Two Novel Approaches for Pavement Performance Modeling: Clusterwise Regression and Mechanistic-empirical Based Probabilistic Procedure

Pavement performance modeling is one of the key challenges facing pavement engineers and researchers. Although many new techniques have been applied to the pavement performance modeling, the accuracy of predictions still needs to be further improved. Two novel approaches were proposed in this work: a clusterwise least squares (CLS) regression method for pavement condition rating and a simplified mechanistic-empirical based probabilistic procedure for fatigue cracking of flexible pavements. In a CLS regression model, several clusters (curves), instead of only one as used in an ordinary least squares (OLS) regression model, are used to fit the modeling dataset. The results of the study show that the proposed model resulted in more accurate predictions than the OLS regression model. In the proposed fatigue cracking model, the cracking area is related to traffic loads through a probabilistic distribution. These two procedures are very useful for researchers and practitioners in the field of pavement management, design, and maintenance. The CLS regression should deserve more attention and efforts from any statistician or person of interest.