인공신경망을 이용한 저주기 피로수명 예측

Abstract

The traditional approach of fatigue life assessment uses Palmgren-Miner Rule as its base. This paper proposes a new method by observing change in material behavior to predict fatigue life. For experiment, austenitic stainless-steel sample was subjected to low cycle fatigue of 0.4% and 0.5% strange range. Towards fatigue life, the material displayed a tendency to soften regardless of strain range. This tendency was characterized as I (Isotropic Softening Factor) and put in to an artificial neural network designed to predict remaining fatigue life. Compared to conventional regression methods, the method proposed in this paper proved to be more accurate by up to 0.171 in coefficient of determination. Also, the returned model was tested in goodness-of-fit through adjusted R^2 and Shpiro-Wilk test. The results showed that the modeling method proposed in this paper could be utilized to predict low cycle fatigue life with high accuracy.