Cutting Force and Tool Wear Prediction
Data-driven and machine-learning models for predicting cutting forces and tool wear in CNC milling.
Overview
This ongoing smart-manufacturing research explores data-driven methods for predicting cutting forces and tool wear in CNC milling. Reliable predictions can support machining-process monitoring, parameter optimization, and predictive maintenance.
Approach
- Developed initial models from 12 large-scale machining datasets, combining nonlinear system analysis with sparse identification.
- Expanded the study to 27 datasets produced through a Taguchi-designed experiment.
- Implemented and compared Random Forest, XGBoost, and Support Vector Regression models.
- Evaluated preprocessing, feature selection, and model behavior against experimental measurements.
Current status
The initial study achieved an R² of approximately 0.5–0.6. The expanded dataset and model comparison are in progress, with the goal of improving generalization and prediction accuracy.