Resistance Spot Welding Quality Evaluation System based on Improved Neural Network Model

Abstract

Resistance spot welding is an important branch of welding technology, the quality of its solder joints directly determines the safety and stability of equipment or system. However, resistance spot welding is a complex process of electrical, thermal, magnetic, force, phase change and other multi-factor coupling, and the quality of solder joints is easily affected by many factors, resulting in low detection accuracy, high equipment cost and poor versatility. Therefore, a quality voltage evaluation system of resistance spot welding based on improved BP neural network algorithm is proposed in this paper to realize real-time monitoring of solder joint quality. The BP neural network algorithm is improved to provide training data and algorithm validation data. The characteristic parameters of electrode voltage are selected as the input end of the neural network. The sampling of characteristic parameters is completed through the detection platform. Finally, the sampling points were used as the training set to train the improved BP neural network algorithm, and the neural network model was obtained and tested. The accuracy of the predicted solder joint quality was as high as 94%, which verified the accuracy of the resistance spot welding quality evaluation method.