Optimization Method of Arc Welding Parameters Based on BP Neural Network Parameter Prediction Model

Optimizing arc welding parameters is of great significance for improving welding quality and controlling welding costs. BP neural network is a relatively mature and widely used network, a parameter prediction model based on BP neural network is proposed and established, the training process adopts the gradient descent BP algorithm, set model learning rate to 0.05, target accuracy to 1 × 10–4, and training step to 30000. Considering the strong correlation between welding parameters, and the line energy is introduced to judge the prediction accuracy of the model, and the accuracy of welding parameters predicted by the overall model is above 90%. The parameters predicted by BP neural network are used to weld the steel sheet, the final results show that the welding parameter design system based on BP neural network can effectively predict the welding parameters, which can realize good welding effect by inspecting with the digital convergence welding seam size calculation model, providing a reliable Digital Fusion Analysis Methodology for improving welding quality and reducing welding cost.