Prediction of the cutting tool wear during dry hard turning of AISI D2 steel by using models based on Learning process and GA polyfit

Abstract

In this paper, we are interested in the prediction of flank wear through dry hard turning of AISI D2 steel with a mixed alumina insert. In the machining process, the cutting tool is principally affected by two kinds of wear: flank and crater wear. The latter are criteria for cessation of the tool function. In the absence of a real-time wear sensor, it is necessary to know or track wear with the view to prevent tool damage. For this purpose, the current research focuses on the development of predictive models of flank wear based on Artificial Neural Network (ANN), Gaussian Process Regression (GPR), Support Vector Machine (SVM), and Polynomial Fit using Genetic Algorithm (GAPOLYFITN). The simulation process involves considering input variables including feed (f), cutting speed (Vc), and cutting time (tc); the output is the flank wear (VB). To assess the statistical efficacy of the predictive models, some performance indicators were employed, including the R-squared statistic-R2, Mean Square Error-MSE, Mean Absolute Error-MAE, and Mean Absolute Percentage Error-MAPE. The results, for the present case study, show that the R-squared statistic ranges from 0.85 to 0.99, the MSE is between 0.000046 and 0.000177, the MAE ranges from 0.002958 to 0.009336, and the value of MAPE varies from 3.50 to 9.60%. The predictive capability of GPR and GAPOLYFITN in determining flank wear are the best, as they exhibit high (R2), and lower values of MSE, MAE, and MAPE. The powerful predictive model of flank wear is the GPR because it provides R2 = 0.96, MSE = 4.6e-5, MAE = 0.002958, and MAPE = 3.50%.