Prediction of cutting forces and surface finish in machining of EN08 by machine learning algorithms

This research paper attempts to use various machine learning algorithms in predicting the response variables in the machining of engineering material EN08. Machining experiments have been performed by using L9 orthogonal array of Taguchi method of design of experiments. Turning experiments have been performed with cutting speed, feed rate and depth of cut as process parameters and cutting forces, surface finish as the response variables on CNC horizontal turning center. Cutting forces in three directions F_x, F_y and F_z have been measured accurately with Kistler lathe Tool Dynamometer and data acquisition system. Machine learning algorithms, like linear regression, decision tree, gradient booster, and random forest regression have been applied to the experimental dataset for prediction of response variables. From the analysis, it has been found that the random forest regression has superior performance with the dataset. The Random Forest Regressor achieved the best performance with an R² error of approximately 0.89, demonstrating strong agreement between predicted and experimental values. Although the dataset is limited, cross-validation and repeated experiments were used to improve reliability. There exists a close agreement between actual and predicted values of the response variables cutting forces and surface finish with random forest regressor as compared with other machine learning models used for analysis purposes.