Optimization of vacuum hybrid welding process parameters for YG8 cemented carbide and 42CrMo steel using artificial neural networks

Abstract

The mechanical properties of the joint of YG8 cemented carbide and 42CrMo steel were studied by using CuZnNi filler metal in the vacuum hybrid welding. The parameters for optimizing the shear strength of joints were selected by orthogonal experiment, mainly including the brazing temperature, brazing time, diffusion temperature, holding time and diffusion pressure. The artificial neural network technique is a very practical tool for predicting the controllable parameters in the non-linear model. BP (Back Propagation) neural network was established under the environment of MATLAB software to simulate and predict designed process parameters. Thus, the optimal parameters were predicted by BP neural network and validated by experiments. The results show that the proposed BP model can obtain a non-linear relationship between the mechanical properties and process parameters. The predicted values are in good agreement with the experiments with the relative error of-0.2458% and the mean square error of 0.052%. The optimal parameters of BP neural network were obtained at brazing temperature (A) of 1045°C, brazing time (B) of 15 min, diffusion temperature (C) of 750°C, holding time (D) of 40 min and diffusion pressure (E) of 8 MPa, and the prepared joints showed the better mechanical properties like glossy surface, no apparent deformation, uniform brazing region and good adhesive interface, etc.