Fault diagnosis of a CNC hobbing cutter through machine learning using three axis vibration data

Area of Science:

• Mechanical Engineering
• Data Science
• Artificial Intelligence

Background:

• Traditional CNC tool monitoring methods struggle with real-time, complex machining environments.
• Automated machine learning models are needed for enhanced tool reliability and machining precision.
• Fault diagnosis in CNC hobbing cutters is critical for operational efficiency.

Purpose of the Study:

• To develop a novel machine learning framework for accurate fault diagnosis of CNC hobbing cutters.
• To leverage three-axis vibration data for real-time tool condition monitoring.
• To improve the transparency and reliability of fault diagnosis systems using Explainable AI.

Main Methods:

• Collected a dataset of vibration signals from a CNC hobbing machine under healthy and faulty conditions.
• Extracted statistical features (RMS, Crest Factor, Kurtosis) from vibration data.
• Trained and evaluated various machine learning models (Decision Trees, Efficient Linear, Neural Networks, Ensemble methods), including Explainable AI techniques (LIME, Shapley values).

Main Results:

• The Ensemble model achieved perfect (100%) classification accuracy with low computational cost.
• Explainable AI techniques enhanced the interpretability and trustworthiness of the diagnostic system.
• The proposed framework outperformed existing methodologies in addressing overfitting, data quality, and model explainability.

Conclusions:

• The developed machine learning framework provides a foundation for effective predictive maintenance in CNC machining.
• The Ensemble model's high accuracy and efficiency make it suitable for real-time fault diagnosis.
• Future research should explore multi-sensor data and transfer learning for broader generalizability and deployment.