Machining performance evaluation in turning of hardened steel using machine learning
Nitin Ambhore, Vishal Naranje, Sneha Shelke
Abstract
In hard turning, tool wear is an unavoidable phenomenon, and it is directly related to finished product quality. Optimum tool usage and good surface finish are highly expected in machining. Nowadays, machine learning (ML) techniques are being employed for evaluating machining performance. This study utilizes a machine learning technique for predicting tool wear and surface roughness. The advanced ML algorithms, such as random forest, XGBoost, and CatBoost, have been used. The comparative analysis is presented using various performance indices. For tool wear, RF, XGBoost, and CatBoost establish adequate predictive ability with an R2 value of 0.8575, 0.8604, and 0.9507, respectively, and R2 value for surface roughness is found as 0.8059, 0.8787, and 0.8891, respectively. It is seen that the CatBoost model has better predictive capabilities than the random forest and XGBoost. This outcome highlights the role of CatBoost machine learning algorithms for machining performance and efficiency evaluation.