Efficient Machine Learning-Based Forecasting of Key Control Parameters for Analyzing Honing Stone Cutting Performance

Abstract This study extends previous research on the performance of laser-textured cemented carbide tools and honing stones. We employ a CBN-based honing stone B151 as a benchmark with a specific focus on using machine learning models to predict its machining outcomes, including surface roughness (Ra, Rz), mass change, and diameter change of workpieces. The dataset, derived from a previous series of cutting tests that replicate industrial conditions, ensures the relevance and applicability of our findings to real-world manufacturing environments. Machine learning algorithms are implemented, including linear regression and random forests regression, to predict surface roughness (Ra and Rz), changes in the outer diameter, mass change, and the tangential force coefficient based on variations in applied normal force (Fn). The effectiveness of these models is demonstrated by satisfactory prediction metrics—MAE < 0.02, MSE < 0.04, and R2 > 0.8 for the test dataset across almost all analyzed parameters—and the comparison of these predictions with experimental results shows a deviation of around approximately or less than 10%.