Fundamental investigation of dynamic response data processing method to improve the estimation accuracy of natural frequency using stochastic subsplace identification method

Abstract

In recent years, studies have been reported which aim to detect bridge damage from changes in vibration characteristics. In those studies, variable characteristics are estimated from bridge dynamic responses by utilizing structural identification methods such as ERA and stochastic subspace identification, where estimation accuracy generally decreases due to the influence of noises in the response data. This study aims to improve the estimation accuracy of natural frequencies by processing response data for inputting. Verification using analytical and actual response data showed improvement of identification in accuracy for lower-order modes and robustness for higher-order modes.