Automated Identification of Overheated Belt Conveyor Idlers in Thermal Images with Complex Backgrounds Using Binary...

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Mechanical industrial infrastructures, like conveyor systems in mines, require regular monitoring.
• Manual inspection of extensive conveyor systems is challenging due to length and harsh environmental conditions.
• Current condition monitoring methods for conveyor idlers often rely on vibration and acoustic signals.

Purpose of the Study:

• To propose an automated robotics-based inspection system for monitoring belt conveyor idlers using infrared images.
• To address the limitations of classical image segmentation techniques in detecting overheated idlers in complex backgrounds.
• To improve the accuracy and reliability of overheated idler detection in mining environments.

Main Methods:

• Development of a robotics-based system for infrared image acquisition of conveyor idlers.
• Implementation of image preprocessing stages to enhance infrared image quality.
• Application of anomaly detection and outlier techniques for hotspot segmentation.
• Utilizing a Convolutional Neural Network (CNN) for binary classification of segmented thermal images to detect overheated idlers.

Main Results:

• The proposed method achieved a precision of 0.9740 and an F1 score of 0.9782.
• This represents a significant improvement over previous research with precision of 0.4590 and F1 score of 0.6292.
• The CNN-based classification accurately identifies overheated idlers even in complex backgrounds with thermal interference.

Conclusions:

• The automated robotics-based inspection system using infrared imaging and CNN classification is highly effective for monitoring conveyor belt idlers.
• The proposed method offers a substantial advancement in the condition monitoring of mining infrastructure.
• This technology enhances safety and efficiency in mining operations by enabling reliable detection of potential equipment failures.