Research on the prediction method of material mechanical parameters based on PSO optimized BP neural network

Engineering needs to periodically determine the mechanical properties of the dangerous parts of the in-service equipment, and the two-hole micro-shear method has a very broad application prospect because of its advantages such as portable equipment and on-site detection. However, there are problems such as low precision [1], complicated calculation methods, and so on, so further research and improvement is needed. In this paper, neural networks and optimization algorithms are innovatively introduced into the prediction of mechanical performance parameters of the double-hole method, and 900 labeled data sets are generated with the help of the finite element model to establish a parameter prediction model based on PSO-BP. Through the modification of network parameters, training and testing, the model output and the target value have a small error, which achieves the research goal of accurate prediction of its mechanical performance parameters. And relatively speaking, the prediction accuracy using PSO-BP prediction model is higher. Compared with the traditional two-hole method, the accuracy of the determination is improved, which lays a solid foundation for the promotion of the two-hole method and reaching the application of productization.