High-precision weld width detection in laser transmission welding via crow and wolf optimized neural networks

Abstract

Laser transmission welding is widely used in the welding scene of highly transparent plastic parts. It could heat the surface of the lower layer through the upper welded parts, perform effective welding, and have a small heat-affected zone. However, due to the complexity and uncertainty of the welding process parameters, many defects will be generated in the welding, which will greatly reduce the quality of polymer welding. Based on the combination of artificial neural network (ANN) and optimization algorithm, this paper proposes a method to obtain high-precision weld width data through process parameters, so that the optimization of parameters can have a reliable prediction model and reduce trial and error experiments to obtain the most accurate parameter design.

First, a multi-layer perceptron model is designed to learn the nonlinear relationship between process parameters and weld width. The model can quickly predict the weld width in the welding result from the matching of parameters. Secondly, in order to improve the prediction accuracy of the model for weld width, a crow and wolf optimization algorithm based on gray wolf optimization is proposed. The algorithm is used to optimize hyperparameters in model, improve the performance of the model, and thus enable high-precision prediction. Finally, the model optimized by the new algorithm is compared with the models using other swarm intelligence algorithms. The experimental results show that the method proposed in this paper could predict the weld width with high accuracy, and the R² value is greater than 0.9.