Transmission Line Vibration Damper Detection Using Multi-Granularity Conditional Generative Adversarial Nets Based on...

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Galloping phenomenon in transmission lines poses risks, necessitating effective vibration damper detection.
• Current Convolutional Neural Network (CNN)-based visual detection methods lack real-time capabilities.
• Manual detection of vibration dampers is time-consuming and inefficient.

Purpose of the Study:

• To develop an advanced image generation model for accurate and real-time vibration damper detection.
• To overcome the limitations of existing CNN-based methods in speed and precision.
• To introduce a novel approach for enhancing visual inspection of critical infrastructure components.

Main Methods:

• Proposed DamperGAN, a multi-granularity Conditional Generative Adversarial Network for vibration damper detection.
• Employed a coarse-grained module for initial low-resolution image generation.
• Utilized Monte Carlo search and an attention mechanism to refine details in a fine-grained module.
• Introduced a multi-level discriminator with a multi-task learning mechanism.

Main Results:

• DamperGAN successfully generates high-resolution images for vibration damper detection.
• The model demonstrates superior performance in image resolution and quality compared to mainstream baselines.
• Experiments on the self-built DamperGenSet dataset validate the model's effectiveness.

Conclusions:

• DamperGAN offers a significant advancement in automated visual detection of vibration dampers.
• The proposed model addresses the need for real-time and high-quality detection in transmission line monitoring.
• This work provides a robust solution for improving the efficiency and accuracy of infrastructure inspection.