Exploratory analysis of LTPP faulting data using statistical techniques

Abstract Long-Term Pavement Performance (LTPP) database that stores considerable and free-to-access pavement information provides beneficial resources to researchers to develop transverse joint faulting prediction models that are very useful for joint concrete pavement design, rehabilitation and management. To develop accurate and robust faulting prediction models, the investigation of the LTPP faulting data is a prerequisite. This study conducted an Exploratory Data Analysis (EDA) of LTPP data by performing statistical analysis and graphically displaying the relevant factors and their correlation with faulting. This analysis was conducted based on two parts of the LTPP historical data, i.e., the pre-repair and post-repair faulting. For the pre-repair data, the relevant factors in faulting are classified into four categories, namely, traffic repetition, pavement information, local climate and material properties. To better examine the effect of the relevant factors in faulting, the descriptive statistics of factors were calculated and the grey relational analysis and the simple linear regression with one variable at a time were performed. In the regression testing, the P-value shows the significance of the individual relevant factors but it is likely to contradict the realistic relationships when the spurious correlation occurs. Through a thorough investigation, the study illustrated the rationale of the occurrence of the spurious correlation. For the post-repair data, the LTPP maintenance data were examined to evaluate the effectiveness of the individual maintenance treatment by calculating the faulting reduction after applying the treatment.