Real time system identification in smart structures using wavelet transform based sparse component analysis

Abstract

Recently, output only based real time identification of structures has attracted attention. One of the powerful offline system identification is the sparse component analysis (SCA) methods which are essentially a subset of blind source identification methods. This methods by transferring the dynamic response of the structure from time domain to frequency domain results in sparsity of the data and identifies the modal parameters of the system. In transferring data to the frequency domain, issues such as high data volume and continuous need for changing the time windows according to the variations in the input data result in limitations for real time identification. To overcome these problems, in this study, a wavelet transform-based sparse component analysis (WT-SCA) method is proposed for real time identification of structures. Then, using the proposed WT-SCA method and a semi-active tuned mass damper (STMD), an algorithm is proposed for developing a smart structure in a way that if any changes tend to occur in modal parameters of the structure due to a strong environmental excitation, the mechanical characteristics of the STMD using the WT-SCA method are retuned such that the system is stable and quite robust against the variations. The performance and accuracy of the proposed method are evaluated using numerical examples. Results demonstrate that the WT-SCA identities the real-time system state of the system with acceptable accuracy and reduces the dynamic response of structures equipped with STMD device effectively.