Surface Roughness Characterization and Inversion of Ultrasonic Grinding Parameters Based on Support Vector Machine

With surface roughness restricted by grinding parameters, the characterization of roughness parameters and the inversion of grinding parameters are of great significance for improving surface performance and realizing active surface machining. This research proposes a combination of statistical theory and data-driven analysis to solve the above problems. Pearson correlation analysis and multivariate variance analysis indicate the correlation characterization parameter set (CPS) consists of Sa, Vmp, Vvv, and Sz and that there are differences in the influence of grinding parameters on the parameters in CPS. Adjustment of support vector machine (SVM) core parameters makes it possible to construct expansion parameter set (EPS) optimal inversion models. By designing pseudo-surface random roughness parameters and grinding experiments, the reliability of inversion models is verified. The results show: (1) The better generalization of inversion model indicates skewness Ssk and kurtosis Sku in EPS have important implications for the optimal inversion model and surface characterization and (2) The data-driven model based on support vector machine provides machining guidance for obtaining the expected ultrasonic grinding surface.