A probabilistic estimation approach for the failure forecast method using Bayesian inference

Positive-feedback mechanisms such as fatigue induce a self-accelerating behavior, captured by models displaying infinite limit-state asymptotics, collectively known the failure forecast method (FFM). This paper presents Bayesian model parameter estimation approach to fully nonlinear FFM implementation and compares results classic linear regression formulation, including uncertainty model. process is demonstrated in cyclic loading crack propagation application, both on synthetic data set full experiment. A novel ”switch point” included formulation account for nonstationary changes growth parameter.