Surface quality optimization of laser cladding based on surface response and genetic neural network model

Abstract

Abstract The model was established to optimize the laser cladding process parameters, the coating surface topography can be predicted and controlled. Taguchi and Box-Behnken (BBD) experiments were used to carry out the experimental design of laser cladding multi-channel lap. 316 L stainless steel coating was cladded on the surface of 45 steel substrate. The genetic algorithm-back propagation (GA-BP) neural network and response surface methodology (RSM) models were established respectively. The prediction accuracy of the two models was compared. The coupling effect between cladding process and multi-channel lap forming quality was analyzed. The relationship between cladding process parameters, such as laser power, feeding speed, scanning speed and overlap ratio, and surface roughness of coating was studied. The experimental results show that: The root mean square error (RMSE) and absolute mean deviation (AAD) of the GA-BP model are smaller than those of the RSM model. The coefficient of determination R2 of the GA-BP model is closer to 1 than that of the RSM model. The minimum roughness predicted by GA-BP model is 20.89 μm, which is lower than that of RSM model (35.67 μm). The final findings: in the optimization of process parameters of laser cladding, overlap ratio and scanning speed has significant effects on coating surface roughness. The GA-BP model of the coating surface roughness prediction accuracy is better than the RSM model. The prediction and control of the coating surface roughness are realized by GA-BP model, for the precise forming of the laser cladding coating surface, which provides theoretical basis and technological direction.