摘要

This paper presents a novel damage detection approach using output-only responses. A statistical analysis is proposed to identify mode shapes of the structure from measured dynamic responses with the aid of empirical model decomposition (EMD). Then the Fourier series fitting method is employed to modify the gapped smoothing method (GSM) for an effective damage detection. In order to identify the mode shape of a target structure with measurement noises, EMD with intermittency criteria is applied to separate dynamic responses of all grid points on the target structure into individual modal responses. Astatistical analysis is proposed to evaluate mode shapes from modal responses. After that, the curvature mode shapes (CMS) which has been investigated as a promising feature for damage identification are obtained by a central difference approximation. The improved GSM with Fourier series fitting method uses mode shapes from all locations to construct the baseline data of the healthy state. Numerical studies are conducted for methodology illustration and validation. Effects of broader damage, multiple damage and measurement noise level are investigated in detail. The accuracy and effectiveness of the proposed method are demonstrated and verified by comparison with original GSM method.