Damage characterization using CNN and SAE of broadband Lamb waves

Area of Science:

• Materials Science
• Mechanical Engineering
• Signal Processing

Background:

• Lamb wave-based methods are crucial for structural health monitoring (SHM) and nondestructive testing (NDT).
• Deep learning algorithms like CNN and SAE show promise for analyzing Lamb wave signals for damage detection.
• Current methods often use narrowband Lamb waves, which have limited damage information, restricting model performance.

Purpose of the Study:

• To develop a broadband Lamb wave deep learning algorithm for improved damage localization and quantification.
• To overcome the limitations of narrowband signals by leveraging richer damage information contained in broadband Lamb waves.
• To enhance the accuracy and effectiveness of SHM and NDT through advanced signal processing and deep learning.

Main Methods:

• A deep learning algorithm utilizing broadband Lamb wave signals was proposed.
• Various mode selections, signal processing techniques, and deep learning algorithms were employed for feature extraction.
• Fusion of results from different extraction methods was used to fully utilize broadband information.

Main Results:

• The broadband Lamb wave approach demonstrated superior performance in damage localization and quantification compared to narrowband methods.
• The fusion strategy effectively utilized the rich information present in broadband signals.
• Experimental validation confirmed the high accuracy and effectiveness of the proposed method.

Conclusions:

• Broadband Lamb wave signals contain richer damage information, leading to improved SHM and NDT.
• The developed deep learning algorithm effectively extracts and fuses features from broadband signals for accurate damage assessment.
• This method offers a promising advancement for reliable structural health monitoring.