Enhanced sparse component analysis for operational modal identification of real-life bridge structures

• Published in: Mechanical systems and signal processing 2019-02, Vol.116, p.585-605
• Main Authors: Xu, Yan, Brownjohn, James M.W., Hester, David
• Format: Article
• Language: English
• Subjects: Blind source separation; Non-stationary signals; Operational modal identification; Sparse component analysis
• ISSN: 0888-3270; 1096-1216
• DOI: 10.1016/j.ymssp.2018.07.026

•Propose a novel blind source separation method for operational modal analysis.•Validated on ambient and non-stationary vibrations from in-operation bridges.•Evaluated by comparing with two classic operational modal analysis methods.•The method could capture mode shape changes due to heavy truck passage.•The method could identify low-energy and closely-spaced modes. Blind source separation receives increasing attention as an alternative tool for operational modal analysis in civil applications. However, the implementations on real-life structures in literature are rare, especially in the case of using limited sensors. In this study, an enhanced version of sparse component analysis is proposed for output-only modal identification with less user involvement compared with the existing work. The method is validated on ambient and non-stationary vibration signals collected from two bridge structures with the working performance evaluated by the classic operational modal analysis methods, stochastic subspace identification and natural excitation technique combined with the eigensystem realisation algorithm (NExT/ERA). Analysis results indicate that the method is capable of providing comparative results about modal parameters as the NExT/ERA for ambient vibration data. The method is also effective in analysing non-stationary signals due to heavy truck loads or human excitations and capturing small changes in mode shapes and modal frequencies of bridges. Additionally, closely-spaced and low-energy modes can be easily identified. The proposed method indicates the potential for automatic modal identification on field test data.