Reliability Analysis of Fatigue Life Prediction from the Viscoelastic Continuum Damage Model

The fatigue resistance of asphalt concrete materials plays an important role in the service life of a pavement. Several empirical and mechanistic models and efforts are under way toward improving these models for accurate prediction of fatigue life. The reliability of predictions with these models has not received substantial attention. This study used the simplified viscoelastic continuum damage model to analyze the reliability of fatigue predictions. Modulus and fatigue tests were conducted on a standard dense graded asphalt mixture to characterize the model and create deterministic fatigue life predictions. Monte Carlo simulations were then used to calculate the reliability of the fatigue predictions, given the variation in input parameters. The analysis was conducted for combinations of three experimental failure criteria and two model failure criteria at two strain levels for a total of 12 study cases. Differences in reliability between the combinations of failure conditions are identified and discussed. There are three major findings of this study. First, depending on the selection of failure criteria, the prediction errors could range from +/- 22% to +/- 52% at a 50% level of reliability, depending on the exact combination of failure criterion and strain level. Second, with an experimental failure criterion of 50%, modulus reduction shows better reliability regardless of the model failure criterion chosen to predict fatigue life. Third, the model failure criterion based on constant pseudostiffness at failure was found to be more reliable than was the pseudostrain energy release rate criterion when characterization was performed with three tests.