Estimation of Pavement Crack Initiation Models by Combining Experimental and Field Data

Abstract

Development of deterioration models for pavements is an essential part of maintenance and rehabilitation planning. Often, when pavement-deterioration models are developed for developing countries or states that do not have regularly scheduled condition surveys, the only available data are experimental. Experimental data sets fail to include environmental variables and cannot capture properly the aging process of pavements, so the estimated models suffer from biases. This paper describes the development of a pavement-crack-initiation model by combining experimental and field data to correct for such biases. The American Association of State Highway Officials (AASHO) Road Test is used as the experimental data set, which is combined with field data from the Washington State Department of Transportation. The two models are first estimated separately, correcting for endogeneity biases that may exist in the field model. Joint estimation is used next to quantify the bias in the experimental data set and estimate the parameters of the combined model. This study shows that joint estimation can lead to more robust crack-initiation models compared to those estimated separately by the two data sets.