Prediction Models for Transverse Cracking of Jointed Concrete Pavements

The main objective of this study was to develop improved prediction models for transverse cracking of jointed concrete pavements with the Long-Term Pavement Performance database. The retrieval, preparation, and cleaning of the database were carefully handled with a systematic and automatic approach. The prediction accuracy of the existing prediction models implemented in the recommended Mechanistic–Empirical Pavement Design Guide (NCHRP Project 1-37A) was found to be inadequate. Exploratory data analysis indicated that the normality assumption with random errors and constant variance by using conventional regression techniques might not be appropriate for this study. Therefore, several modern regression techniques, including the generalized linear model and the generalized additive model, along with the assumption of Poisson distribution, were adopted for the modeling process. The resulting mechanistic–empirical model included several variables—such as pavement age, yearly equivalent single-axle loads (ESALs), accumulated ESALs, annual precipitation, freeze–thaw cycle, annual temperature range, stress ratio, and percent steel—for the prediction of transverse cracking. The goodness of fit was further examined through significant testing and various sensitivity analyses of pertinent explanatory parameters. The tentatively proposed predictive models appeared to agree reasonably with the pavement performance data, although their further enhancements are possible and recommended.