The surface roughness is a main performance parameter to characterize the quality of the macroscopic morphology in laser cladding multi-lap forming of key parts. In this paper, both the Taguchi and the Box-Behnken Design (BBD) experiments were used to conduct the laser cladding multi-channel lap experiment. The relationship between the process parameters regarding laser power, powder feeding speed, scanning speed, lap rate and the surface roughness of the coating was illustrated via experiments. The neural network model (Genetic Algorithm-Back Propagation, GA-BP) and response surface methodology (RSM) optimized by genetic algorithm were established, respectively. Furthermore, the optimal cladding parameters were optimized, and the coupling effect between the cladding process and the forming quality of the multi-channel lap was revealed. It is found that the root mean square error (RMSE) and absolute average deviation (AAD) of GA-BP model were both smaller than those of the RSM model. The coefficient of determination R-Squared of GA-BP model was closer to 1. Moreover, the minimum roughness of 20.89 μm by optimizing the process parameters through the genetic neural network model was lower than the surface response optimization of 35.67μm. The genetic neural network model optimized by genetic algorithm can improve the prediction accuracy on the surface macro-quality of multi-channel lap cladding layer, which provides a theoretical basis and technological direction for the improvement of the surface quality of key parts.

No abstract

No abstract