Multi-objective optimization of laser brazing with the crimping joint using ANN and NSGA-II

Laser brazing process with crimping butt is sensitive to welding parameters, and it is difficult to acquire a good quality of welding joint. To achieve good welding parameters (welding speed, wire feed rate, gap), this paper addresses the multi-objective optimization of bead profile, namely sum of left side and right side of bead geometry and subtraction between top width of bead and bottom width of bead profile. Back propagation neural network was used to predict goals with average error of 9.95 and 8.54 %; non-dominated sorting genetic algorithm was adopted to acquire a Pareto set, and verification experiments demonstrated that relative errors were controlled within 3.97 %. Meanwhile, the importance from welding parameters on goals was ranked by signal-noise ratio and interactions between each parameter. Therefore, a novel multi-objective optimization method was proved to be feasible and would be useful to guide the actual welding process of laser brazing.