CNN-LSTM-Based Damage Localization of Plate Structure.

Area of Science:

• Structural Health Monitoring
• Artificial Intelligence in Engineering
• Signal Processing

Background:

• Conventional methods struggle with accurate feature extraction from time-domain signals.
• Imprecise damage localization is a key challenge in plate structure identification.
• Existing techniques often fail to effectively pinpoint damage locations.

Purpose of the Study:

• To develop an innovative damage localization approach for plate structures.
• To overcome the limitations of traditional damage identification methods.
• To accurately predict the coordinates of damage locations using deep learning.

Main Methods:

• Integrated Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks.
• Converted one-dimensional signal data into two-dimensional grayscale images for CNN analysis.
• Utilized LSTM's memory mechanisms for feature integration and regression prediction.

Main Results:

• The proposed method demonstrated strong performance in accurately localizing artificial damage on aluminum plates.
• CNN effectively extracted features from the converted image data.
• LSTM successfully integrated features for precise damage coordinate prediction.

Conclusions:

• The joint deep learning approach offers an effective solution for plate structure damage detection.
• This method accurately predicts damage coordinates, surpassing traditional techniques.
• The study validates the effectiveness of integrating CNN and LSTM for structural health monitoring.