4. Blind Source Separation: A Generalized Modal Identification Tool for Civil Structures

Over the last decade, Blind Source Separation (BSS) has evolved as a powerful tool for system identification of flexible structures. Several numerical and experimental studies have been proposed to demonstrate its effectiveness in dealing with noisy full-scale data. The author has recently developed a suite of methods that enhance the capabilities of BSS in addressing issues associated with decentralized implementation, autonomous processing, nonstationary excitations, and the presence of narrowband excitations in ambient vibration measurements. The basic idea of the algorithms proposed by the author is to cast the problem of identification within the framework of underdetermined BSS invoking sparsifying transforms. The resulting partial mode shape coefficients are combined to yield complete modal information. The transformations are undertaken using Stationary Wavelet Packet Transform (SWPT), yielding a sparse representation in the wavelet domain. Both numerical and experimental studies demonstrate the potential of these methods. The speaker will introduce this suite of methods and some examples where these methods have successfully been applied.