Prediction of Cutting Force and Surface Roughness in Turning Using Machine Learning

Accurately predicting cutting force in turning is crucial for optimizing machining processes, improving tool life, and enhancing safety. Additionally, predicting surface roughness is essential for enhancing the surface finish of machined parts and reducing post-processing costs by selecting the appropriate cutting conditions. In this paper, to predict cutting force and surface roughness in turning, four machine learning methods are developed and compared: Multiple Linear Regression (MLR), Random Forest Regression (RFR), Gradient Boosting Regression (GBR), and Bagging. At first, the experimental machining data taken from the literature are divided into two sets. The first one, known as training data (75% of total experimental data), is used to learn the relationships between the cutting conditions (inputs) and the cutting force and the surface roughness (outputs). Then, the second set, called testing data (25% of total experimental data), is then employed for validation of regression models. Finally, several metrics are used to evaluate the performance of each method. The best coefficient of determination R2 for testing data is 99.49% for the cutting force model (GBR method) and 95.8% for the surface roughness model (RFR method). These results demonstrated the reliability of the machine learning models developed to predict the cutting force and the surface roughness in turning.