Optimal Surface Grinding Regression Model Determination with the SRP Method

Abstract

The construction of the regression models used to control machining processes is the objective of many experimental studies. Therefore, the effectiveness of the machining process control largely depends on the regression model’s accuracy. This study was conducted to determine the optimal regression model of surface grinding. Accordingly, eight different surface grinding regression models were constructed, including one model without data transformation and seven models that utilized various data transformations. The seven data transformations employed entailed square root transformation, logarithmic transformation, inverse transformation, exponential transformation, asinh transformation, Box-Cox transformation, and Johnson transformation. The process of determining the optimal model was carried out considering five parameters: R2, R2(adj), R2(pred) (predicted R2), MAE (Mean Absolute Error), and MSE (Mean Squared Error). SRP (Simple Ranking Process) was the optimization method followed to identify the best regression model. The Box-Cox transformation was recognized as the most accurate surface grinding regression model.