Development of a Machine Learning-Based Damage Identification Method Using Multi-Point Simultaneous Acceleration...

Area of Science:

• Structural Engineering
• Machine Learning Applications
• Vibration Analysis

Background:

• Accurate structural damage assessment is crucial for safety and maintenance.
• Existing vibration-based methods lack sufficient accuracy for practical application.
• Reliable damage evaluation is needed to prevent structural failures.

Purpose of the Study:

• To develop an accurate structural damage evaluation method using multi-point acceleration data.
• To leverage supervised machine learning, specifically Random Forest, for improved damage detection.
• To enhance the practical usability of vibration analysis in structural health monitoring.

Main Methods:

• Measuring acceleration at multiple points on a structure.
• Utilizing Random Forest, a supervised machine learning algorithm, for data interpretation.
• Employing key parameters: maximum response acceleration, standard deviation, logarithmic decay rate, and natural frequency.
• Implementing a three-step Random Forest process for diverse damage type evaluation.

Main Results:

• The proposed method significantly improves the accuracy of structural damage assessment.
• Cross-validation confirmed the robustness of the Random Forest model.
• Vibration tests on a damaged specimen validated the practical effectiveness of the approach.

Conclusions:

• The developed multi-point acceleration measurement and Random Forest interpretation method offers a highly accurate solution for structural damage evaluation.
• This approach addresses the limitations of traditional vibration-based techniques.
• The findings pave the way for more reliable structural health monitoring and maintenance strategies.