Clay Soil 2.76 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Total Composite Soil 9.24 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 9.47 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 5.21 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 6.26 <sup>×</sup> <sup>10</sup>−<sup>4</sup> Sandy Soil 9.60 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 9.91 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 5.90 <sup>×</sup> <sup>10</sup>−<sup>5</sup> 6.32 <sup>×</sup> <sup>10</sup>−<sup>4</sup> Soft Clay Soil 1.31 <sup>×</sup> <sup>10</sup>−<sup>4</sup>

**Table 6.** Results of environmental load (shallow excavation: H = 15 m).

**Figure 4.** *Cont.*

**Figure 4.** Relationship between environmental impact factor and environmental load by soil condition (excavation depth 15 m, 40 m): (**a**) composite soil; (**b**) sandy soil; (**c**) soft clay soil.


**Table 7.** Results of environmental load (deep excavation: H = 40 m).


**Table 7.** *Cont.*

First, in the composite soil condition of shallow excavation (as shown in Table 6), the environmental load of the H-Pile+Earth Plate construction method was the highest, at 6.26 × <sup>10</sup>−4, which shows that the impact of the environmental load was great due to the use of wood. Next, the environmental loads were high in the S.C.W, C.I.P, and Sheet Pile construction methods, in that order. In the environmental impact factor, the H-Pile+Earth Plate construction method showed the highest ecological toxicity, and the other three construction methods (C.I.P, S.C.W, and Sheet Pile) showed the highest environmental load in the order of global warming and resource depletion. In the composite soil condition of deep excavation, (as shown in Table 7), the environmental load of the H-Pile+Earth Plate construction method for the earth-retaining wall was 1.68 × <sup>10</sup>−<sup>3</sup> (the highest), and the environmental load of the other construction methods was high in the following order: S.C.W, C.I.P and Sheet Pile. Considering the environmental impact factor, the H-Pile+Earth Plate construction method had the largest environmental load for ecotoxicity, and the environmental load of the other three construction methods was high in the order of global warming and resource depletion.

Second, in the sandy soil condition of shallow excavation (as shown in Table 6), out of the four construction methods for the earth-retaining wall, the environmental load of H-Pile+Earth Plate was the highest (6.32 × <sup>10</sup>−4), and the environmental load was high in the order of S.C.W., C.I.P, and Sheet Pile. When compared by environmental impact factor, the H-Pile+Earth Plate construction method had the highest environmental load for ecotoxicity, and the environmental loads of the other three construction methods were high in the order of global warming, resources depletion, and ecotoxicity. The impact of global warming and resource depletion was greater than that of the other environmental impact categories. Moreover, in the sandy soil condition of deep excavation (as shown in Table 7), the environmental load of the H-Pile+Earth Plate construction method out of four construction methods for the earth-retaining wall was 1.71 × <sup>10</sup>−3, followed by the remaining three in the order of S.C.W, C.I.P, and Sheet Pile construction method. According to the environmental impact categories, the environmental load of ecotoxicity in the H-Pile+Earth Plate construction method was the highest, and the environmental load of the other three construction methods was high for global warming, resource depletion, and ecotoxicity, in that order.

Third, in soft clay ground in shallow excavation (as shown in Table 6), the environmental load of the Sheet Pile construction method was 1.31 × <sup>10</sup><sup>−</sup>4, and the environmental load was high in the order of global warming and resource depletion among all categories of environmental impact. Additionally, the environmental load of the Sheet Pile construction method in soft clay ground was much higher than that of the other soil conditions, and the worse the condition of the soil, the greater the associated environmental load because of the need for more input resources (e.g., reinforcing materials, etc.). In soft clay ground in deep excavation (as shown in Table 7), the environmental load for the Sheet Pile construction method was 3.33 × <sup>10</sup>−4, and according to the environmental impact factor, the environmental load amount was associated with global warming and resource depletion

in that order. Compared to the Sheet Pile construction method in other soil conditions, the Sheet Pile construction method in soft clay soil had a higher environmental load.

#### 4.2.2. Evaluation Results of Environment Cost

Tables 8 and 9 and Figure 5 show the results of identifying and evaluating major environmental impacts through list analysis and impact assessment results following the purpose and scope for cases where the excavation area is of medium scale (50 × 50 m) in shallow excavation (15 m) and deep excavation (40 m) depending on the ground conditions. On this basis, in order to evaluate the environmental and economic impacts of the earth-retaining wall, evaluation of environmental economic feasibility was conducted by applying the environmental cost per unit of pollutants based on the environmental impact factor to the characteristics results of the environmental load amount for the eight categories previously calculated.


\* Environmental cost(E-Cost) unit: KRW 1 million.



**Table 9.** *Cont.*

\* Environmental cost(E-Cost) unit: KRW 1 million.

**Figure 5.** Relationship between environmental impact factor and environmental cost by soil condition (excavation depth 15 m, 40 m): (**a**) composite soil; (**b**) sandy soil; (**c**) soft clay soil.

First, in the composite soil condition of shallow excavation (as shown in Table 8), the H-Pile+Earth Plate construction method showed the highest environmental cost for ecotoxicity at KRW 90.9 million, and the other three construction methods had the largest environmental costs due to global warming. Thus, when it comes to the expected total environmental costs at the construction stages for each installation condition of the construction methods for the earth-retaining wall considering all environmental costs corresponding to the eight environmental impact categories, the total environmental cost of the H-Pile+Earth Plate construction method is the highest (KRW 128.3 million), and the total environmental costs are high in the order of S.C.W, C.I.P, and Sheet Pile construction method. Furthermore, the environmental costs of the S.C.W and C.I.P construction methods are quite similar. In the composite soil condition of deep excavation (as shown in Table 9), in the environmental cost calculation conducted by analyzing the environmental economic feasibility, as was carried out for the shallow excavation, when it comes to the total environmental costs expected in the construction stage of the installation condition for the earth-retaining wall, the H-Pile+Earth Plate construction method had the highest costs (KRW 346.6 million), and the environmental cost associated with ecotoxicity was the highest. For the other three construction methods, excluding the H-Pile+Earth Plate construction method, the highest environmental cost was associated with global warming.

Second, in the sandy soil condition of shallow excavation (as shown in Table 8), when it comes to the expected total environmental cost at the construction stage of the corresponding earth-retaining wall, the H-Pile+Earth Plate construction method had the largest cost (KRW 130.2 million), with the largest share of that cost due to ecotoxicity. Moreover, the three construction methods excluding H-Pile+Earth Plate had the largest environmental costs due to resource depletion and global warming, and it was found that the environmental costs of the C.I.P and S.C.W construction methods are similar. In the sandy soil condition of deep excavation (as shown in Table 9), when it comes to the total environmental cost expected in the construction stage of the earth-retaining wall installation condition, the total environmental cost of H-Pile+Earth Plate was the highest (KRW 355.9 million), and the environmental cost for ecotoxicity was the highest. For the three construction methods, excluding H-Pile+Earth Plate, the environmental cost for global warming was the highest.

Third, in soft clay ground in shallow excavation (as shown in Table 8), the total environmental costs expected in the construction stage of the Sheet Pile installation condition were KRW 48.7 Million, and the total environmental costs in the shallow excavation and medium-sized Sheet Pile installation condition were twice as high as the total environmental costs in other soil conditions. The environmental costs due to global warming account for the largest share. In soft clay ground in deep excavation (as shown in Table 9), the total environmental cost of the Sheet Pile construction method was KRW 123.6 million, which is twice as high as the cost in other soil conditions in a deep and medium-sized excavation (H = 15 m, 50 × 50 m), and the environmental cost associated with global warming was the highest.

## *4.3. Relationship between Excavation Depth, Total Environmental Load, and Total Environmental Cost by Soil Condition*

As shown in Figure 6, the total environmental cost of the H-Pile+Earth Plate construction method was the highest in composite soil, and that cost was higher than the cost associated with the other three construction methods. Moreover, the deeper the excavation depth, the clearer the increase in total environmental cost. We confirmed that the total environmental costs of the C.I.P and S.C.W construction methods were similar, and this tendency remained the same when the excavation depth increased. The total environmental cost in sandy soil was similar to that in composite soil, but the cost in sandy soil was slightly greater. In soft clay soil, the total environmental cost of the Sheet Pile construction method increased as excavation depth increased, and the total environmental cost in soft clay soil was twice as high as that in other soil conditions. Furthermore, the assessments of environmental load and environmental cost were similar.

**Figure 6.** Relationship between total environmental load and total environmental cost by excavation depth and soil condition: (**a**) composite soil; (**b**) sandy soil; (**c**) soft clay soil.

#### **5. Conclusions**

This study evaluated the combination of excavation depth and soil condition in medium-sized excavation ground in order to examine the effect of construction methods on environmental economic feasibility for an earth-retaining wall during soil excavation. LCA analysis of the construction stage of the earth-retaining wall was conducted in consideration of eight environmental impact categories, the criteria for selecting the construction method for the earth-retaining wall considering the environmental costs of each construction method were reviewed, and the following conclusions were obtained as a result of this research:

1. If a calculation is conducted after calculating the environmental load by list analysis of the construction stage, this affects the selection of the construction method for the earth-retaining wall, so it is possible to select an optimal construction method for an earth-retaining wall considering stability and economic feasibility in various soil conditions via selection of a construction method that considers environmental loads in line with international trends.


This study considered only the environmental effect in the determination of the retaining wall. Therefore, research should be conducted on the effect of various cost conditions on sustainability in order to be applied to the site.

**Author Contributions:** Conceptualization, Y.S., S.L. and J.-G.H.; methodology, J.-G.H. and G.H.; software, Y.S. and S.L.; validation, Y.S., J.-Y.L. and G.H.; formal analysis, Y.S., J.-G.H. and G.H.; investigation, S.L. and J.-Y.L.; resources, Y.S. and S.L.; data curation, J.-G.H. and G.H.; writing original draft preparation, Y.S.; writing—review and editing, J.-G.H. and G.H.; visualization, J.-Y.L. and G.H.; supervision, J.-G.H.; project administration, S.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the MSIT (Ministry of Science and ICT), Korea, under the ITRC (Information Technology Research Center) support program (IITP-2020-2020-0-01655), the MSIT(NRF-2019R1A2C2088962) and the X-mind Corps program (2017H1D8A1030599) from the National Research Foundation (NRF) of Korea, the Human Resources Development (No.20204030200090) of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korean government, and the Korea Agency for Infrastructure Technology Advancement funded by the Ministry of Land, Infrastructure and Transport of the Korean government (19SCIP- B108153-05).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request to the corresponding author. The data are not publicly available as they form part of an ongoing study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Environmental and Economic Optimization of a Conventional Concrete Building Foundation: Selecting the Best of 28 Alternatives by Applying the Pareto Front**

**Ester Pujadas-Gispert 1,\*, Joost G. Vogtländer <sup>2</sup> and S. P. G. (Faas) Moonen <sup>1</sup>**


**Abstract:** This research optimizes the environmental impact of a conventional building foundation in Northern Europe while considering the economic cost. The foundation is composed of piles and ground beams. Calculations are performed following relevant building Eurocodes and using life cycle assessment methodology. Concrete and steel accounted for the majority of the environmental impact of foundation alternatives; in particular, steel on piles has a significant influence. Selecting small sections of precast piles or low-reinforcement vibro-piles instead of continuous-flight auger piles can reduce the environmental impacts and economic costs of a foundation by 55% and 40%, respectively. However, using precast beams rather than building them on site can increase the global warming potential (GWP) by up to 10%. Increasing the concrete strength in vibro-piles can reduce the eco-costs, ReCiPe indicator, and cumulated energy demand (CED) by up to 30%; the GWP by 25%; and the economic costs by up to 15%. Designing three piles instead of four piles per beam reduces the eco-costs and ReCiPe by 20–30%, the GWP by 15–20%, the CED by 15–25%, and the costs by 12%. A Pareto analysis was used to select the best foundation alternatives in terms of the combination of costs and eco-burdens, which are those with vibro-piles with higher concrete strengths (low reinforcement), cast in situ or prefabricated beams and four piles per beam.

**Keywords:** ground beam; LCA; prefabrication; vibro-pile; eurocode; precast prestressed concrete pile; continuous flight auger pile; eco-costs; life cycle assessment; economic

## **1. Introduction**

## *1.1. Background*

To keep the global temperature rise preferably at no more than 1.5 degrees Celsius by the end of the century [1], it is necessary to diminish global emissions by more than 50% by 2030 and work towards carbon neutrality by 2050. The construction sector accounts for 36% of final energy use and 39% of carbon dioxide (CO2) related to energy and processes [2]. To date, the focus has been on reducing the energy consumed during the use of buildings. However, embodied energy related to the materials, construction, maintenance, and end of life of buildings is becoming increasingly important [3]. Life Cycle Assessment (LCA) has proven to be a suitable tool to reduce the environmental impact of buildings [4,5]. Nevertheless, uncertainties in the LCA calculation must be minimized and reliable benchmarks must be provided for evaluating buildings [6,7]. The LCA can be carried out at various levels of the system, such as for portions, components, or the entire building [8]. Nonetheless, the foundation, which is the lowest part of the building and in contact with the soil, is rarely assessed despite its considerable impact at an aggregate level, leaving ample room for improvement [9,10]. Consequently, rigorous studies are required to optimize foundations, thus reducing the emissions from the construction sector.

Vogtländer, J.G.; Moonen, S.P.G.(. Environmental and Economic Optimization of a Conventional Concrete Building Foundation: Selecting the Best of 28 Alternatives by Applying the Pareto Front. *Sustainability* **2021**, *13*, 1496. https:// doi.org/10.3390/su13031496

**Citation:** Pujadas-Gispert, E.;


Academic Editor: Sunkuk Kim Received: 12 December 2020 Accepted: 20 January 2021 Published: 1 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

## *1.2. Deep Foundations*

A deep foundation is a type of foundation that depends on the deeper layers of the soil, while a shallow foundation is based on the surface layers of the soil [11]. Deep foundations tend to have a greater environmental impact than shallow foundations due to the greater amount of materials used [12,13]. A typical deep foundation for buildings and other civil works consists of piles, which are vertical structural elements driven or drilled deep into the ground at a building site. Piles tend to work in groups braced by ground beams, pile caps, and slabs. A conventional foundation in Northern Europe is composed of piles and ground beams. The foundation can be either built directly on the ground (cast in situ) or built in a factory (prefabricated) and then transported and installed on site [14]. During the installation, the joints between precast components are connected with steel reinforcing bars and sealed with mortar. Cast in situ beams can be built into different types of formwork (e.g., removable or non-removable, different materials and forms, etc.), and a blinding or plastic foil is often required at the bottom. Below are some of the most commonly used types of piles in Northern Europe [11]:


**Figure 1.** Precast prestressed concrete pile.

**Figure 2.** Fundex pile.

**Figure 3.** Continuous flight auger (CFA) pile.

**Figure 4.** Vibro-pile.

Concretes for precast foundations tend to have higher strengths than cast in situ concrete because precast components are built in a factory where production conditions are more controlled and it is easier to obtain higher strengths. The nomenclature for concrete strength from the Eurocode consists of a capital C followed by two compressive strengths, one measured using concrete cylinder test specimens and the other using concrete cube specimens (e.g., C25/30) [15].

## *1.3. Previous Literature*

Previous studies have reported that concrete and steel account for the highest greenhouse gas (GHG) emissions during the construction of deep concrete foundations (66–90%), followed by equipment usage (12–19%) and transportation (9–16%) [14,16–19]. In addition, piles account for the majority of the environmental impact in a foundation compared to a concrete raft or pile caps. Therefore, it is desirable to reduce the environmental impact of foundation materials, particularly for piles.

The decisions made at the design stage are important because they affect the long-term impact of a foundation [20,21]. For instance, the use of precast concrete piles instead of CFA piles was shown to halve the environmental impact of a foundation, fully counteracting the extra emissions from prefabrication (higher concrete strength, more transport, and installation). The same study also reported that prefabrication may be more expensive than conventionally cast in situ methods, although this is closely related to the nature of each work (e.g., number of units, location). Conversely, Luo et al. [16] reported that precast concrete piles can create 5% more GHG emissions than bored piles, with materials and transportation being points to be optimized. However, the emissions from the construction of precast concrete piles were closely related to the area, cost, and number of piles in a foundation [22].

The combination of variables in the design of a foundation also showed reductions in the GWP. For instance, the selection of the level of prefabrication, concrete strength, type, and calculation codes of a foundation reduced the GWP impact of the foundation by up to 50–60% [14,17]. For this reason, this research also incorporated these variables and others derived from common practice with the intention that the resulting recommendations could be easily implemented in future foundation designs. Furthermore, the study also included some types of piles which are used in Northern Europe (PPC, Fundex, and vibro-piles) that have yet to be environmentally assessed. The economic cost was also included, since the implementation of a new solution in the construction sector depends significantly on its economic cost [16,23,24]. Additionally, a Pareto analysis was performed on building foundations to select the best alternatives in terms of the combination of costs and ecological burdens. Furthermore, the study focuses solely on the foundation and its environmental optimization [9,10], which is normally approached as part of a building [7]. In this respect, prior studies [14,16,17] optimized environmentally part of a foundation (i.e., a pile cap with piles, a footing, and only piles of a foundation); the present study aims to go one step further by considering the entire foundation of the building.

#### *1.4. Objectives*

The research is based on a real case and aims to optimize a conventional foundation for Northern Europe in terms of the environment by considering the economic cost and studying the variables of the prefabrication level (fully precast, semi-precast, and cast in situ), concrete strength of cast in situ piles (C20/25, C25/30, C30/37, C40/50), type of pile (precast prestressed concrete pile (PPC), continuous flight auger (CFA), Fundex, and vibro-pile), and the number of piles per ground beam (3 and 4 piles). The specific objectives are (i) to conduct a structural analysis to determine the dimensions of foundation alternatives; (ii) to calculate and analyze the environmental burden using LCA; (iii) to calculate and analyze the economic cost; (iv) to apply a Pareto analysis (the so-called Pareto front) to select the best solution(s), given a combination of the eco-burdens and costs; and (v) to assess the influence of the study variables on the environmental burden and the economic cost of a foundation and, in doing so, define specific design conclusions and recommendations.

## **2. Materials and Methods**

The integrated methodology applied to determine the environmental influence of the study variables includes a selection of equivalent alternatives (Section 2.1), an explanation of a case study (Section 2.2), a definition of a functional unit (FU) (Section 2.3) as well as system boundaries (Section 2.4) and quantitative model (Section 2.5), an explanation of the foundation design (Sections 2.6 and 2.7) and LCA (Section 2.8), and a compilation of the data sources used (Section 2.9).

#### *2.1. Selection of Equivalent Alternatives*

The type of foundation and the study variables are common in usual Northern Europe practices. The abbreviations used to designate the study alternatives are shown in Table 1.

**Table 1.** Abbreviations used in the study.


Example: 1-C600-20/I3.650 is foundation number 1 and is composed of C600-20 piles and I3.650 beams. The C600-20 pile is a continuous flight auger pile 600 mm in diameter with a concrete strength of C20/25. I3.650 is a cast in situ beam supported by three piles with a width of 650 mm.

#### *2.2. Case Study*

The reference project is a neutral energy housing project in Vianen (The Netherlands) (Figure 5) which is composed of 16 buildings with similar characteristics. The foundation of building 13 was selected for assessment (Figure 6). The foundation is composed of precast concrete piles and ground beams (Figure 7). The main characteristics of the foundation of the reference project are shown in Supplementary Material S1. The soil is composed of loose layers of sand that increase in resistance to levels of 8–10 m where the piles are embedded. The reference project has been adapted to better analyze the influence of the variables on the study results. For instance, the prefabricated concrete walls were turned into sand-lime brick and the partition walls into two 120 mm sand-lime bricks. Table 2 shows the building materials and permanent and variable loads considered in the study. Finally, wind load was not considered, given that it does not alter the results of the study.

**Figure 5.** Aerial photo of the reference project [25].

**Figure 6.** Foundation of building 13 of the reference project in Vianen.

**Figure 7.** A foundation of the reference project [25].



#### *2.3. Functional Unit*

The functional unit (FU) is a conventional foundation that consists of concrete ground beams and piles considering different levels of prefabrication, types of pile, concrete compressive strength, and number of piles per beam for a useful life of 50 years.

#### *2.4. System Boundaries*

Figure 8 shows the phases of the LCA and the elements considered in each of them. The phases comprise steps from the extraction of raw materials to the construction of the foundation on site. The various transports are included in the corresponding phases (e.g., the transport of raw material, products, waste, soil). The levelling of the ground prior to building the foundation was not considered because it is very specific to each work and does not alter the study comparison. Nevertheless, the excavation of each foundation has been considered because it is different depending on the alternative. The pumping of concrete and the transport of the machines to the site were excluded because a preliminary

study showed that their environmental impact was low. The use phase was also excluded because well-designed foundations tend to not require maintenance. Similarly, the end of life was not considered because recycling or reusing foundations is not the norm, although there is great potential in this area [7]. The piles are usually left installed in the ground, although precast ground beams, which are a very recent foundation element, have not yet reached the end of their useful life. However, in cases where the precast ground beams were built in temporary buildings (10 and 15 years), which is not the usual case, at the end of their useful life, they were (i) reused in the same building, (ii) reused in another building. or (iii) dismantled and demolished. In cases (i) and (ii), the end of life was considered in the design stage of the beams.


**Figure 8.** Life cycle diagram and system boundaries of the construction of the foundation alternatives.

#### *2.5. Quantitative Model for Environmental Impact Category Calculation*

Below the quantitative model used to calculate the eco-burdens for each category is shown [17].

$$E = \sum\_{i=1}^{n} \sum\_{j=1}^{m} p\_i \,\mathcal{Q}\_{ij} \tag{1}$$

where *E* is the single indicator score (i.e., Eco-costs 2017, ReCiPe 2016 Endpoint World (2010) H/A, GWP 100-year 2013 Intergovernmental Panel on Climate Change and cumulated energy demand) of the FU; *n* is the total number of reference flows; *i* is the reference flow (i.e., material, diesel, formwork, or transportation); *m* is the total number of phases in the life cycle (i.e., 4); *j* is the phase of the life cycle (i.e., raw material extraction, production, earthworks, or construction); *pi* is the combined factor for characterization, normalization, and weighting per unit of reference flow (e.g., factor per kg or m2 material); and *Qij* is the quantity of reference flow in an FU phase.

## *2.6. Structural Design of Ground Beams*

Loads were calculated according to NEN-EN 1990+A1+A1/C2:2019/NB:2019 [26] and NEN-EN 1991-1-1+C1+C11:2019/NB:2019 [27]. Reinforced concrete beams (cast in place and precast) were designed following NEN-EN 1992-1-1+C1:2011/NB:2016+A1:2020 [28]. Calculations of the beams were performed with Technosoft Balkroosters [29]. All the beams were designed for an XC3 environmental class (moderate humidity) according to Table 4.1 in [28], with concrete covers of 35 mm for cast in situ beams and 30 mm for precast beams complying with article 4.4.1 in [28]. In addition, cast-in-place beams were designed to be cast in a common removable timber formwork [30] so that all the alternatives could be compared without insulation. Additionally, they were designed to be built over a 50 mm concrete blinding [28]. The formwork was considered to be used 5 times. The concrete strength considered for cast in situ beams was C30/37, that for concrete blinding was C12/15, and that for precast beams was C40/50, which is aligned with normal practices. However, different strengths of concrete for ground beams were not considered because a preliminary study showed that increasing the strength of concrete in

beams minimally increases their resistance, because beams work mostly in bending rather than in compression. Ground beams were dimensioned following the usual Dutch practice. The height was set at 500 mm for all the ground beams (cast in situ and precast). For the cast-in-situ ground beams, a minimum width of 500 mm for transverse walls and 400 mm for longitudinal walls was established so that the width of the beams matched the width of the walls. However, for the precast floor beams, this was not necessary because the consoles, which are elements that protrude from both sides of the top of the beam, allow the beam to be adjusted to the width of the wall. It should be mentioned that consoles were not included in the material calculation of the study due to their low impact on the results. Finally, various PPC pile and wide beam foundation options were discarded from the study because they were oversized and unrealistic for reasons of economic cost, transportation, and installation.

## *2.7. Geotechnical and Structural Design of Piles*

The bearing capacity of the piles was calculated following NEN 9997-1+C2:2017 [31]. All the piles were designed for an XC4 environmental class (cyclic wet and dry) according to Table 4.1 in [28]. The normal loads to the piles were 510 kN for the four-pile beams and 760 kN for the three-pile beams. Concrete covers were designed to be 70 mm for CFA piles, 50 mm for Fundex piles, 40 mm for vibro-piles, and 30 mm for PPC piles, which is in line with article 4.4.1 in [28]. The concrete strengths considered for cast in situ piles were C20/25, C25/30, and C30/37, complying with NEN-EN 1536:2010+A1:2015 [32], and C35/45 for PPC piles according to product specifications. In most cases, the heads of the piles were not cut off. However, in Fundex piles, it has been considered that the pile head is cut off (1 m), which aligns with the usual practice. Afterwards, the heads of all types of piles were rebuilt with mortar, and a reinforcing bar was placed on top of each pile to connect them with the precast ground beams, which was not necessary for the cast-in-situ beams, since they were built directly on top of piles.

In terms of reinforcement, all the pile types were reinforced with the minimum amount of reinforcement, which is specified in Table 9.6 from [28], and an additional reinforcement was also calculated to resist the bending moment at the head of the pile (38 kNm in three-beam piles and 25.5 kNm in four-beam piles). This bending moment arises from considering a 50 mm eccentricity in the structural calculations to cover possible unforeseen events, such as construction misalignments and horizontal forces. As a result, a larger pile diameter may be required to arm this bending moment. Furthermore, the minimum diameter considered for the longitudinal reinforcement was 12 mm in all piles [28], while for stirrups it was considered to be 6 mm in CFA piles, 8 mm in vibro-piles, and 5 mm in Fundex piles according to article 10.2.4 in [33]. The reinforcement for PPC piles was retrieved from product specifications. Finally, most of the piles were reinforced at all lengths, except CFA piles, which were reinforced only at the three superior meters [33], and an additional bar of Ø20 mm was arranged in the center of the CFA pile to compensate for the weakness of the superior layers of the soil (article 7.1.7 in [32]). Regarding cast in situ piles, the minimum cement content considered was 375 kg/m3, complying with Annex D of [15], and a reduction in the pile diameter was applied to cover the uncertainty of building a pile directly on the ground following article 2.3.4.2 in [28].

The higher the concrete strength in the piles, the less steel reinforcement is required. It is worth mentioning that a higher concrete strength was not considered once the cast in situ piles reached the minimum steel reinforcement because it did not add information in the study. As previously mentioned, the environmental impact of the cubic meter of concrete is the same for the different concrete resistances in this study following the Dutch regulations. Therefore, rather than looking for the best option that may depend on each case (loads, soil, etc.), it is intended to detect the influence of the study variables on the environmental and economic results of the alternatives to consider them in future designs and codes of foundations.

## *2.8. Life Cycle Assessment*

The LCA method was applied to determine the environmental impact, as defined in international codes [34,35]. The software SimaPro version 9.2 [36] was used for the calculations. Since LCA is used here to determine whether "system A is better than System B", so-called "single score methods" are applied, as has been recommended by the Society of Environmental Toxicology and Chemistry [37] and the Joint Research Centre of the EU [38]. The calculations were performed for 4 indicator types: (i) carbon footprint, as a "single issue method"; (ii) ReCiPe, as a "damage-based method"; (iii) eco-costs, as a "monetized prevention-based method"; (iv) and embodied energy, as a "single issue method". In their sectors, these four indicators are the most applied methods in science. The eco-costs comply with [39]. Although monetization in LCA is not very common, the advantage of eco-costs is that they are so-called "external costs" (i.e., costs for our society that are not incorporated in the price of a product), so they have a direct meaning to architects, business managers, and governmental policy makers. Recently, there have been increasing applications of eco-costs in the building industry—e.g., for concrete construction [40] and beams [41]. Eco-costs are also applied in full cost accounting (FCA), which is also called true cost accounting (TCA). The basic philosophy behind TCA is that the external costs (=environmental burden) of a product should be added to the economic costs to enable a fair comparison in product benchmarking between a cheap but polluting product and a "clean" product. Another way to address the issue of "ecology versus economy" is to display the external costs and the economic costs in a two-dimensional graph and determine the Pareto front (being the best solution). Section 3.4 explains how such a Pareto analysis works in practice.

## *2.9. Data Sources*

The economic and construction data were mainly provided by leading foundation and concrete companies in the Netherlands. Data from Vroom Funderingstechnieken [25] provided the quantities of diesel needed to install the various types of piles (CFA, Vibro, Prefab, and Fundex) and precast beams and the quantities of sacrificial steel for Fundex and Vibro piles. The specifications for the PPC piles were provided by a Dutch precast concrete company. Mebin B.V. [42] provided concrete dosages for cast in situ foundations. Most of the economic data were obtained from the EcoQuaestor database [43], except for the cast in situ concretes, which were provided by Mebin B.V. The installation and removal of machinery on site were not considered in either the environmental or economic costs due to their marginal influence on the study results.

Environmental data were retrieved from the Ecoinvent v.3.5 [44] and Idemat database [45]. The various piles used different types of steel. Nonetheless, the same steel was considered for all reinforcements (steel reinforcement with a working process) because not all steels were found in the consulted databases. In this regard, the importance of retrieving the data from the same databases should be remarked on to ensure a fair comparison. Please see Supplementary Material S2 for more information on the materials/processes and quantities introduced in SimaPro. The transport distances for materials and components were obtained from the literature and are summarized in Table 3. Note that these distances are for a trip by truck; however, in the study two trips by truck were considered (one to deliver the product and then another to return to the empty truck to the factory).


#### **Table 3.** Transport distances used for calculation.

#### **3. Results and Discussion**

The results of the research are presented and discussed in the subsections below: structural results of the foundation alternatives (Section 3.1), environmental results of only piles (Section 3.2.1), environmental results of foundation alternatives (piles and beams) (Section 3.2.2), economic results of the foundation alternatives (Section 3.3), and the economic-environmental results of the foundation alternatives (Pareto front) (Section 3.4). Please find absolute values of the environmental and economic results in Supplementary Material S5.

## *3.1. Structural Results*

Table 4 shows the study alternatives along with the main characteristics of the alternatives for later conducting the environmental and economic analysis. Please consult Supplementary Materials S3 and S4 for more information on the structural results.


**Table 4.** Main characteristics of the foundation alternatives.


**Table 4.** *Cont.*

Terminology: foundation alternative number − type of pile (continuous flight auger (C), Vibro (V), Fundex (F), precast prestressed concrete (P) pile) + pile diameter/side (mm) − pile concrete strength (C20/25 (20), C25/30 (25), C30/37 (30), C35/45 (35), C40/50 (40))/Cast in situ (I) and precast (P) beams + (3) and (4) piles per beam + width of the beam (mm). \* Reinforcement all the length of the pile.

#### *3.2. Environmental Results*

3.2.1. Piles

Figure 9 compares the GHG emissions of various piles from alternatives with cast in situ beams and three piles per beam. The piles from foundation alternatives with precast beams and four piles per beam are not shown, as they display a similar trend.

If we observe the environmental impact of each element in the construction of a pile (e.g., reinforcement, transport, etc.), concrete and steel play an important role in the environmental results of all piles, representing 65–95% of the impact. In addition, pile driving (i.e., drilling, driving, etc.) can represent up to 20% of the GWP and CED from pile construction, as is the case with Fundex piles, given that for this type of pile diesel is needed not only for piling but also for an external unit for pumping concrete. Additionally, the sacrificial steel in the Fundex pile type can represent up to 15% of GWP, up to 25% of eco-costs, and 20% of ReCiPe and CED. Transportation in precast piles accounts for 15–20% of the environmental impact. However, the transportation of waste derived from installation in all piles has little impact on the environmental results.

If we compare the environmental impact between the piles, it can be observed that piles with the least amount of reinforcement obtained the best environmental results, namely precast piles with the smallest cross-sections (e.g., P250-35) and vibro-piles with the highest concrete strengths. On the other hand, the types of piles that obtained the worst environmental results were those with large amounts of concrete and/or steel. These included CFA piles (e.g., C600-20), vibro-piles with low concrete strengths and/or large amounts of reinforcement (e.g., V305-20), and Fundex piles with larger cross-sections (e.g., F460-20).

If we analyze the influence of the study variables on the environmental results of piles, we can see that prefabricated piles with the smallest cross-sections obtained the best environmental results (e.g., P250-35). Additionally, the increase in concrete strength in vibropiles led to reductions in GHG emissions, which is in line with a previous study [17]. Finally, four-pile ground beam piles obtained better results than three-pile ground beam piles because the latter have more steel reinforcement to compensate for higher buckling loads.

**Figure 9.** Relative impact of piles from beams with three piles considering the indicators of GWP, Eco-costs, ReCiPe, and CED. Terminology: type of pile (continuous flight auger (C), Vibro (V), Fundex (F), precast prestressed concrete (P) pile) + pile diameter/side (mm) − pile concrete strength (C20/25 (20), C25/30 (25), C30/37 (30), C35/45 (35), C40/50 (40)).

#### 3.2.2. Foundation Alternatives (Piles and Beams)

Materials are the main contributor to environmental impact (85–95%) in all indicators of the foundation alternatives, which is aligned with previous studies [14,17]. Figure 10 shows some of the relevant environmental results to allow for a discussion of the influence of study variables.

In terms of prefabrication, foundation alternatives with small cross-section precast piles (PPCs) obtained the best environmental results in the study (e.g., 9–28). However, the use of prefabricated beams instead of cast in situ beams increased the environmental impact of the foundation alternatives by up to 5% in terms of eco-costs, ReCiPe, and CED and 10% in terms of GWP. This is because concrete in precast beams has a greater impact (more cement) than concrete in cast in situ beams. In this sense, concrete has a special effect on the GWP indicator, while steel is on the eco-cost indicator. Furthermore, precast ground beams require transportation to and installation at the building site, which far outweighs the impact of concrete blinding and the larger volumes of concrete in cast in situ beams. Most likely, as the beam becomes wider (precast or cast in situ), the amount of concrete and steel increases, and consequently, the environmental impact of the beam is higher. However, the prefabrication of ground beams might be interesting from a cradle-to-grave perspective [48], as they can have a prolonged service life (from reuse).

The design of the piles has an important effect on the environmental results of the hole foundation. Nevertheless, the type of pile itself is not a guarantee that the foundation alternative is sustainable, although the reduced amounts of concrete and particularly steel are sustainable. Nevertheless, it should be noted that some types of piles use fewer materials and resources than others to support the same load. The foundation alternatives that resulted in the lowest environmental impact compared to the worst environmental result were those with small cross-section PPC piles (9, 28), which obtained up to 55% lower environmental impacts, and vibro-piles with low amounts of reinforcement, which obtained up to 45% smaller impacts (e.g., 21→4Ø12). Surprisingly, these alternatives

have the lowest amounts of steel reinforcement in the piles examined in this study. In contrast, the foundation alternatives that inflicted the greatest environmental impact were those with vibro-piles with large amounts of reinforcement (e.g., 11→8Ø20), CFA piles (e.g., 1→diameter 600) with large amounts of concrete, and Fundex piles with moderately high amounts of concrete and steel, which includes the sacrificial steel (e.g., 8→ diameter 460 + 7Ø12 + 160 kg tip).

**Figure 10.** Relative impact of the eco-costs, ReCiPe, GWP, and CED of relevant study foundation alternatives. Terminology: foundation alternative number − type of pile (continuous flight auger (C), Vibro (V), Fundex (F), precast prestressed concrete (P) pile) + pile diameter/side (mm) − pile concrete strength (C20/25 (20), C25/30 (25), C30/37 (30), C35/45 (35), C40/50 (40))/cast in situ (I) and precast (P) beams + (3) and (4) piles per beam + width of the beam (mm).

The increase in concrete strength in vibro-piles from C20/25 to C40/50 reduced the eco-cost, ReCiPe, and CED impacts by up to 30% and GWP by 25% (e.g., 2–4). Similarly, the increase in concrete strength from C30/37 to C40/50 reduced the eco-costs and ReCiPe impacts by up to 25%, and GWP and CED by up to 20% (e.g., 3–4). Surprisingly, vibro-piles that differed only in concrete strength had a similar impact from concrete but a different impact from steel (e.g., 3 and 4). This is explained because the concrete for cast in situ piles must have a minimal cement content of 375 kg/m3 according to Dutch regulation [15], which assimilates the impact of concrete between cast-in-situ piles. However, the impact of steel between these piles is different because as the higher concrete strength is, the less steel reinforcement is required, since concrete contributes more to resisting the forces.

In terms of the number of piles in beams, four-pile foundations obtained 20–30% fewer impacts in eco-costs and ReCiPe, 15–20% in GWP and 15–25% in CED compared to those with three piles (e.g., 3–21). This is because less steel is required to compensate for the bending moments and shear forces in the beams and buckling forces in the piles.

Given the same beam, alternatives with vibro-piles that fit the beams better (i.e., larger piles) obtained 25–40% lower eco-costs and ReCiPe and up to 30% lower GWPs and CEDs compared to smaller diameter piles (e.g., 11–14). This is because a larger diameter pile has a higher capacity since the axial forces are distributed over a larger surface and therefore the internal forces are smaller, reducing the required steel reinforcement.

## *3.3. Economic Results of Foundation Alternatives*

Piles accounted for 40–60% of the cost of foundation alternatives. Figure 11 shows some of the representative economic results of the study to allow for a discussion of the study variables. Foundation alternatives with piles with larger concrete pile crosssections incurred the highest economic costs (e.g., 1, 10, 16, 17), while foundations with low-reinforcement vibro-piles with precast beams (e.g., 15, 26) and cast in situ beams (e.g., 6, 21) were the most economical options (up to 40% cheaper). The increase in the strength of concrete in vibro-piles considerably reduced the cost of the foundation by 7–10% from C20 to C30 and from C30 to C40 and up to 15% from C20 to C40. This is because as the strength of the concrete increases, less steel reinforcement is required, although the cost of concrete is slightly higher. Moreover, foundations with four piles per beam were 7–12% more economical than foundations with three piles because the former have less steel reinforcement (e.g., 3–21). Alternatives with vibro-piles that fit the beams better (i.e., larger pile for the same beam) were up to 20% cheaper. Finally, foundations with the same piles and with precast or cast in situ beams obtained similar economic costs (approximately 5% up and down).

**Figure 11.** Economic cost of relevant study foundation alternatives. Terminology: foundation alternative number − type of pile (continuous flight auger (C), Vibro (V), Fundex (F), precast prestressed concrete (P) pile) + pile diameter/side (mm) − pile concrete strength (C20/25 (20), C25/30 (25), C30/37 (30), C35/45 (35), C40/50 (40))/cast in situ (I) and precast (P) beams + (3) and (4) piles per beam + width of the beam (mm).

## *3.4. Environmental-Economic Results of Foundation Alternatives (Pareto Front)*

The best choices on the basis of both economy and ecology are shown in Figures 12 and 13. In a one-dimensional system, there is one best choice, but in a two-dimensional system the best choices are given on a line: the so-called Pareto Front. A solution at the Pareto Front has no better alternative in the sense that there are no alternatives that score better on eco-costs and, at the same time, on costs. Such a solution is called Pareto Optimal, or Pareto Efficient [49]. There are many examples of this mathematical concept of multi-objective optimization (MOO), often related to the costs of energy conservation systems in the building industry [50], the refurbishment of buildings [51], and industrial processes [52]. The method is used to select the best solutions out of a cloud of alternatives and has the advantage that such a selection is still free of subjective choices.

**Figure 12.** Pareto front of all foundation alternatives, eco-costs vs. costs. Note that the dotted line also depicts true costs = eco-costs + costs = constant (e.g., solutions 26 and 21 have the same true costs).

**Figure 13.** Pareto front of all foundation alternatives, 100-year GWP vs. costs.

A final choice of the best solution on the Pareto front, however, is a matter of a subjective choice—in this case, the relative importance of the eco-costs versus the costs (i.e., "how important is the ecology compared to the economy?"). A recent approach is to minimize the "true costs", where "true costs" = "eco-costs" + "costs" [53], and where "true costs" = "constant" is a straight line in Figure 12. In our case, it is a co-incidence that the Pareto front falls along this straight line. That means that, in this case, an additional subjective criterion must be applied to make a final choice. When the carbon footprint in Figure 13 is expressed in terms of money (e.g., the (future) price of carbon allowances, or the "eco-costs of carbon footprint"), the additional criterion of minimizing the true costs leads to a final choice.

The foundation alternatives that obtained the lowest results from this perspective were those composed of vibro-piles with a medium section (305 mm), low reinforcement (4Ø12), higher compressive concrete strengths (C30), four-pile cast in situ beams (21), and four-pile precast beams (26). It is worth mentioning that the same alternative with greater

concrete strength (C40) was not considered because the amount of steel in the pile (21) is already the minimum established by regulations and therefore would have obtained similar environmental results. Second, the best-rated foundation alternatives were those with vibro-piles with a wide cross-section (356 mm), low reinforcement (5Ø12), the highest compressive concrete strength (C25), four-pile precast beams (15) and three-pile cast in situ beams (6). Surprisingly, these alternatives (21, 26, 6, 16) have the lowest amounts of steel reinforcement in cast in situ piles. Prefabricated piles with small cross-sections have even reduced amounts of reinforcement (9, 28). The alternatives with prefabricated piles of small cross-sections obtained the best environmental results (Section 3.2.2.), although from an environmental and economic perspective the aforementioned alternatives with vibropiles obtained better results because vibro-piles are cheaper (according to the database consulted). In contrast, the alternatives that obtained the worst results were the CFA piles with three and four piles per beam cast in situ (1, 17), as they required the highest amounts of concrete. Second, the worst-rated foundation alternatives were those with Fundex piles (8) and PPC piles (10, 16) with large pile cross-sections and highly reinforced vibro-piles (11). It should be mentioned that the larger the section of a pile was, the greater the width of the beam would be, thus increasing the amount of concrete.

## **4. Conclusions**

An assessment has been presented, from an environmental and economic perspective, of the construction of a conventional building foundation composed of piles and ground beams according to the variables of level of prefabrication (fully precast, semiprecast, and cast in situ), the concrete strength of cast in situ piles (C20/25, C25/30, C30/37, C40/50), the type of pile (precast prestressed concrete (PPC), continuous flight auger (CFA), Fundex, and Vibro piles), and the number of piles per beam (3 and 4 piles). Some of the main conclusions of the study are summarized below.


• From an environmental and economic perspective (Pareto front), foundations should have low amounts of concrete and especially steel in the piles. Thus, it is recommended to use piles with reduced cross-sections, as their width also conditions the amount of materials in beams. However, reducing the pile cross-section may increase the amount of steel reinforcement in piles, and then the use of a higher concrete strength can moderate steel amounts. Vibro-piles with higher concrete strengths with cast in situ or prefabricated beams and four piles per beam are the most recommended alternatives from this perspective. Alternatives with CFA piles, Fundex piles, and PPC piles with large cross-sections and three piles per beam are the least recommended.

This paper has highlighted that changing certain common variables in the design of a foundation can significantly reduce the environmental and economic cost of the construction of a foundation. However, it must be considered that each type of pile has its optimal application that depends on many factors, such as the type of soil, loads, regulations, and tradition. Besides this, the economic cost of construction can be variable depending on the location and size of the work and the construction company, among other things. However, an attempt was made to minimize this uncertainty—on the one hand, by starting from a real case that was built, and on the other hand by using the databases and resources conventionally used in real practice. Future research could consider the study variables to optimize other constructive elements (slabs, etc.) and include other variables that lead to a significant reduction in environmental impact and economic cost. Likewise, it is interesting to consider other more sustainable materials (geopolymer concrete, biobased reinforcement, etc.) as well as other optimized designs [48]. This research aims to influence future buildings and codes to contribute to the improvement of environmental sustainability in the construction sector.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2071-105 0/13/3/1496/s1.

**Author Contributions:** E.P.-G. was responsible for data curation, investigation, writing—original draft, and writing—review and editing; J.G.V. was responsible for data curation and writing—review and editing; S.P.G.M. was responsible for supervision and writing—review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We would like to thank the following people for their advice, without whose help this work would never have been possible: Viola Friebel, Justin van der Eerden, Leo Dekker, Tim de Jonge, and Rawaz Kurda. We also want to thank Vroom Funderingstechnieken, Mebin B.V., and EcoQuaestor for the help provided.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


## *Article* **Towards Effective Safety Cost Budgeting for Apartment Construction: A Case Study of Occupational Safety and Health Expenses in South Korea**

**Kanghyeok Yang 1, Kiltae Kim <sup>2</sup> and Seongseok Go 1,\***


**Abstract:** The construction industry has experienced a lot of occupational accidents, and construction work is considered one of the most dangerous occupations. In order to reduce the number of occupational injuries from construction, the South Korean government legislated the occupational safety and health expense law, requiring companies to reserve a reasonable budget for safety management activities when budgeting for construction projects. However, safety budgets have not been spent based on the risk of accidents, and a large amount of the safety budget is spent either in the beginning or late stages of construction projects. Various accident risk factors, such as activity types, previous accident records, and the number of workers on a construction site, need to be considered when determining the safety budget. To solve such problems, this study investigated the expenditure trends of occupational safety and health expenses for 10 apartment construction projects in South Korea. This study also proposed an accident risk index that can be incorporated with the project costs, schedule, the number of workers, and historical accident records when budgeting for the safety costs. The results from the case study illustrate the limitations of the current planning strategy for safety expenditures and demonstrate the need for effective safety budgeting for accident prevention. The proposed safety cost expenditure guideline helps safety practitioners when budgeting for the occupational safety and health expenses while considering accident risk and the characteristics of safety cost expenditures in practice. The outcome of this research will contribute to the development of regulations for the budgeting of safety costs and help to prevent occupational injuries by providing a reasonable budget for safety management activities in an apartment construction project.

**Keywords:** occupational safety and health expenses; construction safety; safety cost expenditures; apartment construction

## **1. Introduction**

The construction industry in South Korea has rapidly grown over the last few decades [1], with infrastructure and residential facility constructions to accommodate the rapid expansion of the major cities. However, the construction industry has experienced a lot of occupational injuries, and construction work is considered one of the most dangerous occupations due to the dynamic and temporary nature of the workplace [2–4]. Specifically, most construction work takes place outdoors and work conditions (e.g., temperature, humidity, and light conditions) and the number of required workers frequently changes, which increases the difficulty of safety management. According to the Korea Occupational Safety and Health Agency (KOSHA), fatal injuries of construction workers have been increasing since the year 2000. In 2018, the construction industry experienced the highest number of fatal accidents, accounting for 49.95% of total fatalities in South Korea [5]. The safety of construction workers is a global issue. In 2011, the construction industry employed almost

Towards Effective Safety Cost Budgeting for Apartment Construction: A Case Study of Occupational Safety and Health Expenses in South Korea. *Sustainability* **2021**, *13*, 1335. https://doi.org/10.3390/su13031335

**Citation:** Yang, K.; Kim, K.; Go, S.


Academic Editor: Sunkuk Kim Received: 10 December 2020 Accepted: 19 January 2021 Published: 27 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

7% of the world's workforce, while the industry recorded 30–40% of the world's fatal injuries [6]. There are many different contributing factors associated with the occurrence of fatal accidents [7–9], but one of the major issues is the lack of appropriate countermeasures to reduce the risk of accidents in construction environments [10]. The South Korean government legislated the occupational safety and health expense law to ensure companies secured a minimum safety management budget, the size of which depends on the size and type of the construction project. The budget for safety costs for a general construction project must equal or exceed 1.86% of the total material and labor costs. Construction projects that have budgets less than 500 million Korean republic Won (KRW) are required to have a higher ratio (e.g., 2.93%) compared to that of projects with more than 500 million KRW of construction costs, to help protect workers in smaller construction projects.

The safety expense law aims to enhance the safety of construction workplaces and restricts the use of the safety budget to certain types of expenses. Specifically, the safety budget is only available for performing safety management activities such as purchasing personal protective equipment, safety education, safety consulting by experts, and so on. The safety expense law also includes a regulation that requires a construction firm to spend a certain amount of the safety budget according to the progress of the construction project. This regulation is adopted to effectively protect workers from an accident by requiring the firm to spend money on safety, but the requirement is insufficient for accident prevention. The regulation enforces construction firms to spend more than half of the budget before completion of 70% of the construction project. Under these conditions, safety budgets are spent either at the early or the late stages of the construction project. In addition, safety related studies argue that safety-cost planning in practice does not consider the risk of ongoing activities, which is not suitable to effectively prevent accidents at construction sites [11,12]. To reduce the accident risk and protect construction workers from accidents, safety costs should be allocated based on the risk of the on-going construction activity. In addition, the safety budget needs to be allocated while considering various safety-related risk factors such as the number of workers, historical accident records, and other site conditions. In short, there is a definite need to analyze how to effectively use the safety budget to decrease the risk of accidents and to advance the safety of construction sites.

To address current issues in safety-cost planning, this study first investigated the budgeting and execution of safety and health expenses by conducting a case study analysis of 10 apartment construction sites in South Korea. The results from this case study illustrated the current problems in safety-cost budgeting and executions in practice. In addition, data on factors related to accident risk (e.g., cost, schedule, number of workers on-duty) were also collected from case study sites to comprehensively assess the accident risk during construction projects. The accident risk index was lastly proposed to consider the abovementioned risk factors (i.e., construction schedule, construction costs, number of workers, and historical accident records) in the effective planning of budgets for safety costs. The recommendation for the expenditure of the safety budget is presented to facilitate the outcome of this study and help safety practitioners perform effective safety management activities. The remaining sections of the manuscript are organized as follows. The research background reviewed the previous research on accident prevention, safety management activities, and safety cost budgeting. The material and method section describe how to compute the accident risk index and analyze the trend of safety cost budgeting in apartment constructions. The remaining sections explain the results of the analysis and the conclusion of this research.

## **2. Literature Review**

The safety of a construction worker is an important issue in many nations, since construction environments are complex and often unsafe due to their dynamic and laborintensive characteristics (e.g., largely relying on a worker's labor and heavy equipment) [13]. In addition, construction works are often placed at elevations that could highly increase the risk of accident. The weather conditions are other factors that could adversely affect

the safety of a worker on a construction site. As a result, the construction industry has recorded a poor safety performance and experienced a lot of fatal and non-fatal injuries [14]. According to Shafique and Rafiq (2019), the construction industry accounted for around 20% of the occupational fatalities that occurred in Japan, United Kingdom, the United States, and Hong Kong in 2017 [15]. The research from the Workplace Safety and Health institute (WSH) illustrated that construction sites in Asia experienced a greater number of fatal injuries compared to the sites in other continents. In South Korea, a large number of fatal injuries also occurred during construction [5]. Among the various types of accidents that can occur, falling from a great height is the leading cause of fatalities [16] and therefore the prevention of fall accidents is a critical issue for the safety of construction workers [17–20]. The Occupational Safety and Health Organization (OSHA) in the United States forced employers to provide a fall protection system that can prevent fall accidents when the work surface is located over 1.82 m (i.e., 6 feet) above the ground or a lower floor. The guardrail, safety net, and personal fall arrest system are the examples of the fall protection systems and the employees should not start their works before the installation of such fall protection systems in the workplace.

The occurrence of accidents are related to the various factors and previous research emphasized the significance of two accident-related factors, which are unsafe work environments and psychological/behavioral characteristics of an individual worker [21–23]. The above-mentioned fall protection systems are used for improving the safety of the workplace by modifying the work environments. However, an individual's unsafe behavior is a persisting issue, since a large portion of construction works are performed by a worker's hand or by manually using equipment. Behavior-based safety is the one solution that can prevent accidents originating from a worker's unsafe behaviors [24]. Several previous studies indicated that more than 80% of accidents could be attributed to a worker's unsafe behaviors [25–27]. Ascending/descending using stairs without holding a guardrail is an example of a worker employing unsafe behavior. Poor housekeeping in a construction site is a result of unsafe worker behavior that can involve neglecting activities such as storing equipment or cleaning the floor after completing a task. A worker's unsafe behaviors are often triggered by factors such as needing to meet excessive production targets, a competitive atmosphere, a tight construction schedule and a lack of available resource [28]. Also, inappropriate safety management activities conducted by the safety manager could strengthen a worker's attitude toward unsafe behavior during the construction process. In short, the prevention of accidents during construction is a complicated issue and it requires various efforts to be addressed including the improvement of the work environment, safety related education, safety observations, and proper safety interventions. Also, financial resources for safety management activities are vital for the success of accident prevention in construction.

Safety management is an important research topic to decide the proper amount of safety costs and to quantity the risk of accidents occurring for effective safety budget allocations. A study by Pinto et al. (2011) analyzed the financial costs of constructionrelated accidents. The occupational injuries assessed did not only badly affect the worker's wellbeing but also adversely affected the cost of the construction projects due to requiring high medical costs [29]. According to the analysis from Everett and Frank (1996), the occupational injuries from non-residential construction projects account for 7.9% to 15% of total construction costs [30]. This research illustrated that the prevention of occupational injuries is essential for both a worker's safety and the success of a construction project. The risk of accidents occurring is commonly defined as the significance of these risky events in terms of the occurrence probability and the severity of a potential injury [31–33]. The previous risk assessment studies utilized the analytic hierarchy process (AHP) technique which is a structured multi-attribute decision method for complex decision making—while maintaining consistency of experts' judgements [34]. The AHP technique has been utilized to rank various safety factors by assessing the severity and the probability of accidents [35] or injuries [36]. Such a risk assessment technique is beneficial for effectivity quantifying the

risk level of accident-related factors, but the process largely relies on subjective decisions, which are prone to being biased. Also, assessment results from the previous studies are not suitable for safety cost budgeting at the project level, since they were conducted to rank different types of hazards. Further research on developing a safety risk index that includes the influences of the factors related to the risk of accidents is essential for effective safety cost budgeting at the project level.

Occupational safety and health expenses were legislated by the Korean government under the law to require securing appropriate budgets for safety management activities. The amount of safety costs required are determined based on the type and size of the relevant construction projects. Specifically, the safety budget is a proportion of the total labor and material costs. Table 1 presents the 15 categories for safety cost budgeting, which is classified by the type of works being conducted (i.e., 5 different construction types) and the total amount of construction costs (i.e., 3 different cost ranges). The usage of the safety cost is limited to the (1) labor costs of safety managers, (2) costs for protective equipment, (3) costs for personal protective equipment, (4) external safety inspection or consulting fees, (5) costs for safety education, (6) health care fees for workers, (7) safety technology fees, and (8) costs for the safety organization to be established in the construction headquarters. The safety costs play an important role in enhancing the safety level of the construction site; however, the allocations of the safety budget are still not optimal in terms of the prevention of accidents in the construction industry. For example, the risk levels for construction works are different depending on the stage of the construction project. As previously described, the risk level of falling accidents is not significant at the initial stage of the construction project, since the excavation and the foundation works are the main construction activities being completed at this stage. Also, the most of construction works at the late stage of the construction project are the finishing works, which are generally performed when the structural works of the building are completed. As a result, the risk of falling accidents at the late stage would not be significant compared to during the middle stage of the construction project. A safety cost expenditure guideline is beneficial to effectively allocate the safety budget for accident prevention and prevent occupational injuries. In this context, this study firstly investigated the expenditures of safety budgets using data from apartment construction projects and proposed an accident risk index and safety cost expenditure guideline to enhance the safety of construction workplaces and protect construction workers.


**Table 1.** Occupational safety and health expense rates by type and size of construction projects.

\* KRW: Korean Republic Won. \*\* defined by the Occupational Safety and Health Act in South Korea.

## **3. Materials and Methods**

## *3.1. Occupational Safety and Health Expenses in Case Study*

This research study analyzed the expenditures on occupational health and safety expenses in apartment construction projects. The number of housing units, total construction periods, and total construction costs were considered during the selection of the construction sites. A total of 10 construction sites with a similar size, number of construction periods, and construction dates (between 2015 and 2017), were selected for the case study analysis (See Table 2 for details) to avoid possible distortionary issues hindering comparisons, such as inflation and temporal material shortages during the construction process.

**Table 2.** Information on the apartment construction sites for the case study.


The 2.29% proportion of material and labor costs for construction projects were applied for the occupational health and safety expenses in construction sites for the case study. In the analysis, safety expenditures were categorized as (1) labor costs for safety professionals, (2) costs for protective devices and facilities for safety activities, (3) costs for personal protective equipment, and (4) costs for other safety related activities (e.g., safety consulting, safety education, and so on). The details of the safety budgeting and expenditures are summarized in Table 3.

**Table 3.** Budgeting and expenditure for occupational safety and health expenses from 10 case study sites.


Most of the safety budget was spent on the labor of safety professionals, protective devices, and personal protective equipment. On average, 47.9% of the safety budget was spent on the labor of safety professionals. The costs for protective devices and personal protective equipment were determined to be 33.2% and 12.7%, respectively. The other safety activities were 6.2% of the total safety budgets. As shown in Table 3, the labor costs and protective device/safety facilities were the two major components of the safety budget (81.1%) while other safety activities accounted for only a small portion of the budget (18.9%). Considering the fact that the labor costs would be spent evenly during the entire construction period, the expenditure trends (as shown in Figure 1) revealed that safety management related activities, including installation of the protective devices and purchasing the personal protective equipment, can be performed irregularly and this would increase the risk of accidents occurring on construction sites. The analysis results demonstrate the need for better safety budget planning to enhance the safety level of

the construction sites. This study introduced the accident risk index to be incorporated with construction site information (i.e., current progress, the number of workers, and construction costs) and historical accident records (i.e., the number of accidents associated with each activity) in safety cost budgeting. The accident risk index can be calculated in accordance with the construction progress (i.e., 0% to 100%) to determine the appropriate expenditure for occupational safety and health expenses at the project level.

**Figure 1.** Expenditures on occupational safety and health expenses in apartment construction.

## *3.2. Accident Risk Index*

The proposed index in this study considers: (1) the number of major accidents, (2) the number of workers on duty, (3) the required working time for the construction work, and (4) the amount of progress payments needed to measure the risk of accidents occurring. These factors are decided based on the fact that the risk level of an accident occurring during a certain stage of the project on a construction site is related to the number of workers on-duty, the progress of the project, the amount of payments that have been made linked to the progress of the project, and previous accident records. In fact, the factors related to the occurrence of accidents are numerous and the selection of these factors is highly related to the risk assessment level. Specifically, risk assessment could be performed with the various risk factors being assessed at different management levels (e.g., the task, activity, and project) and the level of risk management would affect the type of risk factors to consider. As examples, the task location, age of the workers, previous injury records, and the levels of experience could be included when managing the risk of accidents occurring at the task level. However, this study specifically aimed to assess risk management at the project level and the corresponding risk factors that are the information available at that level. The research from Gurcanli et al. (2015) utilized the total construction cost, number of required workers, required construction time, and the risk of completing various activities to decide a reasonable safety budget amounts and budget allocation for construction projects [37]. Similar to the previous study, four attributes were utilized in this study (i.e., the progress ratio, cost ratio, worker ratio, and risk ratio) and such attributes are measured by Equations (1)–(4). The progress ratio is the proportion of the required working time for an activity to the total construction time. The cost ratio is the proportion of construction costs for an activity to the total construction costs. The worker ratio is the proportion of the number of required workers for an activity to the total number of construction workers. The risk ratio is the proportion of the number of major accidents while completing an activity to the total number of major accidents in historical accident data. The accident risk

index, representing the risk level of accidents at a certain period by combining the four aforementioned attributes, is calculated by Equation (5):

Progress Ratio(i) = Construction Time(i)/Total Construction Time × 100 (1)

Cost Ratio(i) = Construction Cost(i)/Total Construction Cost × 100 (2)

Worker Ratio(i) = Construction Worker(i)/Total Construction Worker × 100 (3)

Risk Ratio(i) = Number of Major Accidents(i)/Total Number of Major Accidents × 100 (4)

Accident Risk(i) = [Progress Ratio(i) + Cost Ratio(i) + Worker Ratio(i)] × Risk Ratio(i) (5)

where Construction Time(i), Construction Cost(i), Construction Worker(i), and Number of Major Accidents(i) are the construction time, construction costs, number of workers, and the number of major accidents for each activity or period(i), respectively.

The statistical data published by the KOSHA was utilized to calculate the risk ratio by measuring the number of major accidents that occurred during construction. The major accidents were defined by the KOSHA as accidents that resulted in a fatality or an illness requiring medical care after more than 3 months, or accidents where injuries affected more than 10 workers at once. This study utilized the historical accident data collected between 2014 and 2016 and a total of 118 major accidents were recorded. Laborers were recorded to have the largest number of major injuries (i.e., 27 accidents) while scaffolders and carpenters had the next highest numbers (i.e., 21 and 17 injuries, respectively). This study utilized the progress percentage, representing 10 different levels of construction progress (e.g., 10%, 20%, and 100%) from the beginning to the end of the construction project, to ease the implementation of the proposed accident risk index.

#### **4. Case Study Data Analysis**

#### *4.1. Trends of Safety Expenditures and Computations for Progress, Cost, Worker, and Risk Ratios*

The expenditure on safety costs was reorganized corresponding to the construction project progress (e.g., every 10% progress) to investigate the problems for the expenditures in practice. The expenditures from 10 case study sites were then analyzed and detailed results are presented in Figure 1 and Table 4. As shown in Figure 1, more than 30% of safety budgets were spent before 20% of the projects' completion had occurred and relatively small amounts of safety budgets were used for 30% to 70% of the projects' progress. However, according to the historical accident data from the KOSHA, more than 50% of the major accidents occurred between 30% and 70% of the projects' completion (as shown in Figure 2). Such facts illustrated the problem of current safety cost expenditures and the need for better safety budget planning while considering the risk of accidents occurring to effectively prevent a major accident on a construction site.



**Figure 2.** Number (**left**) and proportion (**right**) of major accidents that occurred between 2014 and 2016.

Data for attribute computations from the construction sites were also reorganized. The required time for the completion of an apartment construction was 730 days on average. The largest amount of project time (i.e., 146 days) was spent on the 0% to 10% progress stage, which includes the excavation and foundation works. The second largest time (i.e., 102 days) was required for the 80% to 90% progress stage, which is the period for the fishing works. The cost ratio was analyzed similarly, and the results showed that costs were evenly spent during the whole construction progress. The largest cost ratio is 12.5% at the end of the project (i.e., the 90% to 100% stage) and the lowest cost ratio is 6.0% between 20% and 30% progress.

The worker ratio, which is the ratio of the number of workers on-duty to the total number of workers, was further analyzed to investigate the change of the required workforce corresponding to the construction progress. On average, a total of 726 workers participated in an apartment construction project and the construction progress from 50% to 60% employed the largest number of workers, which was 18.2% of the total number of workers (See Table 5 for details). This construction period was the moment when both structural and finishing works were performed simultaneously. Similar to the historical accident records, construction works between 40% and 80% progress employed 74.1% of the total number of workers but corresponding safety expenditure was only 40.1% (See Figures 1 and 2 for details). This fact might be one of the reasons why the construction industry recorded a high rate of accidents, considering the comparably low safety cost expenditure for these periods.


**Table 5.** Number of workers on-duty corresponding to construction progress.

The risk ratio was lastly computed from the historical accident records collected between 2014 and 2016. The largest risk ratio is observed during the 40% and 50% construction progress stage that accounts for 20.34% of the major accidents (See Figure 2 for details). The construction period from 40% to 80% showed a higher accident risk (i.e., 55.08%) compared to other periods but safety cost expenditure during this period (i.e., 29.01%) was relatively small, as previously described. These facts demonstrate the existing safety cost expenditure problem and illustrate the necessity of an accident risk index for better safety budgeting for apartment construction projects.

The computed ratios associated with the accident risk are summarized in Table 6. Also, the ratios and the trend of safety expenditures are visualized in Figure 3. The trend of the safety expenditure seems to be similar to the progress ratio, while the worker ratio shows a similar pattern to the risk ratio. These results would imply that the current expenditure of the safety cost is related to the required construction time, while the accident risk has a relationship with the number of workers on-duty. These results also indicate the need for an accident risk index to comprehensively assess various safety related factors and better safety budgeting for construction safety.

**Table 6.** Results of the progress ratio, cost ratio, worker ratio, and risk ratio computations.


**Figure 3.** Visualization of the computed attributes and safety cost expenditure from case study sites.

## *4.2. Accident Risk Index for Safety Cost Budgeting*

The accident risk index was computed to determine the construction periods containing a high risk of accidents and to effectively budget safety costs. The computation results illustrated that the 40% to 50% progress stage has the highest accident risk (i.e., 717.73), which is equivalent to 22.27% of the total accident risk. The second largest accident risk was observed at the 60% to 70% progress stage and this period contained the second largest number of workers on-duty. The analysis results indicated that construction periods

between the 40% to 90% progress stage accounted for 74.08% of the total accident risks. The results also demonstrated the importance of enhancing the safety management efforts during these periods (See Table 7 for details).


**Table 7.** Accident risk index and recommended schedule for the safety cost expenditure.

The differences between the accident risk index and safety expenditures were investigated to suggest a guideline for better safety budgeting. The government's safety law forced companies to spend more than 50% and 70% of their construction safety budgets before 70% and 90% completion of their construction project, respectively. Considering the computed risk index figures during construction projects, more detailed recommendations are necessary to effectively utilize the safety budget and increase the effectiveness of the safety management activities to avoid major accidents. As shown in Figure 4, a huge gap between the computed accident risk ratio and the safety expenditures from analysis of the case studies was found in terms of safety cost executions. Considering that the accident risk index represents the risk level of construction environments during certain periods, the safety expenditure guideline rate is calculated by finding the average of the accident risk ratio and the expenditure ratio to consider the importance of completing various on-site safety management activities. For example, safety facilities should be built at the initial stage of the construction projects and purchasing protective equipment, including personal protective equipment, should be completed before the start of certain construction works. Therefore, this study proposed a safety cost expenditure guideline (shown in Table 8) that consider both the characteristics of construction projects and the risk of accidents occurring for on-going construction projects. Considering that current legislation forces companies to spend more than 50% of their budgeted safety costs before 70% completion of their construction projects, the recommended guideline could help to budget safety costs for high risk periods by considering the accident related factors on construction sites.

**Figure 4.** Visualization of the accident risk ratio, safety cost expenditure (Case Study), and safety cost expenditure guideline.


**Table 8.** Comparison of the accumulated accident risk ratio, accumulated expenditure ratio from the case study analysis, expenditure guideline rate provided by the occupational safety and health expense law, and expenditure guideline rate recommended by this study.

## **5. Conclusions**

The safety of construction workers is one of the important management factors for ensuring the success of the construction projects. This study investigated the expenditure for safety and health expenses by conducting a case study with 10 apartment construction projects in South Korea. The safety expenditures derived from the case study revealed that most of the mandated safety costs were spent during the initial and the last stages of the assessed construction projects and these expenditures did not correlate well with the risk level of accidents occurring during different stages of apartment construction. At the 50% progress stage, the highest accident rate and the lowest safety expenditure rate were observed, and this is a significant problem for the prevention of accidents on construction sites. In addition, the Occupational Health and Safety Expense Law does not have a detailed expenditure guideline covering the 0% to 40% construction progress stage, although almost 30% of the major accidents occur during this stage. To address these problems, this study proposed an accident risk index that can incorporate the construction schedule, construction costs, the number of workers on-duty, and historical accident records in the safety cost budgeting. The safety cost expenditure guideline was also developed by combining the characteristics of the safety cost expenditure in practice and the risk level of accidents occurring corresponding to the construction schedule. The proposed accident risk index would offer information about the risk level of on-going construction activities. The recommended expenditure guideline helps to understand the required safety management efforts for the accident preventions corresponding to the construction schedule. Considering that the legal expenditure guideline would not provide any information about the risk of accidents occurring, the proposed guideline in this study will help safety practitioners to perform effective safety cost budgeting while considering the accident risks and enhancing the level of safety management for apartment constructions.

However, several limitations remained to be assessed by future research. There are many contributing factors (e.g., the construction methods, size of construction projects) to the occurrences of accidents on construction sites, but this study utilized only small sets of attributes for the computation of the accident risk index and the development of the safety cost expenditure guideline. Future research would be essential to investigate the relationships between the occurrence of accidents and various accident-related factors in apartment construction. In addition, this study performed safety cost budgeting corresponding to the construction schedule. Another important issue is how to utilize the safety costs for enhancing the safety of construction sites. The effectiveness of safety management activities needs to be further investigated to find the optimal use of safety budgets to increase construction safety.

**Author Contributions:** Conceptualization, K.Y., K.K., and S.G.; data curation, K.Y., K.K., and S.G.; writing original draft, K.Y., K.K., and S.G.; writing review and editing, K.Y., K.K., and S.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2019R1G1A1100365).

**Data Availability Statement:** The data presented in this study are available upon request.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Using Recycled Material from the Paper Industry as a Backfill Material for Retaining Walls near Railway Lines**

**Karmen Fifer Bizjak \*, Barbara Likar and Stanislav Lenart**

Department of Geotechnics and Traffic Infrastructure, Slovenian Building and Civil Engineering Institute, ZAG, 1000 Ljubljana, Slovenia; barbara.likar@zag.si (B.L.); stanislav.lenart@zag.si (S.L.) **\*** Correspondence: karmen.fifer@zag.si; Tel.: +386-41-39-55-51

**Abstract:** The construction industry uses a large amount of natural virgin material for different geotechnical structures. In Europe alone, 11 million tonnes of solid waste is generated per year as a result of the production of almost 100 million tonnes of paper. The objective of this research is to develop a new geotechnical composite from residues of the deinking paper industry and to present its practical application, e.g., as a backfill material behind a retaining structure. After different mixtures were tested in a laboratory, the technology was validated by building a pilot retaining wall structure in a landslide region near a railway line. It was confirmed that a composite with 30% deinking sludge and 70% deinking sludge ash had a high enough strength but experienced some deformations before failure. Special attention was paid to the impact of transport, which, due to the time lag between the mixing and installation of the composite, significantly reduced its strength. The pilot retaining wall structure promotes the use of recycled materials with a sustainable design, while adhering to government-mandated measures.

**Keywords:** paper sludge ash; deinking sludge; paper industry; backfill material; retaining wall

## **1. Introduction**

According to information published by the European Aggregate Association, the demand for European aggregates is 3 billion tonnes annually [1]. About half of natural (virgin) material is consumed by the construction industry, which also generates a large amount of waste material [2]. Undoubtedly, virgin material can partially be replaced by other materials, such as recycled industrial material, including material made from paper industry waste. This recycled waste can be substituted for virgin aggregates that are used in various applications in the building sector in huge quantities, especially for roads and earthworks [3]. Of course, the mechanical and environment criteria for recycled materials according to the national legislation must be satisfied.

Globally, 420 million tonnes of paper and paperboard are produced annually [4], and production is growing. The production processes result in significant waste generation; 11 million tonnes of solid waste are generated per year in Europe [5]. Approximately 70% of this waste is from paper recycling, for example, deinking sludge [6]. According to the Integrated Pollution Prevention and Control Directive 1996/61/CE [7], the paper industry is required to minimize the amount of waste and develop more sustainable technologies for waste treatment. There are also EU waste management legislative measures and policies [8] implementing a waste hierarchy, with landfilling being the least desirable option and recycling the most, supported by increased taxes for landfilling. Recycled paper residues are a potential material that could be substituted for virgin raw materials from a technical and economical point of view [9,10].

Examples of pulp and paper industry residue implementation have been presented by other authors [11–14], but in general, most paper industry waste is burned in power plant boilers or landfilled. The production process with different fillers, pigments, and

**Citation:** Bizjak, K.F.; Likar, B.; Lenart, S. Using Recycled Material from the Paper Industry as a Backfill Material for Retaining Walls near Railway Lines. *Sustainability* **2021**, *13*, 979. https://doi.org/10.3390/ su13020979


Received: 2 December 2020 Accepted: 14 January 2021 Published: 19 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

coagulates influences the type of paper ash. Also, the technology and temperature in the boilers have an effect [15,16].

If paper sludge ash were to be used only as a binder in the construction industry, some problems due to the presence of lime would be observed, but it could be very useful for the stabilization of road structures or as a backfill material [17]. Different mixtures of paper ash and paper sludge have been tested on a laboratory scale. A mixture of sand, paper fly ash, paper sludge, and cement has been used in laboratory research [18]. The mixture reached a compressive strength of 0.8 MPa, which is high enough for use as a backfill material for a foundation structure, a structural fill, or a hydraulically bound layer in a road structure. If the paper sludge ash is mixed with recycled concrete aggregate (RCA), the mechanical properties are improved, especially the resistance to acid and sulfate attacks [19]. Highly plastic clay soil was stabilized with paper sludge ash [20–22] and the compressive strength increased enough (0.7, 1 MPa) for the mixture to be used without any other additives for a pozzolanic reaction. For mining backfill material, a mixture of paper sludge ash and sewage sludge ash [23] was prepared. Both materials were mixed and calcinated at high temperatures. In addition to paper sludge ash, paper sludge can be used. Paper sludge was used with marine-washed sand, aggregate, and Portland cement [24]. With this mixture, a compressive strength of 8 MPa was achieved. Remediation of contaminated soil by red mud was used with paper ash as a binder material [25].

Most of the research is related to laboratory tests, but some field results have also been published. A road subgrade was stabilized in a length of 250 m with a mixture of paper sludge and cement in Portugal [26]. The installed mixture achieved an unconfined compressive strength of 4.5 MPa. Paper fly ash was used for gravel road stabilization of a hydraulically bound layer [27] in Spain. The hydraulically bound layer reached an unconfined compressive strength of 5 MPa.

Paper sludge ash is also used in the cement industry as supplementary cementitious material in mortar [28,29], concrete manufacturing [30,31], and the brick industry [32].

When using ash as a building material, particular attention should be paid to the impact on the environment. Studies have demonstrated the wide applicability of ash, but it is necessary to carefully investigate the potential environmental impact and use technology that is appropriate for individual recycled materials [33,34].

Investors and designers find it difficult to decide to use recycled material in construction due to a lack of knowledge about the material, technology of installation, high cost of production, and often a negative attitude towards all new materials [1]. The objective of this research is to develop a new backfill material (composite) used behind retaining walls from the residue of paper industry production and promote the use of the recycled material with a pilot structure. Especially in mountainous regions, landslides represent a threat to roads and railways, which must be reduced by slope stabilization with different retaining walls. A new composite must have high enough unconfined compressive strength and shear properties, but at the same time has to allow elastic deformation before cracking. Until now, paper sludge ash and deinking sludge have been used in different mixtures, usually in mixtures with soil and other binders. At present, deinking sludge ash and deinking sludge are mixed together as a new composite in a precise ratio and compacted under strict conditions behind a retaining wall. This type of composite has not been tested in the laboratory so far, nor has its use been validated in field tests.

None of the studies to date have dealt with changing the strength characteristics of the material during mixing and installation. Here, the time of transport of the material from the place of mixing to installation is crucial. In the study, we found that, over time, the strength properties decrease significantly, which may be crucial for the stability of the retaining wall. The study notes that the materials in the laboratory must also be tested in terms of installation time in order to provide the designer with relevant data regarding the geomechanical characteristics of the composite.

The pilot retaining wall structure promotes the concept of a circular economy from idea to laboratory tests, installing the structure, and monitoring it over a long period of time. The new composite and the technology were tested at a construction site and then monitored over a longer period of time. This gives us information about the details of construction and proves that the structure is stable, usable, and meets all the technical and environmental standards available to investors, designers, and contractors. Pilot structures could help suppliers, investors, designers, and contractors identify the factors hampering the use of recycled materials in the construction sector as well as provide strategies that can be adopted to form an economical and sustainable product.

## **2. Materials and Methods**

## *2.1. Material Used*

Deinking sludge ash (DSA) and deinking sludge (DS), used in this study as raw materials, represent the main waste from recycled deinking paper pulp production at a paper industry company, VIPAP Videm Krško d.d., in Slovenia. The DSA is a combustion residue formed in a steam boiler during the incineration of DS. It consists of a mixture of bottom ash (approx. 90 wt%) and fly ash (approx. 10 wt%). VIPAP recycles around 600 tonnes of paper daily. Annually, 25,000 tonnes of DSA and 67,000 tonnes of DS are produced. According to the European Waste Catalogue (EWC), DSA is classified as 10 01 01, while DS is classified as 03 03 05 [35].

## *2.2. Methods*

#### 2.2.1. Testing of Raw Materials

The bulk chemical composition of the DSA and DS was determined by a Wavelength Dispersive X-ray Fluorescence (WD XRF), using a Thermo Scientific ARL PERFORM'X Spectrometer (Waltham, MA, USA). Analysis of loss on ignition (LOI) and the total chloride content of DSA was performed according to SIST EN 196-2 [36]. The physical and mechanical properties of the raw materials were tested according to the standards in Table 1.

**Table 1.** Physical and mechanical properties of raw materials.


Particle size distribution of DSA was measured by laser diffraction analysis (particles < 400 μm) using a CILAS 920 Particle Size Analyser (Cilas, Orléans, France).

#### 2.2.2. Preliminary Laboratory Tests of Composite

In order to design a backfill material for a retaining wall structure for the stabilization of a landslide near a railway line, several mixtures consisting of different ratios of DSA and DS were tested. Among them, two mixtures (Table 2) with sufficiently good geomechanical characteristics and suitable properties for compaction and installation were tested in detail.


**Table 2.** Mixing proportions of the investigated mixture composites and their designations.

DSA: deinking sludge ash; DS: deinking sludge.

Geomechanical tests were performed in an accredited geomechanical laboratory, according to SIST EN ISO/IEC 17025 [37].

The components were mixed in a 20 L planetary mixer. Two kilograms were mixed for 2 min until a homogeneous mixture was obtained. Mixtures were compacted at the maximum dry density according to the SIST EN 13286-2 [38]. In order to prevent evaporation, the composites were stored and cured in a climatic chamber at 90% RH and 22 ◦C.

Compressive strength was tested according to SIST EN 13286-41 [39] immediately after compaction and after one, four, seven, 28, and 50 days of curing.

Freezing/thawing tests were performed according to Slovenian technical specification TSC 06.320 [40]. According to the specification, the composites were exposed to 12 cycles of freezing at −23 ◦C and thawing at 20 ◦C in a climate chamber.

The shear characteristics of the composites were tested according to the SIST EN ISO 17892-10 [41] directly after compaction and after seven days of curing. The permeability of the composite was tested in a triaxial cell, according to SIST EN ISO 17892:11 [42] under a pressure of 50 kPa.

In order to investigate the impact of the transport time to the construction site, the time delay between mixing and compacting was taken into account. The tests were performed with two different testing procedures:


After moistening and mixing, the mixtures were compacted in the following time intervals: immediately, and after 4, 8, and 24 h. After seven days, an unconfined compressive strength test was performed on each specimen.

## 2.2.3. Test of Pilot Structure

In the region between Ljubljana and Novo Mesto, there is a landslide risk that endangers the safety of a railway line. The instability of the slope was already evident from the geological and geomechanical mapping of the site [43]. Subsequently, after carrying out a detailed geomechanical investigation of the railway zone, the final pilot structure location was selected in the exact location shown in Figure 1.

Construction work started in August 2018. The length of the structure was 50 m, with a height of 1.5 m. The retaining wall was made of gabions and backfill material with composite. The width of the backfill material between the gabions and landslide slope was between 2 and 3.5 m.

The mixture was prepared at the VIPAP facility from DPA and DS and transported to the construction site, located 70 km from VIPAP's facility. Mixing was performed by a stirrer (TERREX, Norwalk, CA, USA) for at least 15 min. The composite was transported to the construction site and subsequently compacted to the required density (Figure 2) in a 30 cm thick layer by a compactor (BOMAG, Boppard, Germany). For the whole structure, 100 t of the mixture was compacted into nine layers (Figure 3).

**Figure 1.** Location of the pilot structure.

**Figure 2.** Compacted layer of the composite.

**Figure 3.** Compacting the last layer of the composite.

Special attention was paid to the drainage system since the composite is impermeable. A drainage system was installed under the backfill and at the contact point between the backfill and the landslide slope.

Each compacted layer was tested at the construction site for dry density and moisture content (γd, w) according to TSC 06.711 [44] by a nuclear soil moisture density gauge (TROXLER, Ettenheim, Germany) and for dynamic deformation (Evd) modules according to TSC 06.720 [45] (2003), by a dynamic plate (ZORN, Stendal, Germany). In order to perform the leaching procedure, according to SIST EN 1744-3 [46], samples were collected from the last layer. The leachates were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) and UV-Vis spectrophotometry. Samples were also taken to verify the shear characteristics of the composite according to the standard SIST EN ISO 17892- 10 [41].

## 2.2.4. Long-Term Monitoring

For long-term monitoring, an automatic station was placed at the construction site (Figure 4). It collects measurements from the weather station, inclinometers, piezometers from boreholes, and moisture and temperature probes installed in the backfill material behind the retaining wall made from gabions.

**Figure 4.** Monitoring system.

A soil moisture probe (Imko, Ettlingen, Germany) was installed 90 cm below the surface, and temperature sensors were installed 30, 60, and 90 cm below the surface of the composite.

For environmental monitoring, a water tank was built at the end of the retaining wall. The water flowing from the drainage system between the gabions and the composite (backfill material) was collected in a plastic tank and taken for chemical analyses.

The displacements that can occur above the retaining wall structure were monitored with a manual inclinometer (Interfels, Ulm, Germany), while those of the entire structure (specifically those of the gabions and the foundation) were monitored by a laser scanner (Leica, Aarau, Switzerland).

## **3. Results and Discussion**

Preliminary investigations of two types of composites were performed in the laboratory, and the composite with the most suitable properties was selected for implementation.

#### *3.1. Raw Material*

Chemical analyses showed that the major oxide of DSA (Table 3) was CaO, accounting for almost 50 wt%. More than a quarter of DSA mass was lost after the LOI treatment. The most abundant oxides of DS were CaO (accounted for almost one-third) and the measured LOI was more than 50 wt%. High values of LOI were due to high fibre or other organic compounds. Elements such as Si, Al, and Mg were in concentrations below 0.25 wt%. Higher CaO content in DSA and DS was due to the use of fillers and coatings in the deinking process of the production of new paper from used paper. Also, higher CaO led to the higher pozzolanic reactivity of the composite made from DSA and DS.



LOI: loss on ignition.

The physical and mechanical properties of DSA and DS are presented in Table 4. DSA is a dry material, while the water content of DS ranges between 45% and 50%. In comparison with DS, the specific gravity of DSA was higher by about 20%. A standard Proctor test (SPP) showed that the optimal water content (wopt) and maximal dry density (γd,max) were higher for DS. The unconfined compressive strength of DSA is between 300 and 500 kPa, which is in the range of very stiff soil according to the criteria for virgin materials. DS is a softer material in the range of stiff soil. Both materials were nonplastic.

**Table 4.** Physical and mechanical properties of raw materials.


*3.2. Results of Preliminary Laboratory Tests*

3.2.1. Unconfined Compressive Strength—qu

The unconfined compressive strength decreased (Figure 5) with a higher quantity of DS in the composites and increased with curing time. The tests performed immediately after compaction showed relatively similar values of qu (0.2–0.3 MPa), independent of the composition. After one day of curing, composites with higher percentages of DSA showed higher qu values, accounting for the more intense hydration process in those composites.

**Figure 5.** Uniaxial compressive tests results.

The results of the vertical deformation at the sample failure during the uniaxial compressive test are shown in Figure 6. With an increasing content of DS in the composite, the vertical deformations increased. This is reflected in the rapid breaking and very small elastic deformations, i.e., the brittle deformation behavior of the D80/20. By contrast, ductile behavior was observed in the D70/30.

**Figure 6.** Deformations after uniaxial compressive tests.

The curing time has a strong influence on the vertical deformation (εA) at failure. While this is not as obvious with D80/20, in the case of D70/30, the ε<sup>A</sup> at failure increased rapidly in one day and then decreased over time. After 28 days, the ε<sup>A</sup> was almost the same for both composites. This points to the fact that the failure mode of mixtures changes from ductile to brittle between the first and fourth days of curing.

Even after a retaining wall is built, some deformations can still occur in the structure because of the stabilization processes in the landslide mass, especially in the early days of construction. Because of this, it is desirable for a backfill material to possess higher elasticity and, at the same time, have enough strength to prevent landslides.

#### 3.2.2. Time Effect

The results of the investigation of the impact of the delay between mixing and compacting (Figure 7) showed that qu decreased with the increased time between them. The highest values of qu were seen in samples prepared at the wopt (Figure 8) and compacted

immediately after mixing. For samples compacted 4 h after mixing, higher qu values were measured for composites prepared at the wmax. The hydration process dried the mixtures and after 4 h resulted in a lower qu for mixtures with lower water content. After 24 h, no difference could be observed in qu. Based on the results, it can be concluded that the mixture compacted immediately must be moistened to the wopt. For a mixture that will be compacted after 4 h due to transport, it is more appropriate to moisten it to the wmax.

**Figure 7.** Unconfined compressive strength of samples with delay in compaction time. wopt: optimal water content; wmax: maximal water content.

**Figure 8.** Results of standard proctor tests (SPP) for the investigated composites.

The results of these tests are essential to determine the methods of mixing, transport, and installation of the mixture on the construction site and to assess the maximum distance between the composite production site and the composite construction site. The results from the laboratory and field tests showed that the mixture has to be moistened above the wopt before transport. It is a very important requirement for the installation procedure. If the mixture at the construction site does not have a high enough water content, it will not be properly compacted, and the strength of the installed layer will be lower than it has to be, according to the legislation. Three to five percent of the water is consumed during the hydration process of DSA.

## 3.2.3. Standard Proctor Tests (SPP)

The results of the SPP showed that the optimal water content (wopt) increases slightly with a higher percent of DSA in composites (Figure 8). A higher percentage of DSA required more water for the hydration process in the composite. A similar trend was observed for maximal dry density as it increased from 0.96 Mg/m<sup>3</sup> in D70/30 to 0.99 Mg/m<sup>3</sup> in D80/20.

The main difference between the investigated composites and the natural gravel material is the density. The γd,max for a natural gravel material is 2.3 Mg/m<sup>3</sup> on average, while the investigated composites show a much lower γd,max. This physical characteristic enables the use of the investigated composites in low-bearing-capacity foundation soils. Such a backfill material is light and does not cause large settlement, as does the heavy virgin gravel material. This is one of the important advantages of using the investigated composites as a backfill material for geotechnical structures in regions with difficult geotechnical conditions.

#### 3.2.4. Frost Resistance

According to the Slovenian National Technical Specifications for roads, TSC 06.320 [40] on freeze/thaw resistance, the freeze ratio has to be more than 0.7 (the ratio between the unconfined compressive strength of exposed samples and curing samples at atmospherecontrolled conditions).

Both composites are above the limit value of 0.7. Based on these results, it can be concluded that this material could be used as a backfill material even in the top meter below the surface. That is the depth of freezing at the pilot structure's location.

## 3.2.5. Shear Properties

The results of a friction angle (f) and cohesion (c) analysis showed that shear characteristics increase with the decrease in DS content of composites, but after 28 days, the differences were not so obvious (Figure 9). The shear characteristics are higher in comparison with the natural gravel backfill material conventionally used for retaining wall structures. The friction angle for gravel is around 30◦ with no cohesion. The high shear properties of the composite allow for the construction of a thinner retaining wall than would be needed if virgin gravel material was used. These characteristics are very important, especially for retaining structures near railway lines, where there is generally not enough space for large geotechnical structures.

**Figure 9.** Angles of friction and cohesion for the mixtures.

#### 3.2.6. Composite Used for Backfill Material

Based on the results of the laboratory tests, the composite D70/30 was chosen for use as a backfill material behind the retaining wall structure. The composite has a high enough qu even when compacted 4 h after mixing, and at the same time allows small deformations before cracking. It is ductile enough to sustain small deformations behind the retaining wall in case of landslides. The composite has high enough freeze/thaw resistance to withstand extreme weather conditions. Shear properties are higher in comparison with virgin material (gravel). Because of these properties, the retaining wall structure can be lower and cheaper. Instead of 2 m, the gabion structure that was built had to be only 1.5 m high.

## *3.3. Results from the Pilot Structure* 3.3.1. Field Tests

According to the technical specification TSC 06.711 [44], the average compaction of the last layer has to be at least 95% γd,max. The results of the compaction tests of the installed composite D70/30 showed that only the first three layers were compacted below 95% γd,max (Figure 10). The reason for the lower compactness in the first layers is the uncompacted drainage layer under the composite. In all other layers, a higher compaction degree was measured. Measurements confirmed that the quality of the installed composite was within the requirements of the technical specifications.

**Figure 10.** Results of the measurements with the neutron probe.

3.3.2. Laboratory Tests of Samples from the Pilot Retaining Wall

Shear tests of the samples taken from the composite at the construction site reach high enough shear properties. The results are similar to the results from the preliminary laboratory investigation; even cohesion increases significantly after 28 days. In the design project [43] for the stability analysis, the design parameters were lower, especially for cohesion, which means that the safety factor of the structure was higher. Shear characteristic values obtained from the in-built composite are gathered in Table 5.


The new composite had significantly higher shear characteristics compared to a natural gravel backfill material. The shear properties of the composite are similar to the shear properties of soft rock. If virgin gravel material is used as the backfill, a larger retaining wall structure should be designed to prevent landslide movement, as was confirmed in the project design [43].

A leaching test of the samples taken from the structure showed (Table 6) that none of the components in the water exceeded the limits established by Slovenian legislation (UL RS, No. 10/14, 22 February 2014). The installed mixture does not have an adverse environmental impact because it is impermeable, and water from the composite does not leach hazardous substances. For the composite, very low permeability (only 2.2 × <sup>10</sup>−<sup>10</sup> under the load of 50 kPa) was measured in the laboratory.


**Table 6.** Results from the leaching tests.

## *3.4. Results of the Long-Term Monitoring*

3.4.1. Stability of the Landslide

To ensure the safety of passengers and cargo, the landslide was stabilized with a retaining wall structure (Figure 11). Deformations of the slope above the railway line were measured with an inclinometer. The measurements did not indicate any displacements because the retaining wall successfully stopped the landslide (Figure 12).

For the retaining wall structure, measurements were taken by a laser scanner. The measurements showed that the foundation of the retaining wall was stable. Both the measurements of the retaining wall and the measurements of the landslide deformations showed that the landslide was successfully stabilized.

**Figure 11.** Retaining wall with gabions and the composite as the backfill material.

**Figure 12.** Horizontal displacements in borehole above the retaining wall.

The composite used has to be sustainable and resistant to weather changes. The results of the temperature measurements showed that the weather does not influence the temperature of the composite significantly (Figure 13). Some correlation with the outside temperature (T) can be observed in the sensor that is 30 cm below the surface (T1), while the temperatures at 60 cm (T2) and 90 cm (T3) below the surface remained more or less stable despite the fluctuation of the outside temperature.

**Figure 13.** The temperature outside and in the backfill composite in the retaining wall structure (26 March 2019–10 May 2019).

The water content of the composite at the time of compaction was 57% and is currently around 47%. Comparison with the precipitation showed that the composite is impermeable because there was no correlation between the water content and precipitation (Figure 14). The temperature and water content sensors showed that the weather does not have an influence on the composite quality.

**Figure 14.** Water content in backfill material in the retaining structure vs. precipitation (26 March 2019–10 May 2019).

## 3.4.2. Environmental Monitoring

Samples were taken from the water tank and tested in an accredited chemical laboratory. The results show that the composite did not have a negative influence on the water quality since the concentration of none of the tested parameters exceeded the limits set by Slovenian legislation UR RS No. 98/15 (Table 7).

**Table 7.** Results of the chemical analysis of the water from the drainage system.


## **4. Conclusions**

A new composite from paper residues was developed as a backfill material for retaining wall structures. Preliminary tests were performed in an accredited geomechanical laboratory, and the results were later verified with field and laboratory measurements on a pilot structure.

A composite has to have geomechanical properties that are high enough to be used as a backfill material behind a retaining wall structure in an unstable area near a railway line threatened by landslides. Several mixtures with different contents of DS and DSA were initially tested to choose the optimal composite with the proper geomechanical properties. Two of them were studied in detail within the present research to estimate the effect of their composition upon particular basic characteristics.

No studies to date have dealt with the changing strength characteristics of material during mixing and installation. The time of transport is an important parameter for using this mixture as a backfill material. Results from the laboratory showed that the mixture has to be moistened above the wopt before being transported and used at the construction site within 4 h.

Based on the laboratory results, it was decided to use composite D70/30 as the backfill material for the retaining wall. The composite is a new mixture that has not yet been tested in a laboratory or installed in the retaining wall structure at the construction site. The composite with 30% DS had a high enough qu and shear strength but allowed small deformations before failure. The high shear characteristics of the composite allowed for a slimmer retaining structure to support the unstable slope behind it. At the same time, the DS in the composite enabled ductile behavior of the structure and prevented it from brittle failure.

In 2018, the composite D70/30 was used as a backfill material of the retaining wall structure built by gabions in the south part of Slovenia, near the railway line, for landslide stabilization. All laboratory and field tests confirmed the physical characteristics measured in the research phase and confirmed that the environmental requirements are being reached.

At the construction site, the material was installed in 30 cm layers. Each layer was compacted and controlled to reach the optimal moisture and maximum density. Before, during, and after the construction, landslide stability was assessed, and environmental monitoring was performed. With the test results from the pilot structure presented in this paper, the technology of mixing and compacting was improved. Mixing is usually not a problem in the laboratory, but at a construction site, a large quantity of material has to be mixed with the proper technology. The monitoring system confirmed that the retaining wall with the composite stabilizes the landslide near the railway. The composite was an impermeable material, and precipitation did not influence its stability. As the area covered with such an impermeable layer was not very wide, the impact on the groundwater recharge was limited. However, one should consider such a structure as a groundwater barrier that disturbs shallow groundwater flows. Thus, the use of this kind of structure should be combined with a properly designed drainage system, which minimizes the effect of groundwater disturbance. In the case of the retaining structure presented within this paper, vertical and horizontal drainage were employed to enable effective drainage of water behind the retaining structure.

On the other hand, the low density of the composite from paper residues also has several advantages. There is great potential for its use in construction on soft ground. In such a case, the utilization of very light materials, like the new composite, could prevent large settlement.

The pilot retaining wall structure represents a practical case for future investors, designers, and construction companies to encourage them to use recycled materials from the paper industry instead of virgin materials. The new composite and its installation technology were successfully tested. The presented circular case between the paper industry and the construction sector shows the advantages for both sides. Instead of disposing of the waste in a landfill for a very high price (in Slovenia, 80–150 €/tonne), the paper manufacturer processes the waste into a composite for a retaining wall structure. Contractors, meanwhile, get a composite that is cheap and has better deformation properties than the virgin material.

## **5. Patents**

The Slovenian Technical Approval STS—1870011 was granted for this product.

**Author Contributions:** Conceptualization, K.F.B. and S.L.; methodology, B.L., K.F.B. and S.L.; formal analysis, B.L. and K.F.B.; investigation, K.F.B., S.L. and B.L.; resources, K.F.B., S.L. and B.L., writing—original draft preparation, K.F.B.; writing—review and editing, S.L. and B.L.; supervisor, K.F.B., funding acquisition, K.F.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Union, co-financed by the Horizon 2020 Research and Innovation Programme under grant agreement no. 730305, project Paperchain and the ARRS financial support from the research core funding No. P2-0273, Building structures and materials.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data available in a publicly accessible repository.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Modeling Building Stock Development**

**Antti Kurvinen 1,\*, Arto Saari 1, Juhani Heljo <sup>1</sup> and Eero Nippala <sup>2</sup>**


**Abstract:** It is widely agreed that dynamics of building stocks are relatively poorly known even if it is recognized to be an important research topic. Better understanding of building stock dynamics and future development is crucial, e.g., for sustainable management of the built environment as various analyses require long-term projections of building stock development. Recognizing the uncertainty in relation to long-term modeling, we propose a transparent calculation-based QuantiSTOCK model for modeling building stock development. Our approach not only provides a tangible tool for understanding development when selected assumptions are valid but also, most importantly, allows for studying the sensitivity of results to alternative developments of the key variables. Therefore, this relatively simple modeling approach provides fruitful grounds for understanding the impact of different key variables, which is needed to facilitate meaningful debate on different housing, land use, and environment-related policies. The QuantiSTOCK model may be extended in numerous ways and lays the groundwork for modeling the future developments of building stocks. The presented model may be used in a wide range of analyses ranging from assessing housing demand at the regional level to providing input for defining sustainable pathways towards climate targets. Due to the availability of high-quality data, the Finnish building stock provided a great test arena for the model development.

**Keywords:** modeling; building stock development; mortality of building stock; residential buildings; public buildings; commercial buildings

## **1. Introduction**

It is widely agreed that dynamics of building stocks are relatively poorly known even if it is recognized to be an important research topic. Better understanding of building stock dynamics and future development is crucial, e.g., for sustainable management of the built environment [1]. More advanced and transparent modeling of building stocks also contributes to improving analyses that lean on building stock data. The research fields that are in need of improved information on building stocks include but are not limited to land use planning, energy analysis, life cycle assessment, life cycle costing, mass flow analysis, calculation of green gross domestic product, service life estimation of components, simulation of maintenance and refurbishment, cultural heritage protection, comfort and public health, and resilience. As the fields for which building stock information is relevant are various, different levels of detail in building stock development may be significant for analysis within these fields. For example, depending on the intended use, in some cases, information related to buildings or dwellings is of interest while, in others, a further subdivision according to building types or age bands is relevant. Thus, a simple and transparent modeling approach that is modifiable to several purposes should serve the needs of these various fields.

Research on built environments often has important policy implications, such as contributing to strategies to achieve the goals set by the EU (e.g., energy performance of

**Citation:** Kurvinen, A.; Saari, A.; Heljo, J.; Nippala, E. Modeling Building Stock Development. *Sustainability* **2021**, *13*, 723. https:// doi.org/10.3390/su13020723

Received: 7 November 2020 Accepted: 11 January 2021 Published: 13 January 2021


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

buildings [2,3], low-carbon economy [4,5], and no net land take by 2050 [6]). Despite the importance, many analyses still seem to rely on vague assumptions about the development of building stock. In sustainability-related research, the description of the stock's current state is usually based on approaches using constructed building archetypes (e.g., [7]) or sample buildings (e.g., [8]). As an alternative to those, Nägeli et al. [9] introduced an approach where the idea is to create synthetic microdata on building stocks to describe individual buildings and their usage instead of using aggregate average archetype buildings. Yet another recent approach by Nägeli et al. [10] is an agent-based building stock model which combines a bottom-up building stock model (BSM) with agent-based modeling (ABM) to incorporate the interaction between building owners' decision making and relevant influencing factors. Moreover, previous literature on building stock dynamics covers topics like reconstitution of building stock dynamics [1], mortality of building stock [11], statistical analysis on demolished buildings [12], and vacancy of residential buildings [13].

Even if the abovementioned approaches in the previous literature are enough to provide a relatively good understanding of the current states of building stocks, options for modeling long-term future development of building stocks are limited. To forecast construction demand, researchers have used, e.g., multiple regression analysis [14], a panel vector error correction approach [15], a combination of neural networks and genetic algorithms [16], grey forecasting [17], and Box–Jenkins model [18]. However, these modeling approaches tend to be better suited for predicting short- or medium-term development than for long-term projections. In some of the most closely related studies, dynamic material flow analysis has been applied to model housing stock long-term in the Netherlands [19] and in Norway [20,21]. However, those focus on housing stock alone while our approach also pays attention to other building types. Moreover, there are established practices to assess long-term housing needs in many countries. These include but are not limited to Finland [22,23], Sweden [24,25], Norway [26], Denmark [27], England [28], the US [29], and Australia [30].

To create feasible strategies towards sustainability targets, a better understanding of the development of building stocks is urgently needed. Specifically, various analyses on built environments require long-term projections of building stock development. However, long-term forecasts even at their best include a great amount of uncertainty. Recognizing this inherent uncertainty, we propose a transparent calculation model for modeling building stock development, QuantiSTOCK. Our approach not only provides a tangible tool for understanding the development when selected assumptions are valid but also, most importantly, allows for studying the sensitivity of results to alternative developments of the key variables. Thus, this relatively simple modeling approach provides fruitful grounds for understanding the impact of different key variables, which is needed to facilitate meaningful debate on different housing, land use, and environment-related policies.

The model is particularly developed for modeling the development of Finnish building stock but may also be widely applied to other geographic locations when fitted for locationspecific data. The developed QuantiSTOCK model may be extended in numerous ways and lays the groundwork for modeling the future developments of building stocks. The presented model may be used to a wide range of analyses ranging from assessing housing demand at the regional level to providing input for defining sustainable pathways towards climate targets. Thus, the results should be of interest to a wide range of researchers, policymakers, and community stakeholders who contribute to housing and land use policies.

This work is divided as follows: Section 2 outlines the modeling approach, which is followed by the Section 3 that explains the modeling procedure. In Section 4, the results of the analysis are presented, and thereafter, Section 5 provides the discussion. Finally, concluding remarks are presented in Section 6.

## **2. Modeling Approach**

The quantitative building stock model (QuantiSTOCK) provides a relatively straightforward calculation-based approach to model future development of building stocks. The basic assumptions for modeling are that, logically, (1) population change and (2) mortality of existing buildings are the main drivers for quantitative changes in the building stock. Moreover, (3) gross floor area per capita ratio is an important modeling attribute that captures many overlapping processes, including but not limited to changes in residential density and distribution of housing types, and potential excess of new construction.

Two distinct advantages of the selected method are its comprehensibility and transparency. Comprehensibility refers to the modeling procedure being based on logical attributes that are suggested by common sense, while transparency, in this context, means that the modeling is based on publicly available data and the modeling procedure is clearly described in contrast to various black-box models; thus, the reader can understand how the model is constructed and what the role of the different modeling attributes is.

Relying on publicly available statistics, the model is easy to update when new statistics become available. Moreover, application of the model only requires relatively little effort in comparison to more complicated simulation approaches. Recognizing the fact that publicly available data from Statistics Finland is of an exceptionally high quality and well documented, which is not always the case, it is important that the applicability of publicly available data is evaluated case-by-case. When the reliability of public data is low, it is necessary that the user compiles a consistent dataset for modeling.

As projections of needed new construction cannot be considered an exact science, there is no such model that would produce an exact number of future needs [25]. Thus, it is critical to understand that, due to the great uncertainty about the predictor attributes in the long-term, it is by no means self-evident that the prediction accuracy of more complicated approaches would be any better than the outcome from this stripped-down model. For example, According to Boverket [25], Schmuecker [31] obtains similar results using a more simple method compared to the results from more advanced approaches used in England. Another important aspect of using a relatively simple approach is that it allows transparently, putting into perspective which factors are important relative to the big picture. In contrast, complicated modeling approaches may focus on complicated descriptions of the modeling procedure while the understanding of the critical factors may be blurred.

Here, the selected modeling approach does not aim to produce any exact numbers of future development as it would not be meaningful in terms of long-term projections, but the objective is to picture the potential pathways of future development and help to understand the impacts of these alternative scenarios. Such an approach helps us to understand the relationships between the key attributes and building stock development. Even if a great amount of uncertainty is still present in the selected modeling approach, this strategy combined with relevant sensitivity analyses provides a tangible tool for understanding the boundaries within which future development will fall into.

The structure of the QuantiSTOCK modeling approach is presented in Figure 1. The first step is to define the situation of the building stock at the beginning of the modeling period and then to define the modeling attributes for the future projection. As the modeling parameters are uncertain estimates, sensitivity to their changes is also necessary to be modeled to better understand the boundaries for the actual development. In this study, the sensitivity of building stock development to following key variables is modeled: mortality rate (low, historical, and high), population change (a decrease or an increase of five percent relative to the official population projection), and residential gross floor area per capita ratio (an annual decrease or increase of 0.5 percent relative to that in 2020). The lower and upper limits for the sensitivity analysis are defined by the research group members so that the values should present plausible boundaries for the fluctuation range of the variable. These steps are explained more precisely in the following Section 3, where a detailed description of the modeling procedure is provided.

As the urbanization trend still strongly affects the development of building stocks, allowing regional level heterogeneity is important. However, due to the high uncertainty of modeling attributes, a too fine-grained modeling approach is not meaningful either. In this study, this has been addressed by grouping the Finnish cities into three groups. The first group only includes the fast-growing Helsinki region, while the second group contains other Finnish cities that are growing but still at a slower pace than the capital region. Those include the regions of Tampere, Turku, Oulu, and Jyväskylä. The third group contains the rest of the cities that are, based on the official population projection, non-growing or declining in the study period. This grouping simplifies the modeling but still allows heterogeneity between the regions that are on different development paths.

As an outcome, the QuantiSTOCK model produces a projection of the building stock development in the study period. In this paper, the projection is presented for the above described grouping of cities: (i) fast-growing Helsinki region, (ii) growing regions, and (iii) zero-growth or declining regions. Moreover, an aggregation of the development of the entire Finnish building stock is presented. In addition to the overall development of the building stock, the outcome includes the projected demand for new construction, volume of building stock mortality, and distribution between the existing and new building stock in the study period of 2020–2050.

**Figure 1.** A diagram of the QuantiSTOCK model.

## **3. Modeling Procedure**

The main steps of the QuantiSTOCK modeling approach are illustrated in Figure 1 in the previous section and now follows a more detailed description of the modeling procedure. The starting point for the modeling is the current state of the building stock. In this study, that is the existing building stock in Finland at the beginning of 2020. The building stock data follow the classification of buildings by Statistics Finland [32]. However, industrial and agricultural buildings are excluded from the model, as due to their heterogeneous nature, modeling attempts would not be meaningful in this context. Other

important inputs for the QuantiSTOCK model are the regional distribution of population at the beginning of the modeling period and the regional population projection, which in this study, was available for the period of 2020–2040. To cover the entire study period, the official population projection was extrapolated to reach the end of 2050. The used population projection is described in more detail in Section 3.1. Moreover, gross floor area per capita ratios are calculated based on the situation at the beginning of the modeling period as more precisely described in Section 3.1. The raw data for all the above mentioned input data sets were acquired from the StatFin database [33]. The fourth required input data for the QuantiSTOCK model are mortality rates for different building types, which are defined based on mortality functions that lean on history statistics. As mortality data are not directly available from the statistics, the definition of mortality and creation process of mortality rates are explained in detail in Section 3.2.

When all input data is available, the modeling procedure may start to model the development of building stock in the study period. First, demand for annual new construction is modeled based on the parameters in Equations (1) and (2):

• When demand at the beginning of the year < stock at the beginning of the year:

*Demand for new construction* <sup>=</sup> *Annual change in demand + Mortality + New non-permanently occupied floor area* (1)

• When demand at the beginning of the year > stock at the beginning of the year:

*Demand for new construction*


where annual change in demand = gross floor area per capita ratio × annual population growth + annual change in gross floor area per capita ratio × total population.

Equation (1) is applied when the demand at the beginning of the year is less than the size of existing stock, while Equation (2) is used when the demand at the beginning of the year is greater than the stock at the beginning of the year. Demand is modeled separately for each building type and separately for the three region groups, including (i) the fast-growing Helsinki region, (ii) growing regions, and (iii) zero-growth and declining regions. Furthermore, the regional analysis is also aggregated to describe the development of the entire Finnish building stock.

The next step is to model the size of the building stock at the beginning of next year, which is performed based on the parameters in Equations (3) and (4):

• When demand for new housing construction < 0

*Stock atthe beginning of the year* <sup>=</sup> *Stock at the beginning of the previous year* <sup>−</sup> *Mortality* (3)

• When demand for new housing construction > 0

*Stock at the beginning of the year* = *Stock at the beginning of the previous year* − *Mortality* + *Demand for new construction* (4)

Equation (3) is used when there is no demand for new construction, while Equation (4) is applied when demand for new construction occurs. Again, modeling is performed separately for each building type and separately for the three region groups and, in the final stage, the regional results are aggregated to describe the development of the entire Finnish building stock. Next, a more detailed description of the definitions of the modeling attributes follows.

## *3.1. Population Growth and Gross Floor Area per Capita*

In the QuantiSTOCK model, population growth is a main modeling attribute for the demand of residential building stock. Figure 2 depicts the population projection for the study period of 2020–2050. From 2020 to 2040, it follows the official regional population projection from Statistics Finland while, for the period of 2040–2050, the assumed population growth is extrapolated from the official projection. The impact of urbanization is easy to see in Figure 2. In 2020, the fast-growing Helsinki region, growing regions, and zero-growth and declining regions accommodate 29%, 21%, and 50% of the population, respectively, while, in 2050, it is projected that the respective proportions are 34%, 23%, and 43%. This should notably affect building stock dynamics within these region groups. However, it is important to notice that, even in declining regions, new construction is needed as the existing buildings do not meet all the demand and migration within regions also occurs. In particular, the ageing demographic structure causes moves from more distant rural locations to more attractive locations that are close to district centers and better services.

**Figure 2.** Population projection by region type in Finland for the period of 2020–2050.

Moreover, gross floor area per capita ratio is used to assess how many square meters of each building type are needed. The QuantiSTOCK model operates with gross floor areas instead of more detailed descriptions of building stock units. In the interest of simplicity, this allows for a straightforward approach that still provides important information for multiple purposes. Still, an indicative distribution between (i) detached houses, (ii) semidetached and row houses, and (iii) apartment buildings is reported based on official statistics, with an assumption that the proportions of these different residential building types are assumed to remain at the same level throughout the study period. However, these distributions are reported for information purposes only and they do not affect the modeling procedure where the possible variation in different types of housing units is included in gross floor per capita ratio. Even if this is the case in the base version of the QuantiSTOCK model, alternative approaches such as using headship rate method (e.g., [23,25]) may be incorporated in the model if relevant.

It is also good to notice that, as dwelling densities tend to vary with various factors, such as residential building type and location, we instead use gross floor area ratio per capita as an input in the QuantiSTOCK model. This allows for including uncertainties about various factors into one modeling attribute. Those include changes in residential density and changes in proportions of different residential building types and even potential excess of new construction. In this study, the gross floor area per capita ratio is specified based on the official statistics at the beginning of the study period. For residential buildings, the proportion of gross floor area that is reported to be "non-permanently occupied" is excluded from the ratio. As ratios differ between different regions, a separate ratio is defined for each of the three region types. Furthermore, the ratio is assumed to remain at the same level throughout the study period.

For non-residential buildings, including public buildings and commercial buildings, the gross floor area per capita ratio is specified in the same way as for housing but, with the exception, that the "non-permanently occupied" floor area of public or commercial buildings cannot be distinguished. In the QuantiSTOCK model, different building types are categorized based on the classification of buildings by Statistics Finland [32]. However, we also include office buildings in the group of commercial buildings while our specification of public buildings includes transport and communication buildings, buildings for institutional care, assembly buildings, and educational buildings.

## *3.2. Mortality of Building Stock*

Mortality rates of the existing buildings are yet another central input for the QuantiSTOCK model. As there are different types of mortality, including (A) demolition, (B) alterations to purpose of use, and (C) merger of spaces, it is important to explain what mortality of building stock means in the context of this paper. As the QuantiSTOCK model operates with gross floor areas, the types of mortality that are included are limited to types A and B. This outline is due to data limitations, as demolition of buildings and alterations to purpose of use are visible in building stock statistics while merger of spaces is not. It is also important to notice that types A and B cannot be separated from each other as only the total changes are reported in the official statistics.

To predict the mortality of the existing building stock, mortality functions were constructed based on the official statistics from Statistics Finland: more precisely, Population and Housing Census reports with ten-year intervals between 1950 and 2000 that were acquired from the Doria repository of Statistics Finland [34] that is maintained by National Library of Finland, while the latest cross sections for the years 2010 and 2018 were acquired from the StatFIN database [33]. The collected data account for the size of the stock for different types of buildings by year of building at different cross-sectional years, allowing for construction of separate mortality functions for each classified purpose of use by completion decades. These mortality functions describe the proportion of buildings from their respective decades that still exist at different cross sections of time.

Second, an integrated mortality function for each purpose of use was constructed based on the mortality functions that depict buildings from different decades separately. The first two steps provided information on the differences and similarities between the mortality of different purpose of use classes, allowing further integration of the mortality functions for similarly behaving purposes of use classes. The final integration resulted in two different mortality functions for the entire building stock, including (A) mortality of residential and public buildings, and (B) mortality of commercial buildings (Figure 3). The figure reveals a faster mortality of commercial buildings relative to residential and public buildings. More precisely, the pace of mortality of commercial buildings rapidly increases after the age of 40, and by the age of 70, the majority of commercial buildings do not exist anymore. At the same time, the lifecycle of residential and public buildings is notably longer.

The middle line (black) in Figure 3 depicts the statistics-based mortality of buildings. However, there is no guarantee that the mortality rate of building stock in the future should

follow this history, which makes sensitivity analysis for alternative development paths necessary. To make the sensitivity analysis meaningful, plausible lower and upper limits for mortality development were defined by the members of the research group. In Figure 3, the upper line (green) depicts a low mortality scenario and the lower line (red) denotes a fast mortality scenario. By studying these three alternatives, an adequate understanding of the impact of changes within realistic boundaries should be achieved.

**Figure 3.** Three scenarios for the remaining proportion of the building stock by building types as a function of time: (**A**) residential and public buildings; (**B**) commercial buildings.

The final step is to convert the mortality functions into a usable form in terms of the QuantiSTOCK model. This is done by using a mortality sub-model, where the different variations of mortality functions are combined with the data on the current building stock. To define mortality rates for different building types at different cross sections in the future, the mortality sub-model also incorporates the needed new construction over time. These mortality rates, which are used as an input for the QuantiSTOCK model, are defined separately for (A) residential buildings and public buildings, and (B) commercial buildings. The rates vary between ten-year periods.

## *3.3. Validation of the Modeling Procedure*

To validate the modeling procedure, development of the Finnish building stock in a past period from 2006 to 2019 was modeled using the QuantiSTOCK model. Then, the modeled results were compared to the building stock statistics for the same period to prove that the outcomes are in the expected range. In the test period, the gross floor area per capita ratio annually increased by 0.8% for residential buildings while the yearly increase of gross floor area for non-residential buildings was 1.7%, which was taken into account in the modeling attributes. Figure 4 shows that the calculation-based results from the QuantiSTOCK model seem to correspond well with the actual development of the building stock. This proves that the model is capable of producing accurate results, if the modeling attributes are in line with the actual development. However, the challenge here is to be able to assess the real development of modeling attributes. Given that uncertainty is always present in these assessments, the importance of sensitivity checks should be emphasized in an attempt to find the boundaries for real future development.

**Figure 4.** Modeled development vs. statistics in the period of 2006–2019.

### **4. Results**

In this section, the modeling results for the Finnish building stock in the period of 2020–2050 are presented. The results are provided for the entire building stock, and regional differences are reported in accordance with (i) the fast-growing Helsinki region, (ii) growing regions, and (iii) zero-growth and declining regions. Moreover, sensitivity to changes in the mortality rate, population growth, and floor area per capita ratio is illustrated.

## *4.1. Modeled Development of Finnish Building Stock*

Figure 5 depicts the modeled development of the Finnish building stock in the period of 2020–2050. The results suggest that, in 2050, less than 25 percent of the building stock is built after 2020. This finding confirms the importance of addressing the existing building stocks in strategies to achieve the EU's carbon neutral targets by 2050. Another interesting observation based on the results is that the total size of the Finnish building stock will not increase if the population development is in line with the 2019 population projection from Statistics Finland. Instead, a decrease of two percent is modeled relative to the building stock size in 2020.

In Figure 6, the focus is on a need for new construction over the study period and its distribution into residential buildings and non-residential buildings, including public and commercial buildings. In the study period of 2020–2050, 65 percent of the need for new construction is modeled to be residential buildings, equaling 65 million square meters of gross floor area (if evenly distributed, some 49,000 housing units annually). At the same time, non-residential buildings represent 35 percent of the need for new construction, equaling 35 million square meters of gross floor area.

**Figure 5.** Modeled building stock development 2020–2050.

**Figure 6.** Modeled cumulative building production in the study period of 2020–2050: residential and non-residential buildings.

#### *4.2. Regional Differences*

In Figure 7, the modeled development is presented separately for the three different region groups, including (a) the fast-growing Helsinki region, (b) growing regions, and (c) zero-growth and declining regions. This more fine-grained representation reveals clear differences between the region groups, making the interpretations more meaningful. The modeling reveals that building stock is growing by 17 percent (19 million square meters of gross floor area) in the fast-growing Helsinki region and by 6 percent (5.5 million square meters) in other growing regions, while in zero-growth and declining regions, the total stock decreases 14 percent (33 million square meters). Despite the decreasing total stock, new construction is also needed in the zero-growth and declining regions due to migration

from more distant locations to district centers, where the ageing population has better access to services, increasing demand for housing in regionally central locations.

**Figure 7.** Modeled building stock development in different groups of regions: (**a**) fast-growing Helsinki region; (**b**) growing regions; and (**c**) zero-growth and declining regions.

Figure 8 reveals the distribution of the modeled new production into residential and non-residential buildings. In the fast-growing Helsinki region, the proportion of residential new production is 65 percent, equaling 33 million square meters of gross floor area (some 25,000 housing units annually). In the group of other growing regions, the proportion of new residential building production is 68 percent, equaling 19 million square meters of gross floor area (some 14,000 new housing units annually). In the zero-growth and declining regions, the proportion of residential buildings is 62 percent, equaling 13 million square meters (some 10,000 new housing units annually).

**Figure 8.** Modeled cumulative need for new construction in different regions: (**a**) fast-growing Helsinki region; (**b**) growing regions; and (**c**) zero-growth and declining regions.

## *4.3. Sensitivity Analysis*

As modeling of future development always leans on assumptions, it is important to recognize the key variables and explore how the modeling results are affected if values of these variables vary. This kind of sensitivity analysis allows a better understanding of the boundaries for actual development as modeling only one potential development scenario could result in faulty conclusions. Below, the sensitivity of building stock development to mortality rate, population change, and residential gross floor area per capita ratio is illustrated, ceteris paribus.

## 4.3.1. Sensitivity to Mortality Rate

Figure 9 depicts the impact of mortality rate on the development of building stock. The panels reveal that mortality rate has a notable impact on the structure of the future building stock. Specifically, in the low mortality scenario, the proportion of new construction since 2020 is 13 percent; in the history-based scenario, it is 23 percent; and in the high mortality scenario, it is 35 percent of the stock. However, based on the results, changes in the mortality rate alone do not seem to affect the total size of the building stock but the size of the stock is the same in all three scenarios at the end of the study period. This is because of the assumption that all mortality has to be replaced with new production within the region. The reasoning behind this is that, if this proportion of the building stock was an essential part of accommodating residents in the region before its mortality, these people need a place to live after the mortality as well if population growth is not negative. Whereas the total size of the building stock starts decreasing, if the population growth turns negative. Then, only the proportion of mortality that has a demand in the market is replaced with new production.

**Figure 9.** Modeled development of building stock with different mortality rates: (**a**) low mortality rate; (**b**) history-based mortality rate; and (**c**) high mortality rate.

#### 4.3.2. Sensitivity to Population Growth

In Figure 10, the impact of changes in population growth is illustrated. The middle scenario (b) is based on the official population projection from StatFIN database [33], and sensitivities to a decrease and an increase of 5 percent (approximately 270,000 residents, equaling to some 9000 residents annually if evenly distributed over the study period) are modeled in scenarios (a) and (c). The figure reveals that the impact is notable on both new production and the total size of the building stock. The total size of the stock is 6 percent (27 million square meters) smaller in 2050 than in 2020 in scenario (a), a little less than 2 percent (8 million square meters) greater in scenario (b), and over 2 percent (11 million square meters) greater in scenario (c). At the same time, the respective proportions of new production since 2020 at the end of the study period are 19 percent, 23 percent, and 28 percent, of which the proportion of housing is some 65 percent in all scenarios.

**Figure 10.** Modeled development of building stock with different population changes: (**a**) the population in 2050 at 5% less than in the official population projection; (**b**) the population in 2050 in line with the official population projection; and (**c**) the population in 2050 at 5% more than in the official population projection.

## 4.3.3. Sensitivity to Residential Gross Floor Area per Capita

Finally, Figure 11 illustrates the sensitivity to changes in residential gross floor area per capita ratio that captures several overlapping processes, including but not limited to changes in residential density and distribution of housing types, and potential excess of new construction. In scenario (a), the residential gross floor area per capita annually decreases by 0.5 percent; in scenario (b), the ratio remains the same as is in 2020 throughout the modeling period; and in scenario (c), the residential gross floor area per capita ratio annually increases by 0.5 percent. The figure reveals that the impact is notable on both new production and the total size of the stock. The total size of the stock in 2050 is modeled to be 9 percent (41 million square meters) smaller, 2 percent (8 million square meters) smaller, and 8 percent (36 million square meters) greater than the total stock size in 2020 in scenarios (a), (b), and (c), respectively. At the same time, the proportion of new production since 2020 in the respective scenarios is 15 percent, 23 percent, and 32 percent, of which the proportion of residential buildings is 38 percent, 65 percent, and 77 percent in scenarios (a), (b), and (c), respectively. The proportion of residential buildings since 2020 varies notably between the scenarios, as only residential gross floor area per capita ratio is changed, while the ratio for non-residential buildings remains the same. Changing gross floor area per capita ratios for both residential and non-residential buildings at the same time would not be meaningful, as the ratios may develop towards opposite directions, making scrutinizing only one change at a time more informative.

**Figure 11.** Modeled development of building stock with different gross floor area per capita ratios: (**a**) gross floor area per capita decreases annually by 0.5%; (**b**) gross floor area ratio through the modeling period as in 2020; and (**c**) gross floor area per capita grows annually by 0.5%.

## **5. Discussion**

In this paper, we introduced a calculation-based model for modeling the quantitative future development of building stocks in the long-term. The previous literature covers various ways to describe the current state of the building stock, including approaches that are based on constructed building archetypes [7], sample buildings [8], synthetic microdata [9], and agent-based building stock model [10]. Moreover, there are numerous papers where advanced forecasting approaches have been applied to produce short- or medium-term forecasts, for example, on construction demand. These approaches include but are not limited to multiple regression analysis [14], a panel vector error correction model [15], a combination of neural networks and genetic algorithms [16], grey forecasting [17], and Box–Jenkins model [18]. As our focus is on long-term modeling, the most closely related studies lean on material flow analysis [19–21] and various assessment approaches to forecast long-term housing needs [22–30]. However, these most closely related studies are usually limited to projections of housing stocks while other building types are excluded from the analysis. The presented QuantiSTOCK model provides a novel contribution to sustainable management of building stocks by combining approaches akin to what has been presented on dynamic material flow analysis in [21], on the assessment of housing

needs in [23,25], and on mortality analysis in [11]. In addition to residential buildings, as opposed to the previous literature, the QuantiSTOCK model also covers other building types. The only excluded building categories are industrial and agricultural buildings as their heterogenic nature would make modeling highly uncertain.

The base version of the QuantiSTOCK model operates in gross floor area units and provides a relatively straightforward calculation-based approach to model future development of building stocks, where (1) population growth, (2) mortality of existing buildings, and (3) gross floor area per capita ratio are the three main drivers for quantitative changes in the building stock. The population growth input is directly based on the official population projection from the StatFIN database [33]. Regarding mortality, the compiled mortality functions are akin to the survival functions in [11], but in our simplified approach, we do not apply any mathematical equations but the curves are rather visually fitted based on the points from history statistics. Next, curves that are relatively similar to each other were merged, which in this study resulted in separate mortality curves for (i) residential and public buildings, and (ii) commercial buildings. Finally, taking also into account the current state of building stock and cumulative need for new construction, the mortality curves are translated into mortality rates. These mortality rates vary with time as the building stock evolves. These rates provide a statistics-based base scenario for the analysis, which is complemented with low and high scenarios that provide boundaries for the range of variation.

In terms of gross floor per capita ratio, it is important to notice that this ratio is a multifaceted modeling attribute that captures many overlapping processes, including but not limited to changes in residential density and distribution of housing types, and potential excess of new construction. Thus, using gross floor area per capita ratio differs from using residential density as a modeling attribute instead. For example, urbanization trends contribute to an increasing proportion of apartment buildings where residential densities tend to be lower than in single-family houses. Additionally, as a result of increased housing prices in urban centers, more and more people may still prefer good locations but choose to consume less floor area, which also leads to lower gross floor area per capita ratio. On the other hand, if excessive new construction occurs in an area, it may seem that residential density has increased. However, in such a case, the actual reason for the higher gross floor area per capita ratio—or at least part of it—would be that more new construction has been delivered to the market in relation to the number of new residents who have moved in. Thus, it is important to properly consider which factors may affect gross floor area per capita ratio in each case. Relative to the methods in assessing housing needs in [23,25], our base modeling approach is a simplified version as QuantiSTOCK does not take into account headship rates for different groups. However, we recognize that this may be necessary for some analyses, for example, if the data allows a more detailed analysis for different age cohorts or the number of housing units is of a particular interest. To address these potential needs, the QuantiSTOCK model is easily modifiable to include such an alternative modeling approach. Another distinct advantage of the QuantiSTOCK approach is that it relies on publicly available statistics, making it easy to update the model when new statistics becomes available. Of course, this is limited to locations where public high-quality data are available. Otherwise, it is advisable to use self-compiled data that is tailored for the analysis. In cases where high-quality data are easily accessible, application of the model only requires relatively little effort, as opposed to more complicated simulation approaches. It is also critical to understand that, because of the great uncertainty about the predictor attributes in the long-term, it is not self-evident that application of more complicated approaches would results in more accurate outcomes than the outcome from this stripped-down model. This is supported by Boverket [25] who reports that Schmuecker [31] obtained similar results using a simpler method compared to more advanced approaches used in England. Another important aspect in using a relatively simple approach is that it allows one to transparently put into perspective which factors are important relative to the big picture. In contrast, complicated modeling approaches may focus on complicated descriptions of the modeling procedure, leaving an actual understanding of the critical factors blurred.

The QuantiSTOCK model seeks to provide a simplified and transparent modeling approach that is easy to understand. In this paper, we use "demand for new construction" and "need for new construction" as synonyms. The demand is considered here as objectively assessed need for new construction that is required to address changes in demographics and building stocks. In other words, the definition for demand is broader than the strict traditional definition in economics. As long-term projections even at the best include a great amount of uncertainty, the focus here is rather on major lines than in trifling matters. It is still good to bear in mind that a simplified approach always requires choices and approximations that may also hide some critical aspects. Therefore, it is critical to understand the impact of the incorporated assumptions as well as to perform adequate sensitivity checks for the development of key modeling attributes. In this study, the research group members defined together the sensitivity checks for our analysis on the Finnish building stock. The challenge in this strategy is to be able to define lower and upper limits for the modeling attributes so that they provide proper boundaries for actual development. In order to succeed in setting these proper limits, expertise in the field of a built environment is necessary and the modeling results are only reliable if the interpreters understand the underlying assumptions. Still, any "black box" approaches do not solve this problem either, as they only tend to increase the risk of unexplainable and unreliable modeling results. However, we recognize that, in some contexts, it may be necessary to increase the degree of complexity in the QuantiSTOCK modeling procedure, for example, to better understand the impacts of changing demographics. It is also important to understand that economic conditions and restrictions from land use planning notably affect the volume and structure of new construction that occurs in real world.

#### **6. Conclusions**

The QuantiSTOCK model is particularly developed for modeling the development of the Finnish building stock, which was used as a development and test arena in this study due to the good availability of high-quality data. However, this relatively simple modeling approach may also be widely applied to other geographic locations when fitted for location-specific data. By being transparent, the model provides fruitful grounds for understanding the impact of different key variables. This is necessary to allow more reliable analyses on the built environment and to facilitate meaningful debate on different housing, land use, and environment-related policies. As the proposed model is relatively easily modifiable, we consider it to have a great potential to be widely applied in various fields.

The modeling using Finnish data revealed that, in the study period of 2020–2050, the total size of the Finnish building stock will not be growing and the size of the stock may even slightly decrease if the population growth is in line with the official population projection; 65 percent of the new production in the study period is modeled to be residential buildings. However, the proportion of buildings that are built since 2020 is less than 25 percent in 2050, which once again is a reminder that measures addressing the existing building stock are critical in an attempt to achieve the EU's carbon neutral targets.

Further examination of the modeling results by region type reveals notable regional differences in the building stock development, confirming the high impact of urbanization. In the fast-growing Helsinki region, other fast-growing regions, and zero-growth and declining regions, the percentage changes of the total building stock size in the period of 2020–2050 are 17 percent, 6 percent, and −14 percent, respectively. However, the decreasing stock size does not directly mean that there is no need for new construction but that migration within these regions from more distant locations to district centers increases the demand in regionally central locations even when the total size of building stock in the region is decreasing.

Due to the uncertainty in modeling future development, sensitivity checks to changes in key modeling attributes are necessary to understand the boundaries of actual building stock development. The sensitivity analysis demonstrated that mortality rate has a notable impact on the structure of the future building stock, as proportions of new construction since 2020 varied from 13 percent to 35 percent. Also, a decrease or an increase of 5 percent (some 270,000 people) in the projected total population for 2050 had a clear impact on building stock development as the proportion of new production in 2050 varied between 19 and 28 percent and the size of the total stock varied between −6 and 2 percent. Furthermore, residential gross floor area per capita ratio was observed to have a high impact on the modeling outcome both in terms of new production and the total size of the stock. An annual decrease and increase of 0.5% in residential gross floor area per capita ratio resulted in the total size of the building stock in 2050 varying between −9 percent and 8 percent relative to the beginning of the study period. At the same time, the proportions of new construction since 2020 varied between 15 and 32 percent.

The introduced QuantiSTOCK model may be extended in numerous ways, and it lays the groundwork for modeling the future developments of building stocks. Some potentially fruitful strands for future work that could help develop the QuantiSTOCK model should be, for example, (i) a more detailed study on the dynamics of building stock development at different scales, (ii) the impact of local conditions on building stock development, (iii) the impact of municipal land use planning on building stock development, (iv) a further and more robust validation of the model using longer time horizon and data from other countries, (v) mortality differences between owner-occupied and rental buildings, and (vi) recognizing if construction techniques and materials have an impact on mortality. Nevertheless, already, today's version of the QuantiSTOCK model may be used in a wide range of analyses ranging from assessing housing demand at the regional level to providing input for defining sustainable pathways towards climate and land use targets. Thus, the results should be of interest to a wide range of researchers, policymakers, and community stakeholders who contribute to creating better built environments.

**Author Contributions:** Conceptualization, A.K., A.S., J.H. and E.N.; methodology, A.K., A.S., J.H. and E.N.; validation, A.K., A.S., J.H. and E.N.; formal analysis, A.K., A.S., J.H. and E.N.; investigation, A.K., A.S., J.H. and E.N.; resources, A.K., A.S., J.H. and E.N.; data curation, A.K., J.H. and E.N.; writing—original draft preparation, A.K.; writing—review and editing, A.K., A.S., J.H. and E.N.; visualization, A.K.; supervision, A.S.; project administration, A.S.; funding acquisition, A.S., J.H. and A.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Academy of Finland (decision number: 309,067, project: RenewFIN—Optimal transformation pathway towards the 2050 low-carbon target: integrated buildings, grids, and national energy system for the case of Finland). The main author would also like to thank the Finnish Foundation for Technology Promotion for funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Publicly available datasets were analyzed in this study. This data can be found here: [https://www.stat.fi/tup/statfin/index\_en.html] and [https://www.doria.fi/ handle/10024/67150].

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **The Contribution of Peripheral Large Scientific Infrastructures to Sustainable Development from a Global and Territorial Perspective: The Case of IFMIF-DONES**

**Virginia Fernández-Pérez <sup>1</sup> and Antonio Peña-García 2,\***


**Abstract:** Large scientific infrastructures are a major focus of progress. They have a big impact on the economic and social development of their surroundings. Departing from these well-known facts, it is not trivial to affirm whether the global contribution to Sustainable Development (SD) is higher when they are built in peripheral and not highly developed provinces instead of capitals and rich areas. Besides the economic impact on depressed areas, other SD-related parameters like the attachment of young and skilled people to their homeland, the avoidance of uncontrolled migrations from rural to dense urban zones, the growth of new focuses of knowledge independent from the lines of research established in the universities of the capitals, the indirect impact of auxiliary infrastructures and others must be analyzed. Concerning the next implementation of the "International Fusion Materials Irradiation Facility—Demo Oriented Neutron Source" (IFMIF-DONES) project in Granada (Spain), one depressed and tourism-dependent zone, an analysis and comparison with similar infrastructures were done and presented.

**Keywords:** sustainable development; global sustainability; scientific infrastructures; Post-COVID-19 Scenario

## **1. Introduction**

Science and Technology have been acknowledged for a long time as key motors to foster progress and well-being. However, the way scientific and especially technological development has been focused has evolved from an absolutely results-oriented philosophy to a more global perspective including not immediate benefits. The first philosophy lasted from the Industrial Revolution in the early 19th century until the third quarter of the 20th century, whereas the new perspective taking Environment and Sustainability as the main factors has been a constant trend for about 40 years.

Although there has been no single inflection point from the first to the current philosophy, we might be able to find one remarkably important milestone in the so-called Brundtland Report: "Sustainable development must meet the needs of current generations without compromising the ability of future generations to meet their own needs" [1].

The recent compilation of the Sustainable Development Goals (SDGs) [2], which summarizes and attempts to channel the efforts toward a better world, is fully applicable to the technological development and the infrastructures built to get it. Thus, besides the expected scientific and technical advances, the actions and projects with a strong technological background should bring highly qualified employment, investments and economic prosperity for the territory contributing to the achievements of the SDGs. Table 1 summarizes the SDGs and their official labels.

**Citation:** Fernández-Pérez, V.; Peña-García, A. The Contribution of Peripheral Large Scientific Infrastructures to Sustainable Development from a Global and Territorial Perspective: The Case of IFMIF-DONES. *Sustainability* **2021**, *13*, 454. https://doi.org/10.3390/ su13020454


Received: 16 December 2020 Accepted: 2 January 2021 Published: 6 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).


**Table 1.** United Nations Sustainable Development Goals (SDGs) [2].

In this scenario, it is really important to compile information and foresee the economic and social consequences of each project in order to support or disregard those decisions dealing with their implementation and the way to do it, including its location, timetable, public investment and other factors.

Thus, the role and responsibility of governments and public administrations have remarkably changed because now it is clearly understood and accepted that [3,4]:


The so-called "Circular Economy", more and more deeply implemented in Engineering processes going from design to decommissioning [5,6], is a good example.

Attending to its significance and dimension (environmental, social and economic) [7], it could seem that Sustainable Development is a holistic concept, and technology focused on the SDGs cannot be understood out of a global framework. Certainly, SDGs cannot be achieved from isolated perspectives, and some transversal branches of Science and Technology closely related to them are claiming for global visions [8]. However, it is important to keep in mind that a global perspective of technology is fully compatible and even necessary for territorial sustainable development exclusively focused on concrete areas.

This research presents and analyses the situation and potential effects of critical scientific and technological infrastructures on their near geographical framework from the perspective of Global Sustainability and SDGs. The particular case of the "International Fusion Materials Irradiation Facility—Demo Oriented Neutron Source" (IFMIF-DONES) is taken as an example and guide. Besides the main target of this project, that is, providing some key milestones from other perspectives parallel to the path toward clean and almost unlimited energy from nuclear fusion, it seeks to contribute to sustainable regional development in the abovementioned wide contexts of the Brundtland Report and SDGs (see Figure 1).

**Figure 1.** "International Fusion Materials Irradiation Facility—Demo Oriented Neutron Source" (IFMIF-DONES) from a global and territorial perspective. Sustainable energy economically and socially viable attending territorial necessities.

The next section will briefly present the IFMIF-DONES project and its objectives.

## **2. IFMIF-DONES: The Key Milestone toward the Use of Fusion Energy**

#### *2.1. The Situation with Nuclear Power*

The energy inside the matter can be obtained by breaking heavy and unstable nuclei like Uranium-235 of Plutonium-239 (nuclear fission) or by fusing light nuclei like hydrogen into heavier ones (nuclear fusion). The fundamentals of both, fission and fusion, are not complex and have been understood since the 1930s. However, the technology to actually obtain energy from the nuclei is extremely complex. In the case of fission, big and complex centers have been built since the 1940s, and it is nowadays an essential energy source. In Spain, 20% of the produced energy is achieved with just seven reactors that started to work between 1983 and 1988 [9]. This 20% is not enough and additional energy from nuclear origin must be bought from other countries. However, nuclear fusion is a technical challenge, and its control to produce useful energy is still far away. This is a problem because, whilst fission requires the use, management and storage of hazardous materials, fusion is expected to be a cleaner, safer and much more efficient source of energy.

In this general framework where mankind needs more and more energy to ensure economic growth and resources for everyone without harming the environment, both the scientific community and governments around the world are working hard to control nuclear fusion and thus profit from its huge benefits.

As mentioned above, the control of fusion is very difficult from a technical point of view. There are two main limitations: (1) fusing light elements requires working in extreme conditions around 100 million ◦C; and (2) the neutrons arising from the nuclear reaction are extremely difficult to control, and they would quickly damage the whole installation.

Several experimental installations have tried to demonstrate the feasibility of fusion by solving the abovementioned problems one way or another. Among them, the "International Thermonuclear Experimental Reactor" (ITER, www.iter.org) is the most ambitious project up to date. It is currently under construction in Cadarache (France) and, according to the official sources of the Project, it is expected to maintain fusion for long periods of time and test the integrated technologies, materials and physics regimes necessary for the commercial production of fusion-based electricity [10]. The fact that the ITER members are China, the European Union, India, Japan, Korea, Russia and the United States shows the high strategic interest of this experimental infrastructure and the geopolitical importance of fusion control. Figure 2 shows the state of part of the installation in December 2020.

**Figure 2.** ITER under construction. Credit © ITER Organization, http://www.iter.org/.

ITER Organization retains copyright in the pictures and videos. Download and use of the image do not amount to a transfer of intellectual property (https://www.iter.org/ media/www/downloads/av\_terms\_of\_use.pdf).

In parallel to ITER, the "International Fusion Materials Irradiation Facility—DEMO Oriented Neutron Source", known as IFMIF-DONES (www.ifmifdones.org), is the key to solve the second shortcoming from a more specific perspective.

#### *2.2. IFMIF-DONES: What It Is and How It Will Work*

Although the target of this work is not an exhaustive technical description of IFMIF-DONES, this section will briefly describe it so that its potential impact can be situated in the right framework.

The target of IFMIF-DONES is to obtain neutrons like those to be produced in real fusion reactions without reaching the extremely high temperatures of these reactions. In other words, it is an installation where the neutrons that we want to learn how to stop will be produced. The study of their effects on different materials will be studied, and the materials that support that flux of neutrons in the best way will be selected to build the fusion reactors in the future.

To achieve that key milestone in the race to fusion, IFMIF-DONES will have three main parts:


## *2.3. The Situation of IFMIF-DONES: Current Development of the Project*

IFMIF-DONES is expected to be installed in Escúzar, a small town in the Province of Granada (South of Spain). This is a rather economically depressed zone where most people work in agriculture or low-qualified services.

The project is expected to cost about EUR 700 million, which makes it the largest scientific infrastructure ever built in Spain. The costs will be funded by the Central Government of Spain and the Regional Government of Andalusia, the Spanish region where it will be located. Most of these national funds will come from the European Regional Development Fund, which finances lots of infrastructures in poorly developed European countries. Other funds will come from organizations like EUROfusion and other European programs.

In the first phase, already running in November 2020, the first budget of EUR 32.6 million has been transferred to the main implementing institutions of IFMIF-DONES: CIEMAT (Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas) and the University of Granada, both in Spain. These EUR 32.6 million are being used to start the construction and launch the first actions. In addition, EUR 4 million more have been provided by the European Strategy Forum on Research Infrastructures (ESFRI) in this first phase.

It is interesting to remark that, together with Escúzar, there were other candidates in the European Union to host IFMIF-DONES. They were in Poland and Croatia, two countries whose economic level is lower than the European average. It is also the case in Spain. Finally, Escúzar was appointed as an EU candidate to host IFMIF-DONES, but the three potential locations considered by the EU are a good indication of the potential of critical technological infrastructures as a motor for regional development.

## **3. Effects of Peripheral Large Scientific Infrastructures on Social and Territorial Sustainability**

Large scientific infrastructures connect with the terms of territorial social responsibility, sustainable development, social and economic sustainability and even education for sustainability [11]. Thereby, this type of project has impact and potential enough to contribute to change the territory in which they are inserted, and even far beyond.

According to Torjman [12], "human well-being cannot be sustained without a healthy environment and is equally unlikely in the absence of a vibrant economy". In this sense, the total sum of the investment planned for the first phase of construction and IFMIF-DONES will have a very positive socioeconomic and environmental impact because its implementation amounts to EUR 729.57 million, which will be executed over approximately nine years. Thus, the quality management of large projects is linked to sustainability and the development of a society and vice versa [13].

There is a growing recognition that sustainability is more than just the "green agenda", and it is considered that social practices are increasingly more important [14,15]. Nonetheless, integrated approaches to sustainability have failed to examine social sustainability in adequate detail [16]. For this reason, the scope of this research is the social impact of IFMIF-DONES and its consequences on the surroundings of the future installation. We start by defining social sustainability and how the project contributes to it.

There are several definitions of social sustainability with different ranges. From a wider perspective, the Western Australia Council of Social Services considers that "social sustainability occurs when the formal and informal processes, systems, structures, and relationships actively support the capacity of current and future generations to create healthy and livable communities" [16]. Secondly, social sustainability of a city (territorial approach) "is defined as development that is compatible with harmonious evolution of civil society, fostering an environment conducive to the compatible cohabitation of culturally and socially diverse groups and encouraging social integration, with improvements in the quality of life for all segments of the population" [17]. Lastly, regarding businesses, social sustainability is understood more generally as a business that influences individuals' or society's well-being [18,19] or, in other words, a system that meets the expectations of stakeholders without causing harm to the well-being of society and its members [20]. Here, the idea of social sustainability is commonly interpreted as the ability to continue to stay in business through good relations with stakeholders [21].

All these definitions reflect the development that social sustainability covers the broadest aspects of public and private activities and the effects that they have on employees, customers, suppliers, investors, and local and global communities, thus protecting all stakeholders and fairly respecting social diversity [22].

On the other hand, the key to achieving the sustainable development of a territory (territorial sustainability) is an understanding and modeling of its identity [23,24]. Sustainable territorial development can refer to different levels, and in particular to regional and local or relative to a country, continent or the entire world economy. Many researchers of social sustainability have focused mainly on urban studies from both academic and policy perspectives [25,26].

In this work, we review social sustainability from the perspectives of peripheral development. Ezcúzar and its province, Granada, have quite specific socioeconomic indicators as shown in the next sections. Therefore, it is necessary to work in this dependent context due to the influence of the relevance of the project through the perspective of social and territorial sustainability. Table 2 compares both cities in terms of population and economy.


**Table 2.** Comparison between the village of Escúzar and the city of Granada (Province Capital). Source: Instituto de Estadística y Cartografía de Andalucía 2019 https://www.juntadeandalucia.es/institutodeestadisticaycartografia/sima.

> In the next sections, the specific foreseen impact of IFMIF-DONES on the territory and its contribution to social and territorial sustainability, as well as the duly relationship with the SDGs, and the interconnection between all are analyzed.

## **4. The Overall Impact of IFMIF-DONES at Different Levels**

The implementation of a scientific infrastructure like IFMIF-DONES, the largest in Spain ever, has effects at various levels: worldwide, European, national, regional and local.


construction and operation of IFMIF-DONES will be a success for the European scientific policy.


The Province of Granada and its surroundings are characterized by:


The consequence of this configuration in the Province of Granada, where IFMIF-DONES is expected to be built, is a high rate of public/private employees, most of the employees of private companies working in tourism-related jobs, and a socioeconomic and cultural framework where the University of Granada is, by far, the main actor. Other provinces around Granada, like Almería, Jaén or Málaga, have also a big dependence on tourism, but their industry and/or agriculture has more relative weight.

In this particular territorial framework, the potential territorial yields of IFMIF-DONES are:


In spite of the good perspectives, the desired yields of the installation of IFMIF-DONES in the Province of Granada is not immediate and, obviously, not direct for these companies and citizens that are not prepared to offer the needed services. In addition, given that the vast majority of the contracts will be decided after public concurrence, enterprises and workers from other parts of Spain and even from all over the world could be hired. Table 3 below shows a schematic analysis of positive, negative and uncertain points regarding the implementation of IFMIF-DONES in the site of Escúzar.

**Table 3.** Threats, opportunities, weaknesses and strengths regarding the implementation of IFMIF-DONES in the site of Escúzar.


In summary, the yields of the installation of a critical scientific and technological facility like IFMIF-DONES are expected to be very high, but its contribution to really sustainable development and the relevant SDGs are not free of risks and uncertainties and will need big efforts and careful preparation at all the levels.

In the next section, the potential contributions of IFMIF-DONES to the SDGs are presented and analyzed.

### **5. The Impact of IFMIF-DONES on Social Sustainability and SDGs**

Beyond the impact of IFMIF-DONES from the pure perspective of financial profitability and yields, it is interesting to focus on aspects not frequently included when studying large scientific and technological infrastructures. In this sense, the analysis from the perspective of social sustainability is more frequent in other frameworks more related to actions to eliminate poverty, inequality, etc.

Theoretically, social sustainability as a concept covers broad societal issues [27] and has various interpretations in different fields [28]. Laureate Amartya and Sen identified five dimensions in social sustainability—equity, diversity, social cohesion, quality of life and democracy and governance—which have been considered in determining if a business or a project is socially sustainable. However, these have been extended in many studies. It has also been acknowledged that the social dimension of the SDGs needs further development [15].

In this section it is shown that large scientific and technological infrastructures are an efficient instrument for social sustainability through their direct impact on specific SDGs, using Khan's classification [29]. As it is evident from his table, numerous authors point toward similar themes as they remain the primary constituents of social sustainability.

Table 5 below summarizes the main impacts.


**Table 4.** Impact of IFMIF-DONES implementation on socioeconomic indicators and specific SDGs. Adapted from Khan [29].


**Table 5.** Impact of IFMIF-DONES implementation on socioeconomic indicators and specific SDGs. Adapted from Khan [29].

#### **6. Conclusions**

Large scientific and technological infrastructures have impacts on many aspects of human communities. These aspects go from international to territorial levels and also from economic to social perspectives, including a wide variety of intermediate levels.

These impacts are even greater when the infrastructures have deep implications as experimental nuclear fusion facilities. The necessity of cleaner, safer and more efficient energy from the atomic nuclei is a key milestone in the long way to really sustainable development. The current rate of development requires huge amounts of energy whilst the worrying situation of the environment, the negative effects on human health and the lack of more and more natural resources leaves the question of energy as one of the main challenges for the upcoming decades.

In this framework, the "International Fusion Materials Irradiation Facility—Demo Oriented Neutron Source" (IFMIF-DONES) is a unique project to test different material candidates for the construction of future fusion reactors. In other words, without the results of IFMIF-DONES, access to fusion energy will be impossible during this century.

Departing from this evidence and trying to enlarge the spectrum of positive effects at all levels, large projects like IFMIF-DONES should incorporate targets and strategies beyond the classic social responsibility in order to satisfy the necessities of the territories. Thus it is essential to create added economic and environmental value whereas, in parallel, creating social value [29]. This ensures sustainable economic success in the territory.

There is a mutual relationship between territory and project because the project prospers when the society prospers and vice versa: new enterprises, better houses, schools, businesses, etc., that equally generate new projects. This is one of the reasons why all big projects should foster the achievement of the relevant Sustainable Development Goals (SDGs) in a broad and integral perspective.

In this work, based on the forecasts and current advances in the implementation of IFMIF-DONES in a rural zone of the South of Spain, we highlighted that, besides the wellknown financial outputs and profits, large scientific and technological infrastructures are an efficient instrument for social sustainability through their direct impact on specific SDGs.

Furthermore, the asymmetrical social and economic impact of large scientific and technological infrastructures is also a matter of major concern. Whilst life conditions of people are not highly improved in big and developed regions, in small peripheral provinces with low average incomes, the effect on social sustainability, directly contributing to the achievement of certain SDGs, is very remarkable. Thus, people with a medium-to-high socioeconomic level and a wide variety of job opportunities may have some improvements in their conditions, but not great changes. On the other hand, unemployed people without a high level of studies and/or very limited options to change their jobs can experience big changes that will be transmitted to the rest of society through their big change in incomes, way of life and others.

Urban centers could be spaces where the sources of employment are larger than in rural spaces. As technology develops, the workforce has to be more specialized and therefore a greater gap will be generated between regions. Therefore, the migration for years of the rural population toward the cities is narrowing the labor market of these areas dedicated to agriculture, which is reducing the difference in income and opportunities between the rural and urban population. Thus, land attachment could also be favored with this kind of project that produces high returns to the territorial area in which it would be implanted.

As regards asymmetrical impact, institutions should treat laws and regulations of the projects as opportunities for improvement, development and sustainability, insisting on ethical behavior in their interactions with stakeholders [30]. However, trust and collective action are core topics because sustainability derives from accessible and inclusive processes. An integrated perspective on sustainability is thus implicated in more effective social sustainability, which in turn relies upon attention to local contexts and ideas.

Furthermore, the development of a territorial identity seems basic not only for avoiding offshoring risks but also because new implementations require specific characteristics and demanding rules for service quality in the territory. The regional and local identity among citizens, politicians and society in general allows an integrated approach to environmental and social sustainability that represents an added value when considering the attraction of future investments.

In summary, besides its important technical implications and its contribution to the control of fusion energy, IFMIF-DONES is expected to become an engine fostering the development of a depressed area thus demonstrating some hypotheses and becoming a great field experiment in social and economic sustainable development.

**Author Contributions:** Formal analysis, A.P.-G. and V.F.-P.; Funding acquisition, A.P.-G.; Investigation, A.P.-G. and V.F.-P.; Methodology, A.P.-G. and V.F.-P.; Project administration, A.P.-G.; Resources, A.P.-G.; Visualization, V.F.-P.; Writing—original draft, A.P.-G. and V.F.-P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the European Commission as part of the project "DONES Preparatory Phase" (Ref. 870186).

**Institutional Review Board Statement:** Not relevant.

**Informed Consent Statement:** Not relevant.

**Data Availability Statement:** Not relevant.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Priority of Accident Cause Based on Tower Crane Type for the Realization of Sustainable Management at Korean Construction Sites**

**Ju Yong Kim 1, Don Soo Lee 2, Jin Dong Kim <sup>3</sup> and Gwang Hee Kim 1,\***


**Abstract:** Construction safety is a key factor among the many factors related to the sustainable management of construction sites. Although research is underway to reduce potential accidents in the construction industry in Korea, the number of tower crane (T/C) accidents is consistently increasing based on the increased use of such cranes. In this study, the priorities of accident causes for each T/C type were derived and utilized for the sustainable management of construction sites. An analytic hierarchy process (AHP) questionnaire was completed by experts such as construction engineers, construction managers, safety engineers, and T/C operators with more than ten years of field experience. The results of the AHP questionnaire revealed that the leading cause of cab-control T/C accidents is poor operator visibility, while the leading cause of accidents related to remote-control T/Cs is the poor management of lifting objects and control of surroundings. The high-ranking causes derived in this study should be managed and priority measures should be implemented to reduce the number of T/C accidents.

**Keywords:** sustainable construction management; tower crane accident reduction; priority of tower crane accident causes

## **1. Introduction**

The crane has become a major symbol of building construction sites and is often the most prominent piece of equipment at a building construction site based on its size and the key role that it plays at many construction sites [1]. The use of tower cranes (T/Cs) at construction sites has consistently increased since their introduction into the Korean construction industry in the 1980s. According to statistics on construction machinery statuses from the Ministry of Land, Infrastructure, and Transport, in 2015, Korea contained 3408 cab-control (CC) T/Cs and 272 remote-control (RC) T/Cs. Generally, CC T/Cs are used for lifting objects weighing three tons or more and RC T/Cs are used for lifting objects weighing less than three tons. In 2019, the number of CC T/Cs increased by 22% to 4385 and the number of RC T/Cs increased by over 85% to 1845 [2].

One of the major causes of fatalities is the usage of cranes during lifting operations in the construction phase of the construction project lifecycle [3]. As the number of T/Cs used at construction sites has increased steadily, there has been an increase in fatalities and accidents because T/Cs are relatively dangerous and various risk factors are inherent to assembly, lifting, and disassembly works [4]. As accidents at construction sites are closely related to construction time, cost, scope, and company reputation [5], and because construction workplace safety and health are essential elements of sustainable construction management [6], construction accidents must be reduced for the sustainable management of construction sites. Reyes et al. [7] stated that when quantifying the sustainable value of a construction project, health and safety indexes should be considered. Therefore,

**Citation:** Kim, J.Y.; Lee, D.S.; Kim, J.D.; Kim, G.H. Priority of Accident Cause Based on Tower Crane Type for the Realization of Sustainable Management at Korean Construction Sites. *Sustainability* **2021**, *13*, 242. https://doi.org/10.3390/su13010242


Received: 14 November 2020 Accepted: 25 December 2020 Published: 29 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

the government, academia, and practitioners in Korea have made various efforts to reduce T/C-related accidents. To reduce T/C accidents, the Korean government revised the enforcement regulations of the Construction Equipment Management Act in October of 2019, subdividing the safety training programs for construction equipment operators into 19 types and shortening the training cycle of RC T/C operators. Members of academia have also conducted research [8–10] on various causes of accident occurrence to reduce T/C accidents.

Although many efforts have been made to prevent accidents related to CC T/Cs, research on RC T/Cs, which are becoming increasingly common at Korean construction sites, is insufficient, leading to many accidents at construction sites. Every year, the number of T/C-related accidents at Korean construction sites continues to increase. The numbers of fatalities related to T/Cs were nine in 2016, seven in 2017, six in 2018, and eight in 2019 [11]. T/C accidents were officially announced during the first quarter of 2020, when five casualties had already occurred. Figure 1 presents an image from January 2 of 2020, where a 30 m T/C collapsed at a construction site in Incheon, Korea. This accident caused two fatalities and one injury. Kim [12] analyzed T/Cs at Korean construction sites and proposed the following main accident causes: (1) In the case of CC T/Cs, the main causes are equipment age, insufficient work management, violation of work guidelines and safety rules, and lack of communication. (2) In the case of RC T/Cs, the main causes are a lack of knowledge regarding work manuals for installation workers, insufficient checking of the cables used for fixing lifted objects, a lack of simultaneous checking of camera feeds during tying and lifting work, and insufficient checking of the specifications of heavy objects.

**Figure 1.** A 30 m T/C (tower crane) collapse at a construction site in Incheon, Korea [5].

Therefore, this study aimed to identify whether ranking can be utilized for the accident causes' management of T/C types by analyzing the importance of accident causes for each crane type. The results of this study can contribute to reducing construction accidents by identifying management causes for T/Cs that should be considered during the planning process for construction accident prevention and safety management activities.

## **2. Literature Review**

## *2.1. Previous Research*

Since the 2000s, various studies related to T/C accidents have been conducted around the world. These studies can be classified into three major categories: (1) risk analysis for analyzing T/C accidents [13,14], (2) development of management goals or plans by analyzing T/C accidents [14–16], (3) derivation of the major accident causes related to T/C accidents [17–21], and (4) presenting measures for preventing tower crane accidents [22]. Thus far, most studies have focused on the causes or risks of T/C accidents based on specific causes and direct management. Especially, Fang et al. [22] developed a framework for

real-time pro-active safety assistance (RPSA) for mobile crane lifting operation, and Zhang et al. [20] and Zhou et al. [21] presented the tower-crane accident cause system (TCACS) model, which was a quite logical model through system thinking and case analysis to quantify the tower crane accident causes. However, there was a limit to revealing the direct cause of T/C accidents. Therefore, to reduce T/C accidents, it is necessary to manage various accident causes that cause accidents comprehensively. In other words, it is necessary to manage the accident causes that cause accidents with a high frequency or high probability more intensively. Additionally, in Korea, the use of RC T/Cs has increased based on pressure from the T/C union and the revision of the labor laws that limit working hours. Thus far, most research has focused on CC T/Cs, but there is a need to proceed with research to reduce all types of T/C accidents, including RC T/C accidents.

## *2.2. Construction Safety in Sustainability*

The area for sustainability appears to be focusing on limiting environmental impact, reducing energy, and incorporating less harmful material. Additionally, sustainability takes into account the environmental, economic, social and resource impacts of construction as well as the implementation of its principles throughout the lifecycle of building. However, Chandra [23] insisted that sustainable construction safety and health are an integral part of sustainable and environmentally friendly construction efforts. In addition, sustainable construction is defined as the creation of a healthy construction environment and responsible management based on resource efficiency and ecological principles. Rajendran et al. [24] recommended for research investigating the impact of green design and construction on worker safety and health, taking into account the safety and health of workers as well as the safety and health of the end user. Especially, the Leadership in Energy and Environmental Design (LEED) is designed to define eco-friendly buildings by establishing a common standard for measurement or rating systems and to achieve three main objectives: market innovation, design integration, and education on sustainable principles and sharing ideas [6]. Hinze et al. [6] presented the concept that worker safety and concern belong to education—the third plan of the LEED objectives. As mentioned, the sustainability certification, LEED, also includes worker safety in the construction process, and construction safety has become an indispensable item in sustainability.

## *2.3. T/Cs at Construction Sites*

Cranes mainly used in the construction industry are classified into two equipment types: tower cranes and mobile cranes [1]. Mobile cranes can be classified as truck-mounted mobile crane and crawler crane. The basic truck-mounted crane configuration is a "boom truck" featuring a rear-mounted rotary telescopic boom crane mounted on a commercial truck. The crawler crane is boomed on a vehicle with a crawler track set that provides both stability and mobility. For many years, particularly in Korea, T/Cs have been widely used in all types of building construction projects in both urban and rural areas. In high-rise construction, T/Cs are a key type of equipment for moving materials, building elements, and form work components horizontally and vertically [1]. As shown in Figure 2, the major parts of the T/C are the mast, main jib, and counter jib. The mast, which is a steel structure that serves as a pillar supporting the T/C, is constructed on the upper part of the basic mast, which is connected to the mounting configuration. Masts are available as rail-mounted units, stationary units, climbing units, and mobile units. Depending on whether the mast is fixed or slewing, a T/C can be classified as a fixed or a slewing T/C. Additionally, T/Cs can be classified as top-slewing and bottom-slewing T/Cs according to the T/C rotation position. The T/C cabin is attached to the crane structure or installed at a remote location. The cabin of a top-slewing crane is almost always at the top of the mast, often at a significant distance from the work area. In this type of crane, it is important to improve the quality of operator visibility. Bottom-slewing cranes do not have an operator cab attached to the crane structure. The main types of jibs on T/Cs are saddle jibs and luffing jibs. A saddle jib is supported by a pendant in a horizontal or near-horizontal position and the load hook

changes the hook radius by moving along the jib on a trolley. A luffing jib rotates on the jib foot and is supported by a luffing cable. The load-bearing hoist rope typically passes through the sheave of the jib head and changes the hook radius by changing the inclination angle of the jib. A saddle jib typically has a smaller minimum working radius than an equivalent luffing jib, so it can handle loads closer to the crane tower. Many luffing jib cranes have very short counter jibs, which can be advantageous when a crane is operating near obstacles such as other cranes or adjacent buildings. In the case of Korean construction sites, CC T/Cs have dominated in the past, but RC T/Cs have been increasing in use rapidly in recent years. This trend appears to have been partially influenced by pressure from the T/C operator union and construction labor union.

**Figure 2.** T/C configuration.

### *2.4. Types of T/C Accidents*

The 40 tower crane accidents that occurred in Korea from 2015 to October 2020 were analyzed and classified into three categories such as the processes of erection, dismantling, and operation and management. The accident cases based on the analysis of T/C accident reports published from Korea Occupational Safety and Health Agency are the result of analyzing 40 accident cases—the most frequently occurred in the operation and management stage, 11 cases occurred in the erection stage, 7 cases in the dismantling stage, and 22 cases in the operation and management stage as shown in Table 1. Representative examples of each stage are as follows. In the erection stage, accidents related to eccentricity occur based on the operation of a T/C in a state where the telescopic component, mast, and other components are not fully fixed. Telescopic accidents break the balance of the jib while the telescopic mast is lifting or moving in an unstable state because the lower part of the turntable and upper pin of the telescopic cage are not fastened [25]. Collapse accidents occur because the member of the telescopic cage buckles based on improper use or non-use of a balance weight for the bidirectional balance of the jib during telescopic work [26]. Accidents in the dismantling stage are caused by the mast losing balance or the basic anchor being damaged. The position of the bolt hole for fixing the mast may be misaligned, so when an operator attempts to adjust the hole position, the crane may lose its balance and collapse. There can also be deviation between the cage roller and mast, causing a dismantling worker to disable the interlock function that stops crane operation. Forcible manipulation in this scenario can result in an accident [25]. Some accidents are also caused by the inadequate selection of standard lifting positions for dismantling, which causes the connecting part of the jib to split. During the process of T/C operation and management, the causes of accidents are mainly non-compliance with the work procedures suggested by manufacturers and the negligence of management in terms of safety inspections and education.


**Table 1.** T/C accident cases by construction phase/task.

#### **3. Methods**

This study aimed to derive the importance of T/C accident causes and the potential for T/C accidents by quantifying the experiences of experts related to construction accidents. Intensively managing such factors should aid in reducing T/C accidents. The analytic hierarchy process (AHP) technique was adopted to quantify the experiences of experts related to construction accidents in the field. The AHP is a structured decision-making technique developed by Saaty in the early 1970s. It can reflect the knowledge, experience, and intuition of respondents in pair-wise comparisons based on the elements of the hierarchy of decision-making [27]. The experts related to construction accidents for the AHP are construction engineers, construction managers, safety managers, and T/C operators that have more than 10 years of field experience.

As shown in Figure 3, this research process can be divided into the following 4 steps. (1) Identify major T/C accidents through a literature review. (2) Extract first-level phase/task and second-level structure accident causes for the AHP through interviews with related experts. After extracting 11 phases/tasks from the previous literature [12,14,15,18,19,21,28], 5 items were selected by integrating 11 items through interviews with the related experts. For the second-level accident causes, 15 items were selected by conducting a preliminary survey of accident causes extracted from the previous literature (refer to Table 2) as a Likert scale to the related experts, and the results are presented in Table 3. (3) Provide an AHP questionnaire to a total of 44 related experts, such as construction managers, 10 safety managers, 14 construction engineers, and 10 T/C operators. (4) Follow the AHP to derive priority management accident causes for reducing T/C accidents.

## **Figure 3.** Developed model and research process.




**Table 2.** *Cont.*

**Table 3.** Preliminary survey results of accident causes.


**4. AHP Model for T/C Accident Factors**

*4.1. Extract T/C Accident Factors and Structure of the AHP Model*

Major causes were extracted by interviewing related experts, such as construction managers, safety managers, and construction engineers, after arranging the causes of T/C accidents discussed in the previous literature. Expert interviews for classifying the extracted major phases/tasks and causes, and identifying first-level phases/tasks and second-level causes were conducted to structure the hierarchy of the AHP model. The firstlevel phase/task in the AHP model are five categories of dismantling work, lifting work, erection work, prime contractor management, and T/C machinery (Figure 4). Table 3 lists a total of five first-level phases/tasks and second-level accident causes that were extracted from the previous literature based on expert interviews and preliminary surveys. These accident causes were used in our AHP model for T/C accidents.

## *4.2. Structure of the AHP Model*

Figure 4 presents an AHP model for CC T/Cs and RC T/Cs. F1 to F5 in Figure 4 represent the first-level phase/task. Among the second-level accident causes for lifting management, F21 is applicable to both types of cranes, whereas F22 and F23 are relevant to CC T/Cs, and F24 and F25 are relevant to RC T/Cs.

**Figure 4.** AHP structure for T/C accidents.

## *4.3. AHP Survey*

The analysis results for the AHP questionnaire are presented in Table 4. The weights of the first-level phase/task indicate that erection work is the most important phase/task with a value of 0.226, followed by T/C machinery (0.216), lifting work (0.214), and prime contractor management (0.175), while dismantling work is the least important with a value of 0.170 for CC T/Cs. In the case of RC T/Cs, lifting work is the most important phase/task with a value of 0.264, followed by erection work (0.254), dismantling work (0.170), and T/C machinery (0.167). Prime contractor management has the lowest value of 0.146.


\* Wt: Weight; \*\* R: Ranking.

The analysis results for the AHP questionnaire on the second-level causes of T/C accidents are presented in Table 5. The weights of the second-level causes for CC T/Cs appear from largest to smallest as follows: poor operator visibility, improper bolting of the brace/mast/telescopic element, poor subcontracting technology management process, failure to comply with safety rules and work guidelines for erection work, problems with overseas parts procurement, and poor understanding of risk factors during the dismantling process. As shown in Table 5, in the case of RC T/Cs, poor management of lifted objects and control of surroundings are the most important causes, followed by the operator being unable to check the weight and specifications of the lifted objects, failure to comply with safety rules and work guidelines for erection work, insufficient worker skill for erection work, and improper bolting of the brace/mast/telescopic element.




**Table 5.** *Cont.*

## **5. Case Study**

Since it is difficult to measure the rate of reduction in accident causes by applying the results of this AHP analysis to actual construction sites, we propose the reduction rate of accident causes through the scenario of installing a camera and wireless transmitter on the trolley of tower cranes and also placing a safety manager of tower crane (refer to Figure 5). In this scenario, it is assumed that the tower crane operator (both CC T/C and RC T/C), the safety manager of tower crane, and the field office have installed monitors that can check the lifting work of the tower crane. As shown in Table 6, the weight calculated in AHP analysis was converted into the probability of accident causes in order to calculate the reduction rate of accident causes. The probability of erection phase is 22% for CC T/C and 25% for RC T/C. Dismantling phase is 17% for both T/C and Lifting work; is 22% for CC T/C and 27% for RC T/C and prime contractor management; is 17% for CC

T/C and 15% for RC T/C and T/C machinery; is 22% for CC T/C and 16% for RC T/C. As shown in Table 6, in the case of installing the camera to tower crane for tower crane operator, the safety manager of tower crane and field office, most of the accident causes are eliminated, so that the reduction probability for CC T/C is 29% and RC T/C is 30%. In the case of placing the safety manager of tower crane, the reduction probability for CC T/C is 55% and RC T/C is 49%. Although this reduction probability is not the result obtained after applying two cases to the actual construction site, it is believed that the camera attached to tower crane and safety manager of tower crane can eliminate most of the actual causes of tower crane accidents.

**Figure 5.** Safety camera and wireless transmission for preventing tower crane accidents.

**Table 6.** Accident probability and reduction probability applied to T/C camera and manager.



**Table 6.** *Cont.*

## **6. Discussion and Conclusions**

Despite various efforts to reduce accidents related to T/Cs, the number of accidents is still increasing. This study was conducted to help prevent T/C-related accidents by ranking the accident causes related to CC T/Cs and RC T/Cs, as well as the weights of each cause to be used as a reference for management. As shown in the results of our AHP questionnaire, various accidents occur when an operator cannot directly check the status or ties when a lifting object is hidden by other structures or objects, and is largely dependent on the signals and radio communications of other workers. The result of this study is "operator visibility" as first ranking cause in case of CC T/C, and is "operator unable to check the weight and specification of the lifting objects" as second ranking cause in case of RC T/C. In previous study [17], the operator's impact such as "operator proficiency", "operator character", and "employment source (operator)" was dominated. The collective weight of these causes is nearly 24%. The cause of the tower crane "operator proficiency" was suggested as the second-level accident cause. Especially, the comprehensive cause of operator impact was suggested. The operator proficiency is an ambiguous cause of tower crane accidents. In other words, the way to solve this cause is ambiguous. To solve this cause, the operator training cycle in Korea has recently been shortened. In this study, operator visibility and to check lifting objects are very specific causes that can be managed and are suggested solutions at construction sites. Therefore, one could prevent tower accidents by attaching a device like RPSA to the jib or hook for both CC T/Cs and RC T/Cs. Such device could help T/C operators monitor their work from the cabin and make

decisions based on signals from other workers and their own judgment. Additionally, it is necessary to train managers and workers continuously to help them maintain and comply with guidelines and manuals related to erection work, lifting work, and operation.

The major accident cause for CC T/Cs were ranked in descending order of "poor operator visibility", "bad bolting on the brace/mast/telescopic element", and "poor subcontractor technology". Such causes like "bad bolting on the brace/mast/telescopic element" and "poor subcontractor technology" are combined various factors such as the management problem of the prime contractor, the management problem of the subcontractor, and the skill of erecting/dismantling worker. The major accident cause for RC T/Cs were ranked in descending order of "poor management of lifting objects and control of surroundings", "operator unable to check the weight and specifications of lifting objects", and "failure to comply with safety rules and work guidelines". In particular, in the case of poor management of lifting objects and control of surroundings, there is a problem in the function of properly controlling and managing T/Cs because cranes are controlled and managed by numerous operators who have completed the required training for each type of construction work. Therefore, it is recommended to have a separate manager in charge of supervising work using T/Cs at a construction site. Furthermore, the second major accident cause of checking the weight and specifications of lifting objects can be addressed by attaching RPSA to help operators to make informed decisions. This situation is similar to that of a CC T/C.

In previous study [17], the "site-level safety management" is the highest weight cause affecting safety on construction sites with tower crane. Especially, the superintendent effect has "superintendent character" and "site-level safety management" to influence the safety of the crane-related site, a total of over 23%. In other previous study [20], as a result of analyzing the previous literature, the causal factors belonging to "site personnel management" are also very important as the frequency occupies the first to third place. In this study, most accident causes are related to situation/surrounding control using T/Cs and management issues that require workers to comply with work guidelines and rules of erection, dismantling, and lifting works that reflect reality. Therefore, it is important to designate a safety manager of tower crane with sufficient experience and knowledge regarding T/Cs to train T/C operators, engineers, and managers, and to revise, supplement, and manage various instructions and manuals. Additionally, the safety manager of tower crane is expected to provide sufficient help in terms of reducing accidents through consistent safety-based management and exercising practical control authority over T/C erection, telescoping, and lifting operations. In the case of an RC T/C, plans for supplemental education should be prepared as it becomes increasingly easy to obtain an operating license. In the field, it is necessary to establish a reinforced training plan for safety rules during erection, operation, and dismantling.

To prevent T/C accidents, we derived the priorities of accident causes for different T/C types using an AHP questionnaire. It is crucial to reduce accidents for sustainable management at construction sites. If the priorities of T/C accident causes presented in this paper are utilized in various checklists or management plans, and construction site management is conducted accordingly, then sustainable construction management can be realized. In the future, if additional research is performed by narrowing the scope of AHP questionnaire respondents to workers who directly use T/Cs and experts familiar with T/Cs, more realistic results can be derived, which will further reduce T/C accidents.

## **7. Limitations and Future Research**

This research was conducted to derive the priority of tower crane accident causes based on the experience of the related experts of construction sites with tower cranes. However, the effect was not verified by actual application to safety management plans or tower crane checklists at the real construction sites. Therefore, in the future research, it is necessary to apply the priority of accident causes based on expert experience to the

practical example in accordance with the T/C configuration to identify the actual accident reduction level.

**Author Contributions:** J.Y.K. reviewed the existing literature, completed the AHP questionnaire, and conducted the result data coding of AHP questionnaire and analysis.; D.S.L. analyzed the characteristics of each tower crane type and suggested a new tower crane safety model for application of the research results.; J.D.K. selected the subjects of the AHP questionnaire and conducted a AHP survey; G.H.K. conceived the whole this study and conducted a review of the research results. All authors have read and agreed to the published version of the manuscript.

#### **Funding:** No funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to restrictions on right of privacy.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **The Transition from Traditional Infrastructure to Living SOC and Its E**ff**ectiveness for Community Sustainability: The Case of South Korea**

## **Yeonsoo Kim, Jooseok Oh and Seiyong Kim \***

Department of Architecture, Korea University, Seoul 02841, Korea; yeonsooclarekim@gmail.com (Y.K.); ohjooseok@korea.ac.kr (J.O.)

**\*** Correspondence: kksy@korea.ac.kr; Tel.: +(02)-3290-3914

Received: 23 October 2020; Accepted: 5 December 2020; Published: 8 December 2020

**Abstract:** In 2018, the South Korean government began promoting a "livelihood-improving" social overhead capital policy based on the concepts of an inclusive city, smart shrinkage, and the balanced development of metropolitan and provincial cities. Based on a review of the extant literature and relevant policies from South Korea, this study explores this policy's implementation and makes some suggestions for its sustainability. This study compares the current state of South Korea's urban facilities' and the balance of their supply between metropolitan and provincial cities. To discern which type of facility central and local governments should prioritize, this study conducts a stepwise regression analysis and identifies which preexisting facilities influence the facility type proposed by the current policy. Results show that South Korea's living infrastructure is well distributed among metropolitan and provincial cities. However, urban planning shows little consideration for minimizing the distance between facilities and residential zones. In terms of facility types, the supply of education and local community facilities was adequate throughout the country, while culture and art facilities were inadequate. In metropolitan cities, the supply of sports and leisure facilities was insufficient.

**Keywords:** social capital; living environment; living infrastructure; soft infrastructure; living social overhead capital; inclusive growth; inclusive city

## **1. Introduction**

First presented by the United Nations as a theme of its Global Campaign on Urban Governance in 1999 [1], the concept of city inclusion comprises three dimensions: social, economic, and spatial. Of these, spatial inclusion refers to equal accessibility to living infrastructure and public services [2], mainly because people with limited access to living infrastructure and public services experience social exclusion from various social opportunities [3]. Accordingly, the global community has sought sustainable, inclusive growth by ensuring universal access to living infrastructure and public services [4]. In addition to the concept of inclusive growth, there is a growing emphasis on nonphysical infrastructure (or soft infrastructure) beyond the conventional concept of living infrastructure and public services [5]. Unlike physical infrastructure (or hard infrastructure), such as roads and ports, nonphysical infrastructure refers to all services essential for maintaining a nation's economy, health, and culture. From the late twentieth century, European and North American scholars have defined parks, green areas, and community sports facilities as social infrastructure and have actively participated in policymaking and research related to social infrastructure [6–9].

In 2018, South Korea proposed a "livelihood-improving" social overhead capital policy (hereinafter, Living SOC) as a practical alternative to realizing "spatial inclusiveness". Defined as "a small-scale living infrastructure easily accessible to people in the community", the Living SOC reflects the paradigm

of global change while promoting balanced development between South Korea's regions and cities and creating more equitable living standards [10]. While the policy draws on the concept of the "living infrastructure" insofar as it includes the same type of facilities, it also reflects the concepts of "spatial equality" and "equal accessibility" [11]. The Living SOC policy was motivated by South Korea's 2018 Gini coefficient, which showed that the regional disparity between metropolitan and provincial cities in the Living SOC supply across the country was worse than the income disparity between individuals [12]. In general, metropolitan cities refer to large central cities with a population of more than 50,000. They serve as hubs for social and economic activities in the surrounding area, while provincial cities refer to other areas [13]. However, following Article 175 of the Local Autonomy Act, South Korea has defined metropolitan cities as specific cities with a population of more than 500,000 and eligible for special treatment, and this study follows this definition. With the national budget increasing by approximately KRW 10.4 billion per annum [14], the government allocated funds for a national project for balanced development based on the Living SOC—including some KRW 500 billion for startup expenses alone [15]. As such, policies related to Living SOC are increasing in significance.

However, despite relatively refined standards, the Living SOC policy faces the criticism that it is not much different from the previous policy of supplying "living infrastructure" in terms of its exhaustive list of facilities [16]. Urban shrinkage is becoming common in cities worldwide, including over 20 cities in South Korea [17]. Here, urban shrinkage is a concept established through the 2002 in German miniature city project [18,19]. Urban shrinkage does not mean that a city's physical size is getting smaller, but rather an urban phenomenon in which boundaries and infrastructure remain the same, while the population and economy decline [20,21]. Provincial cities experiencing urban shrinkage may suffer from various problems, including poor usability due to a superfluous supply of facilities and their deteriorating conditions, and the difficulty of procuring the financial resources necessary for the upkeep of facilities. Nonetheless, with the area of various convenience facilities intended to sustain people's lives predicted to increase from 2792 km<sup>2</sup> in 2015 to 3842 km<sup>2</sup> in 2040 [22], the significance of adequately supplying such infrastructure has also been emphasized.

As such, South Korea is facing the complicated task of supplying Living SOC equitably and sustainably to cope with urban change and resolve spatial inequality. Accordingly, it is necessary to analyze the characteristics of cities to address urban shrinkage and identify the type of infrastructure. In particular, as the current Living SOC policy overlaps with preexisting living infrastructure in urban areas, it is essential to determine whether the current supply of living infrastructure overlaps with the facilities proposed by the new policy and whether the pertinent facilities are distributed equitably. Extant studies from South Korea [23–25] are limited insofar as they primarily focus on examining the condition of major living infrastructure and strategies for improving accessibility. Moreover, investigating and analyzing the physical and demographic conditions of each city, as well as the current condition of the major urban facilities that can be categorized as Living SOC, will improve the implementation of relevant policies.

Considering the foregoing, this study examines the implementation and sustainability of South Korea's Living SOC policy. To overcome the limitation in the extant research and relevant policies, this study examines the current condition of the living infrastructure established in South Korea before 2018, when the Living SOC was introduced, from the perspective of urban planning and land use. Per the concept of balanced development, this study examines the distribution of the preexisting facilities and analyzes whether the pertinent facilities are evenly distributed between metropolitan and provincial cities. By discerning which types of preexisting facility influence that proposed in the Living SOC policy, this study identifies which type of facilities the central and local governments should prioritize in terms of supply. The findings of this study can facilitate economic stability and sustainability by improving the implementation of relevant policies going forward.

#### **2. Literature Review**

#### *2.1. Social, Soft, and Living Infrastructure*

This study understands the concept of Living SOC to be similar to that of an inclusive city, which was proposed in the 1990s as a solution to a regional imbalance between cities [10]. The concept also draws on practical land and urban planning strategies and policies proposed by countries in Europe, North America, and East Asia [12].

In this respect, the concept of Living SOC is the closest to that of social infrastructure—that is, a composite of resources and facilities—including spaces, services, and networks—that vitalize the local community [26] and preserve the happiness and quality of life of community residents [27]. The concept of social infrastructure is generally contrasted by physical or economic infrastructure. While the physical infrastructure directly supports economic growth, social infrastructure aims to help build the community by providing the necessary social services [28] and improving residents' quality of life [29]. Social infrastructure also contributes to economic development by ensuring the effective utilization of human resources [30]. Social infrastructure can be defined as the physical environment that determines the successful development of social capital [31].

Compared to social infrastructure, living-related infrastructure refers to more specific physical facilities that community members need for daily life, such as houses, parks or green areas, water facilities, parking lots, and hospitals [32]. Meanwhile, living infrastructure refers to physically alive and easy infrastructure for community members to access and utilize [33]. Living infrastructure is similar to the living-related infrastructure insofar as it is defined in the scope of the preexisting infrastructure in close relation to daily life from the perspective of social and natural science. However, in contrast to living-related infrastructure, the concept of living infrastructure emphasizes the sustainability of the local and urban residents by adding to it "being alive" [34].

Infrastructure can also be divided into hard and soft infrastructure which provide both physical and social services. Hard infrastructure is a new category of large-scale infrastructure, comprising the basic urban structures such as roads, ports, electric/energy plants, water supply, and sewage systems [35]. In contrast, soft infrastructure refers to the necessary services for maintaining a community's economy, health, and culture [5].

As such, new definitions of infrastructure transcend the traditional definition of infrastructure as the physical and essential facilities for constructing and operating cities to include those intended to improve community sustainability and improve residents' quality of life. Such a perspective of infrastructure is widely accepted by developed and advanced countries seeking to ensure cities' sustainability and their inclusive growth. Certainly, South Korea's latest policy adopts a social infrastructure perspective—recognizing that improving quality of life by providing facilities and services supporting people's daily lives will positively affect local production. Living SOC differs from conventional SOC (or social infrastructure) in that it tries to provide equal access to essential living services [36]. In this respect, South Korea's current Living SOC policy seeks to minimize the physical distance between residential zones and the daily living services, ultimately making the routes of urban residents more compact.

#### *2.2. Living SOC as a Community-Supportive Infrastructure*

Since the late twentieth century, policies similar to South Korea's Living SOC policy have been established and implemented in several countries in Europe and East Asia [6–9]. To understand similar policies, first, it is necessary to understand the concept of Smart Shrinkage and Compact City, which is one way to achieve an inclusive city [37]. Smart Shrinkage and Compact City is an urban regeneration method that focuses on improving the quality of life of existing urban residents while reducing population and building land use [38]. This concept differs from existing urban regeneration in that it improves the quality of life rather than inducing population inflow and employment growth [39]. Smart Shrinkage and Compact City can be a strategy to prevent the vicious cycle of decline by reorganizing the urban infrastructure to fit a new level of population, such as returning the abandoned neighborhoods of the city to nature, increasing the walking space, and fitting housing prices [40]. Poppers defined this as less planning, less population, fewer buildings, and less land use, and argued that small could be beautiful [41].

To address the population decline and a worsening local economy, Japan implemented the concept of a Compact City in 2014, placing residential zones in proximity to public transportation and necessary service facilities [42]. To be specific, in July 2014, the Ministry of Land, Infrastructure, and Transport and Tourism (MLIT) unveiled the "Grand Design of National Spatial Development towards 2050", with Aggressive Smart Shrinkage as an alternative to population decline [43]. The concept of Aggressive Smart Shrinkage is not a defensive response that prevents the city from shrinking and disappearing if the population decline is inevitable; it is a reduction plan to proactively continue urban function even if the population decreases by predicting the reduction mode. "Grand Design of National Spatial Development towards 2050" proposes a connection between Compact and Network to maintain urban functions even in the face of a declining population [44]. It is a strategy to prevent urban functions' departure by spatially integrating urbanized areas and resolving insufficient urban functions through the interaction of surrounding areas by reinforcing public transportation. Expressly, in a city with a population of 100,000 or more, a 1-km grid range of reach within an hour is set as an urban area, and a high-level regional urban association is established so that the urban area can sustain a population of 300,000 or more. An institutional response that applies the concept of a Compact City is a plan to appropriate its location. It will seek a network that allows access to medical, welfare, and business facilities through public transportation in areas where population reduction is expected lead to a failure in meeting the minimum residential standards. More specific measures include inducing urban functional facilities, such as medical care and welfare in the hub area, inducing residences in areas with public transportation connection, overhauling the walking and vehicle environments in the center, and introducing community buses. Besides, urban function inducement zones were established to enable urban function services to ease regulations and provide subsidies to induce urban function concentration rather than coercion. To ensure the smooth execution of the policy, the Japanese government implemented a "city function initiation zone" initiative by designating residential zones and collecting feedback from citizens to maintain an optimal population density [45]. The city function initiation zone promotes healthcare, business, education, and basic service facilities, thereby providing optimal services for urban residents. Regarding administration, the Japanese government monitors current convenience facilities and provides support in policy and finance for pertinent facilities when there is a shortage of individual convenience facilities.

In a similar context, Germany is implementing a policy to ensure equal living conditions based on the constitution, which guarantees "living conditions with equal value", and the 1965 Federal Space Planning Act [46]. As the industrial structure changed in Germany, manufacturing declined, and suburbanization increased, resulting in a decline of cities. In particular, in the former East German region after reunification, urban shrinkage became more severe as people who lost their jobs moved to the former West Germany or surrounding large cities [47]. Cities such as Dresden, Leipzig, and Cottbus are typical examples. House remain unoccupied in both the old and newly redeveloped areas of some cities. Therefore, the German government has adopted and implemented a strategic plan for Smart Shrinkage at the local level [48]. INSEK, an integrated strategic plan to respond to the smart-reduction problem, is the basis for allocating all subsidies to promote urban regeneration. In Germany, the government has stipulated that subsidies should be made only after establishing INSEK after 2002. INSEK is an integrated plan for urban development, established by each local government, to review the development priorities to designate areas subject to maintenance and areas of focus. The overall direction of urban development is thus set, and specific plans are flexibly adjusted according to the circumstances. More specifically, the law mandates that all 38 provinces identify a hub city with a population of over 100,000 people and equally distribute various living facilities for each hub city so that each region can enjoy similar living standards [46]. The guidelines are divided

into the social infrastructure category, which involves service facilities such as daycares and hospitals, and the technical infrastructure category, including water supply and treatment facilities. The German government has also focused on creating a universal living environment by monitoring changes in regional characteristics [49].

Numerous scholars have researched balanced urban development with the land as the spatial background. For example, Peters et al. [50] examined communities' social infrastructure to determine the degree of smart shrinkage in small towns based on population, land use, and transportation. Similarly, Chang and Liao [51] identified strategies for improving accessibility to urban parks and balanced distribution while highlighting public facilities' spatial equity. Examining the size and shape of a city and the attributes of urban planning, Hodge and Gatrell [52] highlighted the significance of determining the service area and argued that the attributes of urban planning could be a constraint on the related activities of the political, economic, and social systems. Regarding well-balanced development—the ultimate goal of South Korea's Living SOC policy—the extant research demonstrates the significance of the following: the establishment of relevant strategies prioritizing the investigation of urban land use status [53], an appropriate supply of facilities and commercial zones [54], the equal supply of education facilities [55], and hard and soft infrastructure [56]. Moreover, as the supply should meet the demand in terms of the accessibility and availability of these facilities [57], identifying the demand and the current condition of relevant facilities can positively affect the optimal supply of facilities when implementing relevant policy initiatives.

#### **3. Methodology**

#### *3.1. Research Model*

Like other countries, South Korea defines the Living SOC's scope as facilities assisting citizens in their daily lives, including those related to education, healthcare, welfare, transportation, culture, sports, and parks. South Korea also emphasizes the accessibility of these facilities. In addition to promoting the development of different regions, South Korea seeks to develop land equitably and improve citizens' lives. Therefore, this study examines the entire territory of South Korea to derive implications for the Living SOC policy. Data collection, analysis, and interpretation were conducted in two stages, as follows.

First, this study compares metropolitan and provincial cities in terms of population, the average age of urban residents, urbanization ratio, and the area size of each zone according to South Korea's land-use planning (i.e., residential, commercial, industrial, and green zones, respectively), and the total number of major facilities in the category of the Living SOC policy presented by the government. In doing so, this study examines the type of imbalance between metropolitan and provincial cities using independent sample *t*-tests. Various studies have verified and highlighted the validity of the variables mentioned above [58] and the use of independent *t*-test to compare regional differences [59,60].

Second, this study examines how the amount of basic service, convenience, and cultural and sports facilities in each city influences the number of Living SOC. There is already a wide variety of preexisting basic service, convenience, and cultural and sports facilities throughout the country, many of which are included in the Living SOC policy. Accordingly, if the research model is statistically significant, the type of facilities closely related to the current Living SOC policy is already sufficiently distributed from the policy's perspective when the number of the Living SOC is set as the dependent variable and other facilities as independent variables. Therefore, they are relatively unimportant. By analyzing each city throughout the country, this study derives implications for improving the implementation and efficacy of future policy initiatives. To identify the determined balanced development, we divided the cities into metropolitan cities and provincial cities, conducted two rounds of regression analysis, and compared the results. Various studies have highlighted the importance of conducting regression analysis in analyzing the current condition of facilities in a specific area [61–63]. Figure 1 illustrates this study's research process and methodology.

**Figure 1.** Research process and methodology.

## *3.2. Study Areas and Variables*

As noted, the spatial scope of this study is the entire territory of South Korea. All data used in this study are based on the 2020 administrative division of South Korea. A total of 229 regions were used as samples, including 69 autonomous districts, 75 autonomous cities, 82 counties, one special self-governing city, and eight provinces (Figure 2). These samples are the minimum unit of all data used for statistics and include the entire are of South Korea.

**Figure 2.** Classification of research areas and population density (from Statistics Korea).

For the two rounds of regression analysis, we categorized cities into metropolitan cities and provincial cities based on the relevant statutes, resulting in 74 self-governing cities (or self-governing areas) and 155 provincial cities. In this categorization process, we collected data to compare the two categories of cities in terms of population, average age, number of residents per city, and the ratio of residential/commercial/industrial/green zones. This study used open-access data from the Korea Statistical Information Service (KOSIS).

This study used the following two-step method to obtain and classify convenience facilities. First, according to the government's proposal, Living SOC facilities are intended to enhance the convenience of people's lives and encompass culture, sports, education, healthcare, welfare, and park facilities [13]. Specific facilities include community sports centers; outdoor sports facilities, such as baseball parks, soccer fields, gate ball courts, and artificial rock-climbing walls; and cultural and educational facilities, such as libraries, museums, art galleries, parking lots, daycare centers, kindergartens, elementary schools, welfare facilities for elderly, hospitals, highway rest areas for safe traffic, fire or disaster safety facilities, forests, recreation forests, campsites, and urban parks. Among the listed facilities, the supply of Living SOC facilities is provided by the public sector, and statistics are officially totaled by the central government. There are seven types of Living SOC facilities in total: elementary schools, job training schools, libraries, culture centers, post offices, police stations, and fire stations. As the supply of Living SOC facilities is provided by the central and provincial government, other private facilities were not included as variables. However, there is a possibility that, at the local level, critical private facilities were excluded from this process.

Second, the preexisting living infrastructure used as an independent variable in the regression analysis includes public and private facilities proposed in the pertinent policy and statues. Among over 30 facilities, we selected nine facility types with available open-access data from KOSIS as follows: elementary schools, job training schools, libraries, culture and art facilities such as museums, art galleries, and culture centers; sports facilities such as baseball parks, basketball courts, soccer fields, and gyms; and local community facilities such as recreation forests, campsites, fields, and urban parks. There is a possibility that major facilities may have been excluded during the process of limiting the variables to those for which national data exist. This study recategorized the nine selected facility types into facilities with similar functions to ensure commonality between the variables, resulting in the following four categories of facilities: (1) education and empowerment facilities, (2) culture and art facilities, (3) sports and leisure facilities, and (4) local community facilities—Table 1 details the content of each category. The three-year Living SOC plan of the South Korean Government was referred to for the classification of categories.


## **4. Results**

#### *4.1. Comparison between Metropolitan and Provincial Cities*

South Korea's Living SOC policy aims to supply facilities equitably among metropolitan cities and provincial cities. Using the aforementioned methodology, this study quantitatively compares metropolitan and provincial cities' conditions by conducting independent sample *t*-tests. Through *t*-tests, this study examines the difference between the two types of cities to identify the significant differences in the major variables reflecting regional attributes, such as population, average age of residents, urbanization ratio, the level of Living SOC, and the area ratio of residential, commercial, industrial, and green zones in South Korean urban planning-Table 2 details the results of the analysis.


**Table 2.** Comparison of the key variables by region.

\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001. *t* = *t*-value; *p* = *p*-value.

The results of the analysis show that the two groups differ significantly in terms of population, average age, urbanization ratio, and the ratio of residential, commercial, green, and Living SOC. With regard to population, metropolitan cities were found to have a larger population than provincial cities (*t* = 4.239). With regard to age, the average age of the residents in provincial cities was found to be higher than that in metropolitan cities (*t* = −7.305). With regard to urbanization ratio, metropolitan cities were found to have a higher urbanization ratio than provincial cities *(t* = 19.315). With regard to the ratio of residential zones, metropolitan cities were found to have a higher ratio of residential zones than provincial cities (t = 7.421). With regard to the ratio of commercial zones, metropolitan cities were found to have a higher ratio of commercial zones than provincial cities (*t* = 3.329). With regard to the ratio of green areas, provincial cities were found to have a higher ratio of green zones than metropolitan cities (*t* = −8.509). With regard to the Living SOC, metropolitan cities were found to have a higher level of Living SOC than provincial cities (*t* = 2.078). However, with regard to the ratio of industrial zones, there was no significant difference between the metropolitan and provincial cities.

The findings can be summarized as follows. Compared to metropolitan cities, provincial cities had a smaller population, a lower urbanization rate, and a lower ratio of residential and commercial zones. Meanwhile, the average age of the residents and the ratio of green zones in provincial cities were greater than those in metropolitan cities—a common difference between metropolitan and provincial cities. Moreover, with respect to Living SOC facilities, metropolitan cities were found to have a higher number of Living SOC facilities. However, the maximum capacity per facility in metropolitan cities was 1560, whereas the maximum capacity per facility in provincial cities was 1083. This shows that public-initiated Living SOC facilities are successfully supplied in provincial cities throughout the country.

## *4.2. Results of the Regression Analysis of the Living SOC Perspective*

This study conducted a regression analysis to test the independent variables' effect on the current Living SOC of South Korea's metropolitan cities. The independent variables included the supply of education, culture and arts, sports and leisure, and local community facilities, classified based on population, average age, urban planning attributes, the urbanization ratio, and preexisting convenience facilities. Between rounds of analysis, this study employed stepwise regression analysis to derive the results. This method has the advantage of showing an increase in the explanatory power in accordance with the inclusion of each independent variable group by stage. This study examined the effect of the urban planning perspective on Living SOC as well as its implications by applying population, age, and land-use status to Model I. In contrast, in Model II, this study applied all the variables to examine the related facilities' overall effect and compared the results for metropolitan and provincial cities.

Table 3 presents the results of the regression analysis in metropolitan cities. The analysis of the results showed that the regression model was statistically significant in Stage 1 (F = 13.423, *p* < 0.001) and Stage 2 (F = 46.709, *p* < 0.001). Based on the adjusted R2, the explanatory power was 57.7% in Stage 1 and increased to 88.3% in Model II, indicating a high explanatory power. The Durbin–Watson statistic was 1.731, producing an approximate value of 2. This indicates that the residuals can be presumed to be independent. The variance inflation factor (VIF) was also found to be below 10, indicating no problems with the correlation of variables. Accordingly, the majority of current facilities in metropolitan cities belong to the category covered by the Living SOC policy.

In Model I, the population was found to have a significant effect on the dependent variable (*t* = 7.348, *p* = 0.000). This result indicates that cities with large populations are the main recipients of Living SOC. In contrast, in Model II, which includes all the variables, the size of the population was not significant, while the average age of the residents was found to influence the dependent variable. In both Models I and II, land use was found to be insignificant, indicating a need for revision of the current government policy of creating a dense assortment of Living SOC facilities around the residential zones. Of the four independent variables (education, culture and arts, sports and leisure, and local community facilities), education (*t* = 7.327, *p* = 0.000) and local community facilities (*t* = 9.870, *p* = 0.000) were found to affect the dependent variable. In other words, in metropolitan cities with dense populations, the supply of Living SOC increases in areas with a sufficient supply of education and local community facilities.

Using the same method, this study performed a stepwise regression analysis on provincial cities, the results of which are presented in Table 4. The results show that the regression model was statistically significant in both Stage 1 (F = 13.423, *p* < 0.001) and Stage 2 (F = 46.709, *p* < 0.001). Based on the adjusted R2, the explanatory power was 20.9% in Model I and increased to 72.5% in Model II, indicating a relatively high explanatory power. The Durbin–Watson statistic was 2.120, showing no problem with presuming the independence of the residuals. The VIF was also found to be below 10, indicating no issue with the correlation of variables. As such, similar to metropolitan cities, the majority of the current facilities in provincial cities also influence the Living SOC.



198

#### *Sustainability* **2020** , *12*, 227


*Sustainability* **2020** , *12*, 227

Testing the significance of the regression coefficients, this study found that population and the non-urbanization ratio were significant and positive (+) in Model I. Specifically, population (*t* = 4.656, *p* = 0.000) and the non-urbanization ratio (*t* = 2.487, *p* = 0.000) affected the supply of Living SOC. This indicates that, in provincial cities, Living SOC is generally established in non-urban areas rather than residential zones. In contrast, the aforementioned independent variables were not statistically significant in Model II. Of the four independent variables (education, culture and arts, sports and leisure, and local community facilities), all except for culture and art facilities were found to affect the dependent variable in Model II. In this respect, the supply of Living SOC was found to be sufficient in terms of education facilities (*t* = 2.828, *p* = 0.005), sports and leisure facilities (*t* = 6.307, *p* = 0.000), and local community facilities (*t* = 11.067, *p* = 0.000).

### **5. Discussion**

The extant literature reflects the growing significance of the inclusive city concept—a global trend in the development of Living SOC policy, particularly with respect to improving the spatial equality and quality of life in communities. In this regard, this study aligns with extant urban planning theories, including notions of Compact City and Smart Shrinkage. Examining current urban planning in South Korea, this study found that various living infrastructures lack a connection to residential zones, irrespective of city type (i.e., metropolitan or provincial cities). In terms of facility type, this study found that the supply of facilities related to local communities and education is adequate. The supply of culture and art facilities is insufficient throughout the country, and the supply of sports and leisure facilities in provincial cities is relatively adequate. Based on these findings, this study identified the Living SOC's scope in terms of land planning to ensure the balanced development of metropolitan and provincial cities. This study also discerns the actual disparity between metropolitan and provincial cities and how this might be resolved. The results of this study can be summarized as follows.

First, establishing necessary community support facilities that are easily accessible by foot—as advanced by notions of Compact City and Living SOC—is a significant government policy in many countries, including South Korea. In Germany, the federal government operated a support program to solve urban shrinkage and established a plan (INSEK) to cope with various problems caused by urban reduction, such as vacancies in residential areas. In Japan's case, it has recognized the problems of public infrastructure operations due to aging and the occurrence of vacant homes. However, since many large residential areas had already been built in suburban areas with a population density below the stipulated number, they chose centralization as the solution. In this respect, particular emphasis is placed on the need to establish such facilities in residential zones. This study shows that the independent variables related to current land use—namely, the supply of education, culture and arts, sports and leisure, and local community facilities—did not significantly influence the dependent variable, Living SOC. Accordingly, the future policy needs to provide Living SOC centering on the residential zone(s) within a city.

Second, an analysis of South Korea's metropolitan and provincial cities reveals that population size significantly impacts Living SOC. In other words, the population was the most significant sociodemographic factor impacting the supply of preexisting education, culture, sports, and community facilities. This finding is consistent with the country's current Living SOC policy.

Third, it is crucial to increase the supply of facilities related to culture and the arts and sports and leisure facilities. Regression analysis results show that these two types of facilities did not significantly impact the dependent variable in provincial cities or metropolitan cities with dense populations. This indicates the need to ensure the sustainability of the Living SOC policy in South Korea by supplying related facilities in the future. Given the non-significant impact of such facilities in provincial cities, establishing culture and art facilities may be critical to the successful realization of the Living SOC policy in terms of providing equitable supply and access to facilities in provincial and metropolitan cities.

Fourth, it is essential to perform a business feasibility review regarding the sustainability of the pertinent facilities, particularly as South Korea's Living SOC policy is heavily funded by the central and local government and reliant on the financial soundness of local governments. For instance, despite the result of the analysis showing the lack of sports facilities in metropolitan cities, policyholders should have a lengthy deliberation when establishing large-scale sports facilities in a metropolitan city. Because most local governments of South Korea have insufficient financial independence, consideration of efficiency should come first.

Finally, the supply of Living SOC is based on the current condition of provincial cities. South Korea's land-use system consists of urban areas (16.6% or 17,614 km2), comprising residential, commercial, industrial, and green zones; and non-urban areas (83.4% or 88,448 km2), comprising management (areas requiring systematic management), farming, and natural environment protection zones [64]. It is worth noting that the non-urban areas of provincial cities are some five to nine times larger than those of metropolitan cities, and that such non-urban areas house many residential buildings. Therefore, future policy should direct local government to establish facilities based on accessible "distance" by considering the land use and the location and density of the residential buildings. However, achieving this requires conducting micro-level research of each city or region rather than the entire country.

## **6. Conclusions**

Establishing and executing the Living SOC policy in terms of land and urban planning is key to improving South Korea and its cities' sustainability. Policies related to cities serve as the guidelines in establishing ordinances and the duties of the local government. Therefore, to ensure the policy's sustainability, its development and implementation must reflect the local community's attributes from the perspective of balanced land development. This study is significant in that it examines the entire territory of South Korea, but further and more localized research is necessary.

Using currently available data, this study focused on current urban planning conditions and the impact on Living SOC policy implementation—a hitherto unexamined topic.

However, this study has limitations that need to be considered. First, the analysis of facilities was hindered by the difficulty of obtaining data due to data source limitations. This study focused on quantitative aspects, such as the number of infrastructures. Thus, this study excluded some qualitative aspects regarding quality of life. Further research based on a survey, which includes qualitative aspects, is necessary. Second, this study failed to consider the disparity among cities regarding urban planning, population, and social and economic status. In the same context, this study failed to consider accessibility, which is a crucial Living SOC concept. Accordingly, future studies should examine the Living SOC of each region, considering each facilities' accessibility.

Nonetheless, it is expected that the findings on South Korean policy, including those of this study, will prove useful to the global community and provide baseline data for follow-up studies in South Korea. It is expected to be a barometer to prevent budget waste through the reckless introduction of Living SOC facilities in the future. In particular, this study used quantitative research models to ensure objectivity in assessing policies, which would help the Korean government when they supply Living SOC. Indeed, it conforms to the "benefit responsiveness" section of the Nakamura and Smallwood polities, widely used in policy evaluation. Therefore, it can be used to determine how much Living SOC has been beneficial in improving the residents' quality of life in the area.

**Author Contributions:** Conceptualization, Y.K. and J.O.; methodology, Y.K.; software, Y.K.; formal analysis, Y.K.; data curation, Y.K.; writing—original draft preparation, Y.K.; writing—review and editing, J.O.; visualization, J.O.; supervision, S.K.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**

1. UN-Habitat. *The Global Campaign on Urban Governance*; UN-Habitat: Nairobi, Kenya, 2002.


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Energy and CO2 Reduction of Aluminum Powder Molds for Producing Free-Form Concrete Panels**

**Donghoon Lee <sup>1</sup> and Sunkuk Kim 2,\***


Received: 27 October 2020; Accepted: 14 November 2020; Published: 18 November 2020

**Abstract:** Free-form design may enhance the architectural value of buildings in terms of aesthetic and symbolic effects. However, it is difficult to reuse the mold of free-form concrete segments, so they are manufactured for single use. Manufacturing these molds is a time-consuming process that requires a lot of manpower. To solve these problems, there have been numerous studies on the use of phase change materials (PCMs) to make the molds. PCM molds represent a new technique of producing free-form panels using a computerized numeric control (CNC) machine that employs low-cost material to produce free-form concrete panels. However, PCM molds require a substantial amount of time and energy during fabrication because repeated heating and cooling cycles are required during panel production, and this process increases the CO2 emissions. Thus, the purposes of this study were to develop composite molds using aluminum powder to improve PCM mold performance and to conduct experiments to quantify the reduction of energy use and CO2 emissions. As a result of cooling experiments, it was found that the aluminum powder mold had an energy reduction effect of 14.3% against the PCM mold that had been produced only with paraffin wax, and CO2 reduction effect of more than 50% against the conventional mold.

**Keywords:** free-form building; free-form concrete panel; aluminum powder; composite PCM mold; CO2 emission reduction

## **1. Introduction**

Free-form design is being increasingly adopted in monumental buildings to improve aesthetic and symbolic effects. However, the molds used for the production of free-form panels are 3–10 times more expensive than conventional molds [1]. This is because free-form concrete segments are not produced in fixed shapes, which make it difficult to produce a mold using conventional materials like metal, wood, and synthetic resins; it is also impossible to reuse the produced mold [2]. Presently, molds for the production of free-form concrete segments (FCS) are only used one time. Due to this, the construction of free-form buildings requires longer construction times, is significantly more expensive, and also produces more CO2 emissions compared to conventional construction. However, there have been no studies on the development of practical production technologies for free-form concrete segments to realize the environmentally-friendly construction of free-form buildings [3].

Since free-forms are composed of various irregular curved surfaces, they require more construction time and resources than ordinary molds. There has been a wide range of technologies applied to the production of free-form concrete segments [4–6], but the production of free-form concrete segments is not fit for sale [2,6].

Recently developed phase change material (PCM) molds can be semi-permanently reused, and it is easy and quick to produce free-form concrete segments (FCSs) with this new technique [7]. Here, PCM refers to a material that changes its phase from liquid to solid and vice-versa, depending on the temperature. PCM molds are used in the solid-state for FCS production, and it changes the mold back to a liquid state for reuse. However, heating and cooling must be repeated every time concrete segments are produced, so there is great concern about the environmental impact, including energy consumption and CO2 emissions arising from energy consumption. It is absolutely necessary to evaluate energy consumption as well as CO2 emissions when developing new molding materials [8]. Thus, the purposes of this study were to develop aluminum powder molds to improve PCM mold performance and to verify the reduction in energy consumption and CO2 emissions achieved using these molds. This study is limited to frequently used conventional PCM molds, plywood (wooden) molds, and the newly developed aluminum powder molds. We also analyzed the effect of these techniques on energy consumption and CO2 emissions. This study was conducted in five steps, as described below:


## **2. Consideration of Previous Studies**

Steel, wood, expandable polystyrene (EPS), and plastic were used to produce free-form segments in many of the free-form buildings that have recently been completed [6,9–12]. Latorre [13] developed a "pneumatic system" to produce free-form domes, yet each material had to be individually produced, and it was difficult to produce shapes other than domes. Verhaegh [6] developed a "fabric formwork" using fabric forms, however many of the molds were used for shape constraints, and a great deal of manufacturing time was needed. This method failed to improve construction duration and cost when compared to conventional methods. Mandl et al. [14] and Lindsey and Gehry [15] conducted studies on the formwork made with EPS using computerized numeric control (CNC) technology and Toyo Ito and Associates [16] developed a wooden system form. Franken and ABB [17] used digital forms with CNC and acrylic glass to produce free-form concrete. However, these molds cannot be reused, generate a lot of waste, and require an extremely long manufacturing time. Researchers at the University of Southern California hoped to come up with a building process using a robot automation system, but additional studies are required due to limitations in the production of large segments. To address the problems with free-form molds, CRAFT (Center for Rapid Automated Fabrication Technologies) [18] are conducting a study on the 3D printing of buildings. However, the concrete deposited using a 3D printer nozzle needs time for curing, and it is not yet ready for commercialization because the size of the 3D printer is limited.

The need for developing economical, variable mold, and free-form concrete production technologies is on the rise and has focused on the reuse of molds and the reduction of production time [3]. Recognizing this need, Oesterle et al. [19] developed a reusable wax mold. PCM changes from solid to liquid and vice-versa depending on the type of external stimulation (temperature, electric current, etc.). Therefore, PCM molds can be used to make shapes and can be reused to reduce waste generation. Paraffin, one of the PCMs used in this study, is an alkane hydrocarbon with a chemical formula expressed as CnH2n<sup>+</sup><sup>2</sup> (n ≥ 19). It does not dissolve in water, but it does dissolve in ether, benzene, or ester. Paraffin is a mixture of hydrocarbon molecules comprised of 20 or 40 carbon molecules. It is a soft, white solid without color that is derived from petroleum, coal, or oil shale. The melting point of paraffin ranges from 47 ◦C to 64 ◦C, and the density is 0.9 g/cm<sup>3</sup> [20–22].

However, Oesterle et al. failed to analyze and discover solutions to address the increased energy consumption, increased CO2 emission arising from the energy consumption, long freezing times, crystallization effects, strength issues, solidification shrinkage, and cracking. The study focused only on the realization of wax mold shapes and did not comprehensively explain the device or technology. In particular, the study had some limits in that it lacked specific verification of the shrinkage in wax molds, freezing time, and energy consumption.

Lee et al. [4] developed PCM molds using a CNC machine and suggested a production process of FCP (Free-form Concrete Panel). The production process was composed of a total of nine stages, as illustrated in Figure 1. In the stage wherein PCM molds were produced, liquid PCM was first filled into a CNC machine to produce a shape. After shaping, the hot liquid PCM was cooled to solidify the PCM. The solid PCM was then separated from the CNC machine and was attached to the frame on the side to produce a PCM mold. Production of the FCP involved pouring concrete into the PCM mold, followed by curing to produce FCPs [7]. However, the study failed to suggest solutions to various problems, including the extensive cooling energy of the PCM mold and the effects of crystallization and cooling time.

**Figure 1.** FCP (Free-form Concrete Panel) production process with phase change material (PCM) mold (Lee et al., 2015).

## **3. Aluminum Powder Mold**

As examined in Section 2, there have been great efforts to develop free-form concrete panel molds. In addition, there have been some studies on the development of PCM molds that can be used to make free shapes. However, there are no case studies on the production of concrete panels for free-form buildings using novel PCM molds. This is because newly developed molds may be less practical or have some problems. A wide range of shapes can be produced with PCM molds, yet two problems still exist. First, PCM cooling takes a long time and requires a lot of energy. Second, the low strength of the PCM may result in crushing or sinking of the mold when concrete is poured into it. A light aluminum with outstanding thermal conduction was chosen as a material to improve the performance of the PCM molds in this study. The aluminum powder mold (c) shown in Figure 2 is a free-form mold mixed with aluminum powder (a) and paraffin wax (b); this mixture is produced using a CNC machine.

Aluminum, the main ingredient of the mold, is the most common chemical element found in the earth after oxygen and silicon. The heat of fusion of aluminum is 94.6 cal/g, and it has a melting point of 660 ◦C. As listed in Table 1, aluminum has a specific gravity of 2.71, and it is fairly light, only 1/3 of the weight of copper or iron. In addition, its thermal conductivity is 204 W/m·h·K, which is more than three times higher than that of iron (67 W/m·h·K) [23,24]. In this study, we developed free-form molds by mixing the aluminum powder with paraffin wax. The aluminum powder mold could be used to produce any shape in the heated, liquid state using a CNC machine. Then, it was cooled and used as a robust mold in the solid-state. The thermal conductivity of the aluminum powder mold was high, which solved previous problems related to the high consumption of cooling energy and long cooling times in other PCM molds. When the aluminum/paraffin wax PCM mold was cooled, the surface was first cooled and then frozen, letting down the energy efficiency. However, the thermal conductivity of the aluminum powder mold could solve the problem.

**Figure 2.** Ingredients of the aluminum powder mold.


**Table 1.** Specific heat and thermal conductivity of main metal and paraffin wax [25].

Powder molds have been used for casting metal molds for a long time. They are reusable and are able to produce a wide range of shapes. However, the shape can be easily crushed or deformed. When paraffin wax is added to these powder molds, there is no change to the powder, which is densely packed after freezing. The molds are movable, and they are highly resistant to compression. As shown in Table 2 these types of molds can be reused semi-permanently, yet melting energy and cooling energy must be added during each reuse. Aluminum powder with high thermal conductivity is a suitable material for solving this problem.


**Table 2.** Characteristics of the mold materials.

## **4. Analysis of Energy Consumption and CO2 Emission**

### *4.1. Analysis of Energy Consumption*

Since aluminum powder molds and PCM molds are plate-type molds, they release heat mostly through the upper and lower parts. We used the experimental apparatus shown in Figure 3 to study the heat release of molds for the analysis of energy consumption. A movable heating device was used for heating and cooling of the materials, and a cooler along with other auxiliary devices, were used to maintain a constant cooling temperature. Furthermore, the surface upper and lower parts were exposed so that the heat was released from those parts, and the sides were insulated. The volume of the mixture used in the experiments was 600 mL, and the mixture was sufficiently heated to 90 ◦C or above and then air-cooled to 20 ◦C. Aluminum powder (3 μm) was mixed with commercial-grade paraffin wax. After melting the paraffin wax and mixing it with the aluminum powder, we found that it became saturated at a volume ratio of approximately 30:17. At this ratio, the ingredients were mixed so that the paraffin wax filled in the small spaces between the aluminum particles.

**Figure 3.** Schematic diagram of the experimental apparatus.

There was a significant difference in temperature profile during the cooling of paraffin wax molds and aluminum powder molds, as shown in Figure 4. It was instructive to compare the cooling times from a heated state of 60 ◦C to the temperature where the paraffin wax congealed (40 ◦C). This cooling step took 50 min for the Al powder-based material and took 255 min for the paraffin wax, a more than five-fold difference. This result implied that a CNC machine that produces shapes of melted molds could make five times more molds in the same amount of time.

**Figure 4.** Temperature variation of the molds.

The temperature variation in the liquid state was determined by the removal of sensible heat. When the upper/lower cooling surfaces were exposed to air that was below the phase change temperature, the liquid started to solidify. At this point, natural convection began to be restricted. There was no difference in the output of sensible heat through the thin solid surface layers between the two mixtures in the early stages of cooling. However, as shown in Table 3 and Figure 5, as the solid layers thicken, the transfer of sensible heat inside the PCM molds dropped significantly as the liquid near the cooling surface solidified. The PCM mold must be completely frozen to be used as a mold, so long cooling times are required to freeze all of the paraffin in a solid-state.


**Table 3.** Temperature variation and heat output of the PCM mold.

**Figure 5.** Reduced heat output owing to the solid layer.

This phenomenon did not occur in the case of the aluminum powder molds because the inner thermal energy was quickly transferred. Figure 6 shows the temperature variation inside the surface of the molds. The freezing temperature of both materials was about 40 ◦C. The difference between the time required for the inner surface to freeze was about 10 min for the aluminum powder mold and more than 110 min for the paraffin wax mold. As shown in Figure 6a,b, the temperature difference between the surface and inner part of the PCM mold after 160 min of cooling was 15 ◦C or above. After 260 min, the temperature difference in the PCM mold was 10 ◦C or above. In contrast, the aluminum powder mold showed a temperature difference between the surface and inner part of around 10 ◦C in the earlier stage, as shown in Figure 6c, but the temperature difference drastically declined after 40 min of cooling, as shown in Figure 6d. This was because the inner heat was immediately transferred to the air since the heat transfer efficiency of the material itself was high.

**Figure 6.** Temperature variation on the inside and on the surface of the molds.

In order to freeze and cool 1 L of hot paraffin wax from 80 ◦C to 40 ◦C, 56.7 kcal must be removed from the system, as shown in Table 4. This estimation was reached by adding the 25.2 kcal required to cool the wax and lower the temperature and the 31.5 kcal required for the phase change (freezing). A total of 49.125 kcal was needed to cool 1 L of the aluminum powder mold. The specific heat of this material was 0.215 kcal/kg, which was significantly lower than that of the paraffin wax (0.7 kcal/kg), and there was no phase change between 80 ◦C and 40 ◦C, which implied that there was no need to account for latent heat. Although the specific gravity of aluminum powder mold was bigger, it had a 14.3% lower required cooling energy.


**Table 4.** Heat consumption of molds (based on 1 L).

The PCM molds required around 150 L of paraffin wax to produce 1 m<sup>2</sup> of concrete panels; 8505 kcal of heat was required to melt this amount of paraffin at 80 ◦C. As shown in Table 5 one kWh of electric power could generate 2150 kcal of heat, so 3.96 kWh was needed to heat 1 m2 of the PCM mold. Assuming that the efficiency of the cooler was 67% and its heat loss is 50%, this process required 11.88 kWh of electric power to cool 150 L of the PCM mold at 40 ◦C. Using the same method, the energies required for heating and cooling the aluminum powder mold (1 m2) were 3.43 kWh and 13.71 kWh, respectively. We, therefore, concluded that there was an energy reduction effect of around 2.1 kWh as compared with the PCM mold.


**Table 5.** Power consumption of PCM mold and aluminum powder mold.

## *4.2. Analysis of CO2 Emissions*

The molds that are used in conventional free-form building projects have different curvatures, sizes, and shapes, and it is therefore not possible to reuse them. In addition, they are frequently twisted and damaged during the separation of the mold. The increased use of temporary materials leads to increased CO2 emissions and costs. To compare the energy consumption of conventional methods where numerous temporary wooden boards were used with the use of PCM and aluminum powder molds, we evaluated CO2 emissions from the manufacture of the materials to the completion of the molds. As shown in Table 6, we found that there was a great difference in CO2 emissions between the aluminum powder mold that can be reused semi-permanently and with conventional plywood.



The energy consumption and CO2 emissions required to produce each material can be estimated using interindustry analysis. Interindustry analysis uses an interindustry transaction matrix to construct an input coefficient matrix that can be used to estimate the total energy input for manufacturing the materials. This study applied this method to calculate all CO2 generated associated with panel production. Lee et al. [26] used interindustry analysis to analyze the energy needed for plywood manufacturing, and they found that 3.37 kg-CO2/kg of plywood was emitted. Considering the unit weight of plywood, 21.2 kg-CO2 is emitted per m2. Here, a small amount of electric power was used for assembling the mold, but this was not included for calculation since it was extremely small.

Since PCM molds can be reused semi-permanently, CO2 emissions were estimated based on 150 liters of paraffin wax (the amount needed to produce 1 m<sup>2</sup> of concrete panels). In addition, heating to 80 ◦C and cooling to 40 ◦C were repeated with 150 L of paraffin wax whenever 1 m2 of concrete panels were produced. The total electric energy input during this process was analyzed. In the case of PCM molds, 1584 kWh of electric power in total was used. In the case of the aluminum powder molds, 1371 kWh of electric power was used. A coefficient of 0.424 kg-CO2/kWh [27,28] was used to estimate the total CO2 emissions, which were 1069.6 kg-CO2 and 1065.83 kg-CO2, respectively, for PCM molds and aluminum powder molds. Therefore, compared with CO2 emission of conventional wooden molds (2226 kg-CO2), PCM and aluminum powder molds could be used to reduce the emission to less than 50%. Taking into consideration the waste generated after removal of a single-use wooden

mold, there should be a greater reduction of CO2 emissions from aluminum powder molds that do not generate any waste.

## **5. Conclusions**

In this study, we developed an aluminum powder mold to improve energy consumption and reduce the time required for heating and cooling PCM molds. Our mold utilized the effect of phase change, so it was a type of PCM mold. Since heating and cooling should be repeated for the production of concrete panels using PCM molds, there was a negative impact on the environment, including increased energy consumption and the corresponding higher CO2 emissions.

Our aluminum powder molds had a high thermal conductivity, which allowed them to reduce the required cooling energy and reduced the cooling time. The surface of the aluminum powder mold was first cooled and then congealed to prevent heat blockage, which reduced the consumption of cooling energy. A mold made only of paraffin wax required 1584 kWh of electric power in total, but the aluminum mold required 1371 kWh of electric power, reducing power consumption by 14.3%. In addition, CO2 emissions that result from using the PCM mold and the aluminum powder mold were 1069.6 kg-CO2 and 1065.83 kg-CO2, respectively. These represent a reduction of 50% compared to the use of conventional wooden molds (2226 kg-CO2). Considering the waste generated after the removal of single-use wooden molds, we believe that there was an even greater reduction in CO2 emissions from aluminum powder molds that do not generate any waste.

Aluminum powder molds could be used to easily produce any shapes, just like conventional PCM molds. The aluminum powder molds could also reduce energy consumption and CO2 emissions relative to conventional molds. Our aluminum powder-based molds could be used to improve the economic feasibility of the construction of free-form concrete buildings and reduce energy consumption and CO2 emissions, which would be helpful in realizing various building designs and creating value. Additional studies related to composite molds mixed with a wide range of materials are needed.

**Author Contributions:** Conceptualization, D.L. and S.K.; writing—original draft preparation, D.L.; writing—review and editing, S.K.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is supported by the Korea Agency for Infrastructure Technology Advancement (KAIA) grant, funded by the Ministry of Land, Infrastructure and Transport (Grant 20CTAP-C151959-02).

**Acknowledgments:** This work is supported by the Korea Agency for Infrastructure Technology Advancement (KAIA) grant, funded by the Ministry of Land, Infrastructure and Transport (Grant 20CTAP-C151959-02).

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Construction Management Solutions to Mitigate Elevator Noise and Vibration of High-Rise Residential Buildings**

## **Yangki Oh 1, Minwoo Kang 1, Kwangchae Lee <sup>2</sup> and Sunkuk Kim 2,\***


Received: 11 September 2020; Accepted: 19 October 2020; Published: 27 October 2020

**Abstract:** In high-rise residential buildings (HRBs), elevators run at a high speed, which causes problems such as change of atmospheric pressure, noise, and vibration. Elevator noise and vibration (ENV) of HRBs causes both mental anxiety and a consistently negative effect for promoting a comfortable residential area. Therefore, a solution for alleviating the ENV of HRBs is essential. To date, studies related to ENV have been mostly conducted in the approach of mechanical and electric aspects. There have been few cases conducted from the perspective of construction management (CM), which integrates design and construction. Therefore, the aim of this study is to propose CM solutions to mitigate the ENV of HRB. For this study, the CM solution is presented after identifying the ENV problems of HRBs through documented research and case measurement. By measuring the noise of HRB that the solution was applied to, the noise level, especially in a range of >125 Hz, was extensively reduced. The result of this study will be used as sustainable guidelines that alleviate ENV problems in the process of design and construction of HRB elevators. It is expected that studies for improving ENV problems that occur in high-rise elevators will increase on the basis of the results of this study.

**Keywords:** elevator; noise; vibration; construction management; high-rise residential building

#### **1. Introduction**

A high-rise residential building (HRB) is a type of housing that has multi-dwelling units built on the same land. This housing has become popular in urban areas because of the increase in land cost [1]. Efficient vertical mobility is a critical component of developing and constructing tall buildings [2]. Advances in elevators over the past 20 years are probably the greatest advances we have seen in tall buildings [3].

However, the elevators of HRB operated at a high-speed cause problems such as changes in atmospheric pressure inside and outside a lift car, noise, and vibration [4–13]. In particular, elevator noise and vibration (ENV) cause both mental anxiety for passengers and a consistent negative effect on promoting a comfortable residential area close to the elevator shaft [11,14–21]. To secure a sustainable living environment, the impacts can be significant issues related to sound quality, sleeping conditions, and enjoyment within residences [14,16–20,22–25].

To solve these problems, multiple studies have been conducted, which mostly focused on mechanical and electric aspects [6,7,10,26–36]. However, it is not easy to solve ENV problems only with machinery solutions for HRB elevators that run at a speed of >90 m/min [21,37–39]. The reason is that not only do noise and vibrations occur in the elevating machinery itself, but there are architectural

problems of elevator shafts or air turbulence because HRB elevators run at a high speed [4,14,37–39]. Thus, identifying solutions after analyzing such causes is necessary. Solutions for ENV problems should be provided in detail in the design phase primarily, and construction should be followed according to the design details. However, we have confirmed that ENV problems continue to occur due to errors or mistakes in the design and construction phase through research and site surveys over the past several years.

This is because there is no CM solution that integrates design and construction for ENV problems. As shown in routines (1) and (2) in Figure 1, even if ENV solutions are provided with the documents including drawings and specifications in the design phase, the details including the precision, quality, and tolerance of construction suitable for the site condition must be determined in the construction phase after design review. As shown in routines (3), (4), and (5) in Figure 1, if design errors or omissions are confirmed after design review, construction details are determined after performing supplementary design. In other words, additional designs can be carried out according to the site situation at the construction phase. Solutions applied to the design and construction phases and integrated management of them are defined as CM solutions in this paper. The aim of this study is to propose CM solutions to mitigate the ENV of HRB.

**Figure 1.** Design and construction integrated construction management (CM) solutions for elevator noise and vibration (ENV) problems.

Figure 2 shows the methods and procedures of this study. First, review the definition of HRB and elevator and ENV sources and transmissions by surveying the design guidelines. Second, analyze the ENV sources and the transmissions of HRBs through documents and examine the site, and then measure ENV as a designated HRB that CM solutions have not been applied to as a case study. Third, propose design and construction solutions that have been confirmed using multiple documents that have been presented and studies that have been previously conducted. Moreover, confirm the effectiveness of CM solutions proposed in this study after measuring ENV of a case HRB that these solutions are applied with. Fourth, discuss the consistent improvement of problems that occur in high-rise elevators as per the results of this study and then describe results in the conclusion.

**Figure 2.** Research process and methodology. HRB = high-rise residential buildings.

## **2. ENV of HRBs**

#### *2.1. Review of HRBs and Elevators*

Emporis Standards defines a high-rise building as a multi-story structure between 35 and 100 m or a building of unknown height from 12 to 39 floors [40]. Korea Land and Housing Corporation (KLHC), Korea, which develops public apartments, classifies low-rise buildings as those with 5–6 floors, mid-rise as those with 7–15 floors, high-rise as those with 16–20 floors, and super high-rise as those with 21–49 floors as per residential building guidelines [41]. For construction at a relatively large scale, residential building projects comprise many buildings with different numbers of floors. Depending on the number of floors and households per floor, the capacity and speed of elevators in each building varies [41]. As shown in Table 1, the capacity comprises many passengers, loading capacity, box size, size of exits that are available, and elevator speed regulated by KLHC guidelines comprising six levels with the range from 60 m/min as a minimum to 180 m/min as a maximum as per the number of floors in a building. For the reference, HRBs in Korea have mostly 12–40 floors, and there is recently a case of a building with >40 floors. This study is precisely processed on HRBs with >12 floors and elevators at a speed of >90 m/min.

**Table 1.** Elevator speed by operating floor [42].


The change of air pressure, noise, and vibration generated while HRB elevators are being operated cause passengers to be discomforted and has a negative effect on the residential environment of nearby residents [4,8,11–13]. As per the building code of many countries, the characteristic noise level, because of a life within an apartment building, should not exceed 30 dB(A) in any bedroom or living room of apartments [15,43–45]. However, for many HRBs, there are multiple cases that noise level exceeds 30 dB(A) in the apartments around elevators [15,19,21]. Many researchers state that any noise problem may be described in terms of a source, a transmission path, and a receiver; furthermore, noise control may take the form of altering any one or all elements. The noise source is the one in which the vibratory mechanical energy originates because of a physical phenomenon such as mechanical shock, impacts, friction, or turbulent airflow [14,46,47].

Both noise and vibration that occur in HRB elevators are said to be noise sources; moreover, the parts of elevating system and structure of building are transmission paths and residents are considered as receivers. Therefore, it is primary solutions in response with the location of the source that reduce mechanical shock, impacts, and friction sources occurring while elevator machinery is being operated, or alleviate the occurrence of turbulent air noise generated while an elevator car is on the move. The secondary solution is controlling transmission paths in which elevators are arranged in isolation in the housing units of HRB, or they are designed to have a buffer space between them. Although many studies have been focusing on primary solutions, the ENV problems of HRBs have not been sufficiently solved [14,15,24,25,27,36]. Therefore, in-depth studies on primary and secondary solutions are required and then applied. The reason is that it is difficult to reduce elevator noise under 30 dB(A) caused at a high speed of it, even if noise and vibration can be partly alleviated by primary solution in the case of HRB elevators. This is because the damage largely escalates if ENV problems occur because the number of HRB residents is less.

## *2.2. Review of ENV Sources*

Fullerton [14] and Ingold [23] specifically described sources, transmission paths, and control of ENV. Torres and Haugen [27] reported a case study regarding noise and vibration because of machine room-less (MRL) elevators of apartment buildings and proposed an approach for alleviation [27]. Based on multiple studies, MRL elevators as gearless synchronous machines reduce electric energy by 50% compared to the geared traction elevators [27,48,49].

Moreover, many researchers published studies on ENV sources and transmissions [14,16,18–21,35]. The noise source is the one in which the vibratory mechanical energy originates because of a physical phenomenon such as mechanical shock, impacts, friction, tonal sound, or turbulent airflow [14,27,34,50–52]. Especially, interactions between the pulley and the cables that suspend the elevator cab show the potential for a tonal sound. These sounds vary in loudness and frequency with the system's speed, with the loudest airborne sounds being attributed to the fastest speed of operation [14,19]. The elevators of HRBs runs at a high speed from 90 to 180 m/min, which caused tonal sound at a considerable level.

In many studies, although ENV sources were introduced, most of them were from the perspective of a mechanical system and little was covered in terms of CM solutions for HRB [14,15,26–36,48]. Thus, in this study, ENV sources and transmission paths of HRB are analyzed. Confirming clearly airborne and structure-borne transmission paths through a study is to obtain clues to CM solutions to alleviate the ENV of HRBs. Several researchers, including Fullerton [14], Ingold [23], and Torres and Haugen [27], have partially presented ENV solutions from a design and construction perspective. However, they did not proceed from the perspective of a CM solution that integrates the design and construction phases as introduced in Figure 1.

## **3. Noise and Vibration Analysis of HRBs**

## *3.1. ENV Source and Transmission Analysis of HRBs*

Figure 3 shows the paths on which noise and vibration generated in the traction elevators of HRBs are transmitted to a residence. ENV occurs only when the machinery primarily runs in a machine room and a hoist way. Noise caused from various sources comprises airborne noise transmitted in the form of soundwaves through air particles, as shown in Figure 3, as well as structure-borne noise transmitted through the slabs, walls, and ceilings of buildings [7,14,17,18,21,27]. These two types of noises are delivered to a space that requires quietness, such as bedrooms and living rooms of residences, and consistently hinders living comfort.

**Figure 3.** The transmission paths of traction ENV [17,18,21].

By considering various studies [14,16,18–21,27,34,35,50–53] and the research of previous years, the sources, causes, and types of traction elevators have been identified, which are organized in Table 2. For machine rooms, noise is generated by operating multiple machine parts along with the high-frequency rotation of motors, meshing frequency of the gear system, on-and-off brakes, and electrical contact switches to control the elevator. For the hoist way, the sources are noise caused in the process of operation of door parts and by the misalignment of the door and the noise caused when an elevator car goes up and down as per the rails and elevator car guide, and from activities such as the rotation of bearings and rollers, passing through rail joints, and interactions between pulleys and cables.

As shown in Table 1, elevators at a high speed of >90 m/min are used in HRBs. At this time, inrush, air friction, and puff noises generated by elevator cars work as sources [8,9,38,39]. Moreover, the operation of devices such as machine cooling fans, car ventilation fans, car arrival signals, and friction and impact of machine parts work as sources. As shown in Table 2, various forms of noises comprise mechanical shock, impacts, friction, tonal sound, or turbulent airflow in a physical sense.



Table 3 shows the specific analysis of paths in which noises generated from various sources, as introduced in Figure 3, are transformed into structure-borne and airborne noises and transmitted to the residence. For structure-borne noise, vibratory noise generated by elevating machinery in the machine room is transmitted to residences via an anti-vibration pad, a machine support frame, a machine room slab, and a hoist way wall. Recently, it is common that an anti-vibration pad is designed double-layered between traction machine and support frame and is designed as a vibration transfer area and the size of the machine support frame are reduced, which leads to the considerable reduction of structure-borne noise transmitted from the machine room compared to the past. For structure-borne noise, vibratory noise generated by an elevator car as well as the ascent and descent of counter weight is transmitted to residences via an elevator car guide, rails, rail brackets, and hoist way.


**Table 3.** Transmission paths of ENV [14,21,39].

Airborne noise produced in a machine room is transferred via a rope hole because Figure 3 shows passing by a hoist wall to residences; moreover, airborne noise is transferred to residences past ventilation openings and neighboring residence windows of the machine. Airborne noise produced in the hoist way is transferred to residences through a hoist way wall. Moreover, there could be airborne noise produced in the hoist way, which can be easily controlled by a >200-mm-thick reinforced concrete (RC) wall and tightly closed concrete placement without any cracks.

### *3.2. ENV Case Analysis of an HRB*

To present CM solutions, it is necessary to analyze ENV problems and the actual condition of the ENV of HRB. We performed measurements of 2 buildings, as Figure 4 shows, as designating one area of apartment complex 'W' located in Seoul known for ENV problems. As shown in Figure 4, the elevator is adjacent to the bedroom or living room and is a traction-type, having a machine room. Figure 4a shows a case that the elevator of Building 'A' is adjacent to a living room of a nearby residence, whereas Figure 4b is a case that the elevator of Building 'B' is adjacent to a bedroom of a nearby residence. As shown in Figure 4c, the vibration was measured with four sensors attached to the wall, and another four sensors were then installed on the upper part of a room floor to measure noise. The measurement time was 50 seconds, and the elevator speed was 60 m/min as the lowest speed of HRB classified in Table 1. Table 4 shows the measurement system used to measure noise and vibrations produced by the elevator with a capacity of 550 kg, i.e., 8 persons.

**Figure 4.** ENV measurement of HRB 'W': (**a**) The living room in Building 'A'; (**b**) the bedroom in Building 'B'; (**c**) the vibration measurement sensors on the living room or bedroom wall; (**d**) the noise measurement sensors in the living room or bedroom space.



Figure 5 shows a graph in which noise and the result of vibratory measurements are simultaneously written; it is a case in which a living room is adjacent to the elevator-like Figure 4a. Figure 5b shows a case in which the bedroom is adjacent to the elevator-like Figure 4b. To date, elevator noise has shown a sound pressure level of 35 dBA on an average when reviewing the result of the noise measurement of the elevator in an apartment building in Korea. Although this noise level is quieter than the sound level criteria in a library, many residents face discomfort because the noise in low-frequency bands has a considerable influence on them. Note that the energy distribution of the general environment noise ranges from 125 Hz to 4 kHz; however, as shown in Figure 5, elevator noise is distributed up to 63 Hz or even up to 32 Hz. In other words, the noise accompanying vibrations is sensed with considerable discomfort even if it is low. The common tendency identified in both graphs is the increase in value of contrast measurement data of the background noise (BGN), and the background vibration (BGV) appears to be high within the range of low-frequency bands of <500 Hz. This corresponds to a figure reported in the guidelines of many previous published studies [16,54,55].

**Figure 5.** ENV measurement results of HRB 'W': (**a**) the ENV measurement results of living room; (**b**) the ENV measurement results of bedroom.

The measurement results in Figure 5 confirmed that the noise in the low-frequency band among noises transferred in the form of structure-borne noise and produced by the operation of the elevator affected the living comfort of residents. For measurements in the living room, the effect appeared to be the most in the range of 250 Hz. Moreover, for measurements in the bedroom, the effect appeared to be considerable in the range of 125 Hz. Regarding noise and vibration, the same tendency was confirmed in both the living room and bedroom. Based on these results, solutions should be presented in low-frequency bands in which noise and vibration produced during the operation has considerable influence on residences so as to reduce the elevator noise.

## **4. CM Solutions**

## *4.1. Design Solutions*

## 4.1.1. Separation of Residences and Elevators

In the ground of residents, to relieve ENV damages of HRBs, solutions applied in the design phase comprise two approaches. First, to separate space that is sensitive to noise such as bedrooms and living rooms from hoist ways or elevator shafts. Second, to arrange a buffer space between the rooms that are adjacent to hoist ways to increase the thickness of machine room slabs and hoist way walls, and to supplement partly building components similar to the detailed changes of hoist ways.

The most efficient approach to relieve the effects of noise and vibration of HRB elevators that run at a high speed is to separate elevators from residences and arrange them. By comparing the results of'noise measurement from the case when an elevator is attached to a living room or bedroom, like Figure 4, with the results of noise measurement from the case when an elevator is separated from residence, the effect can be confirmed. In particular, it is confirmed that noise became as low as 10–20 dB in the band of >125 Hz where the elevator is separated from residences. However, in a limited building area, residence design with high density should be high and if considering design-constraint conditions, such as daylight, view, and space function, there is a case that space that is sensitive to noise like living rooms and bedrooms cannot be separated and arranged. In this case, the second design solutions should be identified.

## 4.1.2. Buffer Space Design

As shown in Figure 6a, because the elevator hoist way is adjacent to bedrooms, noise and vibration are directly transferred if the elevator is operated. The case of toilet has a reduced impact of ENV; the case of a bedroom has a bad influence on rest and sleeping time. For HRBs in Korea, because a hoist way is designed with a 200-mm-thick bearing walls, airborne noise is blocked. However, the shield effect of structure-borne noise is not high. Thus, to relieve the transmission of structure-borne noise, residents feel it can be reduced if a wardrobe is placed similar to Figure 6b. To check the effect of it, one case of noise measurement in an apartment bedroom where there was a wardrobe sharing the wall with the elevator shaft was measured in comparison with another case without the wardrobe in the same condition. In Figure 6b, similar to the case of the installed wardrobe, the installation of wardrobe reduced noise level by ~7.6 dB, although Figure 6b is in the same plane compared to the general rectangular shape bedroom of Figure 6a. Such noise reduction was basically made possible by the effect of sound absorption or sound insulation using sound-absorbing materials such as blanket and clothing, both of which are filled at both sides of the wardrobe and inside of it. Moreover, the installation of the wardrobe made it possible to control the phenomenon of noise in a low-frequency band, known as room mode, which breaks the rectangular shape of a bedroom.

**Figure 6.** Noise reduction design using a wardrobe: (**a**) A floor plan without a wardrobe; (**b**) a floor plan with a wardrobe [17,21].

## 4.1.3. Change of Hoist Way Details

The elevator that runs at 120 m/min should be installed in HRBs with >26 floors (Table 1), and for HRBs with >40 floors, the elevator should run at a high speed of 180 m/min. For elevators that run at a high speed, turbulent airflow is produced, which leads to air friction noise, puff noise, inrush noise, and draft noise [38,39,56]. Design solutions regarding airborne noises are attributed to such a turbulent airflow.

(**1**) Air Friction Noise

As an elevator runs at a high speed in a hoist way (Figure 7a), air is compressed in a car-heading direction and elevator piston effect that increases air pressure [4,8,9,37–39,57,58]. Compressed air travels through the cramped gap between car and hoist way to the upper part of the car, which creates noise known as an air friction noise. The air friction noise gets louder as the area of the gap gets smaller compared to the area of the car with a faster speed of the elevator. This occurs as either a single elevator or double elevators run at a high speed in a hoist way. When two elevators run at the same time and in the same direction, the noise gets even louder. There is not an issue in the case that the elevator runs at 120 m/min in the hoist way for a single elevator, and the elevator runs at 180 m/min in the hoist way for double elevators. Note that the area of the hoist way should be designed to be bigger than normal by 40% at a higher speed such as Figure 7b [38,39].

**Figure 7.** Air friction noise and design solution: (**a**) Air friction noise; (**b**) a design solution for air friction noise.

## (**2**) Narrow-Section Passing Noise

Narrow-section passing noise, called puff noise [38], is a noise that is produced when an elevator goes through protruding steel installed as Figure 8a because of the structural mechanism or the installation of a guide rail. The elevator goes through an RC beam, which creates wind pressure that contributes to the noise. For HRBs, such protruding parts are reported per floor; therefore, the noise repeatedly occurs. To prevent this, architectural designing should be processed without the protruding parts inside the hoist way. However, in cases where it is unavoidable to have protruding parts, it is appropriate to install an air sliding panel upon and underneath the protruding parts to relieve the wind pressure that occurs in that area (Figure 8b). The appropriate angle of the air sliding panel is 4–8◦ [38,39]. This solution is required in the hoist way for a single elevator that runs at a speed of >150 m/min; in other words, with >31 floors in HRBs. The same solution is to be considered in the case that the car speed is >180 m/min in a hoist way for double elevators.

**Figure 8.** Narrow-section passing noise and design solution: (**a**) Noise generation; (**b**) the design solution.

## (**3**) Inrush Noise

It is common to arrange more than two elevators in HRB where there are >20 floors and four householders on each floor. When only one elevator of multiple elevators enters in the single hoist way (Figure 9a), noise is produced by compressed air flow, which is called inrush noise [38]. It is common that the car vibration occurs along with noise when the inrush noise by the high speed is considerable. This is preventable because the cause of inrush noise is rapid air compression when an elevator rushes in. The best approach is not to have the single hoist way; however, in the case where it is unavoidable to have one because of structural constraints, two solutions can be considered. First, similar to Figure 9b, ventilation openings can be placed either on one side or both sides of the single hoist way wall. The size of the opening varies depending on the size and speed of the car; however, there is no trouble if it is designed within 1.5 m2. If it is difficult to place the ventilation opening, the wall is expanded such that the floor area of the hoist way is increased by >40% (Figure 9c) [38,39].

**Figure 9.** Inrush noise and design solution: (**a**) Inrush noise; (**b**) a ventilation opening arrangement; (**c**) expansion of the hoist way floor area.

## (**4**) Draft Noise

Generally, in a high-lift elevator, a pressure difference is generated between the landing and hoist way. Therefore, air flows into and out of the hoist way from the gap around the landing door; at that time, a blowing sound called draft noise is generated [39,56]. In particular, this noise gets worse when the air produced by a heater in winter flows in the hoist way and the elevator ascends. The solution to this is to design an entrance door that is double-layered or a revolving door such that cold air outside does not enter within HRB. Moreover, it should be designed with air-shielding details between and around elevator opening frame of the landing floor such as a landing door, jamb, and sill [56].

## 4.1.4. Other Design Solutions

Note that design solutions should be considered, in addition to design solutions that are suggested, so as to alleviate the ENV that residents of HRBs feel.


## *4.2. Construction Solutions*

Construction solutions are the ones to be applied to at the construction phase and are about the selection of elevating machinery, the location, and the method of installation, in addition to design solutions that present solutions to ENV problems at a planning phase. Similar to Table 5, the problems are classified into structure-borne and airborne noises, and again classified into the machine room and hoist way (shaft), and then construction solutions are analyzed and presented. They are subdivided into either controlling elevator vibration or noise sources and controlling transmission paths.


**Table 5.** Construction solutions to mitigate ENV.

Table 5 shows the construction solutions to mitigate ENV that are obtained through the studies for years after contemplating various research documents [14,18–21,27,38,39,50–52]. Structure-borne noise generated in the machine room primarily occurs via the vibration of the elevating machinery. To alleviate this, elevating motor with power filters with high quality basically are used and brakes, gears, and elevator car should be installed to maintain the precise balance of elevator machinery. Moreover, vibration isolation pads that have high damping performance such as neoprene or rubber should be inserted at fixing points such that the vibration is generated in the traction machine and switchgear cabinet [14,18–21,50–52]. When anti-vibration pads are installed in the traction machine, the first natural vibration frequency of the whole elevator should be maintained at a less than audible frequency band. For the traction machine to work precisely aligned with the height between elevator car and hall landing, elevator machinery and anti-vibration pads should be installed, thus maintaining balance to minimize isolation deflection. For this purpose, close cooperation with an elevator manufacturer and installer is required at the installation phase.

To alleviate structure-borne noise generated in the hoist way or elevator shaft, the machining errors of guide rails are to be maintained within ±2 mm/5 m, and seam surface difference between guide rails of the guide rails should be installed within ±0.05 mm, which alleviates vibration because of the movement of rollers. Vibration-borne noise produced in the hoist way occurs between the guide rollers of the car and rails. To prevent this, brackets should be installed next to the floor slab to fixate the rail (Figure 10). The floor slab edges are inherently stiffer than the shaft walls and will limit the transmission of rail and roller guide interactions from generating the structure-borne noise in adjacent spaces [14,17,21]. Furthermore, anti-vibration pads should be installed between brackets and floor slab edges. For reference, depending on the location of the rail brackets, it is confirmed that the noise difference by ~4 dB on an average occurs because of the measurement of noise while an elevator is running [21].

**Figure 10.** Fastening rail brackets at the edge of a slab.

Measures in correspondence with the construction phase should be considered because airborne noises generated in the machine room primarily occur by the tonal sound of the elevating machinery. It is essential to select motors of good quality to encounter noise sources and precise machining; moreover, the fabrication of brakes, gears, and coupling parts should be performed.

A low-noise cooling fan or self-cooling system in the machine room should be used. As shown in Figure 3, sound insulation cover should be installed around the rope hole as transmission paths of airborne noises in the machine room. Moreover, soundproofing materials should be used to encounter isolation of airborne noise on the interior wall and the ceiling in the machine room. After glass wool as thick as 50 mm is placed on the wall and ceiling in the machine room, it is confirmed that the level of noise generated in the elevating machinery is reduced by 6.2 dB on average. In particular, it is confirmed that the noise level in the range of 1 kHz frequency is reduced by 9.4 dB at the maximum.

Airborne noises generated in the hoist way are primarily related to the elevator door. The primary noise sources of the door are its misalignments or improper installation and alarm sound for arrival on each floor. To solve such problems, the precise installation of doors is required, and the volume of the door enunciator is to be adjusted to <60 dB(A).

### *4.3. Verification of CM Solutions*

In this study, the suggested CM solutions are applied and how much ENV is alleviated should be confirmed. For this purpose, after selecting the HRB project with a 25-story building as a case, the most effective elevators of CM solutions and design solutions, such as separation of residences and elevators as well as the hoist way wall as thick as 200 mm, were applied. One noticeable aspect is that it is designed using MRL elevators [49] that are generally used for buildings of <20 floors. Moreover, after selecting an OTIS elevating machine of good quality at the construction phase, the construction solutions suggested in Table 5 were mostly applied. The case project was completed in 2020, whereas the elevator noise of two buildings was measured in the same method of Figure 4d. The vibration measurement with the sensors attached on the wall was not done without the agreement of the residents (Figure 4c).

As shown in Figure 11a, the staircase was placed between the elevator and the bedroom of residence. Two elevators were placed as per design guidelines because there were four householders on each floor in Building 'A' [41,42]. Figure 11b shows that the elevator is separated from residence using the elevator hall. For this measurement, it was conditional that the elevator runs departing from the second floor as the lowest floor and stopped on the 25th floor. The measurement was made in bedrooms on the 25th floor as the highest floor. Moreover, the measurement time was 90 seconds, and the elevator speed was 105 m/mm, which is the speed of the 25-story HRB classified in Table 1. The same measuring system as in Table 5 was used for measuring noise generated by the elevator with a capacity of 15 persons and 1150 kg.

**Figure 11.** Elevator noise measurement of HRB 'D': (**a**) The 25th floor plan of Building 'A'; (**b**) the 25th floor plan of Building 'B'.

Figure 12 shows the result of noise measurement for the case that elevators and residences were isolated, as seen in Figure 11a,b. Similar to Figure 5, the effect of noise measurements compared to BGN in the range of <500 Hz appears to be extensive. Note that regardless of different conditions, it shows that it was a low-frequency band that had an influence on elevator noise. In particular, the effect of the two cases seemed to be extensive at 63 Hz; moreover, the common aspect is that the effect of low-frequency bands at the central measuring point appeared to be less. The reason is that the low-pitched superposition phenomenon by the room mode was most remarkably noticeable at the corners of rectangular bedrooms. Figure 13 shows the distribution of the sound pressure level by the overlapping and offsetting of sound energy per frequency band in the rectangular room. The low sound energy at 63 and 125 Hz certainly demonstrated the phenomenon of remarkable overlap at the corners of the room [59].

**Figure 12.** Elevator noise measurement results of HRB 'D': (**a**) The measurement results of 'Building A'; (**b**) the measurement results of 'Building B'.

**Figure 13.** Different patterns of spatial distribution of sound pressure level in a room according to the frequency bands [59]: (**a**) The noise measurement results at the center; (**b**) the noise measurement results at the corners.

Figure 14 shows the analysis of the level of noise reduction and noise characteristics as per the frequency band after the average conversion of the measurement results of Figures 5 and 12. The measurement results of Figure 12 show that, in this study, CM solutions that were applied as suggested demonstrate the remarkable reduction of noise in the frequency band of >100 Hz. Figure 12 shows that there was no considerable difference at frequencies of <125 Hz; however, it was confirmed that noise of HRB 'D', in which elevators and residences are isolated in the frequency of >125 Hz, was measured at a noise level lower than that of HRB 'W'. As shown in Figure 13, for HRB 'W', while the average value of the noise level compared to BGN makes a difference within 26 dB(A), the case of HRB 'D' maintains the level of 5 dB(A). This is explained to be the alleviating noise that residents feel with CM solutions suggested in this study. However, regardless of the isolation of elevators and residences in the low-frequency band of <125 Hz, the noise level compared to BGN seems to be high. The structure-borne noise generated by the operation of the elevator is analyzed to be high in the low-frequency band.

**Figure 14.** Noise analysis by frequency band according to the location of elevators and residences.

### **5. Discussion**

To solve ENV problems of a building, multiple studies regarding mechanical solutions have been conducted; in certain studies, solutions were presented from the perspective of design and construction, which are only for residential buildings of less than ten floors. Thus, to alleviate the ENV of HRBs that run at >90 m/min, solutions that consider the characteristics of high-speed elevators and HRB floor plans on the basis of knowledge obtained from the results of their study are necessary. We analyzed the characteristics of noises and vibrations produced in machine rooms and hoist ways through ENV-related studies of HRBs for the past few years and sources vs. transmission paths of airborne and structure-borne noises, and then presented solutions that control the stage of design and construction. A case study was processed to confirm the given solutions. CM solutions applied to the case of HRB 'D', including isolation arrangement of elevators and residences, 200-mm-think RC wall design of hoist way and most other solutions are presented in Table 4; they were then applied to the construction phase. We were then able to confirm considerable improvement compared to HRB 'W' case because of the measurement on noise of HRB 'D'. Vibration measurement was impossible because the residents opposed to sensors being installed on the wall; however, it is assumed that there may have been improvement of vibration corresponding to that of noise considering the HRB 'W' case.

Studies on identifying the effectiveness of alleviating noise should continue with thorough experimentation in each item to clarify the utility of the CM solutions given in this study. The extra application of researches of CM solutions should be processed by the subject area of HRBs presented in Table 1.

#### **6. Conclusions**

There are have been many studies regarding ENV worldwide, which is proof that the more sensitive humans have been to noise, the more comfortable life environment they have demanded. In particular, for HRBs built in many downtown areas where there are external noise sources, residents demand more static environments, while noise is more likely to occur because of the use of high-speed elevators. To solve such as dilemma, we contemplated multiple documents and presented CM solutions to alleviate ENV of HRBs in this thesis by means of the study of years. An HRB project with ENV problems and an HRB project that uses solutions given in this study were processed as a case study, and its effectiveness was verified, with the results as following.

First, BGN and BGV appear to be relatively high in the range of low frequency of <500 Hz if an elevator hoist way is adjacent to the living room or bedroom. It was confirmed that the noise produced at low-frequency bands of 250 Hz for a living room and those of 125 Hz for a bedroom had a considerable impact on the living comfort of residents.

Second, in the case of HRBs to which CM solutions are applied, the effect of the measured noise compared to BGN is within the range of <500 Hz. This shows that what affects elevator noise were low-frequency bands of <500 Hz. CM solutions application effects appeared to be the most in the range of 63 Hz; moreover, a common aspect is the effects of low-frequency band turning out to be little at the central measurement point Ch3. The reason is because it is analyzed that the distribution of sound pressure level by overlapping and offsetting of sound energy per frequency band in the rectangular room was remarkably noticeable.

Third, compared to the measurement results of the two case projects, the elevator noise of HRBs that CM solutions were applied to in the frequency band of >100 Hz are confirmed to remarkably reduce noise. Especially, it is confirmed that the CM solutions-applied noise of HRB 'D' was measured to be relatively low, in the frequency band of >125 Hz. While the average value of noise level compared to BGN in the case of HRB 'W' with ENV problems makes a difference of 26 dB(A), the case of HRB 'D' that the CM solutions were applied to maintain the level of 5 dB(A). Thus, it is comprehended that, in this study, the CM solutions suggested largely alleviated noise.

**Author Contributions:** Conceptualization, Y.O. and S.K.; methodology, S.K.; validation, Y.O., M.K., and S.K.; formal analysis, Y.O. and M.K.; measurement and data curation, Y.O. and M.K.; writing—original draft preparation, Y.O. and K.L.; writing—review and editing, S.K.; visualization, K.L.; supervision, S.K.; project administration, S.K.; funding acquisition, Y.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the grant (20RERP-B082204-07) from Residential Environment Research Program funded by Ministry of Land, Infrastructure and Transport of Korean government.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

## **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Dynamic Optimization Model for Estimating In-Situ Production Quantity of PC Members to Minimize Environmental Loads**

## **Jeeyoung Lim and Joseph J. Kim \***

Department of Civil Engineering and Construction Engineering Management, Green BIM Laboratory, California State University Long Beach, Long Beach, CA 90840, USA; jeeyoung.lim@csulb.edu **\*** Correspondence: Joseph.Kim@csulb.edu; Tel.: +1-562-985-1679

Received: 20 August 2020; Accepted: 1 October 2020; Published: 5 October 2020

**Abstract:** CO2 emissions account for 80% of greenhouse gases, which lead to the largest contributions to climate change. As the problem of CO2 emission becomes more and more prominent, research on sustainable technologies to reduce CO2 emission among environmental loads is continuously being conducted. In-situ production of precast concrete members has advantages over in-plant production in reducing costs, securing equal or enhanced quality under equal conditions, and reducing CO2 emission. When applying in-situ production to real projects, it is vital to calculate the optimal quantity. This paper presents a dynamic optimization model for estimating in-situ production quantity of precast concrete members subjected to environmental loads. After defining various factors and deriving the objective function, an optimization model is developed using system dynamics. As a result of optimizing the quantity by applying it to the case project, it was confirmed that the optimal case can save 7557 t-CO2 in CO2 emissions and 6,966,000 USD in cost, which resulted in 14.58% and 10.53% for environmental loads and cost, respectively. The model developed here can be used to calculate the quantity of in-situ production quickly and easily in consideration of dynamically changing field conditions.

**Keywords:** in-situ production; environmental loads; CO2 emission reduction; life cycle assessment; optimization model; system dynamics

## **1. Introduction**

Due to climate change, problems such as droughts, heatwaves, and rising sea levels are globally occurring [1]. One of the biggest causes of climate change is greenhouse gas [2], and international regulations on greenhouse gas emissions are being strengthened [3]. CO2 accounts for 80% of greenhouse gases, and the problem of CO2 emission becomes more and more prominent [4]. In particular, research on sustainable technologies to reduce CO2 emission among environmental loads is continuously being conducted [5–9]. The construction industry is recognized as a major cause of environmental pollution [10], and it is important to quantify and evaluate the environmental load.

In studies related to the calculation of environmental loads of construction, Tae et al. evaluated the CO2 generated during the life cycle of a building and its economic efficiency to assess the environmental loads and costs of buildings that use plaster board drywall [11]. Priatla et al. proposed a methodology to quantify CO2 emissions by life cycle in water supply construction projects [12]. Park et al. studied correlation analysis between the environmental load computed through life cycle assessment (LCA) using the database of national highway construction cases and the inventory of available information that can be extracted in the road planning stage [13]. Lee et al. developed and validated an environmental load estimating model for the New Austrian Tunneling Method (NATM) tunnel based on the standard quantity of major works in the early design phase [10]. These studies

were conducted to develop decision-support tools using quantified environmental loads, or to evaluate environmental loads by applying life cycle cost (LCC) or environmental valuation methodologies.

In-situ production of precast concrete (PC) members not only reduces the cost by 14.5–39.4% compared to in-plant production, but can also result in equal or enhanced quality under equal conditions [14–18]. Through an experiment in which the amount of CO2 emission reduction was analyzed according to the increase in quantity, it was proved that in-situ production is an eco-friendly technology with a high CO2 reduction effect [19]. However, in order to apply in-situ production to a project, additional research is needed to calculate the optimal quantity considering the environmental load.

However, it is difficult to produce all quantities in situ due to various field constraints, as well as the given time. The in-situ production quantity is affected by various factors, such as lead-time, number of molds, and number of cranes [17,20]. Although quantity is an important factor that determines the in-situ production scale, it is difficult to estimate it because it is indirectly affected by most of the influencing factors. Therefore, a simple method is necessary to calculate the optimal quantity to apply in-situ production in real projects.

Therefore, the objective of this paper is to develop a dynamic optimization model for estimating the in-situ production quantity of PC members subjected to environmental load. By defining various factors and deriving the objective function, the optimization model is developed using system dynamics and then applied to a case project for verification. The model developed here can be used to calculate the possible quantities of in-situ production quickly and easily in consideration of dynamically changing field conditions.

This study is carried out as follows.


## **2. Preliminary Study**

Under equal production conditions, it was verified through experimental studies that in-situ production of PC members secures equivalent or enhanced quality compared to in-plant production while significantly reducing costs [17,18,20,21]. Lee studied the management model and necessary conditions for in-situ production of composite PC members, and suggested cost, quality, process, resources, and safety management as management factors [22]. Park et al. studied the manufacturing technology for in-situ production of ultra-high-strength PC piles and suggested optimal production conditions through experimental production [23]. Won et al. studied the energy efficiency of in-situ production of PC members using steam curing and suggested necessary equipment and production concepts [24]. Lim et al. introduced a detailed management process through a case study of in-situ production and showed that the cost effectiveness ratio increased as the quantity increased [18]. Kim et al. presented the embedded energy efficiency of their proposed precast concrete frames based on the material, structure, and construction characteristics [25]. These studies analyzed specific plans for in-situ production of PC members and items to be considered when planning and showed that they were advantageous in terms of quality and cost of in-situ production.

Recently, CO2 has been one of the most important causes of global climate change [2], and a method that can apply in-situ production of PC members is needed. Several studies were conducted to reduce CO2 emissions of the PC method. Dong et al. showed that the PC method is a way to reduce carbon emissions by comparing to the cast in-situ construction method through experiments using life cycle assessments [26]. Yepes et al. proposed a methodology to optimize CO2 emissions when designing a PC road and developed a hybrid glowworm swarm optimization algorithm [27]. Kim and Chae presented that a considerable amount of CO2 is emitted during the steam curing stage when making PCs, and proposed a method of evaluating CO2 emissions throughout the PC life cycle using life cycle assessment [28]. Lim et al. calculated CO2 emissions with the LCA method through an experimental study of in-situ production and determined that the amount of CO2 emission reduction increases as the quantity of in-situ production increases [19]. These studies were reviewed from the eco-friendly perspective for field application through optimization and evaluation methods for the existing PC method's CO2 emissions. In addition, it was confirmed that it is advantageous to apply in-situ production in terms of CO2 emissions, but more research on minimizing CO2 emissions is needed.

Several studies used system dynamics in the production and installation of PC members to analyze influencing factors. Tan et al. applied the Pull-Driven Scheduling (PDS) simulation technique for the Singapore Light Rail Transit (LRT) project to produce PC members. The installation process was dynamically analyzed [29]. Ballard et al. conducted a study to improve the productivity of in-plant production for PC members using dynamic analysis techniques based on the measured data [30]. Cho et al. expressed the production, transportation, and construction process for PC-structured apartment houses as an Entity–Relationship Diagram (ERD) [31]. Lim et al. analyzed the factors that influence the calculation of the in-situ production volume and applied the developed simulation model to six scenarios to derive the in-situ production volume [20]. These papers showed that the dynamic relationships of various influence factors have been considered, and more research using them is being conducted. Therefore, in order to minimize CO2 emissions, this paper aims to calculate the optimal quantity considering the dynamic relationship of various factors such as lead-time, number of molds, and number of cranes.

## **3. Methodology for the Optimization Model**

The authors describe system dynamics and the Monte Carlo simulation method as applied techniques for developing an optimization model for estimating CO2 emissions, followed by the definition of influence factors for in-situ production of PC members with details in step-by-step processes. Then, a dynamic optimization model is developed to minimize environmental loads in the sequence of a generation model, simulation model, and optimization model.

## *3.1. Applied Techniques for Simulation*

## 3.1.1. System Dynamics

In carrying out in-situ production of PC materials, there is a limit to the complexity of the relationship between these influencing factors, which is why it is difficult to clarify this with general static analysis [17]. Static analysis is used because the one-way independent variable affects the dependent variable, it expresses the causes and effects of temporary events, and it views things from a partial perspective. Therefore, a system dynamics technique is needed as a means to grasp and quantify the dynamic relationship between influencing factors.

System dynamics can be defined as follows: (1) Rather than obtaining estimated values of a one-time event or variable, more attention is paid to what kind of dynamic change tends to occur over time in the variable of interest. (2) All phenomena are viewed from the perspective of an internal and cyclical closed-loop thinking, and are understood to be caused by circular dynamic interactions [32]. (3) The research focuses on the process of change and how it is actually happening. In other words, system dynamics is a methodology for understanding complex systems, and is a research methodology that explains the changes in the system over time, focusing on the causal relationship and the feedback relationship [33,34].

## 3.1.2. Monte Carlo Simulation

The dynamic optimization model was developed using the Powersim Studio 10 Expert program and the simulation was run by utilizing influencing factors such as in-situ production quantity, lead-time, number of molds, and number of cranes. Many values for factors must be derived through simulation, and control ranges for the values need to be set. A random number that follows the probability distribution for each factor is generated using Monte Carlo simulation. Using a computer program is the most effective way to generate a series of random numbers [35]. Monte Carlo simulation is performed by generating 100,000 random numbers, assuming a deviation of 0.1. Various values are presented by random numbers for each variable generated through Monte Carlo simulation. In other words, it attempted to overcome the mathematical limitations of the deterministic method by using the probabilistic method [36].

## *3.2. Definition of Influencing Factors for In-Situ Production*

In-situ production of PC members is carried out in the same manner as in-plant production—applying demolding oil, pouring concrete, finishing the surface of the member, curing, demolding, and yard stocking [21]. In-situ production can be carried out at the same level as the factory by placing the assembled reinforcing bars in the steel formwork, as shown in Figure 1a. As shown in Figure 1b, steam curing is performed using a boiler, and the cured PC member is stacked. All processes are accomplished by establishing a production plan that can be supplied in the just-in-time delivery method of PC members according to the installation plan.

**Figure 1.** Major process of in-situ production: (**a**) manufactured steel mold [20]; (**b**) steam curing [18].

A survey conducted by Lim found that quantity was the most important influence factor for in-situ production [17]. The importance of influence factors was in the following order: number of cranes, number of molds, lead-time, yard stock area, production area, production cycle, erection cycle, material and traffic control, and crane location. In the study, only the factors that directly affect the calculation of the in-situ production quantity were considered by reflecting the results of the existing studies and the field application cases, and five influencing factors were selected, including cost, quantity, lead-time, number of molds, and number of cranes. Figure 2 is a causal loop diagram illustrating in-situ production cost, quantity, lead-time, number of molds, and number of cranes using system dynamics.

**Figure 2.** Causal loop diagram for estimating in-situ production quantity.

Details for each influencing factor are as follows.

1. In-situ production quantity: Since CO2 emissions can be calculated only by the quantity, quantity is the factor that has the greatest influence on CO2 emissions, and is a key influence factor and result of this study. Equation (1) shows how the quantity can be calculated using the in-situ production time, number of molds, and production cycle. In Figure 2, the in-situ production quantity affects the stock volume, and the stock quantity is determined by the difference between the accumulated production quantity and the accumulated installation quantity. As the stock volume increases, the yard stock area increases, and as the yard stock area increases, the in-situ production quantity can be increased.

$$Q\_{STUI} = \frac{T\_{STII} \times N\_{MOLD}}{T\_{PC}} \tag{1}$$

where *QSITU*: in-situ production quantity (unit); *TSITU*: in-situ production time (day); *NMOLD*: number of molds (unit); *TPC*: production cycle time (day).

2. Lead-time: When applying in-situ production, a separate process plan for the site is required. Figure 3 shows that the lead-time is the time of in-situ production in advance before the PC member is installed and is the period from the start of production of the PC member to the start of installation. Considering the curing period, not all PC members can be produced during the installation period, so members must be produced in advance [17,20]. As the in-situ production time increases, the amount of in-situ production available increases, so it is important to secure lead-time, as shown in Figure 2. The lead-time can be calculated using the production cycle, quantity during lead-time, and number of molds, as shown in Equation (2).

$$T\_{LEAD} = \left(T\_{PC} \times Q\_{SLi}\right) / N\_{MTi} \tag{2}$$

where *TLEAD*: lead-time (day); *TPC*: production cycle time (day); *QSLi*: in-situ production quantity during lead-time (unit); *NMTi*: number of mold types (unit); *i*: number of mold types (1, ... , *n*).

**Figure 3.** Calculation of in-situ production time [20].

3. Number of molds: As the number of molds increases, the quantity of production per unit of time is increased, resulting in a strong effect of shortening the time, while the cost of in-situ production increases rapidly. The reason is that the steel mold manufacturing cost is high. PC members of various sizes cannot be produced with molds of the same size. Table 1 shows five mold types classified according to the size of the member. Figure 2 shows that as the in-situ production quantity increases, the number of molds increases, and the number of molds affects the production area. The number of molds also affects the in-situ production quantity, so they affect each other. Since the in-situ production cost indirectly increases as the amount of time increases, it is necessary to calculate an appropriate number of molds through a feedback routine. The number of molds is a key influence factor in calculating the quantity of in-situ production, considering the time and cost. In Equation (3), the number of molds can be calculated by using in-situ production quantity, production cycle, and time.

$$N\_{\text{mod}l} = \sum\_{i=1}^{n} \frac{Q\_{MOLDi} \times T\_{PC}}{T\_{SITul}},\tag{3}$$

where *NMOLD*: number of molds (unit); *QMOLDi*: in-situ production quantity of each mold type (unit); *TPC*: production cycle time (day); *TSITU*: in-situ production time (day); *i*: number of mold types (1, ... , *n*).


**Table 1.** The shapes of precast concrete (PC) members.

4. Number of cranes: The large-scale building covered in this paper has a large floor area but not a high number of floors, making it difficult to use a tower crane, so a mobile crane was used. A crane is used to move the module and to lift and install the PC members. Equation (4) shows that the number of cranes can be calculated by dividing the installation time by the product of the unit usage time per member and the number of installation members, and the number of cranes are an integer equal to or greater than 1.

$$\begin{array}{l} \text{N}\_{\text{crane}} = \frac{(T\_{IL} \times Q\_{SITLI})}{T\_{\text{crec}}}\\ \text{Subject to } \text{N}\_{\text{crane}} \ge 1, \text{ integer} \end{array} \tag{4}$$

where *Ncrane*: number of cranes (unit); *TUE*: unit erection time (day); *QSITU*: in-situ production quantity (unit); *Terec*: erection time (day).

5. In-situ production cost: The cost is less than the in-plant production cost and is proportional to the quantity and number of molds. If the cost is not satisfied, in-situ production cannot be applied, so it is a limiting condition for minimizing CO2 emission. If the total production cost of PC components is high, the in-situ production volume, which is lower than the in-plant production cost, is increased [16]. Figure 2 shows that all influence factors affect cost and can finally be collected. Equation (5) shows that cost can be calculated by the number of mold types, unit mold production cost for mold type, in-situ production quantity, and unit PC member production cost.

$$\mathbb{C}\_{\text{STII}} = \sum\_{i=1}^{n} \mathbf{N}\_{\text{MOLIDi}} \times \mathbb{C}\_{\text{MOLIDi}} + \sum\_{i=1}^{n} \mathbf{Q}\_{\text{MOLIDi}} \times \mathbb{C}\_{\text{PRODDi}} \tag{5}$$

where *CSITU*: in-situ production cost (USD); *NMOLDi*: number of mold types (unit); *CMOLDi*: unit mold production cost for mold type (unit); *QMOLDi*: in-situ production quantity for mold type (unit); *CPRODi*: unit PC member production cost for mold type (USD); *i*: number of mold types (1, ... , *n*).

## *3.3. Dynamic Optimization Model to Minimize Environmental Loads*

The dynamic optimization model was developed sequentially through the generation model and simulation model. The generation model determines one case, and the simulation model derives the control range through simulation. The optimization model is used to find the most suitable value for the field condition among simulated values. This development process is shown in Figure 4.

**Figure 4.** Development process for the optimization model.

## 3.3.1. Generation Model

The generation model determines one case by deriving the in-situ production cost and time for various influence factors. For instance, the 72 columns of in-situ production conducted in this experimental study are a generation model. Using the quantity calculated based on the design drawings, the CO2 emissions for the 72 columns generated during in-situ production of total PC members were calculated. Influencing factors each have one value, and Figure 5 shows the mathematical connection of the relationship between influencing factors and CO2 emissions. That is, an equation defines five influencing factors of in-situ production cost, quantity, lead-time, number of molds, and number of cranes, and calculates CO2 emissions by substituting the influence factors into each equation. This generation model can be expressed as shown in Figure 5. All the influence factors have a dynamic relationship with each other, as shown in Figure 2.

**Figure 5.** Concept of the generation model.

In this study, CO2 emission sources, such as material, oil, electricity, and transportation, are classified and calculated for the analysis of the CO2 emission reduction effect on in-situ production. After defining basic units of CO2 emission or estimation equations for each source in advance, the quantity of CO2 emissions generated by in-plant production is calculated using the quantity of sources. Next, the CO2 emissions by in-situ production are calculated and compared with in-plant production. In order to calculate the material use, each basic unit of CO2 emission per material quantity is used [37]. Concrete 140 kg-CO2/m3 and Steel 3500 kg-CO2/t can be used to calculate CO2 emissions. In the studies of Hong et al., Lee et al. and Lim et al., the CO2 emissions generated in the construction stage of the building were calculated in the same way [19,38–42].

For the CO2 emissions of the erection process, CO2 emissions according to the use of oil and electricity must be calculated. Kim et al. analyzed using the LCA technique and proposed a CO2 emission regression equation at the construction stage [37]. Since this regression equation has a gross floor area as a variable, it is easy to estimate CO2 emission according to oil use and power consumption. First, the equation for calculating the CO2 emissions of the construction work according to the use of oil in the construction stage is the same as Equation (7), which can be calculated using Equation (6), an equation for calculating energy consumption [37]. The oil use of in-situ and in-plant production was the same with the same production conditions. The equation for calculating CO2 emissions at the construction stage according to power consumption is the same as Equation (9), which can be calculated using Equation (8), an equation for calculating energy consumption. CO2 emissions by transportation equipment use are only applicable when moving PC members manufactured at the plant to the construction site. Unlike in-situ production, in-plant production requires transport of members, and basic units of CO2 emission are 0.464 kg-CO2/ton·km and 31.080 kg-CO2/number of units of equipment [37].

$$\mathbf{E}\_{\rm CO} = 0.0017 \times \mathbf{A}\_{\rm f} + 37.5 \tag{6}$$

$$\mathbf{Q\_{CO\_2O}} = \mathbf{E\_{c0}} \times \mathbf{3.06} \tag{7}$$

$$\mathbf{E\_{CE}} = 0.0247 \times \mathbf{A\_{f}}^{0.79} \tag{8}$$

$$\mathrm{Q}\_{\mathrm{CO}\_2\mathrm{E}} = \mathrm{E}\_{\mathrm{ce}} \times 1.64 \tag{9}$$

where *ECO*: energy (oil) consumption during the construction stage; *Af*: total floor area (m2); *QCO2O*: CO2 emissions based on oil use in the construction stage (t-CO2); *ECE*: power consumption in the construction stage; *QCO2E*: CO2 emissions based on power consumption in the construction stage (t-CO2).

### 3.3.2. Simulation Model

The simulation model was created based on the generation model [43]. The five influencing factors, such as in-situ production quantity, lead-time, number of molds, and number of cranes, can derive a range value according to the site conditions. The dynamic optimization model was developed using the Powersim Studio 10 Expert program in this paper and can be simulated using the derived influencing factors. Various values are presented by random numbers for each variable generated through the Monte Carlo simulation by using a number of cases as a result of the influencing factor, cost, and the minimum value, which is the control range of CO2 emissions. The minimum value (Min) and maximum value (Max) can be derived, and Figure 6 is a schematic diagram.

**Figure 6.** Concept of the simulation model.

## 3.3.3. Optimization Model

When the simulation model is developed, an optimization model is created using the derived values. The optimization model calculates one most appropriate value from the Min–Max of CO2 emission, which is obtained from the results of the simulation model. It is possible to derive appropriate values of in-situ production cost, quantity, lead-time, number of molds, and number of cranes, which are influencing factors corresponding to optimal CO2 emissions. Figure 7 is a schematic of the optimization model, and it can be derived using the Powersim Studio 10 Expert program.

**Figure 7.** Concept of the optimization model.

Environmental load assessment measures whether CO2 emission is minimized within the range possible for in-situ production. Environmental loads can be calculated for in-situ and in-plant production. The larger the difference between these values, the more CO2 emissions are minimized. The cost, time, and yard stock area are within the allowable range. Equation (10) is an objective function and boundary condition to minimize environmental loads. Among the various values generated through the Monte Carlo simulation within the range satisfying these constraints, the maximum difference between in-plant and in-situ production for environmental loads is derived. Equations (11) and (12) are methods of estimating environmental loads for in-situ and in-plant production, respectively. They can be calculated by multiplying the quantity by the sum of the values for each item of material use, oil use, electronic use, and transportation equipment use. Transportation equipment use is calculated only in the case of in-plant production, and these values can be accumulated by member type.

$$\begin{array}{c} \text{Maximize } f\_{co2} \ (Q\_i) = Q\_{CO2}P - Q\_{CO2}S\\ \text{Subject to } \mathbb{C}\_{req} \le \mathbb{C}\_{avail} \\ T\_{req} \le T\_{avail} \\ A\_{req} \le A\_{avail} \end{array} \tag{10}$$

$$Q\_{CO\_2P} = \sum\_{i=1}^{n} \left[ Q\_{Pi} \times \left( Q\_{CO\_2M} + Q\_{CO\_2O} + Q\_{CO\_2E} + Q\_{CO\_2T} \right) \right] \tag{11}$$

$$Q\_{CO2S} = \sum\_{i=1}^{n} \left[ Q\_{Si} \times \left( Q\_{CO2M} + Q\_{CO2O} + Q\_{CO2E} \right) \right] \tag{12}$$

where *QCO2P*: CO2 emissions of in-plant production (t-CO2); *QCO2S*: CO2 emissions of in-situ production (t-CO2); *Creq*: required cost (USD); *Cavail*: available cost (USD); *Treq*: required time (month); *Tavail*: available time (month); A*req*: required area (m2); A*avail*: available area (m2); *QPi*: in-plant production quantity of mold types (t-CO2); *QCO2M*: CO2 emissions of material use (t-CO2); *QCO2O*: CO2 emissions of oil use (t-CO2); *QCO2E*: CO2 emissions of power consumption (t-CO2); *QCO2T*: CO2 emissions of transport equipment use (t-CO2); *QSi*: in-situ production quantity of mold types (t-CO2); *i*: number of mold types (1, ... , *n*).

## **4. Development of the Dynamic Optimization Model**

Based on the previously mentioned causal loop diagram and the dynamic optimization model, the cost and CO2 emission simulation models were created using the Powersim Studio 10 Expert program. The mold type derived based on the previously analyzed PC column and PC beam shape was applied.

## *4.1. Cost Model*

The cost simulation model created for one mold type is shown in Figure 8. From the cost model, the in-situ production time is calculated using time-lag, lead-time, installation start date, and installation completion date (A). By entering the production cycle of two days and five working days per week, the number of parts per week is calculated, and the number of units of production is applied to the calculation of the number of molds together with the quantity and the in-situ production time (B). Material cost, labor cost, and equipment cost can be calculated using these calculated values.

**Figure 8.** Cost estimation simulation model (mold type 1 for in-situ production); in-situ production time (A), production time (B), total material cost (C), total labor cost (D), total equipment cost (E).

Material cost is calculated by dividing into PC member production cost and mold cost (C). PC member production cost is calculated using the material cost for the production of one member composed of concrete and rebar, the number of parts produced per week, and the quantity of production, and the mold cost can be calculated using the number of molds and the unit cost of the mold. Labor cost is calculated as the input manpower during the period in which the mold was used. The labor cost is calculated using the number of molds, the amount of manpower, the number of working days per week, and the labor unit cost (D). Equipment cost is classified into curing cost and crane rental cost (E). Curing cost is calculated using the number of molds and curing unit cost. Since the stack quantity is determined by the difference between the production quantity and the installation quantity occurring over time, the crane utilization period is calculated based on the quantity of the yard. The crane rental cost is determined by the crane utilization period, unit rental cost, and number of units. The total production cost for one mold type is calculated by summing the material cost, labor cost, and equipment cost. For the production of PC members, the total in-situ production cost is calculated by summing the production cost for each mold type. The calculated in-situ production cost is then compared with in-plant production to determine whether the cost is reduced.

## *4.2. Environmental Load Model*

The environmental load simulation model developed for one mold type is shown in Figure 9. From the perspective of in-plant production, environmental load is calculated by dividing into material use (B), oil use, electric use (C), and transport equipment use (D). First, in (A) of Figure 9, material use is calculated by summing reinforcement works of high-tensile deformed-bar (HD) 13, HD 16, HD 19, super-high-tensile deformed-bar (SHD) 22, SHD 32, SHD 35, embedded steel, etc. by checking the drawing to calculate the rebar weight of the member corresponding to mold type 1. Material use is calculated using this rebar weight, and the amount of concrete is calculated by checking the drawing, the actual steel form used, and the CO2 base unit of steel and concrete. Oil use and electric use are calculated using the total floor area and number of total members, while transport equipment use is calculated using trailer size, distance from site to factory, CO2 emission base unit of distance, and transport equipment use.

**Figure 9.** Environmental load simulation model (mold type 1 for in-plant production); material use (A), in-plant production material use (B), oil and electric use (C), transport equipment use (D)

Total CO2 emissions are calculated by summing material use, oil use, electric use, and transport equipment use. If the total CO2 emissions for one mold type of in-plant production are calculated, the total in-plant production CO2 emissions are calculated by summing the production cost for each mold type. CO2 emissions for in-situ production are calculated excluding the use of transport equipment in all processes. The calculated in-situ production CO2 emissions can be compared with the in-plant production to determine the degree of CO2 emission reduction.

## **5. Application of the Dynamic Optimization Model**

## *5.1. Estimation of Environmental Loads for the Case Study*

## 5.1.1. Estimation of In-Situ Production Quantity

A case study was selected to apply the developed dynamic optimization model. The case project is located in Cheonan-si, Chungcheongnam-do, Republic of Korea, and has a site area of 53,055.60 m2, building area of 42,406.07 m<sup>2</sup> (246 m long <sup>×</sup> 178 m width), and total floor area of 167,614.82 m2. The case project has a scale of four stories above the ground, with 2–4 stories above the ground with a PC structure, a core structure of reinforced concrete, and a roof structure of steel. Therefore, this study targets the 2–4 floors above ground because they are built with a PC structure. A total of 72 members are produced in-situ in the case project. The members to be constructed are columns, girders, and slabs. However, the members capable of in-situ production are limited to columns and girders that require a small production area. The columns and girders are thin and long, so the production space is not wide, but slabs require a large space, making in-situ production difficult in a limited area.

The number of PC members capable of in-situ production at the case site is 1,004 columns and 1252 girders, a total of 2256 members. Table 2 shows that the quantity per column and the 72 columns, as well as the quantity per girder, are calculated. As the resource input for column production is the same in in-situ and in-plant production, the quantity is calculated the same way. Since the material to be actually put in is the same, in-situ and in-plant production were calculated equally. It was confirmed that each quantity was almost proportional to the quantity of each column, as well as the number of columns. The reason is that all 72 columns have similar size and rebar details.


**Table 2.** Quantity of each column and the 72 columns.

Note. UHD: Ultra-high-tensile deformed-bar.

The material of the mold used in the case site is the same as in the steel mold used in the factory. As shown in Figure 1a, the steel mold applied in the in-situ production is the same as in the in-plant production, and the same specification was ordered for manufacturers and suppliers to the plant. The steel form is not used once, but is reused at least 50 times. Therefore, the amount of CO2 generated during one column's production is calculated and reflected. Steel molds have high durability and high cost. Therefore, if the number of uses is small and reuse is possible, resale is possible. In this paper, each mold was used 36 times and then resold. The quantity of steel forms per column can be calculated as follows. For 72 columns' in-situ production, the purchase cost of two molds was 24,942 USD, and after production, they were resold for 14,000 USD. In other words, the mold cost actually used as input for 36 columns' in-situ production was 5471 USD, and it was determined that 82 columns could be produced when converted. When the total weight of the steel form was 1.297 t and divided by the number of reuses (82 times), the steel form input for production of one column was calculated as 0.016 t.

## 5.1.2. Estimation of Environmental Loads (CO2 Emission)

The CO2 emissions for 72 columns, 1004 columns, and total members were calculated, as shown in Table 3, by using the previously calculated quantities for each column and each beam, as well as using the CO2 emission calculation formula. For 72 columns, they were calculated as 65,155 kg-CO2 for concrete, 505,947 kg-CO2 for steel, and 571 t-CO2 in total. Here, since the same amount of material was input by applying column members with the same size and reinforcement details, in-situ and in-plant production were calculated equally. In the case of in-situ production, if the total floor area of 167,615 m2 and Equations (6) and (7) were substituted, the CO2 emissions from oil use of total members were calculated as 987 t-CO2. When the areas occupied by the production of members and the mold area were converted into the sum of 72 columns, the emissions were calculated as 147 t-CO2. If they were calculated by substituting the same total floor area and using Equations (8) and (9), the CO2 emissions from electric use of total members were calculated as 543 t-CO2, and if 72 columns were calculated, they were calculated as 40 t-CO2.

**Table 3.** CO2 emission comparison of 72 columns, 1,004 columns, and total members (unit: t-CO2).


In the case of in-plant production, overhead and gantry cranes installed in the factory were used, and an area of 41,292 m<sup>2</sup> used in the production of the members was applied. The CO2 emissions from transportation equipment use were applied only in the case of in-plant production; the equipment used for transportation is a 25 t trailer, and the distance from the factory to the site is 97.55 km. That is, in the case of 72 columns, CO2 emissions from in-situ production were calculated as 758 t-CO2, and from in-plant production, they were 936 t-CO2, so CO2 emissions from in-situ production were reduced by 178 t-CO2 compared to in-plant production.

The CO2 emissions from 1004 columns were calculated as 14,470 t-CO2 for in-situ production and 16,891 t-CO2 for in-plant production, so in-situ production decreased emissions by 2421 t-CO2 compared to in-plant production. The quantity of 1004 columns and CO2 emissions are not 1004 times those of one column. The reason is that 72 columns were selected as the largest number of members, because the size of these members and the amount of reinforcement are smaller than those of other columns. As a result of calculating the CO2 emissions of total columns and girders (1004 columns, 1252 girders), CO2 emissions from in-situ production were calculated as 33,699 t-CO2, and from in-plant production, they were calculated as 39,095 t-CO2. CO2 emissions from in-situ production was reduced by 5397 t-CO2 compared to in-plant production, and it was confirmed that CO2 emissions increased as the quantity of in-situ production increased. The CO2 emissions of in-situ production without PC were reduced by more than 13.8% compared to in-plant production. Material use accounts for more than 61.0% of the total CO2 emissions, so the quantity has the greatest effect on the CO2 emissions. Among them, the environmental loads of the members increased as the amounts of reinforcing bars with large basic units of CO2 emissions increased.

Table 4 shows that the costs of in-plant and in-situ production with 72 columns applied were calculated by dividing them into material cost, labor cost, equipment cost, transport cost, and overhead and profit (O&P). The cost applied in this paper was the unit price contracted with the PC company and the cost input for actual in-situ production. Service costs were used for transportation costs, and data from PC factories were used for O&P. When calculating the cost, only direct costs, excluding overhead costs, are calculated. The reason for this is that even if the PC member is produced in-plant, as in in-situ production, on-site management costs and on-site land costs are required, so there is no need to additionally calculate additional overhead costs. In-plant production cost is 200,648 USD and in-situ production cost is 160,544 USD, which reduces construction cost by 40,104 USD for in-situ production compared to in-plant production. When 72 PC columns were produced in in-situ production, the cost of each column was reduced by 20% compared to in-plant production.


**Table 4.** Production cost analysis of in-plant and in-situ production.

## *5.2. Derivation of Control Range (Min–Max) by Factors*

A Monte Carlo simulation was performed using the dynamic optimization model developed in this paper. Through the simulation, the number of possible occurrences for factors such as in-situ production quantity, lead time, number of molds, and number of cranes was derived, and the Min and Max for each factor were derived. It was assumed that the construction time was within the allowable range, and this was considered to comply with the 18 months required by the owner and aimed to reduce environmental loads by 10% within the range of cost reduction by more than 10%.

As a result of deriving the management range of Figure 10a, for in-situ production quantity, the number of members was calculated as Min 1757 members and Max 2256. Column type 1 is Min 704 and Max 736, column type 2 is Min 239 and Max 268, beam type 1 is Min 60 and Max 96, beam type 2 is Min 637 and Max 960, and Beam Type 3 is Min 117 and Max 196. The reason for why the number of members does not decrease below a certain number is that the cost reduction ratio increases as the number of members increases. In Figure 10b, as a result of deriving the Min–Max range of lead-time, Min was set to 3.8 weeks and Max was set to 8 weeks. The reason for why the lead-time does not decrease to less than 3.8 weeks is that only PC members that have been produced can be installed, so it is necessary to secure the production time of the members.

As a result of deriving the Min–Max range of Figure 10c, the number of molds was set to Min 37 units and Max 70 units. Column type 1 is Min 11 and Max 20, column type 2 is Min 5 and Max 10, beam type 1 is Min 2 and Max 7, beam type 2 is Min 16 and Max 26, and beam Type 3 is 3 Min and 7 Max. The reason that the number of individual molds does not decrease below a certain number is that if the number of molds decreases, the timely completion of the in-situ production is impossible. In Figure 10d, as a result of deriving the Min-Max range of the crane, Min was set to two units and Max was set to four units. The number of cranes is the result of calculating the in-situ production cost by rounding up the decimal point of the value derived by the simulation. If a crane is added during construction, it is advantageous to shorten the construction time, but this did not increase above a certain level because the cost and CO2 emissions increase. Since the crane was partially added during construction, the number after the decimal point was derived.

In Figure 11, as a result of deriving the management range of in-situ production cost, Min was set to 615 USD and Max was set to 775 USD. In the case of CO2 emission reduction, Min was set to 4557 t-CO2 and Max was set to 5622 t-CO2. That is, it is possible to save more than 14.38%. These values are the results derived for Min 1757 members and Max 2256 members. As a result of analyzing the management scope as described above, the cost and CO2 emissions according to the fluctuations of each factor are generally proportional to each other; thus, a graph of similar shape was found. Using simulation, project participants can predict the management range for in-situ production volume, number of molds, lead-time, and number of cranes under various conditions. In addition, the management scope can be changed according to the site conditions.

**Figure 10.** Control ranges of influencing factors: (**a**) in-situ production quantity; (**b**) lead-time; (**c**) number of molds; (**d**) number of cranes.

**Figure 11.** Control range of cost and CO2 emissions: (**a**) in-situ production cost; (**b**) CO2 emission reduction.

#### *5.3. Optimization for Estimating In-Situ Production Quantity*

The optimal case is derived by utilizing each management range, such as in-situ production volume, lead-time, mold number, and crane number, which were derived through the Monte Carlo simulation. Lim et al. explained six assumptions about the available area, which is selected as the main influencing factor, and derived the highest cost reduction rate among the scenarios applicable to the case project [20]. However, in this study, factors influencing CO2 emissions are selected and the Min–Max of the influencing factor is derived through simulation. From the derived control range, an optimal value is derived to reduce environmental loads by 10% within a range that can reduce costs by 10% or more.

For simulation, it is assumed that the 18 months required by the client are observed. Table 5 shows the values of the influence factors corresponding to the highest value of the CO2 emission reduction ratio of 14.58%. The quantity is 1,757 members, accounting for 78% of the total quantity, lead-time is 3.8 weeks, number of molds is 37 units, and number of cranes is two units. In-situ production reduced costs by 6,966,000 USD and CO2 emission by 7557 t-CO2 compared to in-plant production. In-situ production costs can be reduced by 10.53% compared to in-plant production. The proposed dynamic optimization model can derive an optimal case in consideration of the CO2 emission reduction ratio and the cost reduction ratio. In addition, it can support decision-making on whether or not in-situ production can be applied to the field.


**Table 5.** Optimization result of estimating in-situ production quantity.

## *5.4. Discussion*

The construction industry has the potential to improve productivity by applying prefabrication for building [44]. However, automation and robotics are still not common in the construction field [44,45]. Prefabrication has been plagued by dependence on conventional methods [46], complex interfacing [47], scheduling complexity [48], underutilization of factory space [44], cost barriers [49], fragmented information [50], and inconsistent quality [51].

The in-situ production proceeds in the order of installation of rebar setting, pouring concrete, curing, and stacking [16], and is performed at the same level as in-plant production. After stacking the cured PC members, quality checks and finishing work are done by the manpower if there is partial damage to the surface. In other words, PC members are manufactured at the factory, but the production process is mostly performed by manpower rather than automation or robotics. In the case of in-plant production, there is no actual mechanization except for the use of overhead and gantry cranes in the factory. It was assumed that the idle time of the crane used for erection of PC members is used during in-situ production in this study. In general, although the same sizes of PC columns and beams are designed, the rebar arrangements are not designed with the same PC members. In order for PC members to be manufactured, checked, and installed in the same manner as in the design drawing, work by manpower is required. Therefore, to apply in-situ production in the future, it is required to design PC members with the same size and reinforcement details.

Some of the members that have been produced in-plant have had to be reproduced due to issues with cracks, size, and breakage, as well as rebar arrangements that are different from the drawings. According to an interview with a PC factory official, it was confirmed that if the factory owner does not obtain more than 20% of the production cost as a profit, the official does not contract, because this does not cover the factory management overhead [17]. Through several studies, it has been possible to secure the quality and economic feasibility of in-situ production [14–17,19]. This study examined in-situ production at a construction site that complied with the PC production and installation guidelines of the KCI (Korea Concrete Institute); it was confirmed that there were no problems with cracks, breakages, size, or strength quality standards.

This study focused only on the factors of in-situ production quantity, lead-time, number of molds, and number of cranes with respect to CO2 emissions. It was assumed that the production time and area were sufficient. However, if various field conditions, such as time and production, are additionally considered, different results will be derived. In-situ production can be applied only to sites where the production area is secured due to concerns about interference with ongoing construction, interference around the site, and safety issues.

In this study, a simulation for in-situ production was performed only for PC columns and beams. This means that slender parts, such as columns and beams, can be produced on-site. PC slabs, such as double-T and rib-plus slabs, occupy a large space during production, so in-plant production is more advantageous than in-situ production [18]. That is, in-plant production can be more advantageous depending on the type and number of members. Long-span and heavily loaded buildings, such as the case project examined in this study, can easily secure a production area due to the long distance between columns. However, different results from this study can be derived depending on the site characteristics and the building size.

## **6. Conclusions**

This paper presented a dynamic optimization model that can dynamically predict, control, monitor, and manage five major influencing factors that affect environmental loads during in-situ production. In order to minimize environmental loads, the in-situ production quantity can be adjusted within the range of fluctuations in influence factors. The effect of the model was verified through the simulation results, and the findings are as follows.

First, the optimization model easily and quickly calculated the range of changes in environmental loads by analyzing the dynamic relationship of influencing factors. According to the simulation of the optimization model, in-situ production can reduce a Min of 4557 t-CO2 and Max of 5622 t-CO2 compared to in-plant production, which resulted in saving more than 14.58%. Second, the optimization model uses simulation results to create the control range of each influence factor to achieve the target environmental load reduction rate within the target cost reduction range. In the case project, the control range of the in-situ production quantity was 1757–2256 members to achieve the target environmental load reduction rate of 10% or more. In other words, project participants can predict the extent to which influencing factors are managed under various conditions. Third, the proposed dynamic model supports decision-making as to whether in-situ production can be applied in the field. The derived optimal case to minimize the environmental loads from the case project resulted in a quantity of 1757 members through the simulation model. In-situ production was able to reduce environmental loads by 14.58% and cost by 10.53% compared to in-plant production. Fourth, material use accounts for more than 61.0% of total CO2 emissions, so its quantity has the greatest effect on CO2 emissions. As the amount of reinforcing bar with a large basic unit of CO2 emissions increases, the environmental loads of the member increase. Overdesign should be avoided.

The model developed in this paper can control the influence factors of in-situ production and can easily and quickly simulate the influences of factors that change during project execution to derive the optimum value according to the situation. When applying in-situ production using this model, environmental loads of CO2 emissions can be calculated at the design stage. Furthermore, it is possible to evaluate whether it is applicable at the initial stage of the project in order to establish and review a construction plan. All these values can be used to analyze economic and environmental feasibility. In addition, they can be used to develop a data management system and build a risk management model for analyzing environmental loads of in-situ production in the future. Further research is needed to calculate the optimal in-situ production quantity considering environmental load through more field applications.

**Author Contributions:** Conceptualization, J.L.; methodology, J.L.; validation, J.L.; formal analysis, J.L.; investigation, J.L.; data curation, J.L.; writing—original draft preparation, J.L. and J.J.K.; writing—review and editing, J.L. and J.J.K.; visualization, J.L.; supervision, J.J.K.; project administration, J.L and J.J.K.; funding acquisition, J.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MOE) (No. NRF-2019R1A6A3A12032427).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

## **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article*
