Development of Predictive Models for Low-Temperature Indirect Tensile Strength of Asphalt Mixtures

摘要

Thermal cracking is the predominant flexible pavement distress in northern climates, causing transverse cracks perpendicular to the direction of traffic. The indirect tensile (IDT) strength test is currently the most widely used method to characterize thermal cracking susceptibility and is required in mechanistic empirical pavement design. When laboratory IDT strength testing data are not available, it is predicted by pavement design software using mixture volumetrics and Superpave performance grade (PG) of the binder. The primary purpose of this study was to examine the IDT strength characteristics of asphalt mixtures commonly used by the Michigan Department of Transportation (MDOT) and to develop improved prediction methods for IDT strength. Laboratory testing of 62 unique MDOT mixtures (a total of 201 samples with replicates) showed that the pavement design software predicted the IDT strength very poorly. Three models were developed to improve the accuracy of IDT strength prediction. First, the current software's predictive equation for IDT strength was locally calibrated. Then, an improved statistical model was developed to predict low-temperature IDT strength, based on the information typically available in job mix formulas. Finally, an artificial neural network (ANN)-based model was developed to further improve the accuracy of the low-temperature strength predictions using information from job mix formulas. All three models showed increased prediction performance when compared with the software's IDT strength prediction.