A Machine Vision-Based Pipe Leakage Detection System for Automated Power Plant Maintenance

Area of Science:

• Industrial Automation
• Machine Vision
• Non-Destructive Testing

Background:

• Detecting colorless liquid effusions in industrial pipework is challenging due to poor visibility and illumination.
• Existing methods struggle with identifying subtle leaks like water in complex environments.

Purpose of the Study:

• To develop an automated inspection system for power plant pipework maintenance.
• To detect colorless fluid effusions by leveraging their reflective and thermal properties.

Main Methods:

• Utilized dual-source imaging (visible and infrared light) to capture effusion characteristics.
• Developed a contour features algorithm integrating infrared and visible light image processing.
• Implemented a decision-making tool based on image contour closure for classification.

Main Results:

• The system demonstrated high classification accuracy, reaching up to 99% in real industrial settings.
• Successfully detected colorless fluid effusions by analyzing reflective and thermal signatures.
• Validated the system's suitability for industrial pipework maintenance requirements.

Conclusions:

• The proposed machine vision system offers an effective solution for automated effusion monitoring in industrial environments.
• Leveraging dual-source imaging and contour analysis significantly enhances leak detection capabilities.
• The system provides a reliable and accurate method for power plant pipework inspection and maintenance.