Advanced Machine Learning for Predicting Tool Wear and Tool Life: A Comprehensive Analysis Using Classification and Regression Techniques
Authors :- Jha, A.K., Jha, A.K., Airao, J., Khanna, N.
Publication :- Recent Trends in Material Processing, Characterization and Applications. AEMTA 2024, Springer, 2025
This work delves into the significant impacts of advanced machine learning (ML) techniques in predicting tool wear and tool life, employing both classification and regression methodologies. We have eight key feature parameters—machining time, speed, power, feed rate, depth of cut, cooling medium, tool life, and tool wear. When forecasting tool life, the model relies on seven other features as its input, whereas for projecting tool longevity, those same seven features serve as input variables. Qualitative cooling and wear data are numerically encoded using label encoder, and dataset standardization is achieved through normalization. For thorough model assessment, the dataset undergoes partitioning into training (80%) and test (20%) subsets. Various models, such as Support Vector Machine (SVM), random forest, and logistic regression, are employed for wear classification. The random forest model achieves the highest accuracy among them. In regression analysis for predicting tool life, various techniques such as multiple linear regression and polynomial regression (degrees 2–11) are employed. Notably, polynomial regression with degree 4 exhibits the lowest mean squared error (MSE), signifying its efficacy in predicting tool life. The study further conducts a comparative analysis of polynomial regression across different degrees, revealing a nuanced relationship between degree and MSE. The findings demonstrate a decreasing trend in MSE up to a certain degree, followed by an increase, showcasing the optimal degree for prediction. The study concludes by emphasizing the potential for further innovation and productivity enhancements in the manufacturing industry through the integration of advanced ML techniques in tool wear and tool life prediction.