Reliability evaluation of stochastic subspace identification and frequency domain decomposition methods in estimating modal parameters of a structure excited by the earthquake

Abstract

One of the most common methods of identifying modal parameters in the field of operational modal testing is the method of identifying sub-random space and frequency domain analysis. Unfortunately, the scope of these methods' application is limited to static signals with a long pick-up time, and if the above conditions are violated, the results will be erroneous; This is in the context that the above two conditions are not met regarding the earthquake signal, and so far the reliability of these methods and their error rate in the face of this group of signals have not been studied. In this regard, in this study, the performance of these methods in earthquake conditions (both conditions are violated) is studied. For this purpose, an numerical model of two two-dimensional frames with different heights (five and ten floors) is created and stimulated by using 20 earthquake records in the near and far fields. Using the obtained results and comparing them with the results of the numerical model, the error values for the modal parameters are obtained; Also, with the statistical study of the errors in the estimation of the frequency of the structure, the probability distribution function of the error and an estimate of the distance of the error are suggested .The results of the study showed that (a) the method of identifying random subspace has a better performance than the method of frequency domain analysis; (B) The random subspace detection method is not able to detect the first modes and is proposed to identify higher modes; (C) the efficiency of the frequency domain decomposition method decreases with increasing structural height; (D) By optimizing and locating the sensors, the performance of the frequency domain analysis method is dramatically improved. However, in the random subspace detection method, the detectability increases with the number of sensors.

Keywords

• operational modal analysis
• frequency domain analysis
• identification of random subspace
• structural health monitoring