Predicting Roughness Progression of Asphalt Overlays
JOINT C­SHRP/ALBERTA BAYESIAN APPLICATION
Prepared for: Canadian Strategic Highway Research Program (C­SHRP)

Prepared by:

Marian H. Kurlanda, M.Sc., P.Eng.

Alberta Transportation & Utilities

 

Lyle Kajner, M.Sc., P.Eng.

VEMAX Management Inc.

October, 1995

 

EXECUTIVE SUMMARY

A Bayesian regression model was developed to predict the roughness progression of asphalt overlays (first rehabilitation cycle) placed on existing asphalt concrete pavements with granular base courses. The model will be applicable in central Alberta. Field data used in the analysis consisted of 311 records from the Alberta pavement management system. In accordance with Bayesian statistical theory, the "prior knowledge" of the modelling system was provided by five pavement engineers from the agency. Their expert judgment was encoded and linked to form the input prior. Combining the field data with the input prior resulted in a model which expresses the riding comfort index as a function of the RCI of the original pavement (RCI perf), a soil factor, overlay thickness, age, initial RCI, and cumulative traffic loading;

Regression Equation

The standard error of the estimate for the resultant model is deemed reasonable, at 0.498 RCI.

The model can be used in Alberta's PMS although some changes in the PMS program would be needed to facilitate its integration into the software.

The authors believe that the Bayesian modelling tool should be further investigated by Alberta Transportation & Utilities, particularly when historical data is limited. Furthermore, the methodology is not limited to pavement engineering problems. Other areas where Bayes can potentially prove useful include traffic engineering, location studies and geotechnical engineering applications.

 

TABLE OF CONTENTS

1.0 INTRODUCTION

2.0 BAYESIAN METHOD

2.1 Model Selection

2.2 Select Dependent Variable

2.3 Select Model Type

2.4 Select Independent Variables

2.4.1 Development of Soil Roughness Factor

2.5 Postulate Functional Form 6

2.6 Assemble Information

2.6.1 Assembling Sample Data

2.6.2 Assembling Expert Judgement Data

3.0 BAYESIAN ANALYSIS

4.0 MODEL RESULTS AND EVALUATION

5.0 CONCLUSIONS

LIST OF APPENDICES

Appendix A ­ Selection of Independent Variables ­ Questionnaire

Appendix B ­ Selection of Independent Variables ­ Experts' Response

Appendix C ­ Development of Soil Roughness Factor ­ Questionnaire and Experts' Response

Appendix D ­ Encoding Package Appendix E ­Expert Judgement Data

Appendix F ­ Pavement Management System (PMS) (Historical) Data

Appendix G ­ Output of the Bayesian Analysis

Appendix H ­ Sensitivity Analysis

LIST OF TABLES

Table 1 ­ Experts Who Participated in Bayesian Project

Table 2 ­ Encoding Intervals Used in Encoding Matrix

Table 3 ­ Files Used in Bayesian Analysis

Table 4 ­ Summary of Models

Table 5 ­ Evaluation Table for Alberta's Bayesian Roughness Progression Model

LIST OF FIGURES

Figure 1 ­ The Bayesian Template

Figure 2 ­ Expert's Ranking of Independent Variable Candidates

Figure 3 ­ Bayesian Analysis Process

Figure 4 ­ Prediction Sensitivity Results for Each Expert's Prior

(Continue)

Return to Main Page