The development of deterioration models is an essential element for pavement management systems and the optimal design of facilities. This paper describes the development of crack initiation models for flexible pavements by combining experimental and field data. A European pavement database, the Performance Analysis for Road Infrastructure (PARIS) project database, which consists of both Real-Time Loading Testing (RLT) data and AcceleratedLoading Testing (ALT) data, is used. Crack initiation models are developed first separately for the two datasets. Subsequently, joint estimation isperformed on the two datasets simultaneously. Biases in the model parameters, caused by the endogeneity of construction variables and by data censoring, are corrected using appropriate econometric techniques. Joint estimation allows for the identification and quantification of the effect of variables that are missing from the field dataset. Moreover, it accounts and corrects for biases in the experimental dataset. The results of the study indicate that endogeneity bias correction leads to more reliable models and joint estimation to models with improved statistical properties and prediction accuracy.
Abstract