Edge Preparation Impact on Tool Wear and Surface Roughness: Experimental Analysis and Machine Learning Predictions

This study investigates the impact of different edge polishing conditions on milling tool performance and surface roughness, focusing on the roughness parameters Ra, Rt, and Rz. Using AISI P20 material and solid carbide straight end mills, the tools were subjected to various edge treatments: standard, HSC 1/300, and H 4/400. Tool wear was monitored through VBmax measurements. Multiple regression models were developed, including Linear Regression, Random Forest, XGBoost, Gradient Boosting, Ada Boost, SVR, NuSVR, Extra Trees, and ANN, utilizing Python's scikit-learn, Keras, and TensorFlow libraries. The models were optimized using Cross Validation and GridSearchCV. Data revealed that polished tools exhibited reduced wear and more consistent surface roughness compared to standard tools. NuSVR emerged as the optimal model for predicting tool wear, while Random Forest and ANN excelled in predicting roughness parameters. Polished tools (H 4/400 and HSC 1/300) demonstrated a 20.49% and 18.62% increase in tool life, respectively. The findings support the utility of machine learning models in optimizing milling processes, enhancing tool longevity, and improving surface quality, benefiting industrial applications through increased efficiency and cost savings.