Predicting tool life and sound pressure levels in dry turning using machine learning models

Dry turning reduces the environmental impact and costs associated with cutting fluids, but it challenges the optimization of tool life due to the generated heat. This study evaluated machine learning models to predict tool life (T) during the dry turning of AISI H13 steel by analyzing cutting speed (Vc), feed rate (f), and depth of cut (ap). Nineteen experiments were conducted using a central composite design (CCD), evaluating machining cost, tool life, sound pressure level (SPL), chip removal rate, and machining force. Statistical analysis included calculations of central tendency, dispersion, and Pearson correlation. Five machine learning models were compared: linear regression, decision tree regression, random forest regression, multi-layer perceptron (MLP) regression, and stochastic gradient descent (SGD) regression. The evaluation metrics were MSE, RMSE, and \(R^2\) $R^2$ . The random forest and stochastic gradient descent models proved to be promising. The random forest model stood out with high \(R^2\) $R^2$ values, indicating greater precision. Linear regression and stochastic gradient descent also performed well in predicting sound pressure levels. The consistency of the random forest model in both training/testing and prediction phases highlights its robustness and adaptability to different tools and machining setups. In contrast, the multi-layer perceptron model exhibited greater variability, reflecting its sensitivity to hyperparameter settings. The results demonstrate the effectiveness of the random forest model in predicting tool life, contributing to reduced operational costs and increased manufacturing efficiency.