Modal parameter uncertainty quantification using PCR implementation with SMIT

Abstract

The System Identification (SID) techniques for output-only systems, combined with the use of Wireless Sensor Network (WSN), provide many opportunities to monitor large scale civil infrastructure. Recently, research associated with uncertainties in the measurement data has been conducted to quantify the level of noise and to improve the performance of SID methods. This paper presents the effect of measurement noise when the data is used in Eigensystem Realization Algorithm (ERA) based methods including Observer Kalman filter Identification (OKID), Natural Excitation Technique (NExT), and NExT using average scheme (NExT-AVG). Each algorithm estimates impulse response for ERA algorithm differently, which results in different noise level in terms of Physical Contribution Ratio (PCR) and affects the accuracy of identification results. In order to compare the effect of noise from each SID methods, modal parameters are estimated using the numerically simulated response from simply supported beam model and wireless sensor data from Golden Gate Bridge (GGB). All identification procedures are supported by Structural Modal Identification Toolsuite (SMIT) which provides a convenient environment to access various SID methods.