Bayesian Methodology for Verifying Recommendations to Minimize Asphalt Pavement Distress

摘要

This report documents the development of a Bayesian methodology to quantify multiple regression models from personal interviews with experienced individuals. Procedures for combining this information with objective (measured) data were also demonstrated. Pilot implementation of the methodology was conducted in six states--Arizona, Colorado, Utah, Florida, Louisiana, and Virginia--to investigate the influence of designer-controlled variables on the occurrence of fatigue cracking. Four designer-controlled variables were considered: asphalt penetration, asphalt content, proportion of asphalt concrete, and base density. A multiple regression prediction model of fatigue cracking was defined in each state. In all six states it was found that the fatigue life cycle increased with increase of the asphalt penetration, asphalt content, proportion of asphalt concrete, and base density. The models allow prediction of the fatigue life cycle for any specified level of reliability and designer-controlled variable values. The results obtained were used as a basis to verify design recommendations made in a previous study, NCHRP 9-4, to minimize premature cracking. Only the recommendation to use harder grades of asphalts (lower penetrations) for thick asphalt concrete layers could not be verified. Further study is needed to resolve this difference.