DeepTool: A deep learning framework for tool wear onset detection and remaining useful life prediction

Area of Science:

• Manufacturing Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Milling tool lifespan directly impacts operational efficiency, reliability, and cost-effectiveness.
• Accurate prediction of tool wear onset and remaining useful life is crucial for optimizing milling processes.

Purpose of the Study:

• To develop a deep learning-based system, DeepTool, for predicting milling tool service life and detecting wear onset.
• To leverage a comprehensive feature extraction process and a self-collected dataset for enhanced predictive accuracy.

Main Methods:

• Utilized a self-collected dataset comprising precise sensor signals from milling tests under varied cutting conditions.
• Employed hybrid autoencoder-LSTM and encoder-decoder LSTM models for advanced predictive modeling.
• Implemented an efficient feature extraction technique, analyzing both time-domain and frequency-domain aspects of sensor data.

Main Results:

• Achieved over 95% R2 accuracy score in estimating tool wear onset and predicting remaining useful life.
• Demonstrated the effectiveness of the deep learning models in extracting relevant information from sensor signals.
• Validated the comprehensive feature extraction process for identifying tool wear indicators.

Conclusions:

• DeepTool provides a robust solution for real-time monitoring and prediction of milling tool conditions.
• The developed system enhances optimization, reliability, and cost reduction in milling operations.
• The study highlights the potential of deep learning and sensor data fusion for predictive maintenance in manufacturing.