Advanced Models for Predicting Aggregate Rutting Behavior
Keywords:Artificial newural networks, Unbound aggregates, Rutting, Permanent deformation modeling, Laboratory stress path testing
Artificial neural network (ANN) based advanced aggregate rutting models have been developed and compared for performance using laboratory test data. The primary goal has been to properly characterize the loading stress path dependent permanent deformation behavior from advanced repeated load triaxial tests that can simulate in the laboratory the varying stress states under actual moving wheel load conditions. The aggregate specimens tested were the Federal Aviation Administration (FAA) specified P209 base and P154 subbase materials also used in the pavement test sections of the FAA’s National Airport Pavement Test Facility (NAPTF). Due to the complex loading regimes followed in the laboratory tests and the full-scale NAPTF testing, the ANN rutting models that altogether considered as inputs the static and dynamic components of the applied stresses and the loading stress path slope produced the greatest accuracy. Such advanced neural network models can better describe the aggregate rutting behavior under actual field loading conditions.