Machine Learning Approaches for Monitoring of Tool Wear during Grey Cast-Iron Turning

Area of Science:

• Manufacturing Engineering
• Materials Science
• Artificial Intelligence

Background:

• Advanced manufacturing requires precise quality control, with tool wear being a critical factor.
• Machine learning offers promising solutions for real-time monitoring and prediction in industrial processes.

Purpose of the Study:

• To compare various machine learning methods for identifying tool wear in the turning of gray cast iron.
• To determine the most effective intelligent system for predicting tool condition based on vibration signals.

Main Methods:

• Experimental investigation involving turning gray cast iron (EN-GJL-250) with carbide inserts at different cutting speeds.
• Acquisition and analysis of vibration acceleration signals to derive diagnostic measures.
• Application of machine learning models: classification and regression trees, induced fuzzy rules, and artificial neural networks.

Main Results:

• Diagnostic measures derived from vibration signals were correlated with tool condition.
• Machine learning models successfully classified tools as usable or unsuitable based on wear criteria (VBc = 0.3 mm).
• Feature assessment identified significant inputs for accurate tool wear prediction.

Conclusions:

• Machine learning, particularly artificial neural networks, demonstrates high effectiveness in predicting tool wear.
• Vibration analysis combined with intelligent systems provides a robust method for monitoring tool condition in manufacturing.
• The study highlights the potential of AI to optimize machining processes and improve product quality.