A Hybrid Algorithm Based on GRNN and Grasshopper Optimization Algorithm for Welding Nugget Diameter Prediction

Abstract

Resistance spot welding (RSW) is applied extensively by automotive manufacturers for assembling the structural and body components of vehicles. The current method of welding quality inspection is off-line inspection after welding, which cannot provide real-time feedback on welding quality and cannot meet the rhythm of modern production. Therefore, the online non-destructive testing technology of welding quality is worth studying. In this study, an RSW quality prediction model is developed using the improved grasshopper optimization algorithm combined with the generalized regression neural network (GRNN) algorithm, in which the actual process parameters including welding current, welding voltage, energy, power, and pulse width are used as inputs to predict the nugget diameter. During the network training process, the optimization algorithm is used for finding the optimum smoothing factor σ of GRNN, chaotic mapping, and non-uniform mutation are added to the traditional grasshopper optimization algorithm to enhance the optimization ability of the algorithm. Through bootstrap sampling, a comparison experiment about the prediction effect of the proposed quality prediction model with earlier methods is carried out, and the analysis of the experimental results leads to a conclusion that the accuracy of the proposed welding quality prediction model is higher.