Fatigue Life Prediction of High-strength Bolts Based on Machine Learning

Due to the influence of many design variables and strong contact nonlinearity, the fatigue life design of high-strength bolts is still an urgent problem to be solved. In order to accurately predict the fatigue life of bolts, this paper combines classic parameter analysis methods with machine learning techniques. First, the numerical analysis results are used to reduce the dimensions of the influence parameters of the bolt fatigue life parameters, and then the machine learning model of polynomial regression(PR) and multi-layer perceptron (MLP) regression is used to establish the bolt stress amplitude and the mapping relationship between the influencing factors, and finally the machine learning model is combined with the graphical programming language to develop a set of windowed analysis tools that can accurately analyze the stress amplitude of the high-strength bolt connection system and predict its fatigue life. The experimental results show that the prediction value error of the PR model is less than 2%, and the error of MLP regression model is also less than 4%.