Nonlinear stochastic modelling of fatigue-related damage in asphalt mixtures with threshold-life analysis

Stochastic fatigue-damage models based on asphalt-mixture fatigue responses provide a probabilistic description of material-level fatigue deterioration relevant to pavement fatigue assessment. However, existing models are still largely formulated under linear accumulation and terminal-failure criteria. They are not yet equipped to connect nonlinear damage evolution with threshold-defined life inference within a common probabilistic framework. This study advances this modelling framework by establishing an explicit section-level probabilistic formulation for fatigue-damage evolution, generalizing stochastic cumulative damage to a nonlinear form, and incorporating threshold-life inference into the same probabilistic framework through censoring treatment and inspection-based loading information. Using laboratory fatigue data from NCHRP 9-44A, the improved stochastic fatigue-damage framework is evaluated for no-rest and with-rest subsets and compared with several cycle-based PSR-related empirical damage representations. Calibration and validation results show that the improved stochastic fatigue-damage framework remains consistent with observations in both subsets. For the no-rest subset, the empirical damage points cluster around the model expectation in the high-damage range, supporting consistency near fatigue failure. For the with-rest subset, the empirical points follow the expected damage path and remain largely within the uncertainty band, supporting consistency over the damage-development stage. Overall, the improved stochastic fatigue-damage framework provides a coherent probabilistic basis for fatigue-damage assessment of asphalt mixtures under grouped laboratory testing settings.