Estimation of Tool Life in the Milling Process-Testing Regression Models

Area of Science:

• Manufacturing Engineering
• Data Science
• Machine Learning

Background:

• Accurate estimation of cutting tool lifespan is crucial for efficient milling operations.
• Sensor data from milling processes require sophisticated models to capture dependencies.
• Industry 4.0 initiatives emphasize intelligent manufacturing and process monitoring.

Purpose of the Study:

• To identify the most effective regression model for estimating cutting tool lifespan in milling.
• To evaluate the performance of Support Vector Regression (SVR), decision trees, and neural networks.
• To contribute to intelligent manufacturing by improving production process monitoring and reducing losses.

Main Methods:

• Utilized experimental data from a Haas VF-1 milling machine with vibration sensors and a Beckhoff PLC data collector.
• Applied regression models, including Support Vector Regression (SVR), decision trees, and neural networks, to analyze continuous sensor data.
• Evaluated model performance based on R² parameters and error values.

Main Results:

• Two-dimensional neural networks with the LBFGS solver achieved the highest prediction accuracy at 93.9%.
• Support Vector Regression (SVR) demonstrated comparable performance with 93.4% accuracy.
• Both neural networks and SVR provided the lowest error values among the tested models.

Conclusions:

• Neural networks and SVR are highly effective for predicting cutting tool lifespan in milling processes.
• The findings support the integration of advanced machine learning models in Industry 4.0 for enhanced manufacturing.
• Accurate tool lifespan prediction enables better service planning and minimizes material and tool damage.