Abstract
The aim of this paper is to provide a novel stochastic life prediction approach capable of predicting the total fatigue life of applied uniaxial stress states from a reduced dataset reliably and efficiently. A previously developed strain energy-based fatigue life prediction method is integrated with the stochastic state space approach for prediction of total cycles to failure. The approach under consideration for this study is the Monte Carlo method (MCM) where input is randomly generated to approximate the output of highly complex systems. The strain energy fatigue life prediction method is used to first approximate SN behavior from a set of two SN data points. This process is repeated with another unique set of SN data points to evaluate and approximate distribution of cycles to failure at a given stress amplitude. Uniform, normal, log-normal, and Weibull distributions are investigated. From the MCM, fatigue data are sampled from the approximated distribution and an SN curve is generated to predict high cycle fatigue (HCF) behavior from low cycle fatigue (LCF) data.
A machine-learning fatigue life prediction approach of additively manufactured metals
