Research on nonlinear prediction model of weld forming quality during hot-wire laser welding

Abstract

In order to realize accurate prediction of the weld forming quality during hot-wire laser welding, in this study, a basic multivariate linear regression model regarding multiple process parameters and the weld forming quality is first established. And a residual analysis is then used to analyze the linear regression model. The results of the residual analysis illustrate that the influences of the hot-wire current, welding speed, and wire feeding rate on the weld forming quality are nonlinear, whereas the influences of the laser power and gap width on the weld forming quality are linear. In addition, the significance of the interactions among multiple process parameters on the weld forming quality is also obtained. Based on the results of the residual analysis, the linear regression model is modified in a targeted manner by adding the quadratic and interaction terms of the corresponding process parameters to establish a multivariate nonlinear regression model of the process parameters and the weld forming quality. Through comparing the goodness-of-fit, standard error, and standardized residual of the improved nonlinear regression model with the linear regression model and verifying the improved nonlinear regression model by verification experiment, it can be concluded that the improved nonlinear regression model has a better fitting precision and prediction precision. The results of the verification experiment indicate that the improved nonlinear regression model can correctly describe the relationship between the process parameters and the weld forming quality and can realize an accurate prediction of the weld forming quality.