CNN-LSTM network-based damage detection approach for copper pipeline using laser ultrasonic scanning

Area of Science:

• Materials Science
• Mechanical Engineering
• Signal Processing

Background:

• Copper pipelines are crucial for industrial transport.
• Early damage detection is vital for pipeline integrity.
• Laser scanning offers non-contact, visualized guided wave detection but struggles with large-area accuracy.

Purpose of the Study:

• To develop an accurate, independent damage detection method for copper pipelines using laser scanning.
• To overcome the limitations of traditional laser scanning methods that rely on surrounding signal comparisons.
• To enhance the reliability of non-destructive testing for industrial pipelines.

Main Methods:

• A novel data conversion algorithm for laser scanning signal preprocessing.
• Implementation of a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) network for damage detection.
• CNN module extracts time-domain signal features; LSTM performs feature extraction and classification.

Main Results:

• The CNN-LSTM method achieved high detection accuracies: 99.9% for 0.5 mm deep cracks and penetrating cracks, 99.8% for corrosion and internal cracks.
• The method accurately identifies damage location and size.
• Experimental results demonstrate superior performance compared to state-of-the-art methods.

Conclusions:

• The proposed CNN-LSTM method provides a robust and accurate solution for non-destructive testing of copper pipelines.
• This technique enhances the reliability of laser ultrasonic guided wave scanning for industrial applications.
• The method's ability to detect various damage types independently marks a significant advancement in pipeline integrity assessment.