Study on the Law of Short Fatigue Crack Using Genetic Algorithm-BP Neural Networks

The initiation and propagation of short crack is a complex process of nonlinear dynamics. The material field has been concerning of the law of short crack and put forward a number of models and predictive equations. But most of them have to meet the problems which are unclear physical parameters meaning, narrow range of applications etc. That is the reason why there is not yet widely recognized quantitative model of crack evolution so far. The introduction of genetic algorithm-BP neural networks could solve those problems by avoiding building the explicit model equation and directly extracting the potential rules from the data. In this paper, the short crack for low cycle is studied under complex stress at high temperature. The material adopted in the experiment is Q245R steel. The initiation, propagation and coalescence of short crack are observed. The comparisons between experiment results and neural network simulation results show that genetic algorithm-BP neural networks simulation method can predict the law of short crack with higher prediction accuracy.