Real-time detection of hotspots in dry-type flexible cable terminals

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Dry-type flexible cable terminals are prone to overheating defects, leading to potential failures.
• Real-time detection of these hotspots is crucial for preventing failures and ensuring operational safety.
• Existing methods face challenges with small target detection, complex backgrounds, and real-time processing requirements.

Purpose of the Study:

• To develop an accurate and efficient model for real-time hotspot detection in dry-type flexible cable terminals.
• To address the limitations of current detection methods, particularly for small targets and complex environments.
• To reduce failure rates by enabling timely identification of overheating defects.

Main Methods:

• Defined four types of hotspots specific to dry-type flexible terminals.
• Proposed an improved You Only Look Once (YOLO) model incorporating Diverse Branch Blocks and Multi-Scale Convolutional Attention.
• Integrated FasterNet Block to optimize model complexity and enhance real-time performance.
• Utilized aerial infrared imaging for hotspot detection.

Main Results:

• The proposed model demonstrated significantly superior performance compared to baseline and other mainstream YOLO models.
• The enhanced model effectively addresses challenges related to small target size and complex backgrounds.
• The system meets the real-time detection demands for aerial infrared inspection of cable terminals.

Conclusions:

• The improved YOLO-based model offers a robust solution for real-time hotspot detection in dry-type flexible cable terminals.
• This technology enhances the reliability and safety of electrical infrastructure through effective defect diagnosis.
• The developed model provides a practical tool for reducing failure rates in critical cable terminal applications.