Prediction of fracture toughness using small specimens based on machine learning

Abstract

In this study, a neural network was used to predict the fracture toughness of standard specimens of 16MnDR steel in the tough-brittle transition zone. The main factors such as specimen plate thickness and geometry type are taken as inputs to the network and the fracture toughness of the standard specimen is taken as output. A BP neural network is utilized to describe the nonlinear relationship between them. The network was trained and tested by using finite element simulation simulation data based on experimental data. The established neural network model was found to have good accuracy in predicting the experimental data and can be used as a method for predicting the fracture toughness of 16MnDR standard specimens.