Defects in Flexible Pavements: A Relationship Assessment of the Defects of a Low-Cost Pavement Management System

Abstract

Pavement maintenance is a key concern for pavement management authority. Countries (especially developing countries) are facing severe funding challenges regarding maintenance schemes. The existing pavement maintenance methods are goal-specific and lack integration of various indicators that are significant for low-cost PMSs. Thus, this paper investigates the possible defects that may occur in flexible pavements as well as the relationships between different defects. A detailed literature review was conducted to identify all possible defects in flexible pavements and key features considered PMSs. A questionnaire was designed to seek expert opinions on the defects and their possible relationships for a low-cost PMS. The data were collected from 283 experts currently working in pavement management authorities and pavement maintenance schemes. Aggregated mean score, box plotting, and the chi-square test were used to analyze the data. It is concluded that bumps/sags (3.17) are major defects reported by pavement experts in Pakistan, followed by fatigue cracks (3.07). Rutting (2.98) and rut depth (2.98) are the third-ranked key defects reported in this study. Depression (2.96), potholes (2.76), longitudinal crack (2.69), edge crack (2.55), roughness (2.51), and deflection (2.50) are also regular defects in pavement maintenance activities in Pakistan. The results are in an acceptable range of the three-mentioned validation methods. The correlation test results show that most of the defects in structural, functional, safety, and serviceability indicators reject the null hypothesis; thus, there are close relationships between these defects observed in flexible pavements. In the last stage, a PMS model is suggested to assist road management authorities in developing countries to make low-cost decisions for effective pavement rehabilitation.

1. Background

Pavement maintenance has advanced from a notion in the 1960s to present-day common and practical implementations in many countries [1]. Pavement maintenance management systems were well-developed in different organizations across the world, but they did not take shape until the mid-to-late 1970s when they incorporated all of these operations into a functioning PMS. The World Bank has carried out a variety of road projects since 1968 and has established many assessment modules, including models of pavement efficiency. They have spread worldwide (to over 100 countries) and need to be tailored to local conditions [2]. Road transport infrastructure deteriorates as a result of aging, vulnerability to weather and traffic, and backlog repairs. As a consequence, one primary issue involves the viability of these infrastructures. The latest developments suggest that there are substantial and growing expenses surrounding sustainable transport networks. As a result, the overall thinking regarding infrastructure management that takes maintenance needs into account is gaining importance [3]. From 2008 to 2028, it is projected that an annual expenditure of $USD 101 billion will be needed to sustain all USA highways and that failure to deliver the funds will further weaken road networks [4]. The USA already invests over USD 184 billion annually to repair and expand its road networks [5]. More than GBP 15 billion is expected to be spent in England to expand its road networks [6].

Flexible pavements experience numerous distress indicators, including cracks, potholes, erosion, etc. It is not easy to identify the existence of various distress indicators. Distress position, scale, severity, and mapping are key problems [7]. Inspectors walking around the segment of a highway have historically carried out thorough manual checks. They are replaced by advanced distress investigation approaches owing to additional time and resources, ranging from contactless high-speed laser sensors to machines that capture video photographs of the pavement’s surface [8]. Later, processing algorithms are used to classify forms of cracks and other situations commonly focused on neural networks [9]. Typically, pavement administrators face problems, such as fault classification and lengthy evaluation turnaround periods [10,11]. The pavement’s structural capability assessment shows the pavement’s residual existence, i.e., the number of load repetitions that it can endure. The structural capacity of a road considers the structural capacities of all road layers and the soil and foundation states [12]. A comprehensive literature review was carried out for this study and the core summaries of the selected recent work are shown in

Table 1. Previous research work summary.

Reference	Objective	Method	Key Achievements	Remarks
[7]	The accuracy of prediction models can affect the decision-making process of maintenance activities.	Recurrent neural networks (RNN) algorithm	Increasing the error rate also contributes to the prediction model, resulting in a higher benefit reduction rate.	Focuses on the validation of the existing model. It is an accurate prediction model.
[8]	Development of the road pavement deterioration model	Recurrent neural network (RNN), deep learning, GIS	Accurate prediction of maintenance timing for the pavement is achieved.	Expertise needed. A complex process; equipment is needed.
[9]	Model to evaluate existing pavement condition	Analytical neural network (ANN)	High coefficient of determination and correlation between the actual data and the data assessed.	Needs an ANN expert to process the model; complex.
[13]	Identification of efficient pavement management systems	Case study, survey, mathematical programming	The transportation agencies can determine future budget needs, funding allocations, and treatment policies to demonstrate the best possible use of pavement management resources.	Fixed for local conditions; road condition assessment models cannot be integrated.
[14]	Proposed PMS for Bangladesh	Multiple regression analysis, SPSS,	Pavement rehabilitation using the deflection value predicted by IRI and the age of pavement sections.	Limited to a few defect classes only.
[15]	Identification of hotspots on the urban road network	Deep learning, ANN, convolutional neural networks (CNN), camera phone, the Google Pixel 2XL	Low-cost hotspot analysis of the pavement distresses.	Needs an ANN expert to process the model; complex.
[16]	Developed a PMS	Multistage stochastic mixed-integer programming model, mathematical programming	The indices are high, indicating the effectiveness of the stochastic solution.	The high computational complexity of the mixed-integer programming models hinders the practical application of the proposed stochastic pavement for large-scale networks.
[17]	PMS with fewer data	Markov hazard model	Examines the applicability (to developing countries) of a PMS with fewer data.	Analyzes the relationship between the pavement design and estimates life expectancy.
[18]	Selecting optimal and rehabilitation strategies for pavements	Fuzzy logic, AHP	It provides decision-makers with an easy and intuitive methodology for the selection of pavement sections.	At network level only and complex to operate.
[19]	PMS for Mexico	Past data, case studies	A useful procedure that allows the collection, analysis, processing, and updating of pavement condition data.	Limited to a few defect classes; lacks multi-attribute analysis.
[20]	Proposed PMS for Budapest	GIS, Pareto solution	Follows multi-variable optimization methods to define the rehabilitation technique	Lacks further refinement, including carrying capacity results and road structure.
[21]	A database management system for a small city	GIS, QGIS	GIS technologies applied to establish and keep better road infrastructure management.	The database was created by a manual collection of pavement conditions. Their limiting factors are often the price (or the scope) and functions that do not necessarily correspond to the needs of a particular road administration.
[22]	PaveM, PMS for the California Department of Transportation	Decision tree, H-chart	PaveM provides a screening of the system’s needs and total funding need estimates.	PaveM provides information to support decision-making, not to make decisions.
[23]	Proposed PMS for Kazakhstan	GIS, GIS web, programing	A methodology for analyzing the collected pavement data for the implementation of a network-level pavement management program.	Focuses on the survey data processing. The Decision-making model is linear.
[24]	Optimization-based decision support system (DSS) for pavement management	Pareto, case study	Enhances the sustainability of the road pavement maintenance decision-making process by including road users and environmental concerns.	The objective functions referring to the indicators that can simultaneously be optimized are limited. Cost and environmental indicators are included only.
[25]	A low-cost video-based pavement distress screening system	GIS, image sensing	The proposed VPADS system aims to provide a computer-aided screening solution for transportation authorities	Eighty percent accuracy in detecting crack and distress features; the work can be further expanded by developing a crowd-sensing inspection. Limited to image assessment only.
[26]	Pavement distress classification and maintenance priorities	GIS, Paver	The system was of great help in identifying, collecting, and displaying pavement condition data.	Limited indicators are added in the model suggested.
[27]	StreetScan PMS	Algorithms, decision trees and mathematical models, PAVER, GIS	Users are able to visualize and interact with the repair suggestions and condition indices	Expertise needed. A complex process, equipment needed
[28]	Design of a PMS	Vehicle operating costs VOC	The traffic is directly proportional to the indirect user costs.	Develop a new PMS (small and mid-sized) to include the indirect operating costs.
[29]	Proposed PMS for Lisbon	GIS, survey, GPS, VIZIROAD, survey	Maintenance and rehabilitation decisions tend to have more reactive attitudes than a planned vision.	Such a system cannot be successfully implemented without the full commitment of the responsible authority and the appropriate know-how of the system designer.
[30]	Maintenance plans in Palestine	Survey and case study	Operation and maintenance manual has a positive impact on the ten pilot municipalities.	Data-driven and handled manually. Lacks managing large plans.

.

PMSs have been the topic of key research studies and a sufficient number of works have already been conducted on PMSs; however, limited attempts have been made on the design of low-cost PMSs. Most PMSs have been designed for developed countries, whereas limited attempts have been made for developing countries, where road management agencies are facing substantial financial challenges in managing the maintenance schemes for their existing road networks.

A PMS is a comprehensive, reliable approach for maintenance and rehabilitation needs, in deciding goals, and the optimum repair period [31,32,33]. The usefulness of PMSs depend on the data being used [34]. Primary data categories include pavement quality scores, costs, roadway building backgrounds and repair, and traffic loading. Identifying and assessing pavement conditions and identifying the causes of erosion significantly emphasize PMSs [35]. A PMS tends to eradicate human subjectivity from the equation when properly handled, creating impartial judgments. The general efficiency of the road infrastructure increases as the usable funds for the system are extended [36].

To preserve the pavement’s infrastructure and support its customers, considerable efforts are made each year, involving labor, money, and other services [37]. Adaptation to global climate change is projected to generate the need for a significant rise in pavement maintenance investments [38,39]. Pavement activities need to be prioritized to minimize the costs of maintenance activities and to maximize the life cycle of the network [31]. To reach this goal, a robust and accurate deterioration model is needed [40]. By reducing the errors in deterioration models, agencies can obtain significant budget savings through timely intervention and accurate planning [41]. Long-term and short-term planning techniques that are possible with deterioration models are even more critical when highway agencies have limited funding [42].

Single-goal or multi-goal optimization techniques are used to control paving. Various strategies used for optimizing a single goal, such as lowering expenses or optimizing the advantages of care in terms of road conditions or life span [43]. The study on pavement maintenance is critical to establish multi-goal optimization methods. Efficient road maintenance preparation is also the primary objective of every road administrator, in order to schedule maintenance operations based on available funds [44]. A PMS is an organized procedure used to administer and sustain roadways, efficiently and economically, based on statistical and quantitative methodologies. Thus, this study examines the previous works conducted on PMS in various countries. A critical review was conducted and the gap in this study is also highlighted. It investigates the methods used for decision-making in various PMSs. This paper also identifies all of the possible defects that may occur in a flexible pavement and evaluates the relationships between the defect classes.

2. Research Methodology

A comprehensive process was designed for this study. As there are various stages involved in the research, a complete research flow is presented in Figure 1.

A comprehensive literature review was conducted for this research and the research gaps were highlighted from previous research. A review of various PMS research studies was conducted and input from these studies was considered when finalizing the list of indicators (defects). A questionnaire was designed in the next phase to seek the expert’s opinion on the defects and the priorities for flexible pavement maintenance plans. The questionnaire was sent to nearly 350 experts (via email and hard form) who were working and had experience in flexible pavement maintenance schemes. A total of 283 questionnaires were successfully received and raw data were cleaned to observe any missing data points.

Based on the nature of the study, the data were collected from experts and professionals working in the National Highway Authority (NHA), Pakistan. The data were collected in three different rounds categorized as per the study objectives. Figure 2 shows the graphical representation of the study area.

A descriptive statistical analysis of the survey was conducted using SPSS, version 25. The response to each question in the survey was imported into SPSS from Google Forms and hard form data in Excel. The complete data from 283 respondents for each question were plotted in the SPSS interface. The aggregated mean score (AMS) was used as a data analysis approach. The AMS has been successfully used numerous times in such data sets [45,46]. The result validation was a key milestone in this research; therefore, a two-tier approach was developed. Standard deviation (SD) and skewness–kurtosis tests were conducted on the data in tier one. Both validated questions separately as distinct entities. In tier two, box plotting was also conducted to observe the results’ possible data outliers and fitness.

A box plot, also known as a box and whisker plot, is a type of graph that displays a summary of a large amount of data in five numbers. These numbers include the median, upper quartile, lower quartile, and minimum and maximum data values [47,48]. The use of box plots in place of single points in a quality control chart can effectively display the information usually given in X-Bar and R charts, show the degree of compliance with specifications, and identify outliers [49]. An example of it is shown in Figure 3.

A box plot is a standardized way of displaying the distribution of data based on a five-number summary (“minimum”, first quartile (Q1), median, third quartile (Q3), and “maximum”). It assists in outlier identification and data grouping density [50]. A box plot is a highly visually effective way of viewing a clear summary of one or more data sets. It is beneficial for quickly summarizing and comparing different sets of results from different experiments. At a glance, a box plot allows a graphical display of the distribution of results and provides indications of symmetry within the data [47].

In the last phase, the defect relationships were analyzed using the chi-square test for each primary indicator, which included functional defects, structural defects, safety defects, and serviceability defects of flexible pavements. The chi-square statistic is commonly used for testing relationships between categorical variables as shown in Equation (1). The null hypothesis of the chi-square test is that no relationship exists between the categorical variables in the population; they are independent [51,52].

$$x_c^2=\sum \frac{{\left(O_i-E_i\right)}^2}{E_i}$$

(1)

where C is the degree of freedom, O is the observed value(s), and E is the expected value(s).

This statistic can be evaluated by comparing the actual value against a critical value found in a chi-square distribution, but it is easier to examine the p-value provided by SPSS. To conclude the hypothesis with 95% confidence, the value labeled Asymp. Sig. (which is the p-value of the Chi-Square statistic) should be less than 0.05 (the alpha level associated with a 95% confidence level) [51]. A low value for chi-square means that there is a high correlation between two sets of data. [52,53].

3. Results and Discussion

Each indicator was analyzed to observe the data set and the specific results of each indicator.

Table 2. Case processing summary of structural indicators.

Structural Defects	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Deflection	283	100.0%	0	0.0%	283	100.0%
Fatigue (Alligator) Crack	283	100.0%	0	0.0%	283	100.0%
Longitudinal Crack	283	100.0%	0	0.0%	283	100.0%
Traverse (Thermal) Crack	283	100.0%	0	0.0%	283	100.0%
Block Crack	283	100.0%	0	0.0%	283	100.0%
Swell/Frost Heaving	283	100.0%	0	0.0%	283	100.0%

shows that the summary of the data collected from experts for structural indicators was only considered for this study.

It can be observed that there are no missing data items and all data points of the expert’s judgment were properly stored from raw data in SPSS. Raw data of each defect were carefully shifted in SPSS format to further assess the results and correlations. A separate box plot assessment was carried out in the next phase to observe the data distribution for all defects considered in the structural indicator category. Figure 4 shows the results of box plotting.

It can be observed that the data are usually distributed and there are no outliers in the data set for all defects in the structural indicators category. The minimum score of deflection defect is 1, whereas its maximum score is 4 given by the experts. The defection defect has major data set scores from 2 to 3 as its mean lies at score 2 in the 25th quartile. The cases of traverse cracks and block cracks are similar. It can be observed that longitudinal cracks have similar data distribution-like deflections, traverses, and block cracks, but the mean lies in the 75th quartile, which is a mean score of 3. The cases of fatigue and swell/frost heaving are different. The minimum score of a fatigue crack is 2 whereas its major data set lies between a score of 3 and 4 with a mean score of 3 in the 25th quartile. Its maximum score is 4 and its 75th quartile has the same score. In the case of swell/frost heaving, the minimum score is 1 and the maximum score is 3, which is also in the 75th quartile. Swell/frost heaving has an average mean score of 2, which is in the 25th quartile.

Table 3. Case processing summary of functional indicators.

Functional Defects	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Rutting	283	100.0%	0	0.0%	283	100.0%
Corrugation	283	100.0%	0	0.0%	283	100.0%
Shoving	283	100.0%	0	0.0%	283	100.0%
Potholes	283	100.0%	0	0.0%	283	100.0%
Patching	283	100.0%	0	0.0%	283	100.0%
Raveling	283	100.0%	0	0.0%	283	100.0%
Bleeding/Flashing	283	100.0%	0	0.0%	283	100.0%
Delamination	283	100.0%	0	0.0%	283	100.0%
Drop-off	283	100.0%	0	0.0%	283	100.0%
Polished Aggregates	283	100.0%	0	0.0%	283	100.0%
Depression	283	100.0%	0	0.0%	283	100.0%
Bumps/Sags	283	100.0%	0	0.0%	283	100.0%

shows the summary of data collected from experts for the functional indicators considered for this study.

It can be observed that there are no missing data items and all data points of the expert’s judgment are properly stored from raw data in SPSS. Raw data of each defect were carefully shifted in SPSS format to further assess the results and correlations. A separate box plot assessment was carried out in the next phase to observe the data distribution for all defects considered in the functional indicator category. Figure 5 shows the results of the box plotting.

It can be observed that the data are almost normally distributed but have few outliers. The minimum score of rutting defects is 2, whereas its maximum score is 4 (as given by the experts). The rutting defect’s major data set scores range from 2 to 4; therefore, its mean lies at a score of 3. Corrugation, raveling, bleeding/flashing, delamination, and drop-off are very similar, with a minimum score of 1, which was in the 25th quartile of scores, and a mean score of 2, which was in the 75th quartile. The maximum score was 3, given by all experts considered in this study. The shoving defect had some variations as its data were wide and there were no specific trends in the data. There could be a reason behind this outlier. It could have been an expert’s misinterpretation of the defect or a lack of coordination and information. Potholes and polished aggregate defects had similar data trends, with a minimum score of 1 and a maximum score of 4, whereas their maximum data sets were between 2 and 3 (in the 25th and 75th quartiles). Despite such close associations, it should be noted that their mean scores were not similar. Potholes had a higher mean score of 3 as compared to polished aggregates, which had a mean score of 2. Patching had different data distributions compared to all defects under functional indicators.

Patching’s 25th score and minimum score were similar, whereas its 75th score and maximum score were also similar. The mean score of patching in this data set was 2, as per the data collected from this study. The defect depression had a minimum score of 2, which was also its 25th quartile. It had a maximum score of 4 but its mean score was 3, whereas its 75th quartile had a score of 3.5. The last defect in this category is bumps/sags, with a minimum score of 2 and a maximum of 4 at the 75th quartile. Its mean score was 3 but there were two outliers observed in the data set for this defect. However, throughout this data set, there was not a large percentage of outliers (it was barely 3.8% of the collected data).

Table 4. Case processing summary of safety indicators.

	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Skid Resistance	283	100.0%	0	0.0%	283	100.0%
Potholes	283	100.0%	0	0.0%	283	100.0%
Edge Crack	283	100.0%	0	0.0%	283	100.0%
Rut Depth	283	100.0%	0	0.0%	283	100.0%
Rail Road Crossing	283	100.0%	0	0.0%	283	100.0%

shows the summary of data collected from experts for the safety indicators considered for this study.

It can be observed that there were no missing data items and all data points of the expert’s judgment were properly stored from raw data in SPSS. Raw data of each defect were carefully shifted in SPSS format to further assess the results and correlations. A separate box plot assessment was carried out in the next phase to observe the data distribution for all defects considered in the safety indicator category. Figure 6 shows the results of box plotting.

It can be observed that the data on defects in the safety indicators varied for each defect. There was one defect with an outlier. The minimum score of skid resistance was 1, which was also its 25th quartile, with a maximum score is 3. The mean score of the skid resistance defect was 2, which was also its 75th quartile. The minimum score of potholes was 1, and the maximum score was 4, whereas the mean score was 2, which was also its 25th quartile. The minimum score, 25th quartile, and mean score of the edge crack were 2, whereas its maximum score was 4. The rut depth was 1, whereas its maximum score was 4, as given by the experts. The defection defect had major data set scores from 2 to 3, with a mean score of 2 in the 25th quartile. The cases of traverse and block cracks were similar. It can be observed that longitudinal cracks had similar data distribution-like deflections, and traverse and block cracks, with a mean in the 75th quartile, which was a mean score of 3. The cases of fatigue and swell/frost heaving were different. The minimum score of a fatigue crack was 2, whereas its major data set was between scores of 3 and 4, with a mean score of 3 in the 25th quartile. Its maximum score was 4, and its 75th quartile had the same score. In the case of swell/frost heaving, its minimum score was 1 and its maximum score was 3, which was also its 75th quartile. Swell/frost heaving had an average mean score of 2, which was its 25th quartile.

Table 5. Case processing summary of serviceability indicators.

	Cases
	Valid		Missing		Total
	N	Percent	N	Percent	N	Percent
Roughness	283	100.0%	0	0.0%	283	100.0%
Slippage Crack	283	100.0%	0	0.0%	283	100.0%
Reflection Crack	283	100.0%	0	0.0%	283	100.0%
Draining	283	100.0%	0	0.0%	283	100.0%

shows the summary of data collected from experts for serviceability indicators considered for this study.

It can be observed that there are no missing data items and all data points of the expert’s judgment were properly stored from raw data in SPSS. Raw data of each defect were carefully shifted in the SPSS format to further assess the results and correlations. A separate box plot assessment was carried out in the next phase to observe the data distribution for all defects considered in the serviceability indicator category. Figure 7 shows the results of box plotting.

Any PMS must explore the frequent defects and problems that the pavement is suffering from; therefore, defect identification was conducted, prioritizing the key defects after the expert’s opinion.

Table 6. Ranking of all flexible pavement defects.

Defects	Statistic	Statistic	Skewness	Kurtosis	Ranking
	Mean	Std. Deviation	Statistic	Statistic
Bumps/Sags	3.1731	0.7598	−0.863	0.952	1
Fatigue (Alligator) Crack	3.0769	0.73688	−0.123	−1.108	2
Rutting	2.9808	0.75382	0.032	−1.205	3
Rut Depth	2.9808	0.72735	0.03	−1.058	3
Depression	2.9615	0.73994	0.062	−1.129	4
Potholes	2.7692	0.83114	0.037	−0.802	5
Longitudinal Crack	2.6923	0.80534	0.158	−0.682	6
Edge Crack	2.5577	0.69771	0.867	−0.447	7
Roughness	2.5192	0.72735	0.408	−0.222	8
Deflection	2.5	0.93934	0.148	−0.826	9
Polished Aggregates	2.4615	0.82751	0.235	−0.412	10
Traverse (Thermal) Crack	2.4038	0.7478	0.05	−0.218	11
Block Crack	2.3077	0.64286	0.079	−0.048	12
Potholes	2.2692	1.01199	0.488	−0.781	13
Draining	2.1154	0.70444	−0.166	−0.92	14
Swell/Frost Heaving	2.0962	0.7478	−0.16	−1.162	15
Shoving	2	0.68599	0	−0.794	16
Patching	1.9615	0.79117	0.069	−1.385	17
Drop-off	1.9038	0.7211	0.147	−1.018	18
Slippage Crack	1.8846	0.70444	0.166	−0.92	19
Raveling	1.8077	0.71506	0.302	−0.965	20
Bleeding/Flashing	1.7692	0.67491	0.313	−0.762	21
Reflection Crack	1.7692	0.70336	0.358	−0.893	21
Delamination	1.7115	0.66676	0.404	−0.719	22
Skid Resistance	1.6731	0.67798	0.511	−0.723	23
Corrugation	1.6346	0.62713	0.457	−0.607	24
Rail Road Crossing	0.2308	0.42544	1.316	−0.28	25

shows the ranking of the defects typically observed in flexible pavements.

It can be observed that bumps/sags are major defects reported by pavement experts in Pakistan, followed by fatigue cracks. This is also reflected in [54]. Rutting and rut depths are the third key defects reported in this study. Depression, potholes, longitudinal cracks, edge cracks, roughness, and deflection are also regular defects in pavement maintenance activities in Pakistan. This has also been observed in similar previous studies [55,56,57,58,59]. The validation of the results was done using standard deviation, skewness, and kurtosis tests. Most of the results are in the acceptance range of the three mentioned validation methods. Skid resistance, corrugation, and railroad crossing were rare types of defects in flexible pavements.

In the latter phase, it is essential to analyze the possible relationship between the defects considered under the four PMS indicators. It is crucial to observe the following question: “Is there any close relation occur or do not occur between the defects” under all indicators for the PMS. So, the chi-square test was conducted to assess the relationships between the defects for flexible pavements.

Table 7. Chi-square test results.

Structural Indicators
	Deflection	Fatigue (Alligator) Crack			Longitudinal Crack		Traverse (Thermal) Crack					Block Crack		Swell/Frost Heaving
Chi-Square	9.231a	14.308a			19.231a		25.692a					38.923a		3.500b
df	3	3			3		3					3		3
Asymp. Sig.	0.026	0.006			0.000		0.000					0.000		0.174
a. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 13.0.
b. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 17.3.
Functional Indicators
	Rutting	Corrugation	Shoving	Potholes		Raveling	Bleeding/Flashing			Delamination		Drop-off	Depression			Bumps/Sags
Chi-Square	2.808a	15.500b	9.846a	16.154b		6.731a	10.654a			11.577b		5.808b	3.962a			31.231a
df	3	3	3	3		3	3			3		3	3			3
Asymp. Sig.	0.046	0.067	0.007	0.081		0.035	0.033			0.063		0.045	0.038			0.041
a. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 17.3.
b. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 13.0.
Safety Indicators
	Skid Resistance			Potholes				Edge Crack			Rut Depth		Rail Road Crossing
Chi-Square	11.115a			10.923a				15.269a			5.115a		15.077b
df	2			3				2			2		1
Asymp. Sig.	0.004			0.012				0.000			0.007		0.000
a. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 17.3.
b. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 26.0.
	Serviceability Indicators
	Roughness			Slippage Crack					Reflection Crack						Draining
Chi-Square	30.000a			7.538a					8.000b						7.538a
df	3			3					3						3
Asymp. Sig.	0.000			0.023					0.068						0.023
a. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 13.0.
b. 0 cells (0.0%) have expected frequencies of less than 5. The minimum expected cell frequency is 17.3.

shows the results of the test.

The correlation test results of structural indicators show that the defection, fatigue cracks, longitudinal cracks, traverse cracks, and block cracks reject the null hypothesis; thus, there are close relationships between these defects in flexible pavements. It is possible that where deflection is observed in flexible pavements, there can also be fatigue cracks, longitudinal cracks, traverse cracks, and block cracks. Similar relationships are also possible between these defects, which rejects the null hypotheses. There was no relationship observed between frost heaving and the defect types in the structural indicator of the flexible pavement’s maintenance management. The correlation test results of functional indicators show that rutting, shoving, raveling, bleeding, depression, and bumps reject the null hypothesis and there are close relationships between these defects. It is possible that, where rutting is observed, there can also be shoving, raveling, bleeding, depression, and bumps. Regarding corrugation, potholes, delamination, and drop-off, no relationship was observed between the defect types in the functional indicator.

Similarly, safety indicators, including skid resistance, potholes, edge cracks, and rut depth rejected the null hypothesis. Hence, there were close relationships between these defects. When skid resistance was observed, there could also have been potholes, edge cracks, and rut depth. The serviceability indicators show that roughness, slippage cracks, and poor drainage reject the null hypothesis; thus, there were close relationships between these defects. When roughness was observed in flexible pavements, there could also have possibly been a slippage crack or poor drainage.

In the end, based on the results of the study, a PMS framework is proposed for developing countries, as shown in Figure 8.

The proposed PMS has a three-tiered decision-making approach. In tier one, the case selection is made. There can be three possible cases for any pavement maintenance scheme, which includes; emergency case (needs-based, non-predictable), routine case (short-term plans, which require less time, effort, and funds), and periodic case (long-term plans, which require more time, effort, and funds). Each case has its own implications and significance. Therefore, this framework will select the case first. In the second tier, the indicators are selected based on the feedback from tier one. Similar to emergency cases, road safety indicators have the highest priority followed by functional and serviceability indicators, as part of similar suggestions given in previous studies [11,60,61,62,63]. Similarly, the indicator selections for other cases are different, as shown in the framework. In the third tier, the model will select the sub-indicators based on the second-tier results. Similar to safety, the sub-indicators are different and the sub-indicators are selected based on their scores. The sub-indicators are classified based on their scores from the expert’s feedback, as shown in Table 6. In the class one sub-indicator category, the defects with scores of more than 3 are selected by the model, based on the frequency and significance of the defect as per the scenario rated by the experts. Likewise, the defects with scores between 2.5 and 3.0 are classified as class two. Different defects lie in class two, as shown in the framework. In class three, in the sub-indicator category, the defects with scores between 2.0 and 2.5 are selected and the defects with scores below 2.0 are classified as the last category by the model.

The model will group the tiers based on the different scenarios required by the pavement maintenance management authority. The model will optimize the decision based on the pavement defect type, frequency, and low-cost solution.

4. Conclusions

PMSs are of key interest but limited attempts have been made for low-cost PMSs globally; limited attempts have been made in Pakistan where existing models cannot work efficiently due to varying local conditions. GIS is mainly used for the graphical representation of pavement maintenance schemes, with the combination of surveys, case studies, machine learning, deep learning, SPSS, and PLS as decision-making tools for PMS indicators. Limited attempts have been made using PLS.

Bumps/sags (3.17) are major defects that have been reported by pavement experts in Pakistan, followed by fatigue cracks (3.07). Rutting (2.98) and rut depth (2.98) are the third key defects reported in this study. Depression (2.96), potholes (2.76), longitudinal cracks (2.69), edge cracks (2.55), roughness (2.51), and deflection (2.50) are also regular defects in pavement maintenance activities in Pakistan.

The correlation test of structural indicators shows that defection, fatigue cracks, longitudinal cracks, traverse cracks, and block cracks reject the null hypothesis; thus, there are close relationships between these defects in flexible pavements. When deflection is observed in flexible pavements, there can also possibly be fatigue cracks, longitudinal cracks, traverse cracks, and block cracks. Similarly, correlation tests of functional indicators show that rutting, shoving, raveling, bleeding, depression, and bumps reject the null hypothesis under one category; thus, there are close relationships between these defects in flexible pavements.

The correlation test of safety indicators shows that skid resistance, potholes, edge cracks, and rut depth reject the null hypothesis under one category; thus, there are close relationships between these defects in flexible pavements. Likewise, the correlation test of serviceability indicators shows that roughness, slippage cracks, and poor drainage reject the null hypothesis; thus, there are close relationships between these defects in flexible pavements.

The proposed model works in three stages—case selection, indicator selection, followed by sub-indicator selection. Each case represents a different real practical scenario; cost is the key feature in the defect selection criteria (and the possible treatment selection). The model has a simple operating method, which is user-friendly with limited numeric calculations. This model will assist the decision-makers in pavement maintenance to make condition-based decisions and utilize limited funds to ensure the functionality and serviceability of roads.

The model can also be used for rigid pavements with some modifications in the sub-indicators (only where the cases and indicators can be the same for both classes of pavements). The model can be extended to any mobile application for easy and rapid decision-making for the users.

5. Future Recommendations

The model can be extended to suitable low-cost rehabilitation techniques for flexible pavements. This study identifies the defects; a detailed study should be conducted on suitable rehabilitation approaches for flexible pavements. The model can also be extended to rigid pavements.