Damage quantification of mining explosion-proof equipment using thermal flame dynamics and improved LSTM

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Quantitative characterization of flame propagation in explosion-proof equipment is lacking.
• Internal non-ignition detonation tests require robust damage assessment methods.

Purpose of the Study:

• To propose a novel damage severity quantification method for explosion-proof equipment using thermal imaging and an improved LSTM model.
• To enhance the accuracy and reliability of damage diagnosis in mining safety applications.

Main Methods:

• Constructed a detonation flame thermal imaging dataset.
• Utilized Faster R-CNN for flame contour segmentation and feature extraction.
• Employed an improved LSTM model with bidirectional fusion and multi-layer stacking for spatiotemporal feature analysis.

Main Results:

• Achieved prediction errors within 2% for damage levels ranging from 5-20 cm.
• Demonstrated the effectiveness of bidirectional fusion and deep feature modeling in capturing flame dynamics.
• Validated the method's accuracy through experimental results with near-zero error distribution shift.

Conclusions:

• The proposed data-driven approach accurately quantifies flame propagation damage severity.
• This method supports safety certification and maintenance of mining explosion-proof equipment.
• Findings have significant implications for improving coal mine safety.