Fatigue crack growth prediction method under variable amplitude load based on dynamic Bayesian network

A method of fatigue crack growth prediction based on dynamic Bayesian network (DBN) was proposed to address the problem of large deviation of prediction results due to less consideration of uncertainties in the current study of fatigue crack growth prediction. The physical state equation of fatigue crack expansion was constructed using the unified fatigue life prediction (UFLP) model using the fatigue crack expansion under variable amplitude load as the specific object. The link between typical uncertainties in the fatigue crack growth process was analyzed, and the degradation model of fatigue crack growth was built based on the DBN. Finally, the particle filter (PF) algorithm was used to input the crack observation data into the dynamic degradation model, and the prediction results were corrected to reduce the influence of uncertainties. A simulation example with uncertainty factor fatigue crack expansion prediction was given based on the existing experimental data on crack growth, and results show that the DBN-based method for fatigue crack growth prediction can achieve better prediction accuracy than existing methods. Copyright ©2021 Journal of Zhejiang University (Engineering Science). All rights reserved.