Using Logistic Regression to Estimate the Probability of Asphalt Binder Failure at Low Temperatures

فاقد چكيده فارسي

Although rheological properties of asphalt binders have been widely used to characterize asphalt mixtures performance in the last 50 years, the cost and availability of the laboratory equipment hampered the spreads of using viscoelasticity to qualify asphalts. During 1987-1992 through the SHRP project a new bitumen grading protocol based on new testing techniques were introduced, which tries to relate rheological properties of binders to pavement performance in the high, intermediate and low service temperature ranges. Each of these ranges has their own failure mechanism and criteria. The goal of the paper is introducing the concept of mathematical modeling of the probability of asphalt failure. The logistic regression approach is used to model the probability of asphalt failure at low service temperatures. The sample set used in the model calibration is selected from a recent laboratory study report, where for each sample the low temperature stiffness and m-values are measured by Bending Beam Rheometer. Some related applications of the resulting model are also discussed.