Using Multi-Objective Optimization to Enhance Calibration of Performance Models in the Mechanistic-Empirical Pavement Design Guide

PDF Version (3.77 MB)

HTML Version of Errata for FHWA-HRT-17-104

PDF Version of Errata

The following changes were made to the document after publication:

FOREWORD

This report documents research that applied Long-Term Pavement Performance Data to develop an improved approach to calibrating the American Association of State Highway and Transportation Officials’ (AASHTO) AASHTOWare® Pavement ME Design performance models.⁽¹⁾ Whereas the current AASHTO guidelines used in the Pavement ME Design software for calibration of the performance prediction models to local conditions (e.g., materials, traffic, and climate) relies on single-objective minimization of bias and standard error (STE), this report investigates the use of multi-objective optimization to enhance the calibration of the performance models.

The multi-objective optimization results in a final pool of tradeoff solutions where none of the viable sets of calibration factors are prematurely eliminated. This report also demonstrates the application of engineering judgment and qualitative criteria to select reasonable calibration coefficients from the final pool of solutions that result from the multi-objective optimization. More reasonable calibration factors result in a more justifiable pavement design when considering multiple aspects of pavement performance. This investigation revealed that simply evaluating the bias and STE is not adequate for a comprehensive evaluation of performance prediction models. This report is intended for pavement engineers and State transportation departments.

Cheryl Allen Richter, Ph.D., P.E.
Director, Office of Infrastructure
Research and Development

Technical Report Documentation Page

SI* (Modern Metric) Conversion Factors

TABLE OF CONTENTS

EXECUTIVE SUMMARY

CHAPTER 1. INTRODUCTION

• BACKGROUND
• RESEARCH OBJECTIVES AND OUTCOMES
• REPORT ORGANIZATION

CHAPTER 2. REVIEW OF LITERATURE ON CALIBRATING THE MECHANISTIC–EMPIRICAL PAVEMENT PERFORMANCE MODELS

• INTRODUCTION
• MECHANISTIC–EMPIRICAL PAVEMENT PERFORMANCE MODELS
• INPUT VARIABLES
• SENSITIVITY ANALYSIS OF MEPDG PERFORMANCE MODELS
  • Sensitivity to Input Variables
  • Sensitivity to Calibration Factors
    
    
• STATE CALIBRATIONS OF MEPDG PERFORMANCE MODELS
• OTHER MEPDG CALIBRATION EFFORTS
• MULTI-OBJECTIVE CALIBRATION STUDIES
• INSIGHTS AND OBSERVATIONS FROM THE LITERATURE REVIEW
• IMPACT ON RESEARCH APPROACH

CHAPTER 3. PREPARATION OF MEPDG INPUTS FROM LTPP DATA

• INTRODUCTION
• LTPP DATA AVAILABILITY
• GENERATION OF MEPDG INPUT VARIABLES BASED ON LTPP DATA
  • Project Information
  • Performance Criteria
  • Traffic Data
  • Climate Data
  • Pavement Structure and Materials Data
  • Layer Thickness and Type of Material
  • New Asphalt Concrete Layer
  • Existing Asphalt Concrete Properties
  • Asphalt-Treated Base and Permeable Asphalt-Treated Base
  • Additional Asphalt Concrete Layer Properties
  • Unbound and Subgrade Materials
  • Pavement Permanent Deformation
    
    
• ASSEMBELED CALIBRATION DATASETS
  • SPS-1 Sections
  • SPS-5 Sections
  • FDOT APT Data

CHAPTER 4. PROGRAMMING METHODOLOGY

• INTRODUCTION
• PROGRAMMING SINGLE-OBJECTIVE CALIBRATION
• PROGRAMMING MULTI-OBJECTIVE OPTIMIZATION FRAMEWORK
• MEPDG PERFORMANCE PREDICTION FOR FUNCTION EVALUATION
  • Total Pavement Permanent Deformation
  • Simulating Permanent Deformation in Asphalt Concrete Layers
  • Simulating Permanent Deformation in Unbound Materials
    
    
• CALCULATION OF THE MULTIPLE OBJECTIVE FUNCTIONS

CHAPTER 5. COMPARISON OF MULTI-OBJECTIVE TO SINGLE-OBJECTIVE CALIBRATION RESULTS

• INTRODUCTION
• SINGLE-OBJECTIVE CALIBRATION RESULTS
• MULTI-OBJECTIVE CALIBRATION RESULTS
  • Scenario 1: Optimizing Major Statistical Outcomes (Two-Objective)
  • Scenario 2: Combining Different Sources of Data (Four-Objective)
    
    
• COMPARISON OF RESULTS
  • Quantitative Comparison
  • Qualitative Comparison
  • Quantitative and Qualitative Comparison

CHAPTER 6. SUMMARY OF FINDINGS AND RECOMMENDATIONS

• SUMMARY OF FINDINGS
• RECOMMENDATIONS FOR STATE HIGHWAY AGENCIES
• RECOMMENDATIONS FOR FURTHER RESEARCH AND VALIDATION

APPENDIX A. DETAILS OF CALIBRATION INPUT DATA

• ASPHALT CONCRETE DYNAMIC MODULUS
• ASPHALT CONCRETE CREEP COMPLIANCE
• RESILIENT MODULUS FOR UNBOUND AND SUBGRADE LAYERS
• RUTTING DATA
  • SPS-1
  • SPS-5
  • APT ARB
  • APT DASR

APPENDIX B. PROGRAMMING CODES

• PROBLEM EXECUTION: SOLVING THE MULTI-OBJECTIVE OPTIMIZATION PROBLEM
• RUTTING CALIBRATION PROBLEM: SETTING UP PROBLEM SOLUTIONS AND OBJECTIVE FUNCTIONS
• CLASS FOR DEFORMATION CALCULATIONS
• CALCULATION OF CUMULATIVE RUTTING IN ASPHALT CONCRETE LAYERS
• CALCULATION OF CUMULATIVE RUTTING IN UNBOUND LAYERS

APPENDIX C. COMPARISON OF SIMULATED RUTTING CALCULATIONS TO ME SOFTWARE RESULTS

REFERENCES

LIST OF FIGURES

Figure 1. Flowchart. The AASHTO recommended procedure for local calibration of MEPDG performance models, steps 1 through 5
Figure 2. Flowchart. The AASHTO recommended procedure for local calibration of MEPDG performance models, steps 6 through 11
Figure 3. Screenshot. Location of sections within the wet, no freeze climate and on coarse subgrades from InfoPaveTM
Figure 4. Screenshot. Location of sections within the wet, no freeze climate and on fine subgrades from InfoPaveTM
Figure 5. Illustration. Pavement structure in Florida SPS-1 test sections
Figure 6. Chart. Average rutting measurements on SPS-1 test sections 120107 to 120109.
Figure 7. Chart. Average rutting measurements on SPS-1 test sections 120104 to 120161.
Figure 8. Illustration. Pavement structure in Florida SPS-5 test sections
Figure 9. Chart. Average rutting measurements on SPS-5 test sections
Figure 10. Chart. Average rutting measurements on SPS-5 test sections
Figure 11. Illustration. FDOT DASR project sections
Figure 12. Chart. Rutting for the four sections tested under FDOT DASR project
Figure 13. Illustration. FDOT ARB project sections
Figure 14. Chart. Rutting for the seven sections tested under FDOT ARB project
Figure 15. Flowchart. Multi-objective calibration framework
Figure 16. Chart. Example comparison of the simulated calculation to Pavement ME software output (on SPS-1 test section 120102)
Figure 17. Flowchart. Framework for comparison of the calibrated performance models
Figure 18. Chart. Dynamic plot of SSE in single-objective optimization on Florida SPS-1 data
Figure 19. Scatterplot. Measured versus predicted single-objective calibration results of rutting models for new pavements on calibration dataset for Florida SPS-1
Figure 20. Scatterplot. Measured versus predicted single-objective calibration results of rutting models for new pavements on validation dataset for Florida SPS-1
Figure 21. Chart. Dynamic plot of SSE in single-objective optimization on Florida SPS-5 data
Figure 22. Scatterplot. Measured versus predicted single-objective calibration results of rutting models for overlaid pavements on calibration dataset for Florida SPS-5
Figure 23. Scatterplot. Measured versus predicted single-objective calibration results of rutting models for overlaid pavements on validation dataset for Florida SPS-5
Figure 24. Scatterplot. The final nondominated solution set for two-objective calibration of rutting models for new pavements on Florida LTPP SPS-1 data
Figure 25. Scatterplot. Measured versus predicted two-objective calibration results of rutting models for new pavements on calibration dataset for Florida SPS-1
Figure 26. Scatterplot. Measured versus predicted two-objective calibration results of rutting models for new pavements on validation dataset for Florida SPS-1
Figure 27. Scatterplot. The final nondominated solution set for two-objective calibration of rutting models for overlaid pavements on Florida LTPP SPS-5 data
Figure 28. Scatterplot. Measured versus predicted two-objective calibration results of rutting models for overlaid pavements on calibration dataset for Florida SPS-5
Figure 29. Scatterplot. Measured versus predicted two-objective calibration results of rutting models for overlaid pavements on validation dataset for Florida SPS-5
Figure 30. Scatterplot. The final nondominated solution set for four-objective calibration of rutting models for new pavements: F1 and F2 are SSE and STE on Florida LTPP SPS-1 data, and F3 and F4 are SSE and STE on FDOT APT data
Figure 31. Chart. The final nondominated solution set for four-objective calibration of rutting models for new pavements: RMSE and STE on Florida SPS-1 and FDOT APT data
Figure 32. Scatterplot. Two-dimensional representation of the final nondominated solution set for four-objective calibration: SSE on Florida LTPP SPS-1 versus SSE on FDOT APT data
Figure 33. Scatterplot. Two-dimensional representation of the final nondominated solution set for four-objective calibration: STE on Florida LTPP SPS-1 versus STE on FDOT APT data
Figure 34. Scatterplot. Measured versus predicted four-objective calibration results of rutting models for new pavements on calibration dataset for Florida SPS-1
Figure 35. Scatterplot. Measured versus predicted four-objective calibration results of rutting models for new pavements on validation dataset for Florida SPS-1
Figure 36. Bar chart. Comparison of the quantitative criteria for the calibrated rutting models on SPS-1
Figure 37. Bar chart. Comparison of the quantitative criteria for the calibrated rutting models on SPS-5
Figure 38. Bar chart. Comparison of the qualitative criteria for the calibrated rutting models on SPS-1
Figure 39. Bar chart. Comparison of the qualitative criteria for the calibrated rutting models on SPS-5
Figure 40. Chart. Predicted and measured rutting deterioration on FL SPS-1 section 120108
Figure 41. Chart. Predicted and measured rutting deterioration on FL SPS-5 section 120509
Figure 42. Flowchart. Framework for implementation of multi-objective calibration
Figure 43. Chart. Comparison of simulated rutting calculations to ME software results for test section 120102 with β𝑟1 = 1.05, β𝑟2 = 0.9, β𝑟3 = 0.85, βGB = 1.0, βSG = 1.0
Figure 44. Chart. Comparison of simulated rutting calculations to ME software results for test section 120102 with β𝑟1 = 1.05, β𝑟2 = 1.15, β𝑟3 = 0.85, βGB = 1.0, βSG = 1.0
Figure 45. Chart. Comparison of simulated rutting calculations to ME software results for test section 120102 with β𝑟1 = 1.0, β𝑟2 = 0.9, β𝑟3 = 0.9, βGB = 1.0, βSG = 1.0
Figure 46. Chart. Comparison of simulated rutting calculations to ME software results for test section 120102 with β𝑟1 = 0.7, β𝑟2 = 1.02, β𝑟3 = 1.06, βGB = 1.0, βSG = 1.0
Figure 47. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 0.51, β𝑟2 = 1.0, β𝑟3 = 0.7, βGB = 1.0, βSG = 1.0
Figure 48. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 0.9, β𝑟2 = 1.0, β𝑟3 = 1.0, βGB = 1.0, βSG = 1.0
Figure 49. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 1.0, β𝑟2 = 0.9, β𝑟3 = 1.0, βGB = 1.0, βSG = 1.0
Figure 50. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 1.0, β𝑟2 = 1.0, β𝑟3 = 0.9, βGB = 1.0, βSG = 1.0
Figure 51. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 1.25, β𝑟2 = 1.04, β𝑟3 = 0.94, βGB = 1.0, βSG = 1.0
Figure 52. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 1.17, β𝑟2 = 1.1, β𝑟3 = 1.05, βGB = 1.0, βSG = 1.0
Figure 53. Chart. Comparison of simulated rutting calculations to ME software results for test section 120502 with β𝑟1 = 1.17, β𝑟2 = 1.1, β𝑟3 = 1.05, βGB = 1.15, βSG = 0.9

LIST OF TABLES

Table 1. Calibration factors in prediction models for rutting and fatigue cracking in flexible pavements
Table 2. Sensitive design inputs for rutting and fatigue cracking models. NSIm±2s values are given in parentheses
Table 3. Elasticity of MEPDG calibration factors in rutting and fatigue cracking models for Washington State DOT flexible pavements
Table 4. Major State efforts for calibration of MEPDG performance models
Table 5. Local calibration factors for MEPDG fatigue cracking and rutting prediction models
Table 6. Available number of test sections for each LTPP climatic region and subgrade type
Table 7. General information on the 52 flexible test sections on coarse subgrade soils in Florida
Table 8. Source and availability of traffic data for the selected 52 flexible sections in Florida
Table 9. Source and availability of structure data for the selected 52 flexible sections in Florida
Table 10. Availability of rutting data for the selected LTPP flexible pavements in Florida
Table 11. General project information
Table 12. Performance criteria
Table 13. Traffic input data sources and default values
Table 14. Climate information
Table 15. Layer thickness and type of material
Table 16. Mixture volumetric data
Table 17. Binder properties
Table 18. Mixture properties
Table 19. LTPP data tables and fields for backcalculated moduli
Table 20. Additional AC layer properties
Table 21. LTPP data sources for unbound materials properties
Table 22. Bedrock material properties
Table 23. C-values to convert the backcalculated layer modulus values to an equivalent resilient modulus measured in laboratory
Table 24. LTPP data source for rutting measurements (wire reference method)
Table 25. AASHTOWare® Pavement ME Design software data files
Table 26. “Verification” of the global rutting model for new pavements on Florida SPS-1
Table 27. “Verification” of the global rutting model for overlaid pavements on Florida SPS-5
Table 28. Single-objective calibration results of rutting models for new pavements on Florida SPS-1
Table 29. Single-objective calibration results of rutting models for overlaid pavements on Florida SPS-5
Table 30. Candidate solutions from the two-objective nondominated front for SPS-1, with minimum difference in skewness and kurtosis between predicted and measured distributions
Table 31. Two-objective calibration results of rutting models for new pavements on Florida SPS-1
Table 32. Solutions from the two-objective nondominated front for SPS-5, with difference in skewness and kurtosis between predicted and measured data distributions
Table 33. Two-objective calibration results of rutting models for overlaid pavements on Florida SPS-5
Table 34. Candidate solutions from the four-objective nondominated front with difference in skewness and kurtosis between the predicted and measured data distributions
Table 35. Four-objective calibration results of rutting models for new pavements on Florida SPS-1 data
Table 36. Final selected calibration factors
Table 37. AAE of calibrated models in predicting the rutting deterioration rates
Table 38. Dynamic modulus for the Florida SPS-1 and SPS-5 experiment test sections.
Table 39. Calculated resilient modulus of unbound materials
Table 40. Average measured rut depth for Florida SPS-1 test sections 120107 to 120111.
Table 41. Average measured rut depth for Florida SPS-1 test sections 120112 to 120105.
Table 42. Average measured rut depth for Florida SPS-1 test sections 120101 to 120161.
Table 43. Average measured rut depth for Florida SPS-5 test sections 120502 to 120565.
Table 44. Average measured rut depth for Florida SPS-5 test sections 120509 to 120504.
Table 45. Average measured rut depth for Florida SPS-5 test sections 120562 to 120564.
Table 46. Average measured rut depth (mm) for FDOT ARB experiment sections
Table 47. Average measured rut depth (mm) for FDOT DASR experiment sections
Table 48. Developed source codes for multi-objective calibration of MEPDG rutting models
Table 49. Range of the calibration factors reported in the literature

LIST OF ABBREVIATIONS