Vibration-based health monitoring of railway bridge ballast layers: A PCA–VMD method

This study presents a robust, high-precision framework for detecting and localizing structural damage in railway bridge ballast layers under the passage of a moving train. Unlike conventional vibration-based methods that suffer from low resolution and severe mode-mixing in noisy environments, the proposed approach leverages a hybrid Principal Component Analysis and Variational Mode Decomposition (PCA-VMD) methodology. Utilizing a verified multibody Train-Track-Bridge Interaction (TTBI) simulation of the Tehran-Qom railway line, the dynamic response is analyzed for both standard ballasted tracks and tracks improved with a Tire-Derived Aggregate (TDA) vibration mitigation layer. In this framework, the PCA stage acts as a baseline-free spatial filter to mathematically neutralize the dominant global bridge kinematics and spatially uncorrelated noise. Consequently, the subsequent VMD stage autonomously isolates the subtle, high-frequency damage transients without suffering from frequency masking. A rigorous comparative analysis demonstrates that while traditional techniques (e.g., EMD, EEMD, and standalone VMD) fail due to severe mode-mixing under realistic measurement noise, the proposed PCA-VMD framework maintains exceptional diagnostic reliability. Validated across varying operational train speeds (20, 50, and 100 km/h) and structural configurations, the methodology consistently achieves a damage localization precision exceeding 90%, establishing it as a highly trustworthy tool for autonomous and modern predictive maintenance.