ST98-1 Building Tomorrow’s Society Bâtir la Société de Demain Fredericton, Canada June 13 – June 16, 2018/ Juin 13 – Juin 16, 2018 A NEWER TIME-FREQUENCY DECOMPOSITION-BASED MODAL IDENTIFICATION TECHNIQUE FOR STRUCTURES Lazhari, Malek 1 , Sony, Sandeep 2 and Sadhu, Ayan 3,4 1 MSc student, Department of Civil Engineering, Lakehead University, Canada 2 PhD student, Department of Civil Engineering, Lakehead University, Canada 3 Assistant Professor, Department of Civil Engineering, Lakehead University, Canada 4 asadhu@lakeheadu.ca Abstract: In last few decades, vibration-based structural health monitoring has gained significant popularity to perform condition assessment of civil structures. A wide range of system identification methods has been developed by different researchers to identify modal parameters accurately from the measured vibration data. One of the time-frequency methods, namely empirical mode decomposition (EMD), has been very popular owing to its basis-free nature and applicability to any nonlinear and nonstationary signals of dynamical systems. However, the EMD results in significant mode-mixing in the separated signals that causes inaccuracies in the estimated modal parameters. In this paper, two different newer classes of EMD methods are explored and compared to undertake ambient modal identification using just single channel measurement. The proposed method is perfectly suitable for automation and has significant potential for real-time monitoring since it uses only one channel of data at a time. The performance of the proposed EMD method is verified using a suite of numerical and experimental studies. 1 INTRODUCTION Large-scale infrastructure such as bridges, buildings, wind turbines, dams and tall towers may lose structural integrity due to exposure to severe earthquakes, strong winds or other operational loads. Structural Health Monitoring (SHM) is an essential tool to evaluate the current state of the structure, predict the future damage, and conduct appropriate maintenance and retrofitting. SHM is primarily consisted of sensor-intensive data collection and signal processing-based system identification followed by condition assessment and hazard mitigation. In this paper, two different classes of signal processing methods are explored as a possible tool for modal identification of civil structures. System identification (SI) addresses the problem of deriving mathematical models to describe dynamical systems based on the observed measurement of civil and mechanical structures (Reynders 2012). Consequently, condition assessment and retrofitting are undertaken based on current modal parameters of the structures. In spite of significant development of a wide range of modal identification methods, time- frequency (TF) analysis is considered as a prominent SI method that shows the variation of modal parameters in time and frequency domain simultaneously. Linear TF methods include short-time Fourier transforms (STFTs), and wavelet transforms (WTs), whereas most quadratic methods are variations of the Wigner-Ville distribution and Cohen’s class distribution. However, none of these approaches leads to a unique transform that can be used in all scenarios independent of its own characteristics. Due to the uncertainty relation that links time and frequency, the results from any transformation depend not only on the intrinsic characteristics of the signal but also on the properties of the chosen transform (Auger et al. 2013). In last two decades, time-frequency domain methods have acquired a considerable interest,