Blind Modal Identification Applied to Output-Only Building Vibration

Abstract

Recently a new nonparametric, output-only methodology for modal identification has been introduced. A general framework for Blind Modal IDentification (BMID) was outlined, employing Blind Source Separation (BSS) techniques to solve the modal identification problem. Examples previously published involved free decay data of a three DOF system and a structural frame. In the structural frame example, the BMID method performed extremely well. However, the structural frame example exhibited favorable characteristics for application of BMID. It remains to be shown that BMID performs well with less favorable data sets. In this paper, the BMID method is successfully applied to ambient response data of multi-story buildings, despite the unique challenges that building data presents to the method. A modal accuracy indicator, known as the Modal Contribution Indicator (MCI) is introduced and utilized to facilitate spurious mode identification. In addition, it is demonstrated that operating on random response data directly yields superior results compared to operating on cross-correlation functions. Resulting modal parameters are compared with those obtained using state of the art output-only methods. Advantages of the method over traditional modal identification are illustrated.