A Novel Fatigue and Creep-Fatigue Life Prediction Model by Combining Data-Driven Approach with Domain Knowledge

Abstract

A high-precision and concise-form life prediction model is of essential significance to the deterministic life design and reliability assessment of high-temperature components in structural integrity field. In this work, a novel fatigue and creep-fatigue life prediction model is developed by combining data-driven approach with domain knowledge. To establish the data set for reliable life prediction, a total of 224 sets of data on high-temperature low-cycle fatigue and creep-fatigue interaction of Inconel 718 alloy have been summarized from our prior works. Furthermore, the extrapolation capacity of the new model applied to generalized loading conditions are validated through 125 sets of data available in literature. Meanwhile, the strain energy density exhaustion model and deep neural network are included for comparisons. Results show that the new model presents the best comprehensive performance according to the Bayesian information criterion that both the prediction accuracy and model complexity are acceptable. Finally, the data sets of three different materials are collected to further evaluate the universality of the new model in this work.

Keywords: Life prediction, Symbolic regression, fatigue, Creep-fatigue, Physical interpretability