Multi-objective Optimization Of MAG Based On Neural Network And Genetic Algorithm

At present,with the rapid development of science and technology,different industrial applications require different welded joints.Therefore,it is the future trend to improve the performance of welded joints more pertinently.This means that it is urgent to carry out the research of multi-objective optimization in the selection of welding process.But the internal relationships between welding process parameters and welding performance presents nonlinear and complex,as well as different welding performance is difficult to achieve the optimal at the same time.Therefore,it is of great research value to explore theories and methods related to the multi-objective optimization of process parameters based on experiments.Three-factor and three-level bead-on-plate and butt welding full-factor test,as well as five-factor and three-level T-joint welding orthogonal test were designed for the butt joint and T-joint of box girder structure in the port machinery super-large components.The effect of welding speed,welding voltage and wire feed rate on the tensile strength,impact energy and welding reinforcement was analyzed for the MAG welding butt joint of box girder structure.Error back propagation neural network neural network(BPNN),radial basis function neural network(RBFNN)neural network and kriging interpolation models were established to predict the tensile strength,impact energy and welding reinforcement.The effect of wire feed angle,the gap between skin and stringer sheets,welding speed,welding voltage and wire feed rate on the depth of penetration and weld leg,fillet weld throat and welding angle deformation were analyzed during MAG process.BPNN,RBFNN,kriging interpolation model and second order polynomial regression model were established to predict the depth of the penetrations in stringer and skin sheets,as well as welding angle deformation.After being compared,BPNN has the best prediction accuracy and stability.Then,multi-objective optimization with the purpose of maximizing the tensile strength,impact energy and minimizing welding reinforcement for butt joint,as well as the multi-objective optimization with the purpose of maximizing the depth of the penetrations in stringer and skin sheets and minimizing welding angle deformation for T-joint were carried out based on the combination of the most suitable prediction models and NSGA-II.Finally,verification tests were carried out on the results of the multi-objective optimizations.The test results showed that the optimized process parameter can meet the goal and the multi-objective optimization strategy proposed in this paper is effective.