Prediction and optimization of phosphorus content in electroless plating of Cr12MoV die steel based on PSO-BP model

Abstract

In this paper, the effect factors of phosphorus content of Ni-P coating on the surface of Cr12MoV die steel were discussed by surface fitting technique. We took the temperature, pH, sodium hypophosphite concentration, and lactic acid concentration as the input parameters of the neural network, and phosphorus content as output parameters. The established BPNN (Back Propagation Neural Network) was optimized by PSO (Particle Swarm Optimization) algorithm and the prediction accuracy of this model was verified. The results show that the increase in sodium hypophosphite contributes to raise the phosphorus content of coating, while the effect of pH is opposite. As input dimension decreases, the accuracy of BP network prediction increases as well. In comparison with other optimization models, the PSO-BP (optimizing the BPNN model by using PSO algorithm) model shows higher prediction accuracy even in the cases of small samples, in which MSE (Mean Square Error) of prediction is 0.0792. According to these reliable results, the optimal process parameter range of different phosphorus content and optimal initial parameters of BPNN have been given.