Machine learning-based genetic feature identification and fatigue life prediction

Considering the nonlinear relationship between variables and fatigue life and the computational burden, a machine learning method integrating the artificial neural network (ANN) and partial least squares (PLS) algorithm was proposed as a framework to identify the genetic features through optimizing fatigue life prediction. Twenty-seven specimens of 316LN stainless steel under uniaxial and multiaxial loadings were used as examples. As results, early fatigue data were proved to be informative for fatigue life prediction. Moreover, five genetic features were identified out of them, and a predicting model was developed. The predicted fatigue life of these samples using only these five genetic features were all located within the 1.5-factor band. This framework can be easily extended to identify genetic features and to predict fatigue life of other materials under different loadings. Therefore, it provides an efficient option in this field to greatly reduce experimental time and cost.

Automated summary

A machine learning method integrating ANN and PLS algorithm was proposed to identify genetic features by optimizing fatigue life prediction. Successful development of a predicting model and identification of the significance of early fatigue data for predicting fatigue life were demonstrated in this study.