QUANTIFICATION OF MOISTURE RELATED DAMAGE IN FLEXIBLE AND RIGID PAVEMENTS AND INCORPORATION OF PAVEMENT PRESERVATION TREATMENTS IN AASHTOWARE PAVEMENT-ME DESIGN AND ANALYSIS By Muhammad Munum Masud A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Civil Engineering—Master of Science 2018 ABSTRACT QUANTIFICATION OF MOISTURE RELATED DAMAGE IN FLEXIBLE AND RIGID PAVEMENTS AND INCORPORATION OF PAVEMENT PRESERVATION TREATMENTS IN AASHTOWARE PAVEMENT-ME DESIGN AND ANALYSIS By Muhammad Munum Masud Moisture increase in pavement subsurface layers has a significant influence on granular material properties that affect the expected pavement performance. In-situ moisture variations in unbound base layer over time significantly depend on water infiltration after precipitation and pavement surface conditions. Consequently, base resilient modulus (MR) is decreased considerably, which leads to premature failure and reduced service life. This study presents Long-term Pavement Performance (LTPP) data analyses for quantifying the effect of moisture infiltration through surface discontinuities (cracks and joint openings) on flexible and rigid pavement performance. The data analyses results show that higher levels of cracking and joint openings will lead to an increase moisture levels within base layer. The MR of the base decreases significantly with an increase in moisture levels. For flexible pavements, the maximum reduction in base MR ranged between 40% to 175% for the pavement sections located in dry and wet regions, respectively. In rigid pavements, the maximum reduction in base MR may vary from 10% to 125% for the pavement sections located in dry and wet regions, respectively. The findings imply that an adequate and timely preservation treatment such as a crack sealing can enhance the pavements service life significantly, especially in wet climates. Therefore, cracks should be sealed when the extent of fatigue cracking is within 6% to 7% and between 10% to 11% for the flexible pavements sections located in wet and dry climates, respectively. In rigid pavements, the joints should be resealed when the damaged joint sealant length exceeds 50 to 75 meters. Copyright by MUHAMMAD MUNUM MASUD 2018 I would like to dedicate my thesis to my wife Maryam, our daughters Abeera and Zara, my elder brother Aamir, and my beloved parents. iv ACKNOWLEDGEMENTS First and foremost, I would like to express my sincere gratitude and deep regards to my academic advisor Dr. Syed Waqar Haider for his constant motivation, monitoring, and guidance throughout the course of thesis work. I would also like to extend appreciation to fellow students Gopikrishna Musunuru and Aksel Seitllari, in guiding me and supporting me with the LTPP Infopave® data acquisition and analysis. Finally, I would like to acknowledge the USDOT funding support by the University Transportation Center at Michigan State University. v TABLE OF CONTENTS LIST OF TABLES……………………………………………………………….…………......viii LIST OF FIGURES……………………………………………………………….……………...ix 1 2 3 4 INTRODUCTION .....................................................................................................1 1.1 BACKGROUND ...................................................................................................................1 1.2 RESEARCH OBJECTIVES..................................................................................................3 1.3 POTENTIAL BENEFITS OF THE STUDY ........................................................................3 1.4 RESEARCH APPROACH ....................................................................................................4 1.5 OUTLINE OF THE REPORT ..............................................................................................4 LITERATURE REVIEW ..........................................................................................6 2.1 SOURCES OF WATER INFILTRATION INTO PAVEMENT LAYERS .........................6 2.2 IMPACT OF MOISTURE ON PAVEMENT PERFORMANCE ........................................6 2.3 MITIGATION OF MOISTURE RELATED DAMAGE ......................................................8 2.4 EXISTING MOISTURE PREDICTION MODELS .............................................................9 2.4.1 Empirical Models ....................................................................................................10 2.4.2 Analytical and Mechanistic Models ........................................................................18 2.4.3 Summary of Existing Models from Literature ........................................................30 2.5 SUMMARY ........................................................................................................................32 DATA SYNTHESIS ...............................................................................................33 3.1 SEASONAL MONITORING PROGRAM (SMP) BACKGROUND ................................33 3.2 DATA SELECTION CRITERIA ........................................................................................33 3.3 DATABASE DEVELOPMENT .........................................................................................34 3.4 DATA ELEMENTS ............................................................................................................34 3.4.1 Pavement Performance Data ...................................................................................38 3.4.2 Subsurface Moisture and Temperature ....................................................................39 3.4.3 Precipitation Data ....................................................................................................41 3.4.4 Ground Water Table Depth .....................................................................................42 3.4.5 Freezing Index .........................................................................................................42 3.4.6 Materials Data .........................................................................................................42 3.5 DATA LIMITATIONS .......................................................................................................44 3.6 AVAILABLE SMP SECTIONS FOR ANALYSIS ...........................................................44 3.7 SUMMARY ........................................................................................................................45 DATA ANALYSIS AND MODELING .................................................................47 4.1 HYPOTHESIS .....................................................................................................................47 4.2 METHODOLOGY ..............................................................................................................50 4.3 DESCRIPTIVE STATICS ..................................................................................................51 4.4 IDENTIFYING SIGNIFICANT VARIABLES ..................................................................56 4.5 DEVELOPMENT OF EMPIRICAL MODELS .................................................................58 4.6 FLEXIBLE PAVEMENTS MODELING ...........................................................................58 vi 4.6.1 Site-Specific Models for Flexible Pavements .........................................................59 4.6.2 ANN Modeling Flexible Pavements .......................................................................60 4.6.3 Impact of Base Moisture on Long-Term Performance ...........................................68 4.7 RIGID PAVEMENTS MODELING ...................................................................................84 4.7.1 ANN Modeling Rigid Pavements ............................................................................86 4.7.2 The Relationship between Base Moisture and Base Resilient Modulus-PCC Sections ...................................................................................................................88 4.7.3 Crack Sealing Application Timings — Rigid Pavements .......................................90 4.8 SUMMARY ........................................................................................................................90 5 CONCLUSIONS AND RECOMMENDATIONS ..................................................93 5.1 SUMMARY ........................................................................................................................93 5.2 CONCLUSIONS .................................................................................................................94 5.3 RECOMMENDATIONS ....................................................................................................96 REFRENCES………………………………………………………………….....………………98 vii LIST OF TABLES Table 2-1 Summary of existing models from literature ................................................................ 31 Table 3-1 LTPP data base tables used to extract data elements ................................................... 37 Table 3-2 Layer structure and TDR/thermistors depths ............................................................... 41 Table 3-3 Base layer material properties ...................................................................................... 43 Table 3-4 Number of available SMP LTPP pavement sections ................................................... 44 Table 4-1 Summary of regional climatic and performance data ................................................... 51 Table 4-2 Correlation matrix flexible pavements sections ........................................................... 56 Table 4-3 Correlation matrix rigid pavements sections ................................................................ 57 Table 4-4 Optimum settings for the flexible pavements ANN model .......................................... 65 Table 4-5 Summary — Change in MR due to moisture variations .............................................. 70 Table 4-6 Summary measured /predicted moisture data and Pavement-ME predicted performance ....................................................................................................................................................... 74 Table 4-7 Proportion of observed WP cracking length ................................................................ 77 Table 4-8 Conversions — Total surface cracking length to % area WP fatigue .......................... 78 Table 4-9 Optimum settings for the rigid pavements ANN model ............................................... 86 Table 4-10 Summary — Change in rigid pavements MR due to moisture change ...................... 89 viii LIST OF FIGURES Figure 2-1 Sources of moisture variations in pavement systems (11) ............................................ 6 Figure 2-2 Subgrade moisture variations and precipitation for Arkansas Site 2 (14) .................. 13 Figure 2-3 Field moisture estimation system diagram (29) .......................................................... 15 Figure 2-4 Model simulation results (5) ....................................................................................... 22 Figure 2-5 Model road construction with material constructions, dimensions, and slopes (34) .. 23 Figure 2-6 Water content distribution 3 days after onset of 1-hour rain event (34) ..................... 26 Figure 2-7 Water content distribution 3 days after onset of 7.5 mm, 1-hour rain event (34) ....... 26 Figure 2-8 Water content distribution after the onset of 7.5 mm, 1-hour rain event (34) ............ 27 Figure 2-9 Comparison of predicted and measured resilient moduli for selected materials (18) . 28 Figure 2-10 Vertical Moduli Distribution Base layer (18) ........................................................... 29 Figure 3-1 Subsurface moisture and temperature measurements ................................................. 40 Figure 3-2 Base material particle size distribution ....................................................................... 43 Figure 3-3 Climatic distribution of SMP LTPP sections .............................................................. 45 Figure 4-1 Impact of cracking and precipitation on base layer moisture change (36-0801) ........ 49 Figure 4-2 Effect of GWT on base layer moisture change (36-0801) .......................................... 49 Figure 4-3 Subsurface moisture variations with depth (36-0801) ................................................ 50 Figure 4-4 Cracking progression with age in flexible pavements sections .................................. 52 Figure 4-5 Cracking progression with age in rigid pavements sections ....................................... 53 Figure 4-6 Precipitation levels in different climates ..................................................................... 54 Figure 4-7 Moisture variations in base layer — flexible SMP sections ....................................... 55 Figure 4-8 Moisture variations in base layer — rigid SMP sections ............................................ 56 ix Figure 4-9 Measured Vs. predicted site-specific models for flexible pavements ......................... 60 Figure 4-10 ANN model flow for flexible pavements SMP sections ........................................... 65 Figure 4-11 ANN model predictions and sensitivity — flexible pavements ................................ 66 Figure 4-12 Effect of precipitation on moisture variations ........................................................... 67 Figure 4-13 Moisture variations with depth in DF/WF region ..................................................... 68 Figure 4-14 Impact of moisture variations on flexible pavements base MR ................................ 71 Figure 4-15 Flexible pavement cross sections .............................................................................. 72 Figure 4-16 Impact of flexible pavements base MR on predicted pavement performance .......... 75 Figure 4-17 Impact of base MR on predicted long-term pavement performance......................... 76 Figure 4-18 Preservation treatment plan thick section (WF climate) ........................................... 80 Figure 4-19 Preservation treatment plan thick section (DNF climate) ......................................... 81 Figure 4-20 Preservation treatment plan thin section (WF climate) ............................................. 83 Figure 4-21 Preservation treatment plan thin section (DNF climate) ........................................... 84 Figure 4-22 PCC surface discontinuities relationship with base layer moisture .......................... 85 Figure 4-23 ANN model flow rigid pavements SMP sections ..................................................... 86 Figure 4-24 ANN model predictions and sensitivity — rigid pavements .................................... 88 Figure 4-25 Impact of moisture variations on PCC sections base MR ......................................... 89 x 1 INTRODUCTION 1.1 BACKGROUND The United States highway system is steadily deteriorating and allocating more resources to rebuild new roadways may not be a practical and cost-effective solution. The Nation’s highway system is the single largest public investment in history having an estimated initial cost of $3 trillion spread over many years (1). Today, the replacement cost could not readily be incurred without severe economic consequences. Therefore, one of the most significant challenges for researchers and engineers is how to minimize life-cycle cost and ensure sound asset management. Delaying maintenance and repairs until major rehabilitation or replacement is necessary lead to extensive and disruptive work that increases the potential for accidents, injuries, and fatalities among motorists and road workers. An alternative to this scenario is sound planning and implementation of highway preservation practices, which would assure structural integrity and safety of pavement assets. Currently, pavement preservation is an increasingly widespread practice among highway agencies interested in extending the lives of their pavements cost- effectively. One major impediment to widespread implementation of preserving the pavement infrastructure by transportation agencies is lack of knowledge on how to select preservation actions and when and where to apply them to get the most benefit at the least cost. In other words, how to use the right preservation action at the right time to the right pavement (2-4). Highway agencies have learned from the practices that if applied at an appropriate time, pavement preservation provides a means for maintaining and improving the functional condition and slowing deterioration of an existing highway system. While pavement preservation is not expected to substantially increase the structural capacity of the existing pavement, it generally 1 leads to improved pavement performance and longer service life. However, still, there are challenges to the success of such practices. These challenges include: (a) identifying good candidate pavements, (b) selecting the best preservation treatments for those pavements, (c) choosing the appropriate treatment application timing, and (d) considering preservation treatments in pavement analysis and design stage. This research specifically addresses the last two challenges, i.e., selection of optimum crack sealing application timings by incorporating preservation treatments in the mechanistic-empirical (ME) pavement analysis and design approach. The AASHTOWare Mechanistic-Empirical Pavement Design Guide (Pavement-ME) software provide methodologies for the analysis and design of flexible and rigid pavements. However, these methodologies and related performance prediction models focus on new structural design and rehabilitation of existing pavements and do not explicitly consider the contributions of pavement preservation treatments to the overall pavement performance. Thus, research is needed to identify approaches for considering the effects of preservation on pavement performance and developing procedures that facilitate incorporation of pavement preservation treatments in the Pavement-ME analysis process. Such procedures will ensure that the contributions of preservation treatments to expected performance and service life are appropriately considered in the analysis and design processes. One of the most influential factors affecting pavement performance is the moisture within the pavement system. The infiltration of water from road surface followed by a rainfall event can be a significant cause of premature pavement deterioration (5, 6). The moisture content of the materials near the pavement edges and in the proximity of surface cracks usually shows higher variations due to rainfall events (7). Water infiltration through cracks and joints is particularly 2 important in the estimation of sublayer moisture content and its effect on the resilient modulus (MR) (8, 9). Accurate predictions of moisture variations can assist in the better estimation of unbound layers MR. 1.2 RESEARCH OBJECTIVES The main objectives of this study are to (a) evaluate the effect of cracking and joint openings on the moisture content in unbound layers, (b) quantify the impact of infiltration and moisture on the stiffness properties of unbound layers, (c) predict long-term pavement performance based on the unbound material properties to evaluate the impacts of preservation treatments, and (d) develop guidelines for optimum crack sealing applications timings for different climatic conditions. 1.3 POTENTIAL BENEFITS OF THE STUDY The results of this research effort will improve and facilitate the implementation of preservation practices in the following manner:  There are no widely accepted guidelines for incorporating pavement preservation treatment in pavement analysis and design process, mainly because of different practices and experiences in different regions. This research will provide guidelines to facilitate estimation of timing for a pavement preservation treatment at the design stage. The research will also provide examples for different States to demonstrate how to apply the developed guidelines for estimating treatment timings to improve its effectiveness in extending the life of an existing pavement. This will help State Highway Agencies (SHAs) to incorporate preservation treatment practices at the design stage. 3  The recommendations developed from this research will be practically-oriented for investment decision making on the highway infrastructure. The recommendations will be specifically designed for application.  The analysis results from this research can maximize the benefits (both short-term and long-term) accrued from the large investment made in the construction and monitoring of the highway network. 1.4 RESEARCH APPROACH The following tasks were identified as a general framework for completion of this research: 1. Literature review. 2. Evaluation of infiltration and moisture models. 3. Availability of performance, climatic, and subsurface moisture content data. 4. Analyze subsurface moisture and performance data. 5. Establish impact of moisture change on unbound layers MR. 6. Develop guidelines for incorporating the preservation treatments in the Pavement-ME design process. 7. Demonstrative Examples. 1.5 OUTLINE OF THE REPORT This thesis contains five (5) chapters. Chapter 1 outlines the problem statement, research objectives, potential benefits, and briefly describes various tasks performed in the study. Chapter 2 documents the thorough literature review, which include sources of water infiltration into pavements, the impact of moisture on pavement performance, mitigation of moisture related damage, and summary of moisture prediction models. The work in this chapter corresponds to Tasks 1 and 2. Chapter 3 describes the SMP LTPP database with a special focus on SMP 4 background. This chapter also discusses the type, extents, and sources of various data types used in this study. The summary of available LTPP SMP sites considered for analysis also presented. The work in this chapter corresponds to Tasks 3. Chapter 4 covers the details of data analysis on flexible and rigid SMP pavement sections, development of moisture content prediction models using Artificial Neural Network (ANN), the impact of moisture on unbound layer stiffness and long-term pavement performance. Last part of this chapter covers pavement preservation guidelines with examples using the Pavement-ME. The work in this chapter corresponds to Task 4 to 7. Chapter 5 documents the conclusions and recommendations based on the analysis. 5 2 LITERATURE REVIEW 2.1 SOURCES OF WATER INFILTRATION INTO PAVEMENT LAYERS Water can enter the pavement-unbound layers through many sources and subsequently affects the in-situ moisture in these materials. The primary sources of moisture variation within a pavement system include external elements such as precipitation, temperature, and the groundwater table. Pavement surface conditions (cracking/discontinuities), drainage, shoulders, edges and pavement cross-section can also facilitate the moisture infiltration (10). Figure 2-1 shows the schematic of water ingress sources. Figure 2-1 Sources of moisture variations in pavement systems (11) 2.2 IMPACT OF MOISTURE ON PAVEMENT PERFORMANCE One of the most influential factors affecting pavement performance is the moisture within the pavement system. As early as 1820, John MacAdam noted that regardless of the strength (thickness) of the pavement structure, many roads in Great Britain prematurely deteriorated due to saturation of pavement subgrade (12). Moisture damage in pavements manifests itself in the form of moisture caused, and moisture accelerated distresses. Moisture caused distresses are essentially induced by moisture, such as asphalt stripping in flexible pavements and durability 6 cracking in rigid pavements. Moisture accelerated distresses are those caused by other factors (like traffic loading), but get accelerated with an increase in moisture (13). Many properties of unsaturated soils such as stiffness, permeability and volume vary significantly with change in moisture content. The increase in moisture content affects the durability and stiffness of soils; consequently, the ability of subgrade to support the upper pavement structure (14, 15). Variation in moisture content in field conditions depends on the climate of a location and can be difficult to interpret (16, 17). It is also known that unsaturated granular material (UGM) exhibits moisture-sensitive and stress-dependent nonlinear behavior in flexible pavements. The in-situ moisture content of unbound pavement materials is significantly affected by weather, groundwater table fluctuations, drainage conditions, soil properties and pavement surface conditions. It is a well-established fact that with an increase in UGM degree of saturation, the resilient modulus (MR) decreases considerably (18, 19). While investigating the pavement response to the varying levels of moisture, Salour and Erlingsson concluded that increase in moisture content of UGM considerably reduces the back- calculated modulus of base layers (20). Various field monitoring studies suggest that change in moisture content can occur after rainfall and it can increase up to 50% in addition to the natural seasonal variation (21, 22). This potential increase in moisture content is often neglected while estimating moisture variation in pavement unbound layers. However, such changes in moisture along with axle loads can accelerate pavement deterioration. Therefore, it is essential to develop a moisture prediction model that can capture both seasonal and temporal moistures changes accurately and later incorporate results in the life cycle assessment of infrastructures (16, 23). The infiltration of water from road surface followed by a rainfall event can be a significant cause of premature pavement deterioration (5, 6). It was also revealed in the past research that moisture 7 conditions are relatively stable at the bottom of the pavement system. However, depending on climatic events, the moisture condition in the upper pavement section can vary between very dry and fully saturated conditions. The moisture content of the materials near the pavement edges and in the proximity of surface cracks usually shows higher variations due to rainfall events (7). Considering water infiltration through cracks and joints is particularly important in the estimation of sublayer moisture content and its effect on the resilient modulus (MR) (8, 9). Accurate prediction of moisture content can assist in the better estimation of pavement unbound layers MR. Water movement within pavement system and affiliated moisture change is a complex phenomenon. Problems triggered by prolonged exposure to excess moisture fall into three main categories (13):  Softening of pavement unbound layers as they become saturated and remain saturated for a considerable time.  Material degradation from interaction with moisture.  Loss of bonds between pavement layers from saturation with moisture. 2.3 MITIGATION OF MOISTURE RELATED DAMAGE Despite considerable research in recent years on moisture-related damage in the pavements, there are still several gaps in knowledge and practice. Pavement researchers are still to reach a consensus, whether to design the roads as permeable, impermeable, or combination of the two. One of the primary concern at the pavement design stage is to protect the base, subbase, and subgrade layers from becoming saturated or even being exposed to prolonged high moisture conditions over time. Many pavement engineers would also add hot-mixed asphalt (HMA) and Portland cement concrete (PCC) to this list because excessive moisture coupled with freezing has 8 badly impacted properties of these materials (13). Four widely accepted approaches to mitigate moisture damage are listed below:  Prevent moisture from entering the pavement structure.  Use of less moisture susceptible materials.  Incorporate design features to minimize moisture damage.  Through effective drainage quickly remove moisture that enters the pavement structure. Many highway agencies use the Pavement-ME for designing and rehabilitating pavements and evaluating their maintenance options. Pavement-ME estimates infiltration through cracks and joints for incorporating the permeable base, separator, and edge-drain design in the design process. It does not consider the water infiltration in the modeling of moisture content within the pavement layers. Therefore, moisture and material properties of sublayers are not assumed to be affected by water infiltration through discontinuities present at the pavement surface. This study will evaluate the effect of infiltration due to cracks/joints on moisture content and resulting resilient modulus of the unbound materials in a pavement system. The Pavement- ME input material properties can be modified to capture the effect of infiltration on predicted performance. Such incorporation of infiltration in the pavement design process can assist highway agencies to adopt proactive pavement preservation practices. 2.4 EXISTING MOISTURE PREDICTION MODELS Moisture determination within the pavements layers is a complex task, especially with the varying site and climatic conditions. Researchers have been working to determine field moisture content based on soil properties, field observations and flow theories. In the process of evolution many empirical and analytical solutions were developed to characterize the change in in-situ moisture content. These methods ranged from straightforward empirical equations to very 9 complex computer-based programs (14). An integrated model was also developed to predict soil moisture content levels and movements within a pavement structure (14, 24). The reliability and application of empirically developed models are limited because most of these models are based on regression analysis with a high standard error. It was also observed by Organization of Economic Corporation and Development (OECD) that the model errors can be very high (i.e., percent of moisture content) (15, 25). On the other hand, the analytical solutions available in the literature are complex. Those are based on differential equations with boundary conditions and include variables like hydraulic conductivity, matric suction, porosity and water table depth. Consequently, application of such models is limited for routine use and analysis. Significant limitations of the available models are their universal or regional application and validation with the different site and environmental conditions. Furthermore, most of the available models do not include the effect of surface discontinuities, pavement structure, or temporal changes due to rainfall or subsurface temperature on the sublayer moisture variations. The past research shows that models were developed to measure the change in stiffness properties due to moisture variation. Only a few empirical and analytical models were available in the literature for unbound layers moisture content prediction. Thus, more research is needed for accurate estimation of unbound layers moisture variations due to surface infiltration. 2.4.1 Empirical Models This section documents the details of empirical moisture prediction models found in the literature. 2.4.1.1 Swanberg and Hansen Model In Minnesota, where the subgrades were primarily clayey silt soils with plastic limit varying from 15 to 30 and densities between 90 to 105 percent of the modified proctor maximum density, 10 Swanberg, and Hansen (26) developed a model to measure the moisture content of highway subgrades using plastic limit. The authors also observed that measured moisture content was about 1 percent higher in spring than in summer. The mathematical form of the relationship is given below: where, w = Moisture content PL = Plastic limit 2.4.1.2 US Navy Model (1) US Navy (15, 27) developed a model which also relates moisture content with plastic limit. They considered 70 airport sites for investigation of sandy and clay subgrades where the groundwater table was greater than 60 cm below the surface and reached to the conclusion that subgrade moisture content exceeded the plastic limit by approximately 2 percent. (2) 2.4.1.3 Kersten Model While investigating subgrade moisture contents in the top 30 cm of subgrade soils below airports pavements in seven states, Kersten (28) concluded that water content for sand and clay soils in damp climates could vary between 80 to 120 percent of the plastic limit (PL) (15). (3) It was also noted that typically clay equilibrium moisture content exceeds the PL, silts are equal to or just under the PL, and sandy soils are less than the PL. Thus, for many subgrade soils, the 11 1.167.4WPL2WPL0.81.2PLWPL lower limit of predicted moisture content varies between optimum moisture content (OMC) and the PL and the upper limit between the PL and 100 percent degree of saturation. 2.4.1.4 Arkansas Highway and Transportation Department (AHTD) - Rao, S Moisture Content Prediction Equations In a study at AHTD, prediction equations were developed to estimate subgrade in-situ moisture content for low-volume pavement design. Data of 18 different sites from 14 counties were collected from 1991 to 1993 (14). Data elements including general site information, soil series and profile information, moisture content at 9 different depths (starting from 18 to 90 inches), average monthly temperature and precipitation were obtained for the analysis. At two different depths, (30 and 90 inches) correlation analysis was developed between moisture content and precipitation, also between moisture content and average monthly temperature. Relatively low correlation coefficients were observed for both variables. Also for different sites great variation was observed in correlation coefficients at different depths (14, 15). The author observed that correlation of moisture content with precipitation was positive and with average monthly temperature was predominantly negative. Average monthly precipitation and moisture content at varying depths were plotted as a function of time as shown in Figure 2-2, limited range of values were observed for moisture content at different sites and depths. 12 Figure 2-2 Subgrade moisture variations and precipitation for Arkansas Site 2 (14) The author considered upper and lower values of moisture content as the upper and lower equilibrium values for moisture content in the subgrade. It was concluded that upper and lower limits of moisture content in subgrade depend on soil properties and vary with depth. However, temperature and precipitation had not much effect. Based on this observation, to estimate upper 13 and lower equilibrium values for moisture content from soil properties, the following regression equations were developed: -  For 18 inches below the pavement surface: (4) (5) (6) (7) R2 R2 = 0.79 = 0.80  For 30 inches below the pavement surface: - R2 = 0.61 R2 where, = 0.74 ELL = Equilibrium lower limit EUL = Equilibrium upper limit P200 = Percent passing No. 200 sieve LL PI = Liquid limit = Plasticity index PERM = Permeability 14 1.081.1132.122.860.174(200)0.173(5) 0.021()0.089()LULUELLPLLPILogPERM1.081.1132.126.450.221(200)0.174(5) 0.024()0.071()LULUEULPLLPILogPERM1.081.1132.131.250.313(200)0.292(5) 0.028()0.075()LAUALAUAEULPLLPILogPERM1.081.1432.139.660.212(200)0.118(5) 0.023()0.059()LAUALAUAEULPLLPILogPERM The subscript and are used for upper and lower limits from the county soil reports, whereas subscript indicates soil properties,12 inches above selected depth. 2.4.1.5 A Systems Approach for Estimating Field Moisture Content Han, Petry, and Richardson (29) developed a system for estimation of moisture content. The system was equipped with five different models, including Swanberg and Hansen (26) , Kersten (28), US Navy (27), Arkansas Highway and Transportation Department moisture predictions equations (15), and volumetric moisture content estimation equations from the SMP (29, 30). The user is asked to input project site data and material characteristics, then it provides a range of estimated moisture contents with a guide to narrow down choice. Degree of saturation is also an output because few resilient modulus prediction equations use a degree of saturation instead of moisture content (29). System structure diagram is shown in Figure 2-3. Figure 2-3 Field moisture estimation system diagram (29) 2.4.1.6 Hedayati and Hossain Data-Based Model In North Texas, a study was conducted to estimate moisture variation in pavement subgrade soils due to seasonal and time-dependent changes in climate. A two-lane HMA road was selected for 15 LUA this study. Hourly moisture at varying depth (0 to 4.5 m) and precipitation data were collected over the period of two years. Based on the overall data analysis a model was developed. The model considered the effect of seasonal trends and temporary rainfall in predicting moisture content of different soil layers (16). (8) Where ϴ is calculated using following equation: - (9) Finally, authors summarized above two equations using the following equation: - where, (10) = Volumetric water content at depth at any time ; = Average soil moisture at depth over time = The domain of moisture variation at any depth over time.Which can be determined as (16, 31) using equation 11. = The surface volumetric water content = Angular frequency (equal to for a perfect seasonal trend) = Time from an arbitrary starting point (day) = Depth(m) d = Damping depth (described below) Co = Phase correction factor Raint = Rainfall defined in time series (mm) 16 00sin()(,inf)aztCftraalld0.639[0.390.053sin(0.01720.2)][0.0134.00058]ztetzRain0.6390.41040.053sin(0.0172).00058ztetRainzt0zas12/3650.0172daytz (11) Damping depth reflects a reduction in soil moisture variation with depth and can be estimated as (16): (12) 2.4.1.7 Fredlund And Xing Equation Fredlund and Xing (32) proposed a model to calculate equilibrium moisture content based on soil suction, and soil index properties, such as Passing #200 (P200), diameter (D60), and plasticity index. This soil water characterization curve model is also used in Enhanced Integrated Climatic Model ( EICM). (13) (14) where, = Volumetric moisture content (%) = Saturated moisture content and = SWCC fitting parameters = The surface volumetric water content 17 .exp()aszd24.8Ddmw()lnexp(1)ffsatcbfChha5ln1()11.4510ln1hhChhwsat,,fffabchs h y = Equilibrium suction as defined in equation 15 = Distance from the ground water table = Unit weight of water (15) 2.4.2 Analytical and Mechanistic Models Moisture content predictions based on empirical equations showed significant variations. Moreover, most of the empirical methods were developed for specific locations, which limited their regional application. Therefore, analytical solutions to predict moisture change were developed. Analytical solutions for moisture infiltration/variation found in the literature were reviewed and summarized in this section. 2.4.2.1 Han-Cheng Dan, Jia-Wei Tan, Zhi Zhang and Lin-Hua He Model for Water Infiltration Rate into Cracked Asphalt Pavement Using flow theory in porous and cracked medium, Dan et al. (5) proposed a model to quantify the water balance between surface and drainage layers in asphalt pavements to estimate pavement infiltration rate (PIR). Since the water can enter into the pavements through linked cracks and connected pores, accordingly it was assumed that the total water inflow infiltrating the pavement structure equals the sum of water flow through surface course and cracks. The total water infiltration quantity can be expressed as: where, (16) Q1 = Water quantity through the porosity of asphalt layer (L2/T); 18 water.waterhy12120()BTotalQQQqqdx Q2 q1 = Water quantity through cracks present in asphalt pavements (L2/T) = Water flow through the micro-segmentation of surface course to the drainage layer (L2/T) q2 = Water quantity through the crack per unit length along the longitudinal pavement (L2/T) Using hydraulic conductivity of the porous medium and equivalent hydraulic conductivity of cracked asphalt layer, and solving integral for simplification (5), final expression obtained by the authors for PIR with full-length transverse cracks is given below: (17) For no crack on the pavement surface, authors expressed the infiltration as: Also, the difference between Im and Is is given by the following equation: K1 and KC expressed by the authors as: where, 19 (18) (19) (20) (21) __1211*I()MCTThkkT__1211*ISTThkT__121*ICCTThkT218gKr312CwKg = Water Infiltration rate per unit width of the pavement incorporating crack and porosity (L/T) = Water Infiltration rate per unit width of the pavement due to and porosity only (L/T) = Difference between and (L/T) = Hydraulic conductivity (L/T) = The equivalent hydraulic conductivity of crack (L/T) = The thickness of surface course (L) = The thickness of drainage layer (L) = Average water thickness (L) = Water density (M/L3) = Gravity acceleration (L/T2) = The porosity of porous media = Uniform radius of microtubules (L) = Crack density, defined as = Crack number with uniform width = Crack distance = Crack opening width (L) = Kinematic viscosity of flow (Pa S) Finally, the expression for infiltration rate with the random crack length of asphalt pavement was presented with following modification: (22) 20 MISICIMISIkCk1T2T__*hgr/CNLNCLw.inCSIII where, is the ratio of crack length to pavement width, expressed as: - where, = Pavement width (L) (23) Authors compared the results of this model with Ridgeway's method (5, 33). The general trends noted are presented in Figure 2-4, and briefly discussed below.  PIR increases with increase in crack width as shown in Figure 2-4(a).  Amount of water seeping through pores is negligible as compared to the quantity of water infiltrating through surface cracks [see Figure 2-4(a)].  As the crack interval increases the PIR decreases considerably, and when it becomes very large, the PIR achieves a relatively stable position as shown in Figure 2-4(b).  PIR significantly increases with increase in transverse crack length as shown in [see Figure 2-4(c)].  Crack open width has a significant effect on infiltration rate. Infiltration rate increased in quadratic polynomial form with an increase in open crack width as shown in Figure 2-4(d).  Thicknesses of pavement surface and drainage layers also impact PIR. However, the behavior of both layers is contrary to each other. With the increase in surface layer thickness, PIR decreases. Whereas, with an increase in drainage layer thickness, PIR increases. The reverse trend by both layers is observed, because the change in hydraulic gradient, which decreases with increase in surface course thickness, and increases with increase in drainage layer thickness [see Figure 2-4 (a) and (b)]. 21 /CLBB (a) PIR vs. transverse crack width (b) PIR vs. average crack distance for different transverse crack lengths (c) PIR vs. transverse crack length for different average crack distances (d) PIR vs. crack open width (e) PIR vs. surface course thickness (f) PIR vs. drainage layer thickness Figure 2-4 Model simulation results (5) 2.4.2.2 Hansson, K, Lundin, L. Charister and Simunek, J. Numerical Model Using Hydrus 2D for Modelling for Water Flow Patterns in Flexible Pavements In this study, proposed by Hansson et al. water flow patterns were simulated in flexible pavements. A numerical code built in Hydrus 2D software to depict simulations of water movement in pavement layers. Primarily, water movement due to rainfall was considered in this study. Special emphasis was given to three processes, the surface runoff followed by an 22 infilterarion through an asphalt fractured zone, the surface runoff with subsequent infilteration in the embankment, and capillary barrier effects between layers within the roads (34). The road section simulated is shown in Figure 2-5. Figure 2-5 Model road construction with material constructions, dimensions, and slopes (34) Authors used Richard’s equation to calculate water flow in unsaturated porous medium (34, 35). Different equations used for calculation of water flow, effective degree of saturation, retention curves, and effective hydraulic conductivity of a fracture zone are summarized below: (24) where, = Volumetric water content = Pressure head (L) = Hydraulic conductivity (L.T-1) = Time = Horizontal coordinate 23 hhkkhkhtxxzzzhktx (25) = Vertical coordinate, positive upward where, = Effective saturation = Residual water content = Saturated water content Authors used Van Genuchten analytical model to characterize retention curve(34, 35). (26) where, are empirical parameters. Following relationship by Van-Genuchten-Mualem (34-36) was used to describe unsaturated hydraulic conductivity. where, (27) (28) = Saturated hydraulic conductivity (L.T-1) = Pore connectivity parameter Finally, the effective hydraulic conductivity of fractured zone was obtained by Parallel plate model (34, 37). 24 zresatrSeSrsat11emnSh1,, Lmandn211mlmeseekSkSS11mnskl where, (29) = Fracture aperture = Distance between fractures = The density of water ( = Gravitational acceleration ( ) = Dynamic viscosity ( ) To visualize flow pattern, numerical simulations in Hydrus 2D were carried out using the particle tracking. Many hypothetical particles were released at different locations on the road surface, both at the embankment and fractured portion. No particles were released at intact asphalt surface because it was considered impermeable(34). Multiple simulations were planned to study the effect of rainfall amount, duration, and fracture conductivity (34).  30 mm rainfall amount was applied for 1,2,4 and 8 hours duration to visualize the effect of rainfall rate.  2.75,7.5 and 30 mm rainfall amounts were applied during 1hour duration to see precipitation amount impact,.  three fracture sizes were used as 0.5,0.1 and 0.01 mm, while studying the effect of varying fracture hydraulic conductivity. The precipitation for this simulation was 7.5 mm during a one-hour rainfall event. Following conclusions were made based on the simulations results: 25 22212fbgkB2b2B31000.)Kgmg29.82.ms311100*10..20KgmsatC  Varying precipitation rate had little effect on traveled particle distances at the end of simulations (i.e., three days after the rainfall), however, with higher precipitation rate, particles travelled farther (34) as shown in Figure 2-6. (g) Precipitated amount 3.75 mm (h) Precipitated amount 30 mm Figure 2-6 Water content distribution 3 days after onset of 1-hour rain event (34)  With small fracture aperture, i.e., 0.01 mm, Kf was considerably decreased and all the infiltration took place through the embankment. Whereas for higher Kf, as in case of 0.5 mm aperture the infiltration and particle movement took place right in the fractured zone as shown in Figure 2-7(34). (a) Fracture aperture 0.01 mm (b) Fracture aperture 0.5 mm Figure 2-7 Water content distribution 3 days after onset of 7.5 mm, 1-hour rain event (34)  It was observed that flow velocities were at peak at the end of a rainfall event, and most of the infiltration took place in initial few hours after the rainfall event. This phenomenon is well explained by particle movement. The particles traveled maximum distance in the 26 first couple of hours after the onset of rain. After three days of rainfall event, the increase in distance traveled was minute [see Figure 2-8]. (a) 2-Hours after the rainfall event (b) 3-Days after the rainfall event Figure 2-8 Water content distribution after the onset of 7.5 mm, 1-hour rain event (34) 2.4.2.3 Fan et al. Numerical Modelling of Unsaturated Granular Materials (UGM) in Flexible Pavements In this study, a new constitutive model for UGM was proposed, which captured both non-linear and moisture-sensitive characteristics of UGM. The proposed model was incorporated into finite element model for the base layer to quantify the influence of moisture content on the pavement performance (18). Lytton model was used to capture explanation of this behavior (18, 38, 39). (30) where, = Vertical modulus = First invariant of the stress tensor = Atmospheric pressure = Volumetric moisture content = Saturation factor, 27 23113kkmoctyaaaIfhEkpppyE1Iapf11f = Matric suction in aggregate base = Octahedral shear stress and = Regression coefficients For Lytton model validation, repeated load triaxial test lab results for three different materials at different moisture contents (at OMC and 1.5 OMC) were compared with predicted modulus. The results amply clarified the moisture sensitive and stress-dependent behavior of UGM, as shown in Figure 2-9. Figure 2-9 Comparison of predicted and measured resilient moduli for selected materials (18) In numerical models, different moisture conditions were simulated to investigate the effect of moisture content of UGM on pavement response. Three cases were considered, a low moisture condition with a degree of saturation 0.7, an optimum moisture condition (OMC) with a degree of saturation 0.85, and a saturated condition with a degree of saturation of 1. The results are shown in Figure 2-10 which indicates moisture content influence on UGM. 28 mhoct12,kk3k Figure 2-10 Vertical Moduli Distribution Base layer (18) 2.4.2.4 Resilient Modulus as a Function of Soil Moisture (EICM) The unbound base and subbase layers are an integral part of a pavement structure. Change in moisture content of subsurface layers can have an impact on the material properties (i.e., MR) of these layers. Consequently, the difference in material properties will affect the structural capacity of the whole pavement structure. The Pavement-ME Design Guide uses the moisture-modulus or 29 Witzack model to determine the variation in MR of the unbound layer with moisture change (40). where; (31) = Resilient modulus at the degree of saturation S (decimal) = Resilient modulus at the maximum dry density and optimum moisture = Minimum of log (MR/MROPT) = Maximum of log (MR/MROPT = -0.5934, 0.4 and 6.1324 for fine grained materials 0.3123, 0.3, and 6.8157 for coarse grained materials = Variation in the degree of saturation expressed in decimal 2.4.3 Summary of Existing Models from Literature Table 2-1 summarizes features, advantages, and disadvantages of different moisture content/infiltration prediction models found in the literature. 30 ()log1expln.()RoptRmoptMbaabMkSSaRMROPTMab, , mabandkOPTSS Table 2-1 Summary of existing models from literature Model Main Feature Advantages Disadvantages Swanberg and Hansen (26) Uses PL to calculate moisture content Single input, simple to use US Navy (27) Uses PL to calculate moisture content Single input, simple to use Kersten Model (28) Uses PL to calculate moisture content Single input, simple to use Only considered soil properties, surface conditions and climatic factors not considered Only considered soil properties, surface conditions and climatic factors not considered Large variation in prediction of moisture content for different soils Additional Comments Developed for SG layer Developed for SG layer of airfield pavements Empirical solution Climatic loading and surface conditions are not included in the final model Data from 18 different sites of Arkansas was used to calculate moisture in SG layer No consideration is given to surface cracking. Since the model is developed based on data from only one specific site, the regional application is limited. Moisture and precipitation data for two years of two-lane HMA road in North Texas was used to develop this model Comprehensive but complex in general for new users Derivation of expression is complex. Moreover, experimental and field investigation not yet validated A complex approach requires expertise in model simulation and defining boundary conditions Software Numerical solution Software-based Rao’s moisture content prediction model (15) Uses index properties like % passing No 200, LL, PI, Permeability Hedayati and Hossain- data-based model (16) EICM (40) Han et al. model (29) Uses one-dimensional partial differential equations as a function of time and depth and in situ precipitation and moisture data to predict moisture variation Comprises of three different models An analytical solution to quantify water balance between surface and drainage layer to estimate pavement infiltration rate Hansson et al. (34) solution Numerical code built in Hydrus 2D Index properties can be determined readily. Briefly discussed the effect of precipitation and temperature Considered seasonal variations and temporal changes comprehensively. Incorporated depth factor gives the flexibility to calculate moisture content for different pavement layers Currently used in MEPDG, results are widely accepted Incorporated surface discontinuities in the model. The final form of the solution is user-friendly Good simulation of water movement in pavement layers, especially with varying rainfall intensity, rain and aperture size (detecting cracks on the surface) 31 2.5 SUMMARY This chapter starts with the brief description of various sources causing the moisture change in pavement-unbound layers, particularly infiltration of rainfall through surface discontinuities (cracking and joint sealant damage). It provides discussion on moisture-related damage to the pavements and different procedures adopted for its mitigation. It was found in the literature that the moisture-related damage is significant, especially for the pavements located in areas with higher precipitation levels. This chapter also documents the moisture content modeling techniques found in literature, followed by a discussion on various empirical and analytical models available in the literature. Subsequently, it elaborates the moisture model used in Pavement-ME, which relates the unbound layers stiffness properties to moisture change. Finally, it provides the summary of moisture models along with pros and cons. Moisture variations adversely affect the pavement performance. Based on the literature review, true quantification of moisture variations within pavement unbound layers is warranted. 32 3 DATA SYNTHESIS 3.1 SEASONAL MONITORING PROGRAM (SMP) BACKGROUND Previous research highlighted that moisture variation within unbound layers is one of the leading factors for premature pavement deterioration (7, 13, 16). Therefore, the hypothesis of this study is that moisture variation in unbound layers, i.e., base layer, can be related to the amount of surface discontinuities (cracking and joint seal damage) in different climatic zones. To validate this hypothesis, an important challenge was to identify the data set documenting the subsurface moisture levels in the base layer. Only SMP study has TDRs installed at different depths in many pavement sections. In addition, the performance monitoring data were also recorded for those pavement sections. The SMP study was designed to characterize the magnitude and impact of temporal variations in pavement response and material properties due to the separate and combined effect of moisture, temperature and frost/thaw variations. It also includes higher monitoring frequency of deflections, longitudinal profile, and distress surveys on 64 SMP LTPP test sites, which were selected from GPS and SPS studies. In addition to performance data, other measurements—including subsurface moisture, temperature, rainfall, and surface elevations— were also recorded at these sites (41). The SMP study has a comprehensive database for subsurface moisture and temperature records. Because of its uniqueness, SMP data were identified as the best available source to quantify moisture damage in flexible and rigid pavements. 3.2 DATA SELECTION CRITERIA Various data elements from the SMP LTPP sections were reviewed and collected for further analyses to accomplish the objectives of this study. Of the particular interest was the data 33 assessment of SMP sites with an unbound base material having sufficient subsurface in-situ moisture, precipitation, and performance time series data. The SMP sections with at least three years or more subsurface moisture data were identified and used in the subsequent analyses. The timing of pavement maintenance actions was also considered for each section to obtain the amount of unsealed cracking and joint seal damage in a month. Time series of all the desired variables, (i.e., subsurface moisture, precipitation, and fatigue cracking) was considered during data analysis. As mentioned above, the SMP flexible and rigid sections with only unbound base layers were analyzed. 3.3 DATABASE DEVELOPMENT The required data were obtained from the LTPP database standard release 30.0. All SMP test sections were assigned with a unique ID by combining state code and SHRP ID. Multiple data buckets for desired variables were downloaded using online Infopave® features. The downloaded data elements were organized in various data tables to create a relational database. 3.4 DATA ELEMENTS The following data elements were identified for the analysis:  Section inventory o Sate code. o SHRP ID. o Site location. o Climatic region o Assign date. o Construction number. o Survey date 34  Pavement structure o Layer type. o Representative layer thicknesses. o Survey width. o Survey length.  Performance data o Flexible pavement sections.  Alligator cracking.  Longitudinal cracking wheel path (WP).  Longitudinal cracking non-wheel path (NWP).  Transverse cracking. o Rigid pavement sections.  Longitudinal joint sealant damage.  Transverse joint sealant damage.  Longitudinal and transverse cracking.  Climatic data o Subsurface moisture content. o Subsurface temperature. o Precipitation (rainfall and snow). o Freezing index. o Groundwater table depth.  Materials data o Sieve size analysis. 35 o Atterberg limits. o Specific gravity. Table 3-1 provides a summary of data types assessed in this investigation, along with the corresponding LTPP data tables containing the required data elements. 36 Table 3-1 LTPP data base tables used to extract data elements Type of data Data elements Relevant LTPP chosen tables Table description General LTPP section information inventory Layer thickness Structure and material type Sieve size analysis EXPERIMENT_ SECTION SECTION_LAY OUT SECTION_LAY ER_STRUCTUR E TST_SS01_UG01 _UG02 Material Atterberg limits TST_UG04_SS03 Specific Gravity TST_UNBOUND _SPEC_GRAV The three key fields that define a unique record in this table are STATE_CODE, SHRP_ID, and CONSTRUCTION_NO, which form the primary backbone of relational links within the LTPP database. This table contains section layout and location information. This table contains combined data from INV_ID, INV_GENERAL, SPS_ID, SPS_GENERAL, and SPS_PROJECT_STATIONS. It contains a consolidated set of pavement layer structure information for all LTPP test sections. This table contains the gradation of unbound coarse- grained granular base, subbase, and subgrade materials. This table contains the Atterberg limit test results for the unbound granular base, subbase, and subgrade materials This table contains the specific gravity of unbound base and subgrade materials. SMP_TDR_AUT O_MOISTURE This table contains the volumetric and gravimetric moisture contents calculated using TDR. Subsurface moisture content Climate Subsurface temperature Freezing index SMP_TDR_DEP THS_LENGTHS SMP_MRCTEM P_AUTO_HOUR SMP_MRCTEM P_DEPTH TRF_ESAL_INP UTS_SUMMAR Y Precipitation CLM_VWS_PRE CIP_MONTH Water table depth SMP_WATERT AB_DEPTH_MA N AC surface distresses MON_DIS_AC_ REV PCC surface distresses MON_DIS_JPCC _REV Performance This table contains information on the physical characteristics of the TDR probes, including the depth at which the probe is installed, the length of the probe, and its installation date. This table contains the vast majority of subsurface temperature data. It includes average hourly temperatures at a series of depths. This table contains the depths at which each temperature probe at an SMP section was installed and the date of installation. Contents of this table include Climate characterizations including average annual precipitation and freeze index, LTPP experimental climate region and the source for this classification. Virtual weather station monthly precipitation statistics and calculated parameters. The fields in this table are populated only when data for 24 or more days are available for a month. This table contains manual observations of the distance from the pavement surface to the water table. A null in the WATERTAB_DEPTH indicates that no water was found in the observation piezometer well. This table contains distress survey information obtained by manual inspection in the field for pavements with AC surfaces. This table contains distress survey information obtained by manual inspection in the field for jointed PCC pavements. 37 3.4.1 Pavement Performance Data Monthly surface distress data were obtained for all the flexible and rigid SMP pavement sections. Flexible pavement sections distress data included extent and severity of unsealed alligator, transverse, longitudinal wheelpath (WP) and non-wheelpath NWP cracking. The total cracking length for a flexible pavement section was calculated in meters by using following equation: where; (1) = Unsealed wheel-path cracking length ( ), includes alligator and longitudinal WP = Unsealed longitudinal cracking length outside wheel-path = Unsealed transverse cracking length = Total cracking length for a flexible pavement section (m) Rigid pavement sections distress data included extent and severity of unsealed longitudinal/transverse cracking and joint sealant damage. It was observed that longitudinal and transverse cracking magnitudes were very low in rigid pavements; therefore, only the length of joint sealant damage was used. While calculating the length of the damaged transverse joint seal, 5%, 25% and 50% of the joint seals were considered damaged for low (less than 10% damage), medium (10% to 50% damage), and high (more than 50% damage) severity transverse joint seal damage, respectively. The total PCC joint sealant damage length in meters was calculated by using Equation (2). (2) 38 lCRK = TotalCRKengthNWPWPLCTC CRKlengthWPmNWPLCmTCmCRKTotalSvySvySvyJSD = . W . 0.05 . W . 0.25 . W . 0.5TotalTranslowTransmedTranshighLongNDJNDJNDJLDJ where; = Number of low severity transverse joints with damaged joint sealant = Number of medium severity transverse joints with damaged joint sealant = Number of high severity transverse joints with damaged joint sealant = Length of longitudinal joints with damaged joint sealant (m) = Survey width (m) = Total length damaged joint sealant (m) 3.4.2 Subsurface Moisture and Temperature Time domain reflectometry (TDR i.e., moisture sensors) and thermistors (temperature sensors) were installed in all the SMP pavements sections to measure the in-situ subsurface moisture and temperature data at different depths. Also, the SMP database has volumetric and gravimetric moisture data at different depths (dry densities were used to convert volumetric moisture to gravimetric moisture content) (41). In this study, gravimetric moisture data were used for further analysis. Subsurface moisture and temperature data at the middle of the base layer were obtained from TDRs and thermistors installed at each site. To estimate the exact depth of moisture and temperature measurements within the base layer, unique section IDs were matched with TDR and thermistor numbers, respectively. For example, if middle depth of base layer is at 30 cm from surface (a=30 cm), then average of the moisture and temperature records, measured using TDR/thermistors located within ± 10 cm (b=10 cm) to the reference point (i.e., moisture and temperature was calculated by averaging the measurements taken by all the TDRs/thermistors installed between the depths of 20 to 40 cm) was calculated. However, often only one TDR or 39 TranslowNDJTransmedNDJTranshighNDJLongLDJSvyWJSDTotal thermistor was encountered within base layer for obtaining subsurface moisture and temperature data. This approach represents the moisture and temperature variations within the base layer. Figure 3-1 is showing the schematic of these calculations. Figure 3-1 Subsurface moisture and temperature measurements Table 3-2 presents the summary of SMP sections layer structure, subsurface moisture, and temperature depth, and the available number of years for the data elements. 40 HMABaseSubbaseNatural subgradeabbDHMADBaseDSubbase A+0.5DBaseb = 10 cmTDR/Thermistor Table 3-2 Layer structure and TDR/thermistors depths Unique ID State description Climatic regions Base type Base thickness (cm) Surface type Surface layer thickness (cm) Subsurface moisture availability (years) Mid of Base (cm) TDR depth (cm) Temperature availability (years) Thermistor depth (cm) WNF GB 01_0101 Alabama WNF GB 01_0102 Alabama DNF GB 04_0113 Arizona DNF GB 04_0114 Arizona DNF GB 04_1024 Arizona DF GB 08_1053 Colorado WF GB 09_1803 Connecticut WNF GB 10_0102 Delaware WNF GB 13_1005 Georgia WNF GB 13_1031 Georgia DF GB 16_1010 Idaho WF GB 23_1026 Maine WF GB 25_1002 Massachusetts WF GB 27_1018 Minnesota WF GB 27_6251 Minnesota DF GB 30_0114 Montana WF GB 31_0114 Nebraska 32_0101 Nevada DF GB 33_1001 New Hampshire WF GB DNF GB 35_1112 New Mexico WF GB 36_0801 New York 46_0804 South Dakota DF GB WNF GB 48_1060 Texas WNF GB 48_1077 Texas WNF GB 48_1122 Texas DNF GB 49_1001 Utah WF GB 50_1002 Vermont 51_0113 Virginia WNF GB WNF GB 51_0114 Virginia DF 56_1007 Wyoming GB WF GB 83_1801 Manitoba WF GB 87_1622 Ontario 04_0215 Arizona DNF GB WNF GB 13_3019 Georgia WF GB 18_3002 Indiana WF GB 27_4040 Minnesota 32_0204 Nevada DF GB 37_0201 North Carolina WNF GB 39_0204 Ohio WF GB WF GB 42_1606 Pennsylvania WNF GB 53_3813 Washington WF GB 83_3802 Manitoba 89_3015 Quebec WF GB 20.1 30.5 19.1 30.5 16.0 13.7 30.5 30.0 22.4 22.4 13.7 44.7 10.2 13.2 25.9 31.5 30.5 21.6 49.0 16.3 21.3 30.5 31.2 26.4 39.6 14.7 65.5 20.1 30.2 15.7 14.2 17.0 16.0 18.3 13.9 15.2 15.7 23.6 14.7 21.8 11.5 12.4 33.8 AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC PCC PCC PCC PCC PCC PCC PCC 18.8 10.7 12.4 18.5 27.9 17.3 22.6 14.0 19.3 28.7 27.7 22.9 19.8 16.3 22.9 19.6 16.3 18.3 26.4 15.7 12.7 23.6 19.1 12.7 9.4 15.2 21.6 10.2 22.4 8.9 11.2 19.3 27.9 22.6 28.4 20.6 29.9 23.4 28.2 PCC+AC 25.5+10.9 PCC+AC 20.3+13.2 PCC+AC 24.9+15.3 PCC+AC 20.9+23.9 3 3 3 4 4 3 5 5 5 4 5 5 5 5 4 7 5 4 5 5 7 7 3 6 5 5 7 4 6 5 7 5 3 3 3 3 3 5 3 5 4 3 8 28.0 25.5 22.0 32.5 35.3 18.4 33.5 36.0 29.5 43.5 35.6 37.3 24.1 23.9 29.0 33.5 34.0 27.5 51.8 25.4 24.0 32.8 32.0 30.5 24.6 26.4 28.8 25.9 22.0 33.8 35.9 24.1 37.8 29.0 30.5 39.9 34.5 45.2 24.9 22.9 35.8 35.3 31.5 29.1 50.9 23.9 23.4 38.9 34.7 25.9 29.2 22.6 54.4 Avg(36.8+52.1) 20.2 37.5 16.8 18.3 27.8 35.9 31.7 35.4 28.2 37.8 35.2 35.5 47.1 35.4 46.4 61.6 18.0 32.0 15.2 20.0 22.6 34.5 30.9 32.0 27.5 35.0 27.9 35.9 54.9 35.7 44.5 55.9 4 3 8 8 4 5 4 5 4 4 5 5 4 5 10 5 7 7 5 6 10 9 5 5 7 5 9 7 7 5 10 5 8 7 4 5 3 10 3 8 4 6 9 32.7 25.8 24.3 Avg (31.6+39.4) Avg (34.0+41.8) 24.5 39.2 Avg (30.3+38) 34.1 36.9 Avg (34.9+42.7) Avg (36.1+43.7) 25.7 37.7 38.5 37.8 37.6 32.2 49.6 Avg (20.9+28.2) 26.4 34.6 Avg (32.6+40.1) 25.7 Avg (27.58+35.2) Avg (22.1+30.0) 55.7 21.4 28.6 16.4 22.7 26.3 33.6 32.6 Avg (26.4+37.6) 29.6 32.6 37.2 36.8 49.2 32.4 48.7 56.4 Avg= average of the moisture and temperature data were obtained from all the available TDRs/Thermistors installed within base layer. 3.4.3 Precipitation Data Pavement performance temporal data were matched to obtain total monthly precipitation amount (i.e., rainfall and snow). Water infiltration followed by snow melting can substantially increase moisture levels within the pavements layers, especially in wet climates. Therefore, total monthly precipitation levels were calculated by adding rainfall and snow. 41 3.4.4 Ground Water Table Depth Capillarity action can also cause moisture change within pavements unbound layers. The depth of groundwater table (GWT) was obtained to isolate the effect of capillarity water, traveling from subgrade to base layer. Further, in the analysis part, GWT depth relationship was assessed with varying base layer moisture levels over time. 3.4.5 Freezing Index Average annual freezing index (FI) data were obtained to keep a record of freezing and no freezing regions while developing moisture prediction models. 3.4.6 Materials Data Material data elements were extracted by following the guidelines from the LTPP Information Management System materials module. Site-specific materials data were available for most of the SMP sites. Materials data needed to calculate base layer resilient modulus (MR) were obtained by combining unique ID and layer numbers. Linked SHRP IDs were used to obtain data for those SMP sections with missing site-specific material data. Sieve size distributions, Atterberg limits and specific gravity data elements were extracted by combining various data tables in the database. Sieve size analysis data were used to obtain D60 (the grain diameter at 60% passing). Figure 3-2 (a) and (b) show the base material particle size distribution for flexible and rigid pavement sections, respectively. Table 3-3 presents the summary of base layer material properties for flexible and rigid pavement sections. 42 (a) SMP flexible pavement sections (b) SMP rigid pavement sections Figure 3-2 Base material particle size distribution Table 3-3 Base layer material properties Unique _ID 01_0101 01_0102 04_0113 04_0114 04_1024 08_1053 09_1803 10_0102 13_1005 13_1031 16_1010 23_1026 25_1002 27_1018 27_6251 30_0114 31_0114 32_0101 33_1001 35_1112 36_0801 46_0804 48_1060 48_1077 48_1122 49_1001 50_1002 51_0113 51_0114 56_1007 83_1801 87_1622 04_0215 13_3019 18_3002 27_4040 32_0204 37_0201 39_0204 42_1606 53_3813 83_3802 89_3015 Climati c Region Surface Layer Base type s WNF WNF DNF DNF DNF DF WF WNF WNF WNF DF WF WF WF WF DF WF DF WF DNF WF DF WNF WNF WNF DNF WF WNF WNF DF WF WF DNF WNF WF WF DF WNF WF WF WNF WF WF AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC PCC PCC PCC PCC PCC PCC PCC PCC+AC PCC+AC PCC+AC PCC+AC GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB GB Percent passing # 200 11.5 11.5 7.1 7.6 9.3 8.9 9.6 19.1 7.8 10.7 7.8 4.0 6.9 7.7 9.9 8.2 6.2 12.4 4.6 14.7 8.1 5.9 7.1 9.3 21.7 8.6 3.4 11.0 11.1 8.6 9.6 7.4 8 25.6 4.1 14 8.9 8.8 13.4 10.2 17.5 10.5 3.7 D60 (mm) 9.2 9.2 7.8 7.4 14.9 7.8 3.8 8.1 9.9 0.9 8.1 43.5 11.2 2.5 1.3 8.3 5.0 9.9 14.3 2.2 17.4 8.6 7.2 12.5 5.4 8.7 26.7 6.6 9.2 6.0 5.0 5.0 7.4 13.8 10.7 2.1 13 7.8 7.7 12.5 3.6 7.9 13.8 Specif ic Gravit y 2.87 2.87 2.72 2.72 2.70 2.65 2.65 2.85 2.65 2.70 2.65 2.65 2.65 2.65 2.65 2.65 2.65 2.70 2.68 2.55 2.83 2.71 2.61 2.60 2.58 2.65 2.65 2.63 2.63 2.65 2.65 2.69 2.71 2.61 2.65 2.65 2.65 2.76 2.74 2.7 2.65 2.65 2.65 PI NP NP NP NP NP NP NP NP NP NP NP NP NP NP NP NP NP NP NP 7 NP NP NP NP NP NP NP NP NP NP 3 NP NP NP NP NP NP NP NP 6 NP NP NP 43 Material type 303-Crushed Stone 303-Crushed Stone 304-Crushed Gravel 304-Crushed Gravel 304-Crushed Gravel 304-Crushed Gravel 302-Gravel (Uncrushed) 303-Crushed Stone 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 309-Fine-Grained Soils 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 302-Gravel (Uncrushed) 304-Crushed Gravel 302-Gravel (Uncrushed) 302-Gravel (Uncrushed) 304-Crushed Gravel 303-Crushed Stone 304-Crushed Gravel 302-Gravel (Uncrushed 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 304-Crushed Gravel 303-Crushed Stone 303-Crushed Stone 303-Crushed Stone 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 304-Crushed Gravel 304-Crushed Gravel 303-Crushed Stone 303-Crushed Stone 304-Crushed Gravel 302-Gravel (Uncrushed) 304-Crushed Gravel 304-Crushed Gravel 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 303-Crushed Stone 302-Gravel (Uncrushed) 304-Crushed Gravel 303-Crushed Stone 303-Crushed Stone 304-Crushed Gravel 308-Soil-Aggregate Mixture (Predominantly Coarse-Grained) 304-Crushed Gravel 303-Crushed Stone 01020304050607080901000.010.1110100Percent passing (%)Particle size (mm)01020304050607080901000.010.1110100Percent passing (%)Particle size (mm) 3.5 DATA LIMITATIONS Since time series of all desired variables (subsurface moisture, cracking, and precipitation) had to be matched on a monthly basis, a considerable amount of data points were not used in further data analysis because either time series did not match or data were not available at required depths. The database was shortened further by eliminating SMP sections with treated bases. Finally, SMP sections with less than two years of temporal data were excluded which further reduced the available number of SMP pavement sections. 3.6 AVAILABLE SMP SECTIONS FOR ANALYSIS SMP pavement sections, which satisfied the data selection criteria, were reviewed for quality, reasonableness, and availability in the light of supporting the moisture variation impact on long- term pavement performance. Because of data cleaning, 32 SMP sections were identified with an adequate amount of data for flexible pavements, and 11 SMP sections for rigid pavements. Table 3-4 presents the summary of data elimination process. Figure 3-3 presents the climatic summary of total and available SMP pavement sections considered for this research. Table 3-4 Number of available SMP LTPP pavement sections Surface type Moisture content Temperature Precipitation AC PCC 43 21 43 21 43 21 Freezing index Performance Sites with granular base Time series mismatch/d ata (less than three years) Number of available sections 43 21 43 21 38 11 6 0 32 11 44 Figure 3-3 Climatic distribution of SMP LTPP sections 3.7 SUMMARY Seasonal Monitoring Program (SMP) study in the LTPP program primarily was designed to investigate the combined impact of temperature, moisture and frost/thaw variations on pavement material properties, response and performance. Data from SDR 30.0 (the most up to date at the time of this study was conducted) were obtained for this study. Flexible and rigid pavements sections with granular bases and at least three years or more of performance and moisture data were considered for the analysis. In flexible pavement sections, the total length of unsealed cracking was calculated by adding lengths of transverse, longitudinal, and fatigue cracking. All severity levels, i.e., low, medium, and high were added while calculating extents of cracking. In PCC SMP sections, while calculating length of damaged transverse joint seal, 5%, 25% and 50% of the joint seals were considered damaged for low (less than 10%), medium (10% to 50%), and high (more than 50%) severity transverse joint seal damage, respectively. Total monthly precipitation levels were calculated by adding rainfall and snow. Subsurface moisture and 45 13169511106598226311024681012141618WFWNFDFDNFNo of LTPP sectionsClimatic regionsAC totalAC availablePCC totalPCC available temperature data at the middle of the base layer were obtained from time domain reflectometry (TDR) and thermistors for each pavement section. Exact depths of subsurface moisture and temperature measurements within base layer were estimated by combining unique section IDs with TDR and thermistor numbers, respectively. Pavement construction numbers were also recorded to quantify the exact amount of unsealed cracking/joint seal damage in a month. Material data elements were extracted by following the guidelines provided in the LTPP Information Management System materials module. Materials data needed for base layer MR calculations were obtained by combining unique ID and layer numbers. Site-specific materials data were available for most of the SMP test sections. Linked SHRP IDs were used to calculate data for those SMP sections with missing site-specific material data. Since time series of all desired variables (subsurface moisture, cracking, and precipitation) had to be matched on a monthly basis, a considerable amount of data points were not used in further data analysis because either time series did not match or data were not available at required depths. With data elimination process 32 flexible, and 11 rigid pavement sections were identified with appropriate data for further analysis. 46 4 DATA ANALYSIS AND MODELING 4.1 HYPOTHESIS The past research has defined that moisture variation within pavement unbound layers is one of the leading factors for premature pavement deterioration (7, 9, 13, 16). This fluctuation in moisture can be estimated by analyzing the subsurface moisture data available in SMP study. After analyzing the moisture and performance data for few SMP pavements sections, it was hypothesized that variation in subsurface moisture, essentially within base layer, can be related to the extents of surface discontinuities in different climatic zones. Subsequently, by looking at the data and through descriptive statistics, a few other factors that may cause potential moisture change within pavement base layers were identified. These factors affiliated with pavement structure, materials, and climate, were used as covariates while estimating subsurface moisture content. Factors initially considered for the analysis are described below:  Pavement age  Surface discontinuities (cracking and joint sealant damage)  Subsurface temperature  Precipitation (rainfall and snow)  Number of wet days  Moisture depth  The thickness of pavement structure above the base  Percent passing sieve number 200  Freezing index (FI)  Groundwater table (GWT) depth 47 Figure 4-1 illustrates the effect of cracking and precipitation on base layer moisture for the SMP section 36-0801 located in WF climate. The data shows that when the pavement section is new with minimal cracking, even with the higher amount of precipitation base layer moisture did not vary much, and only showed a cyclic trend. However, as the cracking extents increased over time, moisture content changed significantly even at lower precipitation levels. This moisture change is accumulative and primarily caused by water infiltration through surface cracks. Fluctuation in groundwater table (GWT) depth may also cause seasonal variations in unbound layers in-situ moisture content. GWT depth records over time were obtained to separate the moisture variations associated with a change in GWT from surface infiltration. Seasonal fluctuations in GWT depth adversely affect deeper layer material properties, essentially up to subgrade and subbase layers, and it will have little effect on the base layer in-situ moisture. GWT and subsurface moisture is plotted for one flexible SMP pavement section located in WF climate as shown in Figure 4-2. It is observed from the relationship that when the pavement is new (initial 3-4 years of service life), the variation in base layer moisture is cyclic, even at times the GWT is very high (i.e., lower GWT depth). On the other hand, when the pavement gets older (7- 8 years of service), the variations in moisture are significant for almost same levels of GWT, or even for very low GWT depth (between 6 and 8 years). This evidence supports the hypothesis that main cause of base layer moisture fluctuation is the infiltration though surface discontinuities followed by rainfall. 48 Figure 4-1 Impact of cracking and precipitation on base layer moisture change (36-0801) Figure 4-2 Effect of GWT on base layer moisture change (36-0801) For the same pavement section moisture profile with depth and age is also shown in Figure 4-3. It can be observed that moisture variations are high at the top of pavement structure, i.e., within 49 0.00.20.40.60.81.002468Normalized dataPavement age (years)CrackingPrecipitationMoisture content00.511.522.5302468101214161820012345678GWT depth (m)Moisture content (%)Pavement age (years)Moisture content (%)GWT depth (m) base and subbase layers, and with an increase in depth, these changes become negligible. A similar trend in moisture change was observed in most of the SMP test sections. Figure 4-3 Subsurface moisture variations with depth (36-0801) 4.2 METHODOLOGY As mentioned before, many external and internal sources can cause the subsurface moisture variations in pavement unbound layers (42). Surface discontinuities such as cracks or joint openings allow water to infiltrate in sublayers. Bottom-up fatigue is a classic example of through cracking that would allow the surface water to infiltrate into the base layer. However, the amount of water infiltration is expected to be more on locations with higher precipitation levels. In this study, the amount of surface cracking (joint seal damage in the case of rigid pavements) in flexible pavements over time was related to seasonal moisture levels at different depths of the pavement structure. The primary objective is to identify the additional amount of moisture in the sublayers due to change in surface cracking extent over time in different climates. Subsequently, material properties (i.e., MR) can be related to different moisture levels. The developed models can assist highway agencies in proactive maintenance practices to mitigate 50 0510152025Moisture content (%)SMP survey (years)24 cm39 cm54 cm70 cm84 cm100 cm113 cm130 cm161 cm192 cm moisture-related damage due to surface cracking. The agencies can estimate the maximum cracking extent at which the cracks should be sealed to reduce the water infiltration rate into sublayers. 4.3 DESCRIPTIVE STATICS Summary of descriptive statics for flexible and rigid SMP LTPP sites is given in Table 4-1. The data extents show that cracking, precipitation, and subsurface moisture levels are very high in wet climates. Table 4-1 Summary of regional climatic and performance data Surface type Climate Cracking (m)/Joint seal damage* (m) Precipitation (rainfall +snow) (cm) Temperature (°C) moisture content Freezing index Gravimetric (%) Max Min Max Min Max Min Max Min AC PCC DF DNF WF WNF DF DNF WF WNF 213 189 461 358 40 159 172 215 0 0 0 0 6 6 4 7 46 14 75 27 3 3 37 32 0 0 2 0 1 0 1 0 33 39 28 38 33 39 28 38 -2 2 -14 4 -2 2 -14 4 12 14 19 23 9 12 28 21 2 2 3 4 8 11 2 4 Max 986 108 Min 215 1 1729 194 76 214 1 1684 32 0 214 1 299 12 * Longitudinal and transverse Joint seal damage in case of PCC pavements. Figure 4-4 shows cracking progression for flexible pavement sections located in different climates. As compared to DF/DNF, greater cracking extents were observed in WF/WNF regions. 51 (a) DF (b) DNF (c) WF (d) WNF Figure 4-4 Cracking progression with age in flexible pavements sections Figure 4-5 shows rigid pavements cracking and joint seal damage progression with age in different climates. Due to a limited number of PCC sections, SMP sections located in DF/DNF and WF/WNF regions were combined. As compared to DF/DNF, much greater cracking extents were observed in WF/WNF regions. Additionally, as compared to longitudinal and transverse cracking, the joint sealant damage extents were significantly high. 52 010020030040050003691215Total cracking (m)Pavement age (years)010020030040050003691215Total cracking (m)Pavement age (years)010020030040050003691215Total cracking (m)Pavement age (years)010020030040050003691215Total cracking (m)Pavement age (years) (b)WF/WNF cracking only (a) DF/DNF cracking only (c) DF/DNF joint damage only (d) WF/WNF joint damage only (e) DF/DNF cracking and joint (f) WF/WNF cracking and joint damage (combined) damage (combined) Figure 4-5 Cracking progression with age in rigid pavements sections 53 05010015020025030003691215Long and trans CRK only (m)Pavement age (years)05010015020025030003691215Long and trans CRK only (m)Pavement age (years)05010015020025030003691215Joint seal damage (m)Pavement age (years)05010015020025030003691215Joint seal damage (m)Pavement age (years)05010015020025030003691215CRK and joint damage (m)Pavement age (years)05010015020025030003691215CRK and joint damage (m)Pavement age (years) Figure 4-6 shows the precipitation extents for flexible and rigid SMP pavement sections located in different climates. As compared to dry climates, higher precipitation levels were observed in wet climates. (a) DF/DNF AC sites (b) WF/WNF AC sites (c) DF/DNF PCC sites (d) WF/WNF PCC sites Figure 4-6 Precipitation levels in different climates Figure 4-7 shows the base layer moisture variations with age for the flexible pavements SMP sections located in different climates. As compared to dry regions, subsurface moisture greatly fluctuated for the SMP sites located in wet climates. 54 01,0002,0003,0004,0005,0006,000Total annual percipitation (mm)SMP survey (year)01,0002,0003,0004,0005,0006,000Total annual percipitation (mm)SMP survey (year)01,0002,0003,0004,0005,0006,000Total annual percipitation (mm)SMP survey (year)01,0002,0003,0004,0005,0006,000Total annual percipitation (mm)SMP survey (year) (a) DF (b) DNF (c) WF (d) WNF Figure 4-7 Moisture variations in base layer — flexible SMP sections Figure 4-8 shows the base layer moisture variations with age for the rigid pavements SMP sections located in different climates. Similar to flexible pavements sections, higher moisture fluctuations are observed for the rigid pavements sections located in wet climates. 55 0510152025Jan-93Oct-95Jul-98Apr-01Jan-04Moisture content (%)SMP survey (month)0510152025Jan-93Oct-95Jul-98Apr-01Jan-04Moisture content (%)SMP survey (month)0510152025Jan-93Oct-95Jul-98Apr-01Jan-04Moisture content (%)SMP survey (month)0510152025Jan-93Oct-95Jul-98Apr-01Jan-04Moisture content (%)SMP survey (month) (a) DF/DNF (b) WF/WNF Figure 4-8 Moisture variations in base layer — rigid SMP sections 4.4 IDENTIFYING SIGNIFICANT VARIABLES Correlation matrix between different variables in flexible and rigid SMP pavements sections is given in Table 4-2 and Table 4-3, respectively. Table 4-2 Correlation matrix flexible pavements sections 56 051015202530Jan-93Oct-95Jul-98Apr-01Jan-04Moisture cotent (%)SMP survey (month)051015202530Jan-93Oct-95Jul-98Apr-01Jan-04Moisture content (%)SMP survey (month)Moisture contentAgeCrackingHMA thicknessP200MC depthFITempPrecipitation10.231-0.069-0.0690.172-0.120-0.1300.1560.0270.00060.31110.31460.01140.07890.05660.02240.69310.23110.4080.0820.0760.0320.1280.041-0.0230.0006<.00010.23140.26980.64570.06170.55090.7368-0.0690.40810.100-0.283-0.0220.486-0.1640.1100.3111<.00010.1441<.00010.7472<.00010.01630.1065-0.0690.0820.1001-0.3550.7520.026-0.1260.1450.31460.23140.1441<.0001<.00010.70660.06610.03360.1720.076-0.283-0.3551-0.281-0.3170.176-0.1330.01140.2698<.0001<.0001<.0001<.00010.00980.0516-0.1200.032-0.0220.752-0.2811-0.0930.0200.0950.07890.64570.7472<.0001<.00010.17340.77070.1646-0.1300.1280.4860.026-0.317-0.0931-0.2770.2670.05660.0617<.00010.7066<.00010.1734<.0001<.00010.1560.041-0.164-0.1260.1760.020-0.2771-0.3800.02240.55090.01630.06610.00980.7707<.0001<.00010.027-0.0230.1100.145-0.1330.0950.267-0.38010.69310.73680.10650.03360.05160.1646<.0001<.0001PrecipitationHMA thicknessP200MC depthFITempPearson Correlation Coefficients, N = 215Prob > |r| under H0: Rho=0Moisture contentAgeCracking Table 4-3 Correlation matrix rigid pavements sections The correlations of moisture content with independent variables were not very strong. However, by running forward and backward model selection in statistical analysis software (SAS) and then by extensively running the genetic algorithm, following variables were identified for accurate estimation of moisture variation in the base layer.  Surface cracking  Moisture depth  P200  Precipitation  FI  Subsurface temperature Moisture depth and HMA/PCC layer thicknesses were highly correlated, therefore considering the relationship with subsurface moisture, only moisture depth was included in 57 Moisture contentAgeJoint seal damagelong and trans Crackingcombined cracking and jointdamgePCC thicknessP200MC depthFITempPrecipitation10.3360.664-0.0380.5930.1550.4250.150-0.184-0.1190.1150.014<.00010.786<.00010.2680.0020.2850.1880.3980.4110.33610.274-0.0380.2390.1480.2800.1690.294-0.2620.3240.0140.0470.7860.0850.2910.0430.2280.0330.0580.0180.6640.27410.2770.9730.3920.3200.285-0.306-0.0120.158<.00010.0470.045<.00010.0040.0200.0390.0260.9310.260-0.038-0.0380.27710.4920.558-0.1390.4620.093-0.0500.0080.7860.7860.0450.000<.00010.3200.0010.5100.7220.9530.5930.2390.9730.49210.4900.2560.369-0.255-0.0230.145<.00010.085<.00010.0000.0000.0640.0070.0650.8700.3010.1550.1480.3920.5580.4901-0.4090.8990.333-0.1480.0500.2680.2910.004<.00010.0000.002<.00010.0150.2890.7200.4250.2800.320-0.1390.256-0.4091-0.427-0.347-0.0450.2560.0020.0430.0200.3200.0640.0020.0010.0110.7490.0640.1500.1690.2850.4620.3690.899-0.42710.340-0.126-0.0290.2850.2280.0390.0010.007<.00010.0010.0130.3680.835-0.1840.294-0.3060.093-0.2550.333-0.3470.3401-0.2690.0550.1880.0330.0260.5100.0650.0150.0110.0130.0520.696-0.119-0.262-0.012-0.050-0.023-0.148-0.045-0.126-0.2691-0.7010.3980.0580.9310.7220.8700.2890.7490.3680.052<.00010.1150.3240.1580.0080.1450.0500.256-0.0290.055-0.70110.4110.0180.2600.9530.3010.7200.0640.8350.696<.0001TempPrecipitationCobined cracking and joint damgePCC thicknessP200MC depthFIPearson Correlation Coefficients, N = 53Prob > |r| under H0: Rho=0Moisture contentAgeJoint seal damageLong and trans cracking further modeling. Freezing index was included as an independent variable to keep a record of freeze and no freeze regions. 4.5 DEVELOPMENT OF EMPIRICAL MODELS As highlighted earlier, the main objective of this study is to investigate the additional amount of moisture in the pavement base layer due to infiltration of water through surface cracks in different climates. SMP data in LTPP is highly scattered due to large variations in climate, material, and pavement structure. With preliminary correlations, significant variables like surface cracking, joint seal damage, precipitation, subsurface temperature, moisture depth, and % passing No.200, were identified which could probably cause a change in base layer moisture content. Different multilinear, nonlinear and polynomial regression techniques were used to develop the relationship between independent variables and subsurface moisture content. However, due to the complexity and great variation within the data, none of these procedures yielded desired results. Finally, Artificial Neural Network (ANNs) were used to model the data and it gave reasonable results with an acceptable degree of error. 4.6 FLEXIBLE PAVEMENTS MODELING This section presents the subsurface moisture prediction models developed to estimate base layer in-situ moisture content. It also documents the potential impacts of subsurface moisture variations on base MR. Subsequently, the influence of base MR on long-term pavement performance in terms of predicted cracking are discussed. Based on the results, appropriate crack sealing application timings are recommended to extend the service life of flexible pavements in different climates. 58 4.6.1 Site-Specific Models for Flexible Pavements As the first step, data from individual SMP pavements sections were used to develop empirical correlations. The site-specific models gave a good insight of the moisture variation phenomenon in the base layer; however, due to a typical climate and material type these models lacked potential of the universal application. In the beginning, separate models were developed for wet and dry climates. Eureqa (genetic algorithm) (43) toolbox was used to establish a relationship for base layer moisture content as a function of surface cracking, precipitation, and subsurface temperatures. Equation (1) shows the model developed for DF/DNF region using data from two SMP sites. where, = Gravimetric moisture content (%) = Precipitation (cm) = Total monthly Cracking (cm) = Average monthly temperature (oC) (1) Equation (2) shows the model developed for WF region using data from one SMP section. (2) Figure 4-9 shows the goodness of fit for both the models. 59 220.0005170.0005177.570.3380.003290.004940.0004580.0691TTMCPCTPPTPTMCPCT6.410.6010.0160.0710.000708MCPCPTCT (a) DF/DNF climate (b) WF climate Figure 4-9 Measured Vs. predicted site-specific models for flexible pavements Figure 4-9 (a) shows that variation of moisture is very small in dry climates. To address greater variability and limited applicability of site-specific moisture prediction models, the scope of data modeling was expanded by adding data from all available flexible SMP pavement sections for further data analysis and modeling. This was a very challenging task because the variety and extents of climatic, material and pavements structure variations. Finally, five independent variables were chosen for ANN modeling. Subsurface temperature data were not used in further analysis due to its insignificance, and to reduce the number of independent variables. 4.6.2 ANN Modeling Flexible Pavements ANNs are computational modeling tools that have lately emerged and found extensive acceptance in many disciplines for handling very complex problems. They can be defined as structures consist of tightly interconnected processing elements (called artificial neurons or nodes) operating in parallel (44, 45). ANNs are capable of solving non-linear problems by acquiring information and restructuring the relationship between independent variables and response variables even when the information and data are complex, noise-contaminated, and incomplete (46). ANN is an information processing system that replicates functioning of a 60 7.07.37.67.98.28.57.07.37.67.98.28.5Predicted moisture content (%)Measured moisture content (%)n = 6 R2 = 0.6905101520250510152025Predicted Moisture content (%)Measured moisture content (%)n = 12R2 = 0.81 human brain by emulating the functioning and connectivity of biological neurons (47, 48). ANN does not need much of detailed description or formulation of the underlying process, and thus widely received by practitioners and researchers, who tend to rely on data. Depending on the network structure, usually, a series of connecting neuron weights are altered to reduce the error between training data outputs and the network predicted outputs (49). When a neuron weight is adjusted, it is said that the neuron is learning. The training is the process through which NN learns. Depending on the complexity of the data and intended use, ANN can be composed of one or more hidden layers (50, 51). More discussion on ANN training can be found elsewhere (47). In the current study, ANN fitting app in MATLAB toolbox was used to establish a relationship for base layer moisture content as a function of surface cracking, precipitation, moisture depth, the percentage passing #200, and freezing index (FI). Since ANN toolbox is equipped with flexible hidden layer and neuron features, very complex trends in the data can be captured by selecting the best layer and neurons combination. Using LTPP SMP data, multi-layer perceptron (MLP) (single input layer, single hidden layer, and single output layer) feedforward-backpropagation artificial neural network (BPNN) was developed with hidden sigmoid neurons and linear output neurons. A feedforward NN consists of series of layers. The first layer has a connection from the network input. Each subsequent layer has a connection from the previous layer. The final layer gives the network’s output. Feedforward networks can be used for any kind of input to output mapping (52). ANNs toolbox in MATLAB provides different features and apps to deal with complex nonlinear systems that are not easily modeled with a closed-form equation (53). The main network architecture is the number of hidden layers and number of neurons (NoN). The selection of these parameters largely depends on the complexity of the data inputs 61 used for training. If the NoN are too low, the network may not capture the real trends in the data. If the NoN are too high, it may over fit the data. There is no exact guide for the choice of the NoN, and the optimum model design is often achieved by trial and error (47, 54). MATLAB ANN fitting app also provides different options for network training and layer activation functions; those are chosen based on available memory, computational speed, and research needs. The aim of the best suitable training function is to train the network at relatively fast speed with high precision. Levenberg - Marquardt backpropagation (trainlm), Bayesian regularization backpropagation (trainbr), and Scaled conjugate gradient backpropagation (trainscg) are widely used network training functions available in MATLAB ANN toolbox, whereas main transfer functions are, Hyperbolic tangent sigmoid transfer function (tansig), Log- sigmoid transfer function (logsig), and Linear transfer function (purelin).To develop ANN model for this study trainlm was used to train the network, tansig and purelin activation functions were used for the hidden and output layer neurons, respectively. Detail description of training and transfer functions and related algorithms is given elsewhere (52). The layer activation and network training functions used for this study are briefly discussed in subsequent paras. 4.6.2.1 Network Training Function ⸺ trainlm Trainlm is a network training function in MATLAB toolbox that adjusts the weights and bias values according to Levenberg-Marquardt algorithm (LMA) optimization. LMA or just LM, also known as damped least-square (DLS) method is used to address non-linear least square problems. It is often the fastest backpropagation algorithm in the MATLAB toolbox, as is highly recommended as a first choice supervised algorithm, though it requires more memory than other 62 training algorithms(52). As opposed to unsupervised training function, a supervised training algorithm requires target (response variable) data (55). Like the quasi-Newton methods (56), the LMA was developed to approach second-order training speed without having to compute the Hessian matrix. The Hessian matrix or Hessian is a square matrix of second-order partial derivatives of a scalar-valued function, or scalar field which describes the local curvature of a function of many variables (57). LMA uses the Jacobian for calculations, therefore the network trained with this function must use either mean squared error (MSE) or the sum of squared errors (SSE) as performance function. When the performance function is SSE (52), as mostly the case for feed-forward networks, then Hessian matrix (H) and the gradient (g) can be estimated as: (3) (4) Where J is the Jacobian matrix contains the first derivative of the network errors concerning the weights and biases, and e is the vector of network errors. The Jacobean matrix is computed through a standard backpropagation technique that is much less complex than computing the Hessian matrix. The Levenberg-Marquardt algorithm uses this approximation to the Hessian matrix in the following Newton-like update (58): (5) When the scalar µ is zero, this is just Newton’s method, using the approximate Hessian matrix. When µ is large, this becomes gradient descent with small step size. Newton’s method is faster and more accurate near an error minimum, so the aim is to shift toward Newton’s method as quickly as possible. Thus, µ is decreased after each successful step (reduction in performance function) and is increased only when a tentative step would increase the performance function. In 63 THJJTgJe11[]TTkkXXJJIJe this way, the performance function is always reduced at each iteration of the algorithm. Further discussion on LMA and its application in NN training is described elsewhere (59, 60). 4.6.2.2 Hidden and Output Layer Transfer Function ⸺ (tansig and purelin) Transfer functions are generally allocated to a network layer to first start the input signal, followed by the calculation of appropriate weight for the output signal such that the relationship between the input and target data can be ascertained (46). Mathematical expressions for mainly used transfer functions in ANN models given in the following equations: (6) (7) (8) Logsig yields output in the range of 0 to 1, tansig yields output in the range of -1 and +1, and purelin yields output in the range of - to + (47, 61). Sigmoid transfer functions are usually used in hidden layers, and linear functions are used in the output layer. The hyperbolic tangent sigmoid transfer function (tansig) is mathematically equivalent of tanh(n). In NN, it is widely used as hidden layer activation function; it runs faster than the MATLAB application of tanh(n) with very small numerical differences. This function is a good tradeoff for NN applications, where speed is important, and the exact shape of the transfer function is not (62, 63). Linear transfer (purelin) is mainly used as the output layer transfer function for function fitting (or nonlinear regression) problems. Logsig is frequently used in output layer for pattern recognition problems(in which decision is made by the network), where the output range is between 0 and 1 (52). Variations in layer activation functions are primarily due to the high ANN sensitivity to the type of data. Likewise, different data inputs would require different activation 64 (2)2tan () = 11nsigne1log () = ()1nsigne () = purelinnn functions in ANN architecture. No single network settings can be universally applied to model different types of problem situations effectively (46). Besides many pros, due to the inherent complexity of ANN models, they are not easy to interpret and understand. Optimum settings for the ANN model that was developed using data from 32 flexible SMP sites are given in Table 4-4. However; the best network configuration may vary from case to case and largely depend on input/output data type and complexity. Figure 4-10 shows the schematic of ANN model developed for the flexible pavements sections. Table 4-4 Optimum settings for the flexible pavements ANN model Network type No of hidden layer Data entries for training, testing, and validation Training function Hidden layer transfer function Output layer transfer function Performance function No of hidden neurons BPNN 1 162,35,35 trainlm tansig purelin MSE 37 Figure 4-10 ANN model flow for flexible pavements SMP sections Figure 4-11(a) shows the goodness of fit for ANN model. Figure 4-11(b), (c), and (d) show model sensitivity to different inputs. The results of the ANN model sensitivity show that with an increase in surface cracking, there is an increase in the base moisture levels. This change in 65 INPUTHIDDEN LAYEROUTPUT LAYEROUTPUTP 5x1WO1x37tansigpurelinbO1x1WH37x5bH37x1nH37x1nO1x1aH37x1aO1x1Y 1x1aH= tansig(WHp+bH)aO= purelin(WOaH+bO) moisture is significant in WF/WNF climates with higher precipitation levels [see Figure 4-11(b)]. Higher the percentage passing 200, higher is the moisture, higher the depth of moisture within the base layer, lower is the moisture [see Figure 4-11 (c) and (d)]. It can be seen in Figure 4-11(d) that effect of moisture content depth is minimal. It is worth noting that when the cracking reaches approximately 70 to100 meters in length, moisture content increases exponentially. (a) Measured Vs predicted (b) Effect of Precipitation and freezing (e) Impact of percent passing 200 (f) Impact of MC depth Figure 4-11 ANN model predictions and sensitivity — flexible pavements To see the effect of precipitation alone, its levels were varied between 0 and 75 cm for ANN predictions in wet regions. The ANN model predictions show that moisture increases with 66 0481216202404812162024PredictedMeasuredy = 0.74x+1.9918R2 = 0.73n =21505101520253035050100150200250Moisture content (%)Cracking (m)DNFDFWNFWF05101520253035050100150200250Moisture content (%)Cracking (m)8%10%12%14%05101520253035050100150200250Moisture content (%)Cracking (m)15 cm20 cm25 cm30 cm increase in precipitation levels up to a certain limit, and then the effect of precipitation becomes negligibly. This implies that after a certain amount of precipitation base layer reaches saturation and a further increase in precipitation may not cause much moisture variation [see Figure 4-12]. Figure 4-12 Effect of precipitation on moisture variations Figure 4-13 shows moisture profiles with depth for two SMP sections before and after a considerable amount of cracking. These profiles help to visualize the overall moisture variation in the pavement system for the sites located in wet and dry regions. It is evident from the moisture profiles that change in moisture is more pronounced at the top few centimeters of the pavement section, i.e., up to base/subbase layers. The relative change in moisture becomes negligible below subbase levels. Figure 4-13 (a) and (b) show moisture variation in base layer for two pavement sections located in WF and DF climates, respectively. More substantial moisture variations in WF climate are mainly because of the higher rainfall and greater extent of cracking for the selected pavement sections. In contrast, the overall change in base layer moisture for the 67 05101520253035050100150200250Moisture content (%)Cracking (m)0 cm5 cm10 cm15 cm20 cm25 cm30 cm35 cm40 cm45 cm50 cm55 cm60 cm65 cm selected site located in DF climate is minimal. It is mainly because of low precipitation in these particular locations [see Figure 4-13 (b)]. (a) 36-0801 (WF) (b) 56-1007 (DF) Figure 4-13 Moisture variations with depth in DF/WF region 4.6.3 Impact of Base Moisture on Long-Term Performance The surface cracks increase the infiltration of water into pavement sublayers. Therefore, the moisture levels will increase in the unbound materials. Moisture variations in base layer will affect the MR of the unbound materials. The SMP sections show that there can be significant variations in base layer moisture, especially in wet climates. The moisture changes can be used to obtain MR of the base material. Subsequently, the calculated MR can be used to predict long- term pavement performance by using the Pavement-ME Design Guide. A brief discussion of these effects is provided next. 4.6.3.1 The Relationship between Base Moisture and Base MR- Flexible Pavements Witzack model (briefly discussed in Chapter 2) was used to calculate the base layer MR due to variation in in-situ moisture. This model needs several inputs, including % passing 200, LL, PL, 68 020406080100120140160180200051015202530TDR depth (cm)Moisture content (%)20011994020406080100120140160180200051015202530TDR depth (cm)Moisture content (%)19971993 D60, and Gs. All the required inputs were obtained from the LTPP database for all the sections. The estimated MR values for base materials with a change in moisture levels based on Witzack model are shown in Figure 4-14. The results show that as the moisture increases, the base MR decreases. Variations in estimated MR are less for the pavements sections located in DF/DNF regions, i.e., approximately 18 to 40 percent as shown in Figure 4-14(a) and (b). For the SMP pavement sections located in WF/WNF climate, the maximum change in base layer MR was approximately 153 to 175 percent [see Figure 4-14(c) and (d)]. Table 4-5 provides the summary of percent reduction in estimated MR based on all the flexible pavement SMP sections located in different climates. 69 Table 4-5 Summary — Change in MR due to moisture variations Section ID Climatic region 08_1053 16_1010 30_0114 32_0101 46_0804 56_1007 04_0113 04_0114 04_1024 35_1112 49_1001 09_1803 23_1026 25_1002 27_1018 27_6251 31_0114 33_1001 36_0801 50_1002 83_1801 87_1622 01_0101 01_0102 10_0102 13_1005 13_1031 48_1060 48_1077 48_1122 51_0113 51_0114 DF DF DF DF DF DF DNF DNF DNF DNF DNF WF WF WF WF WF WF WF WF WF WF WF WNF WNF WNF WNF WNF WNF WNF WNF WNF WNF Minimum MR Maximum MR Reduction in (MPa) 289.0 (MPa) 292.9 297.2 302.1 271.6 252.2 286.0 285.8 276.9 200.3 285.2 296.2 282.6 291.8 287.2 279.7 296.9 291.3 296.4 110.7 284.1 226.8 296.8 281.7 287.2 275.3 267.4 297.4 92.3 259.4 270.3 282.7 297.1 306.7 304.2 284.0 297.2 302.4 301.5 289.8 282.7 303.9 305.8 288.4 296.3 298.7 301.1 304.2 305.6 301.8 304.3 302.3 302.8 302.1 289.5 301.6 289.8 276.7 301.5 233.2 303.1 277.1 294.6 303.3 MR (%) 1% 3% 1% 5% 18% 6% 5% 5% 41% 7% 3% 2% 2% 4% 8% 2% 4% 2% 175% 6% 34% 2% 3% 5% 5% 3% 1% 153% 17% 3% 4% 2% Note: Results based on approximately 8-9 years of measured SMP LTPP data. 70 (a) DF (c) WF (b) DNF (d) WNF (e) Wet region (f) Dry region Figure 4-14 Impact of moisture variations on flexible pavements base MR 71 y = -0.6685x2+ 3.7088x + 299.41R² = 0.99290501001502002503003500510152025MR (MPa)Moisture content (%)y = -1.1033x2+ 8.794x + 287.57R² = 0.99090501001502002503003500510152025MR (MPa)Moisture content (%)y = -0.7282x2+ 4.8178x + 294.69R² = 0.98630501001502002503003500510152025MR (MPa)Moisture content (%)y = -0.477x2+ 0.982x + 308.58R² = 0.98890501001502002503003500510152025MR (MPa)Moisture content (%)y = -0.546x2+ 1.9917x + 304.63R² = 0.97690501001502002503003500510152025MR (MPa)Moisture content (%)y = -1.0369x2+ 8.0903x + 288.53R² = 0.98320501001502002503003500510152025MR (MPa)Moisture content (%) 4.6.3.2 Impact of Flexible Base Resilient Modulus on Long-Term Pavement Performance The moisture variations and its adverse impact on base MR were quantified for the pavement sections located in different climates. While evaluating the impact of base MR on long-term pavement performance, two flexible pavement sections were considered with the cross-section details as shown in Figure 4-15. (b) Thick AC section (a) Thin AC section Figure 4-15 Flexible pavement cross sections The long-term performance was predicted for approximately 14 million ESALs by using the Pavement-ME. Base MR values were varied from 30 to 350 (MPa) in the Pavement-ME runs to compare the performance of both the sections. Climatic data from two different weather stations were used to simulate different climates in the Pavement-ME. Weather stations located in Washington and New York were used to simulate DNF and WF climates, respectively. Figure 4-16(a) shows the relationship between total predicted cracking and MR for thin and thick sections located in WF/DNF climates. It can be observed that with a decrease in base MR, amount of surface cracking increased. As compared to dry regions, slightly higher surface 72 Subgrade (MR=69 MPa)10 cm HMA20 cm Base (MR varied from 30 to 350 MPa)15 cm Subbase (MR=103 MPa)Subgrade (MR=69 MPa)20 cm Base (MR varied from 30 to 350 MPa)15 cm Subbase (MR=103 MPa)20 cm HMA cracking extents were observed in wet regions. In addition, much higher levels of surface cracking were observed in thinner section, essentially because of higher traffic. With the decrease in base layer MR values, total surface rutting also increased, this trend is more pronounced for the thin section as compared to thick section, especially for the pavements located in wet climates [see Figure 4-16(b)]. Figure 4-16(c) and (d) show the impact of base moisture on total cracking and rutting in different climates. The results show that for thin HMA sections in wet regions, if the MR is decreased by 175 percent (from 304 to 110 MPa i.e., the maximum reduction in MR values due to moisture in wet climates), there will be about 102% increase in the long-term total cracking [see Figure 4-16(c)]. Similarly, for thick HMA section in wet regions, the increase in cracking is about 114%. Figure 4-16(d) shows the relationship between predicted rut depths with a change in base MR. In wet climates, a 175% reduction in base MR showed about 17% and 6% increase in surface rutting for thin and thick pavement sections, respectively. Alike, if the MR is decreased by 41 percent (from 282 to 200 MPa i.e., the maximum reduction in MR values due to moisture in dry climates), there will be 35% and 38% increase in long-term total cracking and about 6% and 2% increase in surface rutting for thin and thick sections, respectively. Table 4-6 provides the grand summary of measured/ANN predicted moisture data and Pavement-ME performance data. 73 Table 4-6 Summary measured /predicted moisture data and Pavement-ME predicted performance s n o i t a l u c l a c r o f d e s u e p y t a t a d e r u t s i o M P P T L d e r u s a e M s n o i t c i d e r p l e d o m N N A n o i t c i d e r p l e d o m N N A s n o i t c i d e r p l e d o m N N A g n i k c a r c m 0 0 1 t a g n i k c a r c m 0 6 t a s g n i k c a r c m 0 4 t a Change in measured moisture content LTPP climatic region ) % ( e r u t s i o m m u m i x a M ) % ( e r u t s i o m m u m i n i M ) % ( e r u t s i o m e g n a R Reduction in MR Increase in cracking Increase in rutting based on measured moisture ) a P M ( R M m u m i n i M ) a P M ( R M m u m i x a M Thick section Thin section Thick section Thin section ) % ( R M n i n o i t c u d e R R M m u m i x a M t a ) m ( g n i k c a r c l a t o T R M m u m i n i M t a ) m ( g n i k c a r c l a t o T ) % ( g n i k c a r c n i e s a e r c n I R M m u m i x a M t a ) m ( g n i k c a r c l a t o T R M m u m i n i M t a ) m ( g n i k c a r c l a t o T ) % ( g n i k c a r c n i e s a e r c n I R M m u m i x a M t a ) m m ( g n i t t u r l a t o T R M m u m i n i M t a ) m m ( g n i t t u r l a t o T ) % ( g n i t t u r n i e s a e r c n I R M m u m i x a M t a ) m m ( g n i t t u r l a t o T R M m u m i n i M t a ) m m ( g n i t t u r l a t o T ) % ( g n i t t u r n i e s a e r c n I DF 11.5 6 5.5 297.2 252.2 18% 138 164 18% 523 610 17% 24 24 1% 33 34 2% DNF 13.7 8.1 5.6 282.7 200.3 41% 134 202 38% 549 739 35% 24 25 2% 33 35 6% WF 19.7 3.2 16.5 304.3 110.7 175% 134 281 114% 512 1036 102% 24 26 6% 33 39 17% WNF 22.5 12.5 10 233.2 92.3 153% 135 310 76% 654 1108 69% 24 26 5% 34 40 16% Wet region 19.7 3.2 16.5 304.3 110.7 175% 134 281 114% 512 1036 102% 24 26 6% 33 39 17% Dry region 13.7 8.1 5.6 282.7 200.3 41% 134 202 38% 549 739 35% 24 25 2% 33 35 6% DF 8.1 2.9 5.2 305.0 285.4 7% 135 144 7% 511 544 6% 24 24 0% 33 33 1% DNF 5.4 4.7 0.7 302.2 298.1 1% 136 138 1% 515 522 1% 24 24 0% 33 33 0% WF 10.7 6.2 4.5 299.8 228.0 31% 137 180 32% 519 666 28% 24 25 2% 33 34 4% WNF 10.5 0.0 10.5 307.3 253.7 21% 133 163 22% 507 607 20% 24 24 1% 33 34 2% Wet region 10.7 0.0 10.6 307.3 228.0 35% 133 180 35% 507 666 31% 24 25 2% 33 34 4% Dry region 8.1 2.9 5.2 305.0 285.4 7% 135 144 7% 511 544 6% 24 24 0% 33 33 1% DF 5.4 2.9 2.5 305.0 298.9 2% 135 137 2% 511 520 2% 24 24 0% 33 33 0% DNF 4.8 4.7 0.1 302.2 300.2 1% 136 137 1% 515 518 1% 24 24 0% 33 33 0% WF 8.5 6.2 2.4 299.8 266.4 13% 137 155 13% 519 580 12% 24 24 1% 33 33 1% WNF 4.0 0.0 4.0 307.3 302.2 2% 133 136 2% 507 515 2% 24 24 0% 33 33 0% Wet region 8.5 0.0 8.5 307.3 266.4 15% 133 155 16% 507 580 14% 24 24 1% 33 33 2% Dry region 5.4 2.9 2.5 305.0 298.9 2% 135 137 2% 511 520 2% 24 24 0% 33 33 0% DF 4.3 2.9 1.4 305.0 302.1 1.0% 135 136 1% 511 515 1% 24 24 0% 33 33 0% DNF 4.8 4.7 0.1 302.2 300.2 0.7% 136 137 1% 515 518 1% 24 24 0% 33 33 0% WF 7.7 6.2 1.5 299.8 276.3 8.5% 137 149 9% 519 561 8% 24 24 0% 33 33 1% WNF 1.7 0.0 1.7 307.3 306.3 0.3% 133 134 0% 507 509 0% 24 24 0% 33 33 0% Wet region 7.7 0.0 7.7 307.3 276.3 11.2% 133 149 12% 507 561 11% 24 24 1% 33 33 1% Dry region 4.8 2.9 1.9 305.0 302.1 1.0% 135 136 1% 511 515 1% 24 24 0% 33 33 0% 74 (a) Effect of region and thickness on cracking (b) Effect of region and thickness on rutting (c) Effect of MR on cracking (d) Effect of MR on rutting Figure 4-16 Impact of flexible pavements base MR on predicted pavement performance 4.6.3.3 Demonstrative Examples of Crack Sealing Application Timings — Flexible Pavements The last task of this study was to define optimum timings for effective crack sealing. Few rational assumptions were made to achieve this task. Firstly, to define appropriate crack sealing application timings, the variations in base layer moisture and a corresponding change in MR were estimated when the pavement section was new (at minimal cracking), and when the total surface cracking levels reached 40m, 60m, and 100m (for calculations see Table 4-6). Figure 4-17(a) and (b) show that when total surface cracking is 100m, the maximum reduction in MR is 75 020040060080010001200140016000100200300400Total cracking (m)MR (MPa)10 (cm) WF10 (cm) DNF20 (cm) WF20 (cm) DNF010203040500100200300400Rut depth (mm)MR (MPa)10 (cm) WF10 (cm) DNF20 (cm) WF20 (cm) DNFy = 0.0058x2-5.1124x + 1530.4R² = 0.9779y = 0.0017x2-1.5014x + 434.34R² = 0.937502004006008001000120014000100200300400Cracking (m)MR (MPa)Thin HMAThick HMAy = 1E-04x2-0.0707x + 45.141R² = 0.8268y = 2E-05x2-0.0159x + 27.112R² = 0.395152025303540450100200300400Rut depth (mm)MR (MPa)Thin HMAThick HMA 7% in dry climates. In contrast, the maximum reduction in MR for wet regions is 35%. The findings imply that moisture variation severely affects pavements in wet climates, and it is important to seal the cracks when the extent of surface cracking is low (i.e., between 40m to 60m). For pavements in dry regions, this extent can be tolerated to slightly higher levels of surface cracking, i.e., may be up to 100m. Secondly, bottom-up fatigue is a classic example of through cracking that would allow the surface water to infiltrate into the pavement unbound layers. The observed proportion of WP cracking length (out of total cracking) is shown in Table 4-7. Based on the observed data an assumption was made that if approximately 50% of the total cracking length is within WP, the optimum crack sealing limits were estimated in terms of percentage area fatigue cracking. Based on this assumption, the cracks should be sealed when the WP fatigue is below 6% (average of 5% and 8%) and 11% (average of 8% and 14%) for the pavements located in wet and dry climates, respectively. The detailed conversions for 40m, 60m and 100m total surface cracking to percentage WP fatigue are given in Table 4-8. (a) Reduction in MR (b) Increase in cracking Figure 4-17 Impact of base MR on predicted long-term pavement performance 76 0%5%10%15%20%25%30%35%40%DFDNFDry regionWFWNFWet region(%) Reduction in MRClimatic regions40 m60 m100m0%5%10%15%20%25%30%35%40%DFDNFDry regionWFWNFWet region(%) Increase in crackingClimatic regions40 m60 m100m Table 4-7 Proportion of observed WP cracking length Section ID Climatic region Pavement age (years) WP length cracked (m) Total cracking Proportion of length (m) WP fatigue (%) 08_1053 16_1010 30_0114 32_0101 46_0804 56_1007 04_0113 04_0114 04_1024 35_1112 49_1001 09_1803 23_1026 25_1002 27_1018 27_6251 31_0114 33_1001 36_0801 50_1002 83_1801 87_1622 01_0101 01_0102 10_0102 13_1005 13_1031 48_1060 48_1077 48_1122 51_0113 51_0114 DF DF DF DF DF DF DNF DNF DNF DNF DNF WF WF WF WF WF WF WF WF WF WF WF WNF WNF WNF WNF WNF WNF WNF WNF WNF WNF 7.1 8.6 6.8 5.96 14.2 9.7 8.9 9.1 10.3 7.1 16.4 11.8 4.9 15.1 15.8 10.7 5.4 12.1 13.3 15.33 16.8 16.1 14.1 11.9 9.4 11.8 9.32 10.75 11.24 10.54 9.8 13.3 146.3 79.9 94.1 7.3 78.0 11.6 118.2 157.7 47.9 0.0 51.9 35.9 53.3 152.0 133.0 150.6 95.5 100.9 253.3 139.3 259.7 305.0 238.1 248.8 255.0 67.0 0.0 11.7 81.7 1.8 277.0 305.0 168.7 177.3 160.3 11.5 211.5 82.5 186.1 209.7 51.2 29.6 142.8 66.9 279.6 237.2 490.7 403.3 107.5 268.7 410.0 219.6 456.1 479.6 252.7 257.0 305.0 224.6 192.2 12.2 234.6 6.3 305.0 472.5 Average proportion of WP cracking length 77 87% 45% 59% 63% 37% 14% 64% 75% 94% 0% 36% 54% 19% 64% 27% 37% 89% 38% 62% 63% 57% 64% 94% 97% 84% 30% 0% 96% 35% 29% 91% 65% 55% Table 4-8 Conversions — Total surface cracking length to % area WP fatigue ) m ( h t g n e l n o i t c e S ) m ( h t d i w n o i t c e S ) 2 m ( a e r a n o i t c e s ) m ( ) 2 * 6 7 . ( s h t a p l e e h w o w t h t d i W l a n i d u t i g n o l y t i r e v e s h g i h f o h t d i W ) m ( g n i k c a r c e s r e v s n a r t d n a h t g n e l g n i k c a r c P W f o n o i t r o p o r P ) % ( 152.5 3.65 556.6 1.52 0.019 152.5 3.65 556.6 1.52 0.019 152.5 3.65 556.6 1.52 0.019 152.5 3.65 556.6 1.52 0.019 152.5 3.65 556.6 1.52 0.019 50 60 70 80 90 152.5 3.65 556.6 1.52 0.019 100 Total surface cracking 40m 60m 100m ) m ( e u g i t a f P W h t g n e L 20 24 28 32 36 40 ) 2 m ( e u g i t a f P W a e r A ) % ( e u g i t a f P W a e r A 30.4 5% 36.48 7% 42.56 8% 48.64 9% 54.72 10% 60.8 11% ) m ( e u g i t a f P W h t g n e L 30 36 42 48 54 60 ) 2 m ( e u g i t a f P W a e r A ) % ( e u g i t a f P W a e r A 45.6 8% 54.72 10% 63.84 11% 72.96 13% 82.08 15% ) m ( e u g i t a f P W h t g n e L 50 60 70 80 90 ) 2 m ( e u g i t a f P W a e r A ) % ( e u g i t a f P W a e r A 76 14% 91.2 16% 106.4 19% 121.6 22% 136.8 25% 91.2 16% 100 152 27% Based on the assumptions and results previously discussed, the Pavement-ME predicted performance data were used to simulate the effectiveness of crack sealing application timings. This time, for thin pavement section, the long-term performance was predicted for approximately 4.3 million ESALs by using the Pavement-ME. Separate guidelines for crack sealing application timings were defined for wet and dry climates. A base MR value of 275 MPa, as observed in the field was assumed as the original material property. Subsequently, to simulate the effect of moisture increase based on the field observations, reduced MR values of 90% and 75% of the original MR value were assumed while planning preservation treatment. The above MR values were assumed to characterize the stiffness parameters of base layer after the application of particular preservation treatment. It is also known that preservation treatments cannot restore materials to their original stiffness. However, those can extend the service life of the pavements by retarding the deterioration rate. Comparisons of the Pavement- 78 ME predicted performance was made by considering the base MR original stiffness (275 MPa), 90% of original MR (240 MPa), and 75% of the original MR (205 MPa). These variations in the MR values should be based on the moisture increase observed in the actual materials. Figure 4-18 to Figure 4-21 show the flexible pavements preservation plans based on Pavement-ME predicted long-term pavement performance. Based on the analysis performed in the previous section the optimum crack sealing limits for fatigue cracking were 6% and 11% (approximately) for wet and dry climates, respectively. While developing the preservation plan for wet climates these limits were strictly followed because higher rainfall coupled with higher surface cracking can adversely impact the flexible pavements base layer base MR in wet climates. Figure 4-18 shows an example of a preservation plan (crack seal application timings) by using the Pavement-ME for a pavement section located in WF climate. First crack sealing application was planned as the cracking reached a threshold of 6 percent (as identified for the sites located in wet climates) in about 8 years of service life [see Figure 4-18 (b)]. The application cycle will repeat, once the pavement reaches the same cracking limit at about 16 years as shown in Figure 4-18 (c). The overall effect of crack sealing on cracking progression is shown in Figure 4-18 (d). The results show that the pavement live can be significantly extended at a lower level of cracking when crack sealing is applied at the appropriate times (i.e., 6% cracking). 79 (a) Effect of MR on fatigue cracking (c) Preservation at 8 and 16 years (b) Preservation at 8 years (d) Effect of preservation after 20 years Figure 4-18 Preservation treatment plan thick section (WF climate) Figure 4-19 shows an example of a preservation plan (crack seal application timing) for a pavement section located in DNF climate. As compared to wet climates, only single sealing application was planned as the cracking reached a threshold of 11 percent (as identified for the sites located in dry climates) in about 11 years of service life [see Figure 4-19 (b)]. The WP fatigue threshold for pavement in dry climates was between 11 to 12%. This cracking threshold was reached at about 16 years of service life on the Pavement-ME prediction curve. Ideally, sealing application should have been planned at the 16th year of service life. However, it was planned at the end of the 11th year because for obvious factors (traffic, climate, material, etc.) a 80 051015202530048121620WP fatigue (%)Age (years)WF- 205 MPaWF- 240 MPaWF- 275 Mpa(Original MR)0510152025048121620WP fatigue (%)Age (years)No PreservationPreserved at 8 yrs- Orig MRPreserved at 8 yrs- 90% Orig MRPreserved at 8 yrs- 75% Orig MR0510152025048121620WP fatigue (%)Age (years)No PreservationPreserved at 8 yrs- Orig MRPreserved at 8 yrs- 90% Orig MRPreserved at 8 yrs- 75% Orig MRPreserved at 16 yrs- Orig MRPreserved at 16 yrs- 90% Orig MRPreserved at 16 yrs- 75% Orig MR19.67.29.211.24.14.34.60510152025WP fatigue (%)Preservation treatment application pavement can rarely stay in its actual condition after 16 years of design life. The overall effect of crack sealing on cracking progression is shown in Figure 4-19 (c). It is evident from the results that proactive maintenance/preservation can considerably enhance pavement service life. (a) Effect of MR on fatigue cracking (b) Preservation at 11 years (c) Effect of preservation after 20 years Figure 4-19 Preservation treatment plan thick section (DNF climate) Figure 4-20 and Figure 4-21 show the suggested preservation plan for thin pavement sections located in WF and DNF climates respectively. For thin pavement sections, more frequent crack sealing applications are needed due to higher levels of fatigue cracking at early ages of pavements service life. In wet climates, the sealing applications were applied every 3 to 4 years 81 051015202530048121620WP fatigue (%)Age (years)WF- 205 MPaWF- 240 MPaWF- 275 Mpa (Original MR)0510152025048121620WP fatigue (%)Age (years)No Preservation MR 275 MPaPreserved at 11 yrs- Orig MRPreserved at 11 yrs- 90% Orig MRPreserved at 11 yrs- 75% Orig MR15.35.45.96.3024681012141618WP fatigue (%)Preservation treatment application to maintain the pavement within the tolerable limit of fatigue cracking [see Figure 4-20 (b) and (c)]. Alike, sealing applications were planned every 3.5 and 5 years for the pavements sections located in DNF climates [see Figure 4-21 (b) and (c)]. For thin pavement sections located in WF and DNF climates, the overall effect of crack sealing on cracking progression is shown in Figure 4-20 (d) and Figure 4-21 (d), respectively. The results show that the pavement live can be significantly extended at a lower level of cracking when crack sealing is applied at the appropriate times. Although, seeing the fatigue cracking limits in thin sections, after few years sealing applications may not be a viable option, and one might have to rehabilitate the pavement. Pavement-ME is the current state of the art tool for pavement design and analysis, and its farsighted application will enable to plan preservation right at the design stage. Preservation plans presented in this study by using crack seal treatment can be used as a guideline when moisture variations are only limited to aggregate base material MR. However, to accurately estimate the preservation treatment application timing, stiffness properties of entire pavement structure must be given due importance while predicting long-term performance. 82 (a) Effect of MR on fatigue cracking (c) Preservation after every 4 years (b) Preservation after every 3 years (d) Effect of preservation after 20 years Figure 4-20 Preservation treatment plan thin section (WF climate) 83 05101520253035048121620WP fatigue (%)Age (years)WF- 205 MPaWF- 240 MPaWF- 275 Mpa (Original MR)05101520253035048121620WP fatigue (%)Age (years)No PreservationPreserved every 3 yrs- Orig MRPreserved every 3 yrs- 90% Orig MRPreserved every 3 yrs- 75% Orig MR05101520253035048121620WP fatigue (%)Age (years)No PreservationPreserved every 4 yrs- Orig MRPreserved every 4 yrs- 90% Orig MRPreserved every 4 yrs- 75% Orig MR30.94.14.85.410.816.121.405101520253035WP fatigue (%)Preservation treatment application (a) Effect of MR on fatigue cracking (c) Preservation after every 5 years (b) Preservation after every 3.5 years (d) Effect of preservation after 20 years Figure 4-21 Preservation treatment plan thin section (DNF climate) 4.7 RIGID PAVEMENTS MODELING After quantifying the moisture variations in the flexible pavements base layer and its effect on long-term performance, available rigid pavement sections with granular base layers were also investigated for moisture change. The available number of rigid pavement sections (11 such sections were identified) and data were limited, especially in DF/DNF climate (only two sections). In rigid pavements, same independent variables like flexible pavements were used while developing moisture prediction models, except the total surface cracking, which was replaced with the length of joint seal damage. Initially, total cracking lengths for the rigid SMP 84 05101520253035048121620WP fatigue (%)Age (years)WF- 205 MPaWF- 240 MPaWF- 275 Mpa (Original MR)05101520253035048121620WP fatigue (%)Age (years)No PreservationPreserved every 3.5 yrs- Orig MRPreserved every 3.5 yrs- 90% Orig MRPreserved every 3.5 yrs- 75% Orig MR05101520253035048121620WP fatigue (%)Age (years)No PreservationPreserved every 5 yrs- Orig MRPreserved every 5 yrs- 90% Orig MRPreserved every 5 yrs- 75% Orig MR29.54.35.26.112.017.122.205101520253035WP fatigue (%)Preservation treatment application sections were ascertained by adding the length of damaged joint sealants and length longitudinal and transverse cracking. It was observed that the lengths of longitudinal and transverse cracking for these sections were very low, as compared to the lengths of damaged joints seals. Base layer moisture content relationship with damaged joints and longitudinal and transverse cracking is shown in Figure 4-22. It can be concluded from these relationships that damage joints are the primary cause of water infiltration into rigid pavements. Therefore, only lengths of damaged joints were used while developing ANN model for rigid SMP sections. (a) Long/Trans cracking Vs base layer moisture (b) (Joint damage + long/trans cracking) Vs base layer moisture (c) Joint seal damage Vs base layer moisture Figure 4-22 PCC surface discontinuities relationship with base layer moisture 85 y = -0.0048x + 11.864R² = 0.0003051015202530050100150200250Moisture content (%)Long and trans cracking (m)y = 0.0407x + 7.7158R² = 0.3715051015202530050100150200250300Moisture content (%)Joint seal damage+ long/trans cracking (m)y = 0.0499x + 7.4737R² = 0.4587051015202530050100150200250Moisture content (%)Joint seal damage (m) 4.7.1 ANN Modeling Rigid Pavements Essentially same network settings used earlier for flexible pavements ANN model, were adopted while developing rigid pavements model with minor modifications. In contrast to flexible pavements model, the ANN model developed for the rigid pavements is simple. Also, due to its small size, the data were only used for network training and validation. Optimum setting for the developed ANN model are given in Table 4-9. Figure 4-23 shows the schematic of ANN model developed for PCC sites. Table 4-9 Optimum settings for the rigid pavements ANN model Network type No of hidden layer Data entries for training, testing, and validation Training function Hidden layer transfer function Output layer transfer function Performance function No of hidden neurons BPNN 1 37,0,16 trainlm tansig purelin MSE 5 Figure 4-23 ANN model flow rigid pavements SMP sections Figure 4-24(a) shows the goodness of fit for rigid pavements ANN model. Figure 4-24 (b), (c), and (d) show the ANN model sensitivity to different inputs. The results of the ANN model 86 INPUTHIDDEN LAYEROUTPUT LAYEROUTPUTP 5x1WO1x5tansigpurelinbO1x1WH5x5bH5x1nH5x1nO1x1aH5x1aO1x1Y 1x1aH= tansig(WHp+bH)aO= purelin(WOaH+bO) sensitivity show that with an increase in joint seal damage, there is an increase in the base layer in-situ moisture. Moisture change is significant for higher precipitation levels (wet climate), especially in freezing region [see Figure 4-24(b)]. Higher (%) passing #200, higher is the moisture change, higher the moisture depth within base layer higher is the moisture levels [see Figure 4-24(c) and (d)]. It is worth noting that when the joint seal damage length reached approximately 50m to 75m, moisture increase is substantial. It was observed that rigid pavements ANN model mostly overpredicted the base layer moisture levels. These overpredictions are plausibly caused by, small data size used for the development of ANN model. There could be other potential reasons as well, associated with ANN model settings. 87 (a) Measured Vs predicted (b) Effect of Precipitation and freezing (e) Impact of percent passing 200 (f) Impact of moisture depth Figure 4-24 ANN model predictions and sensitivity — rigid pavements 4.7.2 The Relationship between Base Moisture and Base Resilient Modulus- PCC Sections The estimated MR values for base materials with a change in moisture levels based on Witzack model are shown in Figure 4-25. The results show that as the moisture increases, the MR decreases. For the pavement sections located in dry climates, the reduction in base layer MR is small, i.e., approximately 4 to10 percent [see Figure 4-25(a)]. The main reasons for the lower 88 04812162024280481216202428PredictedMeasuredR2=.87n=5305101520253035400255075100125150175Moisture cotnet (%) Joint seal damage(m)DNFWNFDFWF05101520253035400255075100125150175Moisture content (%)Cracking (m)12%16%20%05101520253035400255075100125150175Moisture content (%)Cracking (m)38 cm46 cm54 cm change in MR are lower levels of cracking/joint seal damage coupled with low precipitation levels in dry climates. For the sections located in wet climates, the maximum reduction in base layer MR was approximately 112 to 127 percent [see Figure 4-25(b)]. This higher variation in MR can be associated with higher precipitation and cracking/joint seal damage levels in wet climates. Table 4-10 provides the summary of percent reduction in estimated MR based on all the rigid pavement SMP sections located in different climates. Due to limited data for rigid pavements, the results may not represent the exact quantifiable moisture variations in these regions. Table 4-10 Summary — Change in rigid pavements MR due to moisture change Section ID 32_0204 4_0215 18_3002 27_4040 39_0204 42_1606 83_3802 89_3015 13_3019 37_0201 53_3813 Climate region Minimum MR (MPa) Maximum MR (MPa) Reduction in MR (%) DF DNF WF WF WF WF WF WF WNF WNF WNF 258.0 246.4 198.9 255.8 247.5 105.7 79.2 290.6 213.4 280.8 122.5 283.2 257.3 283.0 304.8 260.7 224.3 142.0 306.0 227.0 304.1 277.6 Note: Results based on approximately 8-9 years of measured SMP LTPP data. 10% 4% 42% 19% 5% 112% 79% 5% 6% 8% 127% (a) Dry region (b) Wet region Figure 4-25 Impact of moisture variations on PCC sections base MR 89 y = -0.2078x2-4.2392x + 329.59R² = 0.992050100150200250300350051015MR (MPa)Moisture content (%)y = -0.1545x2-7.6293x + 346.18R² = 0.9368050100150200250300350051015202530MR (MPa)Moisture content (%) 4.7.3 Crack Sealing Application Timings — Rigid Pavements Based on the rigid pavements sections data analyses results, it can be concluded that PCC joints should be sealed when the length of damaged joints is between 50 to 75 meters. Because within this range the variations in base layer moisture are small and may not significantly affect the stiffness properties of base material [see Figure 4-22 (c) and Figure 4-24 (b)]. 4.8 SUMMARY This section summarizes the data analyses part of rigid and flexible SMP pavements sections, followed by quantification of moisture-related damage and pavement preservation guidelines. The following is a summary of the findings:  Moisture variation in flexible and rigid pavements base layers significantly impact the pavement performance.  Higher cracking and greater precipitation levels are the primary reasons for greater moisture change in wet climates.  GWT can affect seasonal variation in unbound layers moisture content, but the relationship is not very obvious, especially within base layers.  Subsurface moisture levels significantly vary before and after substantial amount of surface cracking levels.  As compared to dry climates, moisture variations are very high for the pavement sections located in Wet climates, because of higher precipitation levels and greater cracking extents in these regions.  Site-specific moisture prediction models highlight the effect of precipitation and cracking on base layer moisture change. 90  Factors including surface cracking, precipitation, percentage-passing # 200 sieve, and moisture depth, and freezing index can be used to predict base layer moisture levels with reasonable accuracy.  The artificial neural network (ANN) models were developed using SMP data for flexible and rigid pavement sections. The results show that higher levels of cracking and joint openings will lead to an increase moisture levels within base layer. Also, the moisture content increases with higher percentage passing # 200 sieve (P200), and higher precipitation levels, especially in wet climates.  Moisture significantly affected the base layer MR. The observed reduction in MR was up to 41% and 175% for the flexible pavement sections sites located in dry and wet climates, respectively.  Pavement-ME calculated long-term pavement performance results show that with a reduction in base layer MR, surface cracking, and rutting increased significantly.  In wet climates, 175% reduction in base MR showed about 114% and 102% increase in cracking, and 6% and 17% increase in surface rutting for thick and thin sections, respectively.  In dry climates, 41% reduction in base MR showed about 38% and 35% increase in cracking, and 2% and 6% increase in surface rutting for thick and thin sections, respectively.  Timely and effective preservation can substantially enhance the pavements service life.  PCC SMP sites data analysis showed that magnitude of transverse and longitudinal cracking is the minimal and primary cause of moisture variation is damaged joint sealant length. 91  Moisture variations significantly affected the PCC base layer MR. The observed reduction in MR was up to 10 % and 127% for the PCC sites located in dry and wet climates, respectively,  Based on the data analysis results it can be concluded that joint seal damage is the main cause of moisture variation in PCC pavements sections.  Pavement-ME is the current state of the art tool for pavement design and analysis, and its farsighted application will enable us to plan preservation right at the design stage. Preservation plans presented in this research serve as a guideline for the researchers and essentially based on the reduction of base layer moduli only. To accurately estimate the preservation treatment and time, stiffness properties of entire pavement structure must be given due importance while predicting long-term performance. 92 5 CONCLUSIONS AND RECOMMENDATIONS 5.1 SUMMARY Highway agencies have learned that if preservation treatments are applied at an appropriate time, those can help in improving and slowing the deterioration rates for the existing pavements. While pavement preservation is not expected to substantially increase the structural capacity of the existing pavement, it generally leads to improved pavement performance and longer service life. However; still, there are challenges in adoption of such practices. Selection of preservation treatments depends on the pre-existing conditions and other factors contributing to the deterioration of existing roadways. One of the most influential factors affecting pavement performance is the moisture variations within the pavement system, essentially caused by infiltration of rainfall water through surface discontinuities. The SMP study in the LTPP was designed to investigate and quantify the moisture variations, and related damage in flexible and rigid pavements (64). Therefore, the main objectives of this research study were to (a) evaluate the effect of cracking and joint openings on the moisture content in unbound layers, (b) quantify the impact of infiltration and moisture on the stiffness properties of unbound layers, (c) predict long-term pavement performance based on the unbound material properties to evaluate the impacts of preservation treatments, and (d) develop guidelines for optimum crack sealing applications timings for different environmental conditions. This study presents LTPP data analyses for quantifying the effect of moisture infiltration through surface discontinuities (cracks and joint openings) on flexible and rigid pavement performance. Previous research highlighted that moisture variation within unbound layers is one 93 of the leading factors for premature pavement deterioration (7, 9, 13, 16). Therefore, the hypothesis of this study was that moisture variation in unbound layers, i.e., base layer, could be related to the amount of surface discontinuities (cracking and joint seal damage) in different climatic zones. To validate this hypothesis, an important challenge was to identify the data set documenting the subsurface moisture levels in the base layer. Only SMP study used TDRs; those were installed at different depths to record moisture variations within the entire pavement structure. While quantifying the moisture related damage, SMP moisture and performance data from 32 flexible, and 11 rigid pavement sections with granular base layers were used in this study. The Pavement-ME software provide methodologies for the analysis and design of flexible and rigid pavements. However, these methodologies and related performance prediction models focus on new structural design and rehabilitation of existing pavements and do not explicitly consider the contributions of pavement preservation treatments to the overall pavement performance. Thus, research was needed to identify approaches for considering the effects of preservation on pavement performance and developing procedures that facilitate incorporation of pavement preservation treatments in the Pavement-ME analysis process. The procedures and guidelines documented in this study will help the pavement engineers and agencies to ensure that the contributions of preservation treatments to expected performance and service life are appropriately considered in the analysis and design processes. 5.2 CONCLUSIONS Based on the results of the analyses performed, the following conclusions were drawn. 1. Moisture variation in flexible and rigid pavements base layers significantly impact the pavement performance. 94 2. The SMP data can be used to investigate the moisture variations in pavement layers and impact of different climates on moisture variations can be quantified. 3. The rigid and flexible pavements SMP sections data analysis show that there can be significant variations in granular base layer in-situ moisture content. 4. Subsurface moisture levels considerably vary before and after the substantial onset of surface cracking. 5. For wet climates, moisture variations in base layers were very high. Higher cracking and greater precipitation levels are the primary reasons for greater moisture variations for the pavement sections located in wet climates. Relatively, lower cracking and precipitation levels are the primary reasons for small moisture variations for the pavement sections located in wet climates. 6. The artificial neural network (ANN) models were developed using SMP data for flexible and rigid pavement sections. The results show that higher levels of cracking and joint openings will lead to an increase moisture levels within base layer. Also, the moisture content increases with higher percentage passing # 200 sieve (P200), and higher precipitation levels, especially in wet climates. 7. Moisture related damage was very high in WF/WNF climates (153 to 175 percent reduction in MR). It is critical to prevent the unbound layers from moisture related damage due to infiltration, especially before the MR reduction becomes significantly high. 8. Subsurface moisture variations showed relatively less impact on the sites located in DF/DNF climates (18 to 41 percent reduction in MR). For the pavement sites in DF/DNF climates, damage associated with other factors like high temperature is more critical. 95 9. Pavement-ME predicted long-term pavement performance results show that with a reduction in base layer MR, surface cracking and rutting levels were increased significantly. 10. In wet climates, a 175% reduction in base MR (i.e., maximum MR reduction in wet climates) showed about 114% and 102% increase in cracking, and 6% and 17 % increase in total rutting, for thick and thin flexible pavements sections, respectively. 11. In dry climates, a 41% reduction in base MR (i.e., maximum MR reduction in wet climates) caused 38% and 35% increase in cracking, and 2% and 6% increase in total rutting, for thick and thin flexible pavements sections, respectively. 12. Rigid SMP sections data analysis show that moisture significantly affected the PCC base layer MR. The observed reduction in MR was up to 10 % and 127% for the PCC sites located in dry and wet climates, respectively. 13. Based on the data analysis it was concluded that damaged joint sealant length is the main cause of moisture variation in PCC pavement sections base layers. Therefore, damage joints should be sealed when the extents are between 50 to 75 meters. 5.3 RECOMMENDATIONS The following are the recommendations based on the findings of this study: 1. Moisture variation severely affects the flexible pavements performance in wet climates. Therefore, in wet climates, it is essential to apply preservation treatment when the fatigue cracking extent is below 6 to 7 percent. 2. For flexible pavements in dry climates, this extent can be tolerated to slightly higher levels of surface cracking, i.e., may be up to 10 to 11 percent. 96 3. To prevent moisture related damage in rigid pavements, the joints should be resealed when the damaged joint sealant length exceeds 50 to 75 meters. The current Pavement- ME performance models for rigid pavements do not predict damaged joint sealant length. It is recommended for future that damaged joint sealant length may accounted for by indirectly relating it to some other performance measures like joint faulting or IRI. 4. The crack sealing guidelines and moisture predictions models presented in this study can be further improved by including more data to improve pavement preservation practices and, the accuracy of the models. 5. Pavement-ME is the current state of the art tool for pavement design and analysis, and its farsighted application will enable us to plan preservation right at the design stage. Preservation plans presented in this research serve as a guideline for the researchers and essentially based on the reduction of base layer moduli only. To accurately estimate the preservation treatment and time, stiffness properties of entire pavement structure must be given due importance while predicting long-term performance. 97 REFERENCES 98 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. REFERENCES Munnell, A. H., "Policy Watch: Infrastructure Investment and Economic Growth," The Journal of Economic Perspectives, vol. 6, pp. 189-198, 1992. Geiger, D. R., "Pavement Preservation Definitions," FHWA Memorandum, 2005. Davies, R. M. and J. Sorenson, "Pavement Preservation: Preserving Our Investment in Highways," Public roads, vol. 63, 2000. Galehouse, L., J. S. Moulthrop, and R. G. Hicks, "Principles of Pavement Preservation: Definitions, Benefits, Issues, and Barriers," TR News, 2003. Dan, H.-C., J.-W. Tan, Z. Zhang, and L.-H. He, "Modelling and Estimation of Water Infilteration into Cracked Asphalt Pavement," Road Materials and Pavement Design, vol. 18, pp. 590-611, 2017. Espinoza-Melendrez, J., S. Tamari, and A. Aguilar-Chavez, "Simulation of Water Infiltration Below a Damage in the Pavement Surface Course," Ingenieria Hidraulica en Mexico, vol. 20, pp. 77-95, 2005. Salour, F., "Moisture Influence on Structural Behaviour of Pavements: Field and Laboratory Investigations," KTH Royal Institute of Technology, 2015. allela, J., G. Larson, T. Wyatt, J. all, and W. Barker, "User’s Guide for rainage Requirements in Pavements–DRIP 2.0 Microcomputer Program," 2002. Witczak, M., D. Andrei, and W. Houston, "Resilient Modulus as Function of Soil Moisture–Summary of Predictive models," Appendix DD-1, 2000. Petry, T. M., Y.-P. Han, and D. N. Richardson, "Systems Approach for Estimating Field Moisture Content," Presented at Transportation Research Board 85th Annual Meeting, 2006, pp. Officials, T., AASHTO Guide for Design of Pavement Structures, 1993, vol. 1: AASHTO, 1993. Cedergren, H. R., "America's Pavements: World's Longest Bathtubs," Civil Engineering, vol. 64, pp. 56, 1994. "NCHRP 1-37A Guide for Mechanistic Empirical Design Part-3,Design Analysis ", pp. 32, 2004. 99 Hall, K. and S. Rao, "Predicting Subgrade Moisture Content for Low-Volume Pavement Design Using In-situ-Moisture Content Data," Transportation Research Record: Journal of the Transportation Research Board, pp. 98-106, 1999. Rao, S., "Analysis of In-situ Moisture Content Data for Arkansas Subgrades," 1997. Hedayati, M. and S. Hossain, "Data Based Model to Estimate Subgrade Moisture Variation Case Study: Low Volume Pavement in North Texas," Transportation Geotechnics, vol. 3, pp. 48-57, 2015. Fredlund, D. G., H. Rahardjo, and M. D. Fredlund, Unsaturated Soil Mechanics in Engineering Practice: John Wiley & Sons, 2012. Gu, F., X. Luo, Y. Zhang, R. Lytton, and H. Sahin, "Modelling of Unsaturated Granular Matrerials in Flexible Pavements," E-UNSAT 2016, 2016. Lekarp, F., U. Isacsson, and A. Dawson, "State of the Art. I: Resilient Response of Unbound Aggregates," Journal of Transportation Engineering, vol. 126, pp. 66-75, 2000. Salour, F. and S. Erlingsson, "Investigation of a Pavement Structural Behaviour During Spring Thaw using Falling Weight Deflectometer," Road Materials and Pavement Design, vol. 14, pp. 141-158, 2013. Nguyen, Q., D. G. Fredlund, L. Samarasekera, and B. L. Marjerison, "Seasonal Pattern of Matric Suctions in Highway Subgrades," Canadian Geotechnical Journal, vol. 47, pp. 267-280, 2010. Hedayati, M., M. S. Hossain, A. Mehdibeigi, and B. Thian, "Real-time Modeling of Moisture Distribution in Subgrade Soils," Presented at Geo-Congress 2014: Geo- Characterization and Modeling for Sustainability, 2014, pp. 3015-3024. Broun, R. and G. F. Menzies, "Life Cycle Energy and Environmental Analysis of Partition Wall Systems in the UK," Procedia Engineering, vol. 21, pp. 864-873, 2011. Lytton, R., D. Pufahl, C. Michalak, H. Liang, and B. Dempsey, "An Integrated Model of the Climatic Effects on Pavements," 1993. OECD, "Water in Roads,"Prediction of Moisture Content in Road Subgrades," Orgnization of Economic Corporation and Developpment,Paris,France," 1973. Swanberg, J. and C. Hansen, "Development of a Procedure for the Design of Flexible Bases," Presented at Highway Research Board PROCEEDINGS, 1947, pp. Navy, U., "Airfield Pavements, Bureau o Yards and Docks, Technical Publication," NAVDOCKS TP-PW-4 1953. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 100 28. 29. 30. 31. 32. 33. 34. 35. Kersten, M. S., L. Palmer, and W. Campen, "Progress Report Of Special Project On Structural Design of Nonrigid Pavements: Subgrade Moisture Conditions Beneath Airport Pavements," Presented at Highway Research Board Proceedings, 1946, pp. Petry, T. M., Y.-P. Han, and D. N. Richardson, "A systems Approach for Estimating Field Moisture Content," TRB 2006, pp. 20, 2006. Jiang, Y. J. and S. D. Tayabji, "Analysis of Time Domain Reflectometry Data from LTPP Seasonal Monitoring Program Test Sections," 1999. Kodikara, J., P. Rajeev, D. Chan, and C. Gallage, "Soil Moisture Monitoring at the Field Scale Using Neutron Probe," Canadian Geotechnical Journal, vol. 51, pp. 332-345, 2013. Fredlund, D. G. and A. Xing, "Equations for the Soil-water Characteristic Curve," Canadian Geotechnical Journal, vol. 31, pp. 521-532, 1994. Ridgeway, H. H., "Infiltration of Water Through the Pavement Surface (Abridgement)," Transportation Research Record, 1976. Hansson, K., L.-C. Lundin, and J. Šimůnek, "Part 3: Pavement: Subsurface rainage and Reliability in Design and Performance: Modeling Water Flow Patterns in Flexible Pavements," Transportation Research Record: Journal of the Transportation Research Board, vol. 1936, pp. 131-141, 2005. Van Genuchten, M. t., "A Closed Form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils," Soil Science Society of America Journal, vol. 44, pp. 892-898, 1980. 36. Mualem, Y. A., "A New Model for Predicting the Hydraulic Conductivity of Unsaturated Porous Media," Water Resources Research, vol. 12, pp. 513-522, 1976. 37. 38. 39. 40. Freeze, R. A. and J. A. Cherry, "Groundwater," Prentice Hall,Englewood Cliffs,N.J, 1979. Gu, F., H. Sahin, X. Luo, R. Luo, and R. L. Lytton, "Estimation of Resilient Modulus of Unbound Aggregates using Performance-related Base Course PROPERTIES," Journal of Materials in Civil Engineering, vol. 27, pp. 04014188, 2014. Lytton, R., "Foundations and Pavements on Unsaturated Soils," Presented at Proceedings of the First International Conference on Unsaturated Soils/Unsat'95/Paris/France/6-8 September 1995. Volume 3, 1996, pp. "Guide for Mechanistic-Emperical Design of New and Rehablitated Pavement Structures,Appendix DD-1; Resiliant Modulus as Function of Soil Moisture- Summary of Predictive Models," 2000. 101 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. Elkins, G. E., P. Schmalzer, T. Thompson, and A. Simpson, "Long-term Pavement Performance Information Management System Pavement Performance Database User Reference Guide," 2003. Han, Y.-P., T. Petry, and D. Richardson, "Part 2: Evaluation of Subgrade Resilient Modulus: Resilient Modulus Estimation System for Fine-Grained Soils," Transportation Research Record: Journal of the Transportation Research Board, pp. 69-77, 2006. Dubčáková, R., "Eureqa: software review," Genetic Programming and Evolvable Machines, vol. 12, pp. 173-178, 2011. Hecht-Nielsen, R., "Theory of the Backpropagation Neural Network," in Neural Networks for Perception: Elsevier, 1992, pp. 65-93. Basheer, I. A. and M. Hajmeer, "Artificial Neural Networks: Fundamentals, Computing, Design, and Application," Journal of Microbiological Methods, vol. 43, pp. 3-31, 2000. Ismail, H. M., H. K. Ng, C. W. Queck, and S. Gan, "Artificial Neural Networks Modelling of Engine-out Responses for a Light-duty Diesel Engine Fuelled with Biodiesel Blends," Applied Energy, vol. 92, pp. 769-777, 2012. Karadurmus, E., M. Cesmeci, M. Yuceer, and R. Berber, "An Artificial Neural Network Model for the Effects of Chicken Manure on Ground Water," Applied Soft Computing, vol. 12, pp. 494-497, 2012. Alba, R. D. and R. M. Golden, "Patterns of Ethnic Marriage in the United States," Social forces, vol. 65, pp. 202-223, 1986. Haykin, S. and N. Network, "A comprehensive foundation," Neural Networks, vol. 2, pp. 41, 2004. Seitllari, A. and M. E. Kutay, "Use of Soft Computing Tools to Predict Progression of Percent Embedment of Aggregates in Chip Seals," 2018. Hornik, K., M. Stinchcombe, and H. White, "Multilayer Feedforward Networks are Universal Approximators," Neural networks, vol. 2, pp. 359-366, 1989. Guide, M. U., "ANN 7.11. 0.584," R2010b). MathWorks. 52. 53. 54. Neural Networks, vol. 1, pp. 17-61, 1988. Guide, M. U. s., "Neural Network Toolbox," The MathWorks, 2002. Grossberg, S., "Nonlinear Neural Networks: Principles, Mechanisms, and Architectures," 102 55. Sathya, R. and A. Abraham, "Comparison of Supervised and Unsupervised Learning Algorithms for Pattern Classification," International Journal of Advanced Research in Artificial Intelligence, vol. 2, pp. 34-38, 2013. 56. Wedderburn, R. W., "Quasi-likelihood Functions, Generalized Linear Models, and the Gauss—Newton method," Biometrika, vol. 61, pp. 439-447, 1974. 57. 58. Hiriart-Urruty, J.-B., J.-J. Strodiot, and V. H. Nguyen, "Generalized Hessian Matrix and Second-order Optimality Conditions for Problems with C 1, 1 Data," Applied Mathematics and Optimization, vol. 11, pp. 43-56, 1984. Kişi, Ö. and E. Uncuoğlu, "Comparison of Three Back-propagation Training Algorithms for Two Case Studies," 2005. 59. Moré, J. J., "The Levenberg-Marquardt Algorithm: Implementation and Theory," in Numerical Analysis: Springer, 1978, pp. 105-116. 60. 61. 62. 63. 64. Hagan, M. T. and M. B. Menhaj, "Training FEEDFORWARD networks with the Marquardt Algorithm," IEEE Transactions on Neural Networks, vol. 5, pp. 989-993, 1994. Bouabaz, M. and M. Hamami, "A Cost Estimation Model for Repair Bridges Based on Artificial Neural Network," American Journal of Applied Sciences, vol. 5, pp. 334-339, 2008. Version, M., "7.2. 0.232 (R2006a). The MathWorks," Inc. www. Mathworks. com, 2006. Vogl, T. P., J. Mangis, A. Rigler, W. Zink, and D. Alkon, "Accelerating the Convergence of the Back-propagation Method," Biological Cybernetics, vol. 59, pp. 257-263, 1988. Rada, G. R., G. Elkins, B. Henderson, R. Van Sambeek, and A. Lopez, "LTPP Seasonal Monitoring Program: Instrumentation Installation and Data Collection Guidelines," 1995. 103