ON PREDICTING PAVEMENT SURFACE DISTRESS WITH EMPIRICAL MODELS OF FAILURE TIMES

Abstract

A statistical procedure that overcomes common difficulties in the development of distress models from empirical data is described and applied in an example. The method applies failure-time theory, using maximum likelihood methods so that censored data can be included to prevent statistical bias in the predictions. The variability of failure times is represented by a Weibull distribution because that is considered the most appropriate for the concurrent mechanisms of fatigue and weathering and is flexible in shape. An example application of the procedure to a major analysis of Brazilian condition data shows that it permitted the use of a much wider data base than would otherwise be possible and produced important results that quantify strong effects of weathering and variability.