A Wavelet Packet Transform and Convolutional Neural Network Method Based Ultrasonic Detection Signals Recognition of...

Area of Science:

• Civil Engineering
• Materials Science
• Artificial Intelligence

Background:

• Concrete structures are critical infrastructure requiring reliable non-destructive testing methods.
• Ultrasonic testing is a common technique for concrete defect detection, but traditional methods face limitations in accuracy and efficiency.
• Intelligent algorithms offer potential for improved analysis of ultrasonic data.

Purpose of the Study:

• To develop and validate a novel intelligent recognition method for concrete ultrasonic detection.
• To enhance the accuracy and efficiency of concrete defect classification using advanced signal processing and machine learning.
• To compare the performance of the proposed method against existing techniques like stochastic configuration networks (SCN).

Main Methods:

• Wavelet packet transform (WPT) for signal decomposition and feature extraction from ultrasonic data.
• Convolutional Neural Network (CNN) for intelligent classification of concrete defects.
• K-fold cross-validation for robust performance analysis and model evaluation.
• Calculation of precision, recall, and F-score to assess classification performance.

Main Results:

• The four-classifying CNN model achieved over 99% detection accuracy.
• The six-classifying CNN model demonstrated a minimum recognition accuracy of 92.5% on the testing dataset.
• The proposed WPT-CNN method outperformed existing SCN models in recognition performance.
• Analysis highlighted the ongoing challenges in precise and efficient feature extraction and classification of diverse concrete defects.

Conclusions:

• The proposed intelligent recognition method based on WPT and CNN is highly effective for concrete ultrasonic detection.
• The data-driven approach significantly improves the accuracy and efficiency of concrete defect classification.
• Further research is needed to address complex feature extraction and classification for a wider range of concrete defects.