An Improved Kernel Entropy Component Analysis for Damage Detection Under Environmental and Operational Variations

Area of Science:

• Engineering
• Data Science
• Physics

Background:

• Vibration-based structural health monitoring (SHM) methods are prone to false alarms due to environmental factors.
• Conventional techniques like Principal Component Analysis (PCA) struggle with measurement noise, nonlinearities, and non-Gaussian data.

Purpose of the Study:

• To develop a novel, robust damage detection method for SHM that overcomes limitations of existing techniques.
• To enhance the accuracy and reliability of identifying structural damage under environmental influences.

Main Methods:

• A hybrid approach combining Variational Mode Decomposition (VMD) for noise and pattern removal with Dynamic Kernel Entropy Component Analysis (DKECA) was proposed.
• VMD preprocesses modal frequencies to eliminate noise and seasonal effects.
• DKECA, utilizing a time-delay data matrix, identifies principal components that maximize Rényi entropy for sensitive damage detection.

Main Results:

• The proposed VMD-DKECA method demonstrated superior performance in distinguishing actual damage from environmental variations.
• Validation on a simulated 7-DOF model and real-world Z24 bridge data confirmed the method's effectiveness.
• Comparative analysis showed significant advantages over traditional PCA and cointegration techniques.

Conclusions:

• The integrated VMD-DKECA approach offers a more reliable solution for vibration-based structural health monitoring.
• This method effectively mitigates false alarms, leading to more accurate structural damage assessment.
• The findings suggest a promising advancement for ensuring the safety and integrity of structures.