INTEGRATING INSTRUMENTATION DATA IN PROBABILISTIC PERFORMANCE PREDICTION OF FLEXIBLE PAVEMENTS

ABSTRACT

The goal of this research was to develop a methodology integrating instrumentation data with existing mechanistic-empirical performance models for flexible pavements. The methodology is further enhanced with probabilistic features that take into account the uncertainties associated with design parameters. Two types of pavement structures are considered: 1) full-depth structures, including subbase, base, and Superpave-designed HMA layers constructed over subgrade and 2) structural overlays including only Superpave-designed HMA layers. One pavement section per structure type was selected from the instrumented sections of a comprehensive research project called the Superpave In-Situ Stress/Strain Investigation (SISSI), sponsored by Pennsylvania Department of Transportation.

The first task of this project was to simulate pavement response using 3-D viscoelastic-based finite element models. A sensitivity analysis was then conducted to identify site-specific parameters that are required by empirical performance models. The variabilities associated with these parameters were quantified and further considered in a Monte Carlo simulation-based probabilistic approach. The predicted performance measures included the overall pavement functional performance (IRI) and structural performance in terms of individual distresses over a specified analysis period.

The main contribution of this research is not toward the development of new performance prediction models but, rather, the demonstration of the use of instrumentation data for performance predictions. The developed methodology is enhanced with an analytical method to predict pavement responses over time and thus will be ideally suited for situations where sophisticated instrumentation data are not available. In addition, the probabilistic nature of the developed methodology proposes a unique way of assessing the effects of variabilities of design parameters on pavement performance.

 

TABLE OF CONTENTS

LIST OF FIGURES ………………………………………………………………………………………..ix

LIST OF TABLES………………………………………………………………………………………….xiii

ACKNOWLEDGEMENTS……………………………………………………………………………..xvii

Chapter 1  Introduction……………………………………………………………………………………1

1.1 Problem Statement………………………………………………………………………………1

1.2Research Goal and Objectives……………………………………………………………..21.3Research Scope………………………………………………………………………………….21.4Research Hypothesis…………………………………………………………………………..21.5Research Approach…………………………………………………………………………….3

1.5.1 Task 1. Literature Review …………………………………………………………..3

1.5.2 Task 2. Preliminary Data Analysis – Phase I …………………………………4

1.5.3 Task 3. Simulation of Pavement Response Using 3-D Finite

Element Modeling – Phase II…………………………………………………………4

1.5.4 Task 4. Strain Response Prediction – Phase II………………………………..4

1.5.5 Task 5. Sensitivity Analysis to Identify Site-Specific Parameters –

Phase III……………………………………………………………………………………..4

1.5.6 Task 6. Variability Study of Site-Specific Parameters – Phase III…….5

1.5.7 Task 7. Probabilistic Performance Prediction – Phase III………………..5

1.6Research Contributions……………………………………………………………………….5

Chapter 2  Research Background and Literature Review……………………………………..6

2.1Research Background…………………………………………………………………………6

2.1.1Pavement Performance Measures………………………………………………..6

2.1.1.1Fatigue Cracking ……………………………………………………………..7

2.1.1.2Rutting……………………………………………………………………………7

2.1.1.3 Thermal Cracking……………………………………………………………..8

2.1.1.4 Smoothness………………………………………………………………………8

2.1.2Pavement Performance Prediction ………………………………………………9

2.1.2.1 Fatigue Cracking ………………………………………………………………11

2.1.2.2 Rutting…………………………………………………………………………….13

2.1.2.3Smoothness……………………………………………………………………..14

2.2Pavement Instrumentation…………………………………………………………………..15

2.2.1 State-of-the-Art………………………………………………………………………….15

2.2.1.1MnRoad………………………………………………………………………….16

2.2.1.2Virginia Smart Road…………………………………………………………16

2.2.1.3Ohio National Test Road…………………………………………………..16

2.2.1.4 NCAT Test Track……………………………………………………………..17

2.2.2Application of Instrumentation Data……………………………………………17

2.3 Deterministic Approach vs. Probabilistic Approach ………………………………..19

2.4Summary…………………………………………………………………………………………..20

Chapter 3  The SISSI Project……………………………………………………………………………22

3.1 Background………………………………………………………………………………………..22

3.2 Objectives………………………………………………………………………………………….22

3.3 Site Selection……………………………………………………………………………………..23

3.4 Pavement Construction………………………………………………………………………..24

3.5 Pavement Instrumentation ……………………………………………………………………25

3.6 Data Collection…………………………………………………………………………………..27

3.6.1Material Characterization Data……………………………………………………27

3.6.2 Instrumentation Data ………………………………………………………………….27

3.6.3 Traffic Data ………………………………………………………………………………28

3.6.4 Climate Data……………………………………………………………………………..28

3.6.5 Falling Weight Deflectometer Data………………………………………………29

3.6.6 Performance Data………………………………………………………………………30

Chapter 4  Preliminary Data Analysis ……………………………………………………………….32

4.1 Introduction……………………………………………………………………………………….32

4.2 Traffic Data ……………………………………………………………………………………….32

4.2.1 General Information …………………………………………………………………..33

4.2.2 Vehicle Operational Speed………………………………………………………….36

4.2.3 Traffic Growth Factor…………………………………………………………………36

4.2.4 Vehicle Class Distribution…………………………………………………………..37

4.2.5 Monthly Adjustment Factor ………………………………………………………..38

4.2.6Hourly Truck Distribution………………………………………………………….394.2.7 Axle Load Distribution……………………………………………………………….39

4.2.8 Number of Axles per Truck Class………………………………………………..41

4.3 Climate Data………………………………………………………………………………………41

4.3.1 General Analysis of Temperature Data…………………………………………41

4.3.2 Pavement Temperature at Blair……………………………………………………45

4.3.3 Pavement Temperatures at Warren ………………………………………………46

4.3.3.1Review of Temperature Prediction Models …………………………46

4.3.3.2Predicting Warren Pavement Temperatures Using EICM……..48

4.4 Pavement Response Data……………………………………………………………………..51

4.4.1 Processing Response Data…………………………………………………………..51

4.4.2 Typical Strain and Stress Response………………………………………………53

4.4.3 Evaluation of Pavement Response ……………………………………………….54

4.5FWD Data…………………………………………………………………………………………56

4.5.1Analysis Results for Warren……………………………………………………….57

4.5.2 Analysis Results for Blair……………………………………………………………57

4.6 AC Material Characterization Data……………………………………………………….59

4.6.1Mechanical Behavior…………………………………………………………………59

4.6.2 Laboratory Tests………………………………………………………………………..60

4.7 Summary……………………………………………………………………………………………65

Chapter 5  Simulation of Pavement Response Using 3-D Finite Element Modeling..66

5.1 Introduction………………………………………………………………………………………..66

5.2Finite Element Model…………………………………………………………………………67

5.2.1Modeling Strategy…………………………………………………………………….68

5.2.2 Boundary Conditions………………………………………………………………….69

5.2.3 Material Properties …………………………………………………………………….70

5.2.3.1 Bound Materials……………………………………………………………….71

5.2.3.2 Unbound Materials……………………………………………………………73

5.2.4 Simulation of Moving Load ………………………………………………………..74

5.2.5 Element Type…………………………………………………………………………….76

5.2.6 Optimum Element Size……………………………………………………………….77

5.2.7 Model Dimensions……………………………………………………………………..86

5.3 Model Validation………………………………………………………………………………..87

5.3.1 Comparison of FEA and Measured Responses ………………………………92

5.3.1.1 Blair FE Model…………………………………………………………………92

5.3.1.2Warren FE Model…………………………………………………………….93

5.3.2 Comparison of FEA and KENLAYER …………………………………………96

5.3.3 Linearity of Pavement Response………………………………………………….99

5.4 Summary……………………………………………………………………………………………100

Chapter 6  Strain Response Prediction……………………………………………………………….101

6.1 Introduction………………………………………………………………………………………..101

6.2 Research Approach……………………………………………………………………………..101

6.2.1 Exploratory Data Analysis ………………………………………………………….103

6.2.2 Regression Analysis …………………………………………………………………..105

6.2.2.1 Speed Effect on Strain Response ………………………………………..105

6.2.2.2 Temperature Effect on Strain Response……………………………….109

6.2.3 Response Superposition………………………………………………………………112

6.2.4 Demonstration Example……………………………………………………………..113

6.3 Summary……………………………………………………………………………………………116

Chapter 7  Sensitivity Study …………………………………………………………………………….117

7.1 Introduction………………………………………………………………………………………..117

7.2 Overview of MEPDG………………………………………………………………………….118

7.2.1 General Considerations ………………………………………………………………119

7.2.2 Hierarchical Input Level……………………………………………………………..120

7.3 Running MEPDG Software………………………………………………………………….121

7.3.1 Description of MEPDG Input………………………………………………………122

7.3.1.1 Traffic Module …………………………………………………………………122

7.3.1.2 Climate Module………………………………………………………………..122

7.3.1.3 Structure Module………………………………………………………………123

7.3.2 Description of MEPDG Output……………………………………………………123

7.4 Sensitivity Study…………………………………………………………………………………125

7.4.1 Analysis Parameters …………………………………………………………………..125

7.4.2 Analysis Results ………………………………………………………………………..126

7.4.2.1 Longitudinal Cracking……………………………………………………….129

7.4.2.2 Alligator Cracking…………………………………………………………….130

7.4.2.3 AC Rutting ………………………………………………………………………131

7.4.2.4 Subgrade Rutting………………………………………………………………132

7.4.2.5 Smoothness………………………………………………………………………132

7.5 Summary……………………………………………………………………………………………133

Chapter 8  Variability Study…………………………………………………………………………….135

8.1 Introduction………………………………………………………………………………………..135

8.2 Statistical Analysis Approach……………………………………………………………….136

8.3 Distribution Analysis…………………………………………………………………………..138

8.3.1 Probability Distribution………………………………………………………………139

8.3.1.1 Normal (Gaussian) Distribution………………………………………….139

8.3.1.2 Lognormal Distribution……………………………………………………..139

8.3.1.3 Weibull Distribution………………………………………………………….140

8.3.2 Estimation of Distribution Parameters ………………………………………….140

8.3.3 Evaluation of Goodness-of-fit……………………………………………………..142

8.3.3.1 Chi-Square Test………………………………………………………………..144

8.3.3.2 Kolmogorov-Smirnov Test ………………………………………………..144

8.3.3.3 Anderson-Darling Test………………………………………………………145

8.3.4 Findings from Distribution Analysis…………………………………………….145

8.4 Variability Analysis…………………………………………………………………………….149

8.4.1 Construction Variability……………………………………………………………..149

8.4.1.1 AC Layer Thickness………………………………………………………….149

8.4.1.2 Air Voids…………………………………………………………………………151

8.4.1.3 Effective Binder Content……………………………………………………153

8.4.2 Field Variability…………………………………………………………………………154

8.4.2.1 Resilient Modulus of Unbound Materials…………………………….155

8.4.2.2 Ground Water Table Depth………………………………………………..161

8.5 Summary……………………………………………………………………………………………165

Chapter 9  Probabilistic Performance Prediction…………………………………………………166

9.1 Introduction………………………………………………………………………………………..166

9.2 Probabilistic Approach ………………………………………………………………………..167

9.3 Implementation of Monte Carlo Simulation……………………………………………168

9.3.1 Random Number Generation……………………………………………………….169

9.3.2 Sampling Strategy ……………………………………………………………………..170

9.3.3 Optimum Number of Simulations ………………………………………………..171

9.4 Evaluation of Performance Predictions ………………………………………………….173

9.5 Summary……………………………………………………………………………………………179

Chapter 10  Summary, Conclusions, and Recommendations………………………………..180

10.1 Summary………………………………………………………………………………………….180

10.2 Findings and Conclusions…………………………………………………………………..182

10.2.1 Principal Findings…………………………………………………………………….182

10.2.2 Conclusions …………………………………………………………………………….182

10.3 Recommendations……………………………………………………………………………..182

Bibliography ………………………………………………………………………………………………….184

Chapter 1

 

Introduction

1.1 Problem Statement

In the current mechanistic-empirical (M-E) design procedures for flexible pavements (MEPDG 2004), the mechanistic response models are used to predict pavement responses, stresses, strains, and deflections. A response model must account for the effects of climate, traffic, material properties, and pavement structure. The complex interaction of these variables calls for utilizing advanced material and mechanics theories such as viscoelasticity, damage mechanics, and fracture mechanics. Empirical performance models are then employed to predict pavement structural and functional performance from mechanistic responses. Performance prediction models are usually derived from statistically based correlations of field performance with observed laboratory specimen performance, full-scale road test experiments, or by both methods. Unfortunately, most of the existing models do not reflect true field conditions, as is evidenced by the fact that the failure of asphalt concrete (AC) materials occurs much more quickly under a laboratory setting than in a field environment. This difference has been typically accounted for by the use of calibration factors based mainly on engineering experience.

Pavement instrumentation has recently become an important tool for monitoring in-situ pavement material performance and quantitatively measuring pavement response under different environmental and traffic conditions. Instrumentation devices are designed to measure, but are not limited to, strains, stresses, deflections, moisture, temperature, and traffic in the field. The concept of the use of instrumentation data for performance predictions is often discussed but to date has only been studied on a limited level (that is, using environmental data) and, for the most part, in a broad conceptual fashion, with respect to limited performance features. Therefore, it is necessary to investigate the feasibility of integrating instrumentation data in mechanistic-empirical performance prediction.

One major limitation of the existing performance models is that they are deterministic models, which do not consider uncertainties associated with input parameters. Although recent research proposes shifting the effort to consider the uncertainty in the performance model, which implies that a performance model should include all relevant sources of uncertainties, little work has been accomplished in this area. There is a need to apply probabilistic concepts to performance predictions.

1.2 Research Goal and Objectives

The goal of this research is to develop a methodology that can integrate instrumentation data with existing mechanistic-empirical performance models for flexible pavements. The methodology will be further enhanced with probabilistic features that take into account uncertainties associated with input parameters. To limit the research scope, a sensitivity analysis will be conducted first so that site-specific parameters can be identified. These parameters then will be considered in the probabilistic analysis. The output will consist of pavement performance describing the overall pavement functional performance (IRI) and structural performance in terms of individual distresses over a specified analysis period. This unique aspect will allow the pavement engineers to assess the uncertainties associated with input parameters based on the probability of performance that may be predicted.

To achieve this goal, the following objectives should be accomplished for this research:

1. integrate instrumentation data in performance prediction.
2. apply probabilistic concepts to performance prediction.

1.3 Research Scope

The focus of this research will be asphalt-surfaced pavements only. Two types of pavement structures will be considered:

• full-depth structure including subbase, base, and Superpave-designed

HMA layers constructed over subgrade and

• structural overlay including only Superpave-designed HMA layers.

One pavement section per structure type will be selected from the instrumented sections of a comprehensive research project called Superpave In-Situ Stress/Strain Investigation (SISSI), sponsored by the Pennsylvania Department of Transportation. In view of the depth of dynamic and environmental data, pavement sites in Blair and Warren counties will be used in this research.

1.4 Research Hypothesis

The general hypothesis in this research is:

Š With well-defined procedures and appropriate assumptions, instrumentation data can be effectively used in performance prediction for flexible pavements.

1.5 Research Approach

To address the research objectives, a three-phase research approach is presented. An overall research framework is illustrated in Figure 1.1. The seven tasks associated with this project are detailed in this section.

 

1.5.1 Task 1. Literature Review

The focus of the literature review was to identify all the applications of pavement instrumentation data (dynamic, environmental, and traffic) and probabilistic analysis techniques. Available information of several ongoing pavement instrumentation projects, as outlined in the introduction section, was searched, during which any information regarding sensitivity studies on performance prediction-related parameters was also compiled.

1.5.2 Task 2. Preliminary Data Analysis – Phase I

In this task, the instrumentation data collected during Phase I of the SISSI project was carefully reviewed. Analytical procedures were developed to process and analyze different types of instrumentation data: traffic, climate, and dynamic.

1.5.3 Task 3. Simulation of Pavement Response Using 3-D Finite Element Modeling – Phase II

In this task, separate 3-D finite element (FE) models were developed for the Blair and Warren sites to capture pavement responses to loading. Since there were periods of data collection interruption for a specific SISSI site due to the loss of the dynamic and environmental sensors or connection problems, comprehensive validation of the FE model was also conducted such that pavement response predicted from FE analysis could be used to fill missing dynamic data.

1.5.4 Task 4. Strain Response Prediction – Phase II

Although it is possible to perform rigorous 3-D finite element analyses, the computational cost still remains a challenge to predicting the distress/damage accumulation schemes incorporated in the MEPDG. In this task, an analytical procedure was developed to accurately and rapidly predict strain response with known traffic and environment information, particularly axle load, vehicle speed, and pavement temperature. This is the key component of integrating instrumentation data in performance prediction. Further discussion on this is presented in Task 7.

1.5.5 Task 5. Sensitivity Analysis to Identify Site-Specific Parameters – Phase III

Using the information compiled during Task 1, sensitivity analyses using the MEPDG software was conducted to assess the importance of parameters required for performance prediction. The sensitivity study was carried out in two steps using the MEPDG software (version 0.910). First, general parameters that have been reported in published literature were summarized. Second, for each of these general parameters, a detailed sensitivity study was carried out to determine which general parameters would affect site-specific pavement performance. Only site-specific parameters identified from the second step are considered in probabilistic analyses. All MEPDG-required input including traffic, climate, pavement structure, and material properties was obtained from instrumentation data.

1.5.6 Task 6. Variability Study of Site-Specific Parameters – Phase III

As soon as the site-specific parameters were determined, an attempt at identifying sources of variation and quantifying the variabilities associated with them was carried out using instrumentation data. Available information on this task was also researched through the literature so that only minimum statistical analyses would be needed.

1.5.7 Task 7. Probabilistic Performance Prediction – Phase III

Probabilistic performance prediction in this task was performed in two steps. In the first step, Monte Carlo simulation techniques were used to simulate each site-specific parameter based on its probability distribution and variability determined from Task 6. In the second step, pavement responses predicted in Task 4 were fed into the empirical performance models adopted in the MEPDG. With this two-step approach, uncertainties associated with analysis parameters were incorporated systematically within the predicted performance. Finally, probabilistic performance predictions were evaluated by comparison to field conditions and to deterministic predictions.

1.6 Research Contributions

The main contribution of this research is not toward the development of new performance prediction models but, rather, the demonstration of utilizing instrumentation data and probabilistic analysis in performance prediction. The most important characteristics of the developed methodology can be summarized as follows:

• The developed methodology utilizes instrumentation data to predict pavement performance over time.
• The predicted pavement performance is based on probability analyses. With known variabilities associated with input parameters, effects of uncertainties on performance predictions can be assessed.

INTEGRATING INSTRUMENTATION DATA IN PROBABILISTIC PERFORMANCE PREDICTION OF FLEXIBLE PAVEMENTS

Sharing is caring!