Concrete Multi-Type Defect Classification Algorithm Based on MSSMA-SVM.

Area of Science:

• Materials Science
• Engineering
• Artificial Intelligence

Background:

• Non-contact nondestructive testing (NNDT) is crucial for concrete infrastructure assessment.
• Accurate classification of internal concrete defects remains a challenge for existing NDT methods.

Purpose of the Study:

• To develop an automated algorithm for classifying multiple types of internal concrete defects.
• To enhance the accuracy and practicality of NDT for concrete using laser-based vibration analysis.

Main Methods:

• Acquired concrete surface vibration signals using a laser Doppler vibrometer (LDV).
• Applied wavelet packet transform (WPT) for signal decomposition across different frequency bands.
• Utilized energy ratio change rate, energy ratio, and wavelet packet singular entropy as input features.
• Developed a mixed strategy slime mold algorithm support vector machine (MSSMA-SVM) classifier.

Main Results:

• The MSSMA-SVM classifier demonstrated high accuracy in distinguishing between no defect, segregation, cavity, and foreign matter.
• Feature extraction using WPT provided effective data for defect classification.
• The proposed method offers a practical approach for laser-based concrete defect detection.

Conclusions:

• The developed MSSMA-SVM algorithm provides an effective solution for automatic, non-contact classification of concrete defects.
• Laser vibration measurement combined with advanced signal processing and machine learning offers a promising NDT technique.
• This research contributes a practical tool for ensuring concrete structure integrity.