Tool-Wear-Estimation System in Milling Using Multi-View CNN Based on Reflected Infrared Images

Area of Science:

• Manufacturing Engineering
• Machine Tool Technology
• Artificial Intelligence in Manufacturing

Background:

• Accurate tool wear estimation is critical for optimizing machining processes and preventing equipment damage.
• Traditional methods for tool wear monitoring can be time-consuming, intrusive, or lack precision.
• Developing non-contact, automated methods for real-time tool wear assessment is an ongoing challenge.

Purpose of the Study:

• To propose and demonstrate a novel method for estimating tool wear in milling operations.
• To leverage infrared (IR) laser vision and deep learning for precise tool wear measurement.
• To improve machining efficiency through reduced downtime and optimized tool utilization.

Main Methods:

• Utilized an infrared (IR) line laser system irradiating the tool at multiple angles (-7.5°, 0.0°, +7.5°).
• Employed three cameras at 45° intervals to capture reflected IR light from the rotating tool.
• Applied an exposure fusion method to obtain high dynamic range (HDR) images and a multi-view convolutional neural network for wear estimation.

Main Results:

• Achieved a mean absolute error (MAE) in tool wear prediction ranging from 9.5 to 35.21 μm.
• Demonstrated the effectiveness of the IR laser vision and deep learning approach for flank wear estimation.
• Validated the method on SDK-11 workpieces using end milling with a TH308 insert.

Conclusions:

• The proposed IR laser vision and deep learning method provides a viable solution for accurate tool wear estimation in milling.
• This non-contact measurement technique can significantly reduce downtime associated with manual tool inspection.
• The method contributes to increased tool life utilization and overall machining efficiency.