Study On The Structural Damage Identification Based On The Hilbert-huang Transformation

The dynamic response signals are usually non-linear and non-stationary, however, the traditional damage identification methods are in the premise of the linear system. So, in this paper, the Hilbert-Huang transform(HHT), which can analysis the non-linear and non-stationary signal, and the Auto-regressive(AR) model of the time series are used to identify the damage, the main contains of the article are as follows:(1) The research development of the structural damage identification is reviewed, especially for the structural damage identification methods based on the signal processing technology, the scope of their application and the existing problems are also pointed out.(2) The basic theories of the HHT and the AR model are briefly introduced.(3) The damage identification methods based on the HHT and AR model is presented. The intrinsic mode function(IMF) is obtained by decomposing the acceleration response signal with empirical mode decomposition(EMD), the IMF is fitted with AR model, the residual variances of AR model pre and post damage are employed to construct the damage characteristic index, which is used to alarm the damage, the instantaneous phase of the IMF is used to identify the damage extent. The numerical simulation under the white noise excitation and the non-stationary excitation are implemented separately, the influences of the measurement noise are also considered, the effectiveness of the presented method is verified by the experiment.(4) The damage identification methods based on the improved HHT and AR model is presented. The inherent problem existing in the process of EMD is mode mixing. To solve this problem, the Butterworth II filter is employed to filter the original wideband signal into a series of narrow-banded sub-signals. In order to improve the anti-noise ability of the presented method, the autocorrelation function of the narrow-banded sub-signals is decomposed with EMD, then, the gained IMF is fitted with AR model, the residual variances of AR model pre and post damage are employed to construct the damage characteristic index, which is used to alarm the damage, finally, the instantaneous phase of the IMF is used to identify the damage extent. The numerical simulation under the white noise excitation and the non-stationary excitation are implemented separately, the influences of the measurement noise are also considered, the effectiveness of the presented method is verified by the experiment.