The Prediction of LWST Values from DFT and CTM Measurements Using Linear and Nonlinear Regression Analyses

The objective of this work is to develop statistical models to predict Locked Wheel Skid Trailer (LWST) skid numbers from Dynamic Friction Tester (DFT) and Circular Texture Meter (CTM) measurements conducted on asphalt pavement surfaces. The analyses conducted are descriptive as well as analytical. They include all descriptive measures along with linear and nonlinear regressions. For both analyses, DFT measurements at 20 km/h (12.5 mph) and 64 km/h (40 mph) and Mean Profile Depth (MPD) were used to predict LWST skid values. Furthermore, the International Friction Index (IFI) parameters (F60 and SP) were used in an additional analysis to predict LWST skid values. Multiple linear regression techniques were used to identify the significant quantitative predictors. Model selection using stepwise regression showed that DFT64 and MPD are statistically significant predictors. Moreover, regression analysis showed that DFT20 is highly correlated with other predictors and therefore removed due to multicollinearity. Additionally, it was shown that F60 and SP are also significant in predicting the dependent variable with slightly less correlation coefficients. Nonlinear regression technique was also utilized for the same purpose. In both cases, higher correlation coefficients were noticed when using the nonlinear method as opposed to the multiple linear regression method.