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

CHAPTER 1. INTRODUCTION

BACKGROUND

The Guide for Mechanistic–Empirical Design of New and Rehabilitated Pavement Structures (MEPDG) was developed under the National Cooperative Highway Research Program (NCHRP) Project 1-37A.⁽²⁾ This pavement design methodology is based on various performance prediction models that relate mechanistically calculated pavement responses to empirically measured field performance. The performance models in the current American Association of State Highway and Transportation Officials (AASHTO) AASHTOWare® Pavement ME Design software were calibrated based on Federal Highway Administration’s (FHWA’s) Long-Term Pavement Performance (LTPP) data from all North American regions with various material characteristics, traffic patterns, and climatic conditions.⁽¹⁾

Numerous State agencies have adopted the AASHTO recommended calibration procedure developed in NCHRP Project 1-40B for their State or regional conditions.^(3,4) This calibration approach minimizes the sum of squared errors (SSE) between measured and predicted pavement performance. Different States have reported calibration results with varying levels of success depending on the performance models. In this single-objective optimization approach to calibration, the SSE and, if needed, the standard deviation of error, is minimized in two separate steps.

Using a multi-objective optimization approach enables researchers to escape preconception, avoid excessive concentration on only one aspect of the problem, and combine multiple sources of information in an objective manner. In this study, application of a multi-objective optimization approach to enhance calibration of the AASHTOWare® Pavement ME Design software performance models is investigated. To demonstrate the multi-objective approach, a subset of the LTPP data from Florida is selected for calibration of MEPDG permanent deformation (rutting) prediction models for new and rehabilitated flexible pavements. In addition to the LTPP data, some accelerated pavement testing (APT) data were received from the Florida Department of Transportation (FDOT) State Materials Office. This project was awarded by FHWA as a result of a proposal in response to the FHWA Broad Agency Announcement (BAA) on analysis of LTPP data.

RESEARCH OBJECTIVES AND OUTCOMES

The proposed research is aimed at calibration and validation of pavement performance prediction models, which was considered as an objective (research area d.i) in the BAA. Employing LTPP data, this research effort will support strategic objective number 5, “Development of Pavement Response and Performance Models Applicable to Pavement Design and Performance Prediction,” of the current LTPP Strategic Plan for LTPP Data Analysis. Through application of alternative computational tools, calibration of pavement performance models is enhanced. The following enhancements are anticipated as a result of this research effort:

• Increase utilization of the available information in each iteration of the computationally intensive calibration process by simultaneously increasing accuracy and precision (minimizing mean and standard deviation of error) of the performance models.
  
  
• Combine multiple disparate sources of data into the calibration process in an objective manner.
  
  
• Provide engineering judgment and qualitative criteria to select reasonable calibration coefficients from the final pool of solutions that result from a multi-objective optimization.

Using this multi-objective calibration approach, multiple sources of information are incorporated in an objective manner, resulting in a final set of tradeoff solutions. This way, none of the viable sets of calibration factors are eliminated prematurely, and all of the nondominated solutions are included in the final tradeoff front. Exploring the final front might reveal unknown aspects of this calibration problem and result in more reasonable calibration coefficients that could not be identified using single-objective approaches.

If successful, a superior calibration of the ME software performance models is realized using multi-objective optimization compared to conventional single-objective methods. Of particular value, this study demonstrates how State and local agencies can adopt this multi-objective approach to determine more reasonable calibration factors for their pavement networks at all age and distress levels. This contribution can result in more economical and justifiable pavement design considering multiple aspects of pavement performance.

REPORT ORGANIZATION

Following an overview of the corresponding literature in chapter 2, this report discusses the extraction of relevant LTPP and APT data and generating the ME software input files in chapter 3. Chapter 4 explains the programming approach to single-objective (according to AASHTO guidelines) and multi-objective calibration. Chapter 5 presents the calibration results and discusses a comprehensive comparison of the final performance models calibrated through the single-objective and multi-objective approaches. Finally, chapter 6 concludes with insights from the discussion of results and provides some recommendations for future implementation of this multi-objective calibration approach and further research to enhance the methodology.