Performance evaluation of signal decomposition techniques for vehiclescanning- based modal identification of bridges

Abstract

Indirect Structural Health Monitoring (SHM) otherwise known as vehicle-scanning method for bridge health monitoring is increasingly gaining importance due to the advantages such as the usage of single sensor on the moving vehicle without installing any sensors on the bridge, operation of traffic during vehicle-scanning, and better spatial resolution of the bridge due to moving sensor. The extraction of bridge frequencies, mode shapes, and damage using the drive-by method are posing few challenges due to the presence of roughness, measurement noise, and less transmissibility of thebridge characteristics. In order to obtain the bridge parameters and damage features, powerful signal processing tools are being relied on. One of the effective ways to perform feature extraction is through the decomposition of the signalinto useful components. In this paper, three signal decomposition techniques namely, Variational Mode Decomposition (VMD), Empirical Fourier Decomposition (EFD), and Singular Spectrum Analysis (SSA) which are robust in handling mode mixing and measurement noise are employed for extraction of bridge frequencies using the vehicle response. The performances of the three decomposition techniques are compared to evaluate their efficiencies for indirect SHM of bridges, using the drive-by vehicle. The comparative analysis is made and findings are obtained using the numerical simulation of a simply supported beam, with moving vehicle under varied speeds, road roughness, and measurement noise. The results of the numerical investigation show that the three techniques are efficient in decomposing the signal without mode mixing. It was found that, while the speed of the vehicle has no effect on the performance, measurement noise and roughness are affecting the effectiveness of the techniques. Also, SSA was found to outperform the other techniques in extracting the maximum number of mode shapes when compared to the other decomposition techniques from a noisy signal.