Bayesian Belief Network Analysis for Chinese Off-Site Manufacturing Risk

Abstract

The construction industry continues to face challenges such as increased costs, time overruns, and low quality. Off-site construction (OSC) methods are increasingly being adopted as alternatives to traditional construction practices to address these issues, with off-site manufacturing (OSM) representing a key difference in construction methods. However, existing studies have largely neglected the systematic evaluation of OSM risks on quality, cost, and delivery (QCD) outcomes, leaving a significant gap in understanding the complex interdependencies among risk factors. To improve risk management in OSC projects, it is crucial to evaluate the impact of OSM risks on QCD outcomes. This study applies the Bayesian Belief Network (BBN) method to develop an evaluation model that measures the impact of OSM risks on QCD outcomes in OSC projects. The results identify 12 significant risk factors affecting QCD outcomes in OSC projects. Five key risk groups were identified as critical for managing OSM risks. This approach provides a systematic framework for managing OSM risks and optimizing OSC practices in China.

1. Introduction

Off-site construction (OSC), also known as modular construction, has gained attention in recent years [1]. It differs from traditional construction by involving off-site module prefabrication, component transportation, and on-site component installation [2]. Compared to traditional construction methods, OSC offers significant savings in cost, time, quality control, waste reduction, and improved safety [3].

In that case, many countries, such as Singapore [4], the United States [5], the United Kingdom [6], China [7], and Australia [8], have adopted OSC as an alternative construction method. As off-site construction gains more attention, there is increasing demand for guaranteed quality, cost, and delivery (QCD) in OSC projects [7,9,10]. To maintain high QCD standards in OSC projects, it is essential to effectively identify and manage OSC risks [11].

The OSC project process includes design, off-site manufacturing, transportation, and on-site assembly, requiring joint efforts from owners, consultants, manufacturers, transporters, and contractors [12]. This interrelationship means that problems caused by one stakeholder can significantly impact the entire OSC project. As OSC shifts the on-site construction process to off-site manufacturing, the manufacturer’s role becomes crucial from the outset. Given the persistent challenges in OSC projects, off-site manufacturing (OSM) risks must be considered [13,14]. Therefore, it is crucial to understand how OSM risks influence QCD in OSC projects and identify the significant OSM risk factors that need to be analyzed.

Risk management has significantly improved in recent years. For example, Natural Language Processing with SGD has been used to automatically estimate risk levels from text in marble quarries for effective risk assessment [15], and a GPT-based AIR Agent has been developed to automatically analyze subway construction accident reports, enhancing analysis efficiency [16].

Previous studies on OSC risk and risk management mainly focus on stakeholders such as owners [17], consultants [18], and contractors [19]. However, few studies have considered OSC risks from the perspective of manufacturers. This indicates that most existing OSC risk and risk management studies lack effective methods to evaluate the impact of OSM risks on QCD in OSC projects. To address this gap, this research employs a Bayesian Belief Network (BBN) model to evaluate the impact of OSM risks on QCD in OSC projects.

This research methodology involves three key steps. First, semi-structured interviews were conducted with key Chinese OSC practitioners to identify risk factors in the OSM process. Second, a questionnaire was developed based on the identified risk factors, allowing for the determination of significant risks through quantitative analysis. Third, a BBN model was built using the collected data to evaluate how these significant risks influence QCD in OSC projects.

2. Materials and Methods

2.1. Literature Review

A narrative literature review was conducted to identify and analyze previous studies on risk management in OSC. Electronic databases such as Web of Science, Scopus, and Google Scholar were searched using predefined keywords (‘off-site construction’, ‘risk management’, ‘OSM’, ‘quality, cost, delivery’). This narrative approach provided a solid basis for identifying research gaps and justified the subsequent analysis using the BBN model.

2.1.1. Risk and Risk Management in OSC

Risk management is one of the ten key areas in project management [20]. Compared to traditional construction methods, OSC involves distinct risks but offers advantages such as cost, time, and waste reduction, as well as improved safety and quality control [3]. OSM is unique to OSC projects, making its risks and management distinct from those in traditional construction [21]. For example, OSM requires a higher initial capital outlay for manufacturing facilities [22], the component design is fixed during the early development stage [23], and transporting large components becomes increasingly challenging for logistics providers [24].

Many studies have explored risk management practices in OSC projects [3,25,26]. Lee, Jang [27] identified 12 critical risk factors and 50 management activities in OSC, highlighting a significant performance–importance gap that demands intensive management. Hussein, Eltoukhy [28] demonstrated that BIM technology can significantly improve safety management in modular construction through enhanced training, inspection, and site optimization. Chatzimichailidou and Ma [29] identified supply chain risks in the OSC process from 309 articles through a mixed review method. Risk management in OSC projects is generally divided into three steps: risk identification, risk analysis, and risk response [25]. Enshassi, Walbridge [30] proposed a tolerance-based risk framework for modular construction to manage geometric variability and reduce rework, cost overruns, and schedule delays. Li, Al-Hussein [3] identified key modular construction risk factors using fuzzy AHP and simulation, revealing cost and schedule impacts. Darko, Chan [31] reviewed BIM-based risk management in modular integrated construction, examining current integrations and suggesting future automation- and safety-focused research directions. Vithanage, Sing [32] reviewed OSM safety and classified risks under human, organizational, and environmental factors.

This research focuses on evaluating the influence of OSM risks on QCD in OSC projects, with a particular emphasis on risk analysis methods. However, previous studies do not specifically identify or classify OSM risk factors. Therefore, this research investigates several risk factors that occur in the OSM process and analyzes how they influence OSC projects.

2.1.2. The Importance of QCD in OSC Projects

One risk factor could influence at least one project objective [33]. By impacting project objectives, the risk factor may cause significant project issues. Evaluating the influence of risk factors on project objectives is crucial for effective problem control. Evaluating the influence of risk factors helps to identify significant risks and assess how they affect OSC projects, allowing project managers to respond more efficiently. To understand how risk factors influence OSC projects, the concept of QCD is introduced as a project performance criterion. Quality refers to the level of adherence to design and performance standards, cost refers to the efficiency in budget management, and delivery refers to the timeliness of completing and handing over project components. QCD is commonly used as a measure of project success. Atkinson [34] highlighted the importance of QCD for project success, often referred to as the Iron Triangle. Thomas, Barton [35] proposed QCD as a measure of Six Sigma effectiveness in manufacturing. Kojima and Amasaka [36] developed a quality assurance model to prevent defects and achieve QCD in projects.

QCD is often regarded as a direct result of effective project management in many construction projects. Tayur, Ganeshan [37] emphasized the importance of teamwork, cooperation, and effective coordination throughout the construction supply chain. Carpenter and Bausman [38] analyzed the QCD of a public school construction project over a two-year period to understand the inherent risks. Risk factors closely linked to QCD are considered significant. These significant risk factors have a major impact on the success of OSC projects. Given the relationship between OSC stakeholders, significant risk factors from one stakeholder can affect those of others [39]. To understand how significant risk factors influence the OSM process, it is essential to evaluate their impact on QCD in OSC projects.

2.1.3. Risk Analysis in OSC

The traditional probability and impact (P-I) model has been widely used to assess various risks in OSC project management. For example, Arashpour, Abbasi [40] defined risk using the P-I method for theorizing risk analysis in hybrid projects. Wu, Xu [17] used the P-I method to analyze risks in OSC-integrated design and construction project delivery. Yang, Pan [41] used the P-I method for risk analysis in OSC logistics for high-rise building projects. According to the Project Management Body of Knowledge (PMBOK) [42], the P-I model is a standard technique for risk analysis, where each risk is assigned a likelihood (probability) and a potential effect on project objectives (impact). By multiplying or otherwise combining these two dimensions, project teams can prioritize risks, allocate resources, and develop appropriate response strategies.

Other risk analysis methods have also been adopted in OSC projects. Monte Carlo simulation is commonly used for risk estimation. Rausch, Nahangi [43] employed this method for tolerance analysis to reduce the risk of component rework. The multi-criteria decision-making (MCDM) framework has been used for complex decision-making processes, and Shahpari, Saradj [44] developed an MCDM framework to compare the productivity of OSC and traditional construction. Decision support systems (DSS) were used by Wuni and Shen [45] to review the determinant factors in deciding to use OSC methods. The Analytical Hierarchy Process (AHP) is another widely used approach for prioritizing decision alternatives. Li, Al-Hussein [3] deployed AHP to identify risk factors affecting project cost and duration in OSC projects.

2.1.4. BBN in Construction Risk Analysis

The Bayesian Belief Network (BBN) is widely used in risk management due to its ability to graphically represent conditional dependencies between variables [46,47,48]. For example, Zhou and Zhang [49] developed a BBN model for deep foundation pit construction projects to evaluate the risk probability in China. Kabir, Balek [50] used the BBN to prioritize buried infrastructure for maintenance, rehabilitation, or replacement. Xue and Xiang [51] developed a BBN model to analyze key social risks in high-speed rail projects in China.

In the construction project, a BBN could present the interrelationship between each risk factor and update when new information is available [52]. A BBN can combine both objective data and subjective data, which can improve the quality of input data, especially when the historical data are limited and difficult to obtain [53]. Hon, Sun [54] indicated that a BBN to risk management still has room for improvement in dynamic risk management, which should cover all stages of the project.

Although the BBN has been widely applied to manage risks in construction-related research, few articles have used the BBN to analyze risks in OSC construction projects, especially for the OSM process. Although Yu, Man [12] deploys a BBN model for OSC projects, it focuses on stakeholder impacts regarding quality issues, not on OSM risk management. Considering the clear advantages of a BBN, this research uses a BBN model to analyze OSM risk factors and the relationship between QCD and OSM risk factors.

2.2. Methods

In the previous step, a literature review was carried out to compile a preliminary set of risk factors that potentially affect the QCD outcomes in OSM.

Table 1. OSC risk group and risk factor.

Risk Group	Risk Factor	References
Cost	High cost of initial investment	[55]
	High design costs	[56]
	High transport costs	[56]
	High manufacturing costs	[57]
Culture	There is a bias that OSC can only produce low-cost products	[56]
Flexibility	Clients and designers might change their demands	[57]
	The design must be frozen	[58]
Health and safety	Regular mobile cranes are unsuitable for off-site components	[56]
	Heavy OSC components may increase hazards in the event of an earthquake	[25]
Knowledge	Lack of adequate knowledge of OSC	[25]
Supply chain	Components from a foreign country may not comply with the standards	[56]
	OSC developers choose certain suppliers only	[57]

below presents these factors, categorized by risk group. Many of these items may influence multiple aspects of QCD, thus reflecting the range of challenges identified in existing studies.

To achieve the objectives of this study, a Bayesian Belief Network (BBN) model was employed. The BBN has been widely applied to risk management in various fields, including computer engineering [59], civil engineering [60], economics [60], and disaster prevention [61]. In traditional construction, the BBN has proven effective in managing risks during the construction process [62,63,64]. Since OSM involves a multi-stage production process similar to traditional construction, the BBN is a suitable method for managing OSM risks. Additionally, the BBN can effectively analyze complex, multi-factor systems with interdependent risk factors, making it an ideal approach for evaluating the relationship between risks and QCD outcomes in OSC projects. Therefore, a BBN was selected as the primary method for this research.

Figure 1 presents the research flow chart of this study. The study followed a systematic approach consisting of three primary steps:

Literature Review: The existing literature on OSC risk factors was analyzed to identify the most relevant OSM risk factors and compare existing risk management methods.

Data Collection: Interviews and questionnaires were conducted with OSC practitioners to identify significant OSM risk factors.

BBN Model Development: A BBN model was developed based on the identified significant risk factors, and sensitivity analysis was conducted to evaluate the relationship between OSM risks and QCD in OSC projects.

3. Results

3.1. Risk Collection and Validation

3.1.1. Risk Collection—Interview

According to Purdy [65], OSM risk is defined as an uncertain event or condition that, if it occurs, may positively or negatively impact one or more objectives of an OSC project. In this context, OSM risks can result in changes to QCD in the OSM process, such as increases or decreases in the QCD of OSC manufacturing production. The literature on OSM process risks is limited and incomplete [66]. However, as the OSM process is part of the broader OSC project, risks in OSC projects can serve as a reference for identifying OSM process risks [67]. This research tried to identify the OSM risk by recognizing the OSC risk that may happen in the OSM process and influencing the QCD of the OSM process.

The purpose of this phase was to identify the risk factors in the OSM process. Based on Table 1, semi-structured interviews were conducted with key members of the OSC project team. Since OSC projects require cooperation among all participants, OSM risks may arise not only from the OSC manufacturer but also from other stakeholders. Therefore, the interviewees in this study included owners, consultants, manufacturers, transporters, and contractors. This approach is consistent with the methods used in similar studies focusing on OSC risks and risk management [7,21,68]. A total of 25 managers and engineers from various OSC projects were interviewed to identify the OSM risks for further analysis.

The data collected during the interview process were analyzed using NVivo 14 software. A total of 61 risk factors were identified as associated with the OSM process risks. Risk types in this study are classified into three categories: Internal risks, which are internal factors inherent to the manufacturing process. Participant risks relate to the interactions and coordination among various stakeholders. External risks relate to external environmental conditions.

Table 2. Risk factors for OSM process.

Risk Type	Risk Group	Risk Factors
Internal risks	Cost	High off-site manufactory building costs
		Overall working process increase
		Components paid after delivery
		High training costs
		More types of workers
		More regular employees
		High component model and material fees
	Off-site feature	Component is too heavy
		Component support interference
		On-site assembly limitations
	Project management	Lack of cooperation
		On-site construction period adjustment or change
		Lack of risk management method
		Hard to deploy new management method
	Time	Complexity of joint assembly
		Insufficient production time
Participant risks	Owner	Owner changes demand
		Demands not suitable for off-site project
		Owner changes project partners
	Consultant	Design error
		Consultant lacks off-site experience
		Consultant lacks on-site experience
		Consultant lacks responsibility
		Consultant lacks standardization
		Consultant lacks suitable professional software
	Manufacturer	Manufacturer lacks experience
		Manufacturer lacks employees
		Manufacturer lacks responsibility
		Unavoidable errors in component production
		Assembly line lacks control
		Lack of off-site manufactory facilities and equipment
		Producing different types of components at the same time
		Insufficient component yard storage
		Insufficient assembly line
	Transporter	Transporter lacks experience
		Transporter lacks responsibility
		Mistakes arise during the transfer process
		Transportation road problems
		Long transport distance
	Contractor	Contractor has a lack of willingness to participate in off-site project
		Contractor lacks experience
		Contractor lacks responsibility
		Contractor lacks experienced employees
		Lack of on-site assembly standardization
External risks	Environment	Geographical environment
		Manufactory indoor environment
		Seasonal changes
		Natural disaster
	Government policy	Lack of government policy standards
		Local government policy standard differentiation
		Rigid prefabricated rate requirement
		Lack of subsidy and support
	Resource	Low material quality
		Supply delay or not on time
		Lack of material
		Component model lacks standardization
		Manufactory equipment damage
	Society	Contract bidding problem
		Unstable economic situation
		Public society’s prejudice for off-site building
		Inconsistent quality demand for OSC project

presents the OSM process risk that affects the OSC project.

3.1.2. Risk Validation—Questionnaire

After the OSM process risk factors were identified, the next step was to determine which factors were the most significant. A five-point Likert scale questionnaire was then used to validate the OSM risks. The questionnaire was carefully designed based on the 61 risk factors identified in the interviews to assess the significance of each risk factor.

The snowball sampling method was used in the questionnaire process. Similar to the interview process, all OSC participants were considered appropriate targets for the questionnaire data, including owners, consultants, manufacturers, transporters, and contractors. The questionnaire consisted of two sections: personal data and the identification of significant risks. The first section asked respondents to provide information on their experience, company type, and the materials used in their OSC projects. The second section included the OSM process risk factors identified during the interview process and asked respondents to assess the probability and impact of each risk using a five-point Likert scale.

The Statistical Package for Social Science (SPSS, version 29.0) was used to analyze the data generated from the questionnaire. A total of 438 copies of the questionnaire were distributed to Chinese OSC practitioners, who were then asked to distribute the questionnaire to other OSC practitioners. A total of 120 questionnaires were received from the first-tier respondents. After one month of data collection, a total of 436 questionnaires were returned to the researcher as part of the snowball sampling process. More than 64% of respondents have more than 5 years of experience in the construction industry. Although only 26% have more than 5 years of experience in the OSC industry, considering OSC was vigorously promoted in China after the 13th Five-Year Prefabricated Building Action Plan in 2016, it is acceptable for many of the respondents to have less than 5 years of experience in OSC industry [69]. The types of respondents’ company types show that the respondents’ company types are, on average, between the owner (17%), consultant (28%), manufacturer (28%), transporter (4%), contractor (36%), and others (7%); this confirms the reliability of collected data for identifying significant OSM risk factors for a variety of angles. Moreover, most of the respondents (84%) have been involved in the concrete structure OSC project, which implies that this research focuses mainly on the concrete structure OSC project.

The collected data were statistically analyzed to determine the mean values for the 61 OSM risk factors. Since the questionnaire considered both the probability and impact of each risk factor, those factors with mean scores for both probability and impact that were higher than the average mean scores (3.30 for probability and 3.34 for impact) were identified as significant risks. This dual-measurement approach enables the identification of risks that consistently exhibit high levels in both dimensions, which are more likely to have a sustained effect on project performance. The Cronbach’s alpha value was calculated, and the result was 0.991, which exceeds the 0.7 threshold, indicating that the five-point scale measurement was reliable at the 5% significance level [70].

A total of 28 valid significant risk factors were identified, as shown in

Table 3. Significant OSM risk factors.

Risk Type	Risk Code	Risk Factor	Risk Factor Explanation
Internal risk	IR1	High off-site manufactory building costs	The high capital investment for off-site manufacturing, especially manufactory building costs, is a major cost for off-site manufacturers.
	IR2	High component model and material fees	As the component model and material lack standards, the cost is higher than the set price.
	IR3	On-site construction period adjustments or changes	On-site period change results in the manufacturer having to change their production plan.
	IR4	Lack of cooperation	Cooperation includes cooperation between project participants and cooperation inside the manufacturing facility.
	IR5	Lack of risk management methods	The new process in the OSC project requires a new risk management method.
	IR6	On-site assembly limitations	On-site assembly limitation includes extra support for the component, extra steps for the on-site assembly process, and interference for the construction workers.
	IR7	Components paid after delivery	Manufacturer cannot get their funds until the project is finished, which increases their costs.
	IR8	Component is too heavy	Heavy component causes all transport processes to require more time.
	IR9	Complexity of joint assembly	The component joint assembly is still a new technology that needs more technical support.
	IR10	Hard to deploy new management method	Project management methods like Six Sigma and lean production are still relatively new for off-site manufacturers, which requires more time to establish these methods.
Participant risk	PR1	Consultant lacks off-site experience	The consultant has little knowledge about OSC process, which causes the design diagram to be unsuitable for off-site component production.
	PR2	Consultant lacks on-site experience	The consultant does not need to go on-site to learn how to work on-site, which causes the design diagram to be unsuitable for the construction project.
	PR3	Owner changes demand	The owner can change their demand during a traditional construction project. However, the feature of the OSC project result’s changed demands require more time and costs.
	PR4	Demands not suitable for off-site project	Some owners still use traditional construction requirements for OSC projects.
	PR5	Contractor lacks experienced employees	There are two reasons for contractor’s lack of experienced employees. First, there are a few experienced OSC workers. Second, young people are unwilling to become on-site workers.
	PR6	Contractor lacks experience	Only a few contractors have experience with OSC projects.
	PR7	Design error	Design error is caused by a consultant, which include conceptual design errors and design development errors.
	PR8	Consultant lack of standardization	The consultant’s lack of standards results in other participants needing to change in different projects.
	PR9	Manufacturer lacks experience	The off-site manufacturer experience includes product experience, transport experience, manufactory design experience, etc.
	PR10	Lack of on-site assembly in standardization	Lack of on-site assembly standardization causes the on-site assembly time to extend, which leads to the manufacturing times to extend.
External risk	ER1	Contract bidding problems	Current construction contract bidding is the lowest price win the bid, which means the quality may be relatively low.
	ER2	Lack of subsidy and support	Although the government provides subsidies and support for OSC, many companies still think the support is insufficient.
	ER3	Component model lacks standardization	Different component model companies have different standards, which means the components from different component models cannot be assembled.
	ER4	Public society’s prejudice against off-site building	OSC projects require much less time for on-site processes than traditional construction. However, many people think it is too quick; it must not be safe to live in these.
	ER5	Inconsistent quality demands for OSC project	OSC could increase the quality of a building. However, as the off-site manufacturer has a similar production environment as a general manufacturer (car, phone, etc.), some people think the OSC project has a similar quality to a general manufacturer.
	ER6	Rigid prefabricated rate requirement	The government policy gives the off-site company a certain requirement for a prefabricated rate. However, some buildings are not suitable for OSC; to reach the prefabricated rate, the OSC company has to pay extra costs.
	ER7	Unstable economic situation	As the trade war and COVID-19 happened in recent years, the economic situation is unstable, which caused OSC project reduction.
	ER8	Local government policy standard differentiation	Different provinces have different policies, which leads to an OSC company having to change its operation method in a new province.

. These 28 significant risk factors were grouped into three types: internal risk, participant risk, and external risk.

3.2. BBN Model Analysis

After the significant OSM risk factors were identified, the BBN model was employed. Since the effects of the OSM risks are not always simple or direct, the relationships between different OSM risks need to be visually represented for further analysis [25]. The BBN model was constructed to illustrate these relationships.

3.2.1. BBN Model Development

The data collected from the interviews and questionnaires were analyzed using Netica v5.18 software, a type of BBN decision-making tool. By presenting a visual representation of the interactions between the significant risk factors in the OSM process, the relationships and levels of risk could be determined. It also linked the risk factors to the QCD outcomes to assess how these risks lead to cost overruns, quality issues, and time delays in OSC projects.

Figure 2 presents the structure of the BBN model, which is based on the interview results. The interviewees were asked to verify the relationships between the risk factors and QCD. This BBN model consists of 31 nodes, including 28 risk factors that could impact the OSM process objectives—low quality, extra cost, and time delays in OSC projects—and three nodes representing quality, cost, and delivery (QCD). The rectangles represent nodes, which denote the OSM risk factors or QCD, and the arrows represent edges, reflecting the causal relationships between different nodes. This figure illustrates that all significant risk factors can directly or indirectly affect the QCD of the OSM process. The risk factors that directly affect QCD are referred to as proximal factors, while those that indirectly affect QCD are referred to as distal factors.

However, Figure 2 only presents the relationship between the significant OSM risk factors and QCD outcomes in OSC projects; it does not show the effects of each node’s state on the distribution of the other factors. To capture these dynamic interactions, each node in the Bayesian Belief Network was assigned discrete states (e.g., low, medium, or high) that represent different levels of occurrence or impact. In this framework, the Conditional Probability Table (CPT) is not only used to illustrate the static relationships between nodes but also serves as the basis for probabilistic inference and risk propagation throughout the network. To demonstrate the impacts between risk factors and QCD, each node must be assigned an appropriate state. In this research, a CPT was used to reflect the strength of the edges between directly connected nodes [71]. A CPT is a table that provides probabilities for every possible combination of the parent and child states. For example, if node B is a child of node A, the CPT includes a table showing whether node A has occurred or not, along with the corresponding probabilities of node A’s state.

In this research, Figure 2 was presented to the 25 experts who participated in the previous interviews. These experts were asked to identify the conditional probabilities for each node and fill in the CPT for each node.

Table 4. Example for CPT.

IR7
Parent Nodes If		Child Node
IR4	ER7	Low	Medium	High
Low	Low
Low	Medium
Low	High
Medium	Low
Medium	Medium
Medium	High
High	Low
High	Medium
High	High

presents an example of a CPT, where IR7 is the child node, and IR4 and ER7 are the parent nodes. This table illustrates the probability evaluation of the relationships between the nodes. To assist the experts in identifying the nodes, each node was assigned three states: low, medium, or high, representing the probability levels for each node. The probability levels are typically categorized into three ranges: low probability (less than 10%), medium probability (between 10% and 20%), and high probability (greater than 20%). These probability levels are used to assess the likelihood of various risk events occurring [72].

After review, the BBN model for OSM risks and QCD outcomes in OSC projects was developed, as shown in Figure 3. In Figure 3, each node has four characteristics essential for the BBN model: node name, node status, relationships with other nodes, and the CPT.

To assist the OSM project manager in evaluating the effects of each node’s state on the distribution of other nodes, a sensitivity analysis will be presented in the next section.

3.2.2. Sensitivity Analysis for BBN Model

After the BBN model was constructed, a sensitivity analysis was conducted. The sensitivity analysis measures the degree to which any node can influence the beliefs of another node [73].

Table 5. Quality sensitivity analysis.

Node	Entropy Reduction	Percent	Variance of Beliefs
ER5	0.11969	8.07	0.0257942
ER1	0.10619	7.16	0.0214719
ER4	0.10141	6.84	0.0218519
PR9	0.08731	5.89	0.0187870
IR5	0.08731	5.89	0.0187870
IR10	0.05499	3.71	0.0120208
Delivery	0.01980	1.34	0.0041432

,

Table 6. Cost sensitivity analysis.

Node	Entropy Reduction	Percent	Variance of Beliefs
IR2	0.16912	11.4	0.0329685
IR1	0.13547	9.12	0.0310714
PR7	0.02009	1.35	0.0036497

and

Table 7. Delivery sensitivity analysis.

Node	Entropy Reduction	Percent	Variance of Beliefs
PR9	0.23067	15.3	0.0523023
IR5	0.23067	15.3	0.0523023
PR3	0.16621	11	0.0333913
IR10	0.14536	9.66	0.0328930
IR3	0.09125	6.06	0.0175443
IR4	0.08042	5.34	0.0159779
IR7	0.03508	2.33	0.0069591
Quality	0.01980	1.32	0.0039923

show that the risk factors ER5, ER1, ER4, PR9, IR5, and IR10 were identified as sensitive quality-related factors, each having a relatively high degree of impact on the occurrence of low quality. Risk factors IR2, IR1, and PR7 were identified as sensitive cost-related factors, each having a relatively high degree of impact on the occurrence of extra costs. Risk factors PR9, IR5, PR3, IR10, IR3, IR4, and IR7 were identified as sensitive delivery-related factors, each having a relatively high degree of impact on the occurrence of time delays. Quality and delivery were identified as critical factors that significantly affect the results of the final OSM production. Therefore, more attention should be given to quality and delivery in order to reduce risks in OSC projects.

In this research, IR1, IR2, IR3, IR4, IR5, IR7, IR10, PR3, PR7, PR9, ER1, ER4, and ER5 were highlighted as significant OSM risk factors that can substantially affect the QCD outcomes of the OSM process. Based on the BBN model, these factors were also identified as sensitive nodes, underscoring the importance of these risk factors in OSC projects. Quality and delivery were identified as critical factors affecting the objectives of the OSM process, and many of these sensitive factors were directly or indirectly linked to the two critical factors.

4. Discussion

To specify how the significant OSM risk factors affect the OSM process, the viewpoints of the interviewees are summarized. Based on different risk types, all significant and sensitive risks are analyzed in the following section.

Table 8. Sensitive risk factor.

Node	Risk Group	Risk Factor
IR1	Cost	High off-site manufactory building costs
IR2		High component model and material fees
IR7		Components paid after delivery
IR3	Project management	On-site construction period adjustments or changes
IR4		Lack of cooperation
IR5		Lack of risk management methods
IR10		Hard to deploy new management method
PR3	Owner	Owner changes demand
PR7	Consultant	Design error
PR9	Manufacturer	Manufacturer lacks experience
ER1	Society	Contract bidding problems
ER4		Public society’s prejudice against off-site building

presents the risk group of each sensitive risk factor.

Many studies highlight the importance of costs in OSC projects [8,10]. This research emphasizes that cost also has a significant impact on the OSM process. For manufacturers, a large initial investment leads to inflexible capital turnover. For example, material costs, component model fees, and factory building costs make the initial investment for the OSC method considerably larger than for traditional construction methods. Additionally, the ’pay after delivery’ convention, which originates from traditional construction practices, exacerbates the cost problem. This financial pressure results in cost challenges for OSM, which significantly affects the QCD outcomes of OSC projects.

As mentioned in the literature review, few articles have considered OSM risks and risk management, which has led to project management risks in the OSM process. Compared with traditional construction, OSC involves an additional OSM process. Conventional risk management methods, which are designed mainly for on-site construction, do not address the unique challenges presented by OSM and are, therefore, not suitable for OSC projects. Traditional construction risk management methods are not suitable for OSC projects; however, many OSC companies still use traditional construction risk management methods to manage OSC project risks. The cost of new risk management methods for OSC projects has weakened the willingness of OSC practitioners to develop and implement OSC-specific risk management strategies. Manufacturers complained about the lack of cooperation among all OSC practitioners, which resulted in the magnification of OSM risks. For example, contractors frequently change or adjust the on-site construction period, which forces OSC manufacturers to alter their production cycle, leading to other risks such as additional storage requirements for components, component production delays, etc.

The entire OSC project is based on the requirements provided by the owner. If any step does not align with the owner’s demands, it may result in OSM risks. In this case, the owner’s demands should remain consistent throughout the OSC project process. However, many owners frequently change their demands during traditional construction projects, and this habit persists in OSC projects as well. Due to the owner’s lack of experience with OSC, few owners realize that changes in demand can significantly disrupt the processes of other OSC stakeholders.

Consultants play a significant role in the OSM process. Errors in the design scheme can significantly affect the accuracy of the production process. A reasonable, accurate, and detailed design scheme is essential in OSC projects [12]. Reducing design errors will help the manufacturer meet their requirements more easily and reduce the chances of rework. Consultants should also avoid designing arc components or special-shaped components. Otherwise, the manufacturer must purchase special-shaped models to produce these components, which increases the cost of OSM. In this case, consultants should have more experience in OSC projects and communicate more effectively with manufacturers.

Manufacturers have been shown to significantly influence the final quality of OSC projects [74]. One of the reasons for this is the lack of experience among manufacturers. Although many OSC factories have industrialized OSC production lines, few manufacturers have the experience to operate these lines. For example, one manufacturer imported a German OSC production line, but no one knew how to program the line, and only several pre-set component types could be produced. When the production line broke down, no one could fix it, and due to COVID-19, it was nearly impossible to bring German experts into China. This 200 million Yuan production line became an ‘extremely expensive storage yard’. Therefore, the manufacturer’s lack of OSC experience greatly affected the QCD outcomes of the OSC project.

The public has a prejudice that OSC buildings are unsafe, a bias that stems from the 1976 Tangshan earthquake, during which over 95% of the houses in the city collapsed, and more than 600,000 residents were either killed or injured. Many of the victims lived in OSC buildings. As a result, prefabricated components were dubbed ‘coffin boards’ by the public after the earthquake [75]. This historical issue continues to affect the promotion of OSC today. Another issue is the occasional practice of subcontracting in some areas of China. Some contractors may win contracts at prices well below the market rates and then seek to subcontract the work in order to make a profit. This forces manufacturers who aim to produce high-quality components to make a difficult choice: whether to accept lower profits to secure the contract or maintain profits but lose the contract.

5. Conclusions

In this research, a BBN model was used to evaluate the risk factors in the OSM process, considering the relationship between these risk factors and the QCD outcomes of OSC projects. Existing studies on construction risk management have predominantly focused on traditional on-site methods, while the unique challenges and interdependencies inherent in OSM have received limited attention. This research addresses a notable gap in the literature by systematically evaluating the OSM risks and their impacts on project performance, thereby offering a novel framework for risk assessment in OSC projects. Interviews and questionnaires were developed and conducted in China to identify the significant risk factors in the OSM process. The BBN model was developed based on these significant risk factors, and the relationships between risk factors and QCD in OSC projects were evaluated using the BBN model. The analysis identified 12 significant risk factors that were further grouped into six categories: cost, project management, owner, consultant, manufacturer, and society. Notably, the sensitivity analysis revealed that factors such as high off-site manufactory building costs, high component model and material fees, and manufacturer’s lack of experience have a particularly strong influence on project outcomes. The results identified significant and sensitive risk factors for the OSM process, which were categorized into six risk groups: cost, project management, owner, consultant, manufacturer, and society. These findings contribute to the body of knowledge regarding OSM process risks and risk management in China.

Based on the findings, several key recommendations are proposed: Consultants are advised to adopt standardized design practices and minimize the use of complex or atypical components. Manufacturers should optimize prefabrication designs and enhance staff training to reduce production-related risks. Transporters should develop flexible logistics strategies to accommodate production and delivery uncertainties. Contractors need to improve coordination with manufacturers to ensure aligned project schedules. Academics are encouraged to extend research on OSC risk management to a broader international context and consider the development of a comprehensive OSC risk management database.

The limitations of this research should be acknowledged. First, the data collection process may reflect subjective biases, which could reduce the accuracy of the results. Second, the data were collected from mainland China; therefore, the findings may apply to other OSC projects in China with minor modifications. However, these results may not be applicable to other countries, especially developed nations. Therefore, future research should focus on OSM in other countries, and it is also recommended that an OSM risk management database be developed to help practitioners evaluate OSC projects from a more objective perspective. Moreover, further studies employing more complex modeling techniques are warranted to quantify the impact of individual risk factors on QCD outcomes in greater detail, thereby enhancing the understanding of risk interdependencies and refining OSC risk management strategies. In future studies, complementary methods such as Monte Carlo simulation could be integrated to provide a more objective quantification of uncertainties and to further validate the BBN results.