An improved time-varying empirical mode decomposition for structural condition assessment using limited sensors

The safety concerns of transport infrastructure around the world have significantly increased over the past few years and pose a significant threat to the public and economic well-being. Condition assessment of transport infrastructure is currently performed using structural health monitoring (SHM). Over the last few decades, SHM of ageing infrastructure is leveraged using advancements in sensing technologies. However, vibration-based SHM methods that rely on sensor-driven data collection face several practical challenges such as traffic interruptions, availability of limited space, and a limited amount of sensors. In this study, a system identification technique is developed, which can deal with these practical challenges of SHM and provide reliable results. The proposed technique involves a time-varying filtering empirical mode decomposition algorithm that can decompose a single measured signal into its mono-component modal responses. The resulting mono-component signals are then analyzed using a wavelet-based reallocation algorithm called Multi-Synchrosqueezing Transform. The iterative nature of this algorithm is utilized to enhance the time-frequency representation of a nonstationary signal obtained from the ageing structures. The capability of the proposed method to identify the modal parameters is demonstrated using a suite of numerical and full-scale studies under various cases of measurement noise, closely spaced and low energy modes, and damage detection using the limited sensor measurements.

Automated summary

The study introduces a system identification technique to address practical challenges of SHM and provide reliable results. The proposed technique involves a time-varying filtering empirical mode decomposition algorithm for decomposing signals and a wavelet-based reallocation algorithm for signal analysis. The capability of the method in identifying modal parameters is demonstrated through numerical and full-scale studies.