Development and Verification of the Overlay Tester Based Fatigue Cracking Prediction Approach (9-1502-01-8)

Fatigue cracking is one of the major distress modes in the long-term performance of asphalt pavements. Many research efforts have been made to study the fatigue behavior of asphalt pavements. Various fatigue analysis approaches have been developed, and some are in use today. However, many approaches (or models) are inaccurate in predicting fatigue performance of asphalt pavements. Consequently, fatigue cracking prediction continues to be a major concern for pavement engineers. The main objective of this project was to develop and verify the overlay tester (OT) based fatigue cracking prediction approach in which the OT is used to determine fracture properties (A and n) of hot-mix asphalt (HMA) mixtures. This approach was developed based on fracture mechanics. However, not only was the fatigue crack propagation characterized by Paris¿ law fracture concepts, but the crack initiation described by traditional fatigue model was also included in this approach. In this approach, the fundamental HMA fracture properties (A and n) were used to estimate fatigue life of asphalt pavements including crack initiation and crack propagation. Note that the traditional fatigue (crack initiation) model parameters, k1 and k2, were also estimated from the fracture properties (A and n) through the quantitative relationship that was established based on theoretical investigation and historical bending beam fatigue test data (1348 data sets). Field verification of this OT based fatigue crack prediction approach was demonstrated with performance data collected from the Federal Highway Administration's Accelerated Loading Facility (FHWA-ALF) test program. The proposed approach has been successfully verified by analyzing five FHWAALF fatigue test lanes. The ranking of the predicted fatigue life for the ALF test lanes had a very good agreement with the measured fatigue performance data under the ALF loading. However, further field validation using other data sets is still needed.