Structural Damage Identification Under Earthquake Based On Hilbert-Huang Transform

Structural health monitoring and damage detection is a hot research direction in the area of engineering and academe circle. A review of present research on structural damage detection is presented in this thesis as well as its application. Traditional structural damage detection research based on the vibration mostly supposes that the system is in stationary state before and after damage so dynamic properties of the structure can be compared.The notion for damage detection lodged in this thesis is however to identify instantaneous characters of structure namely the instantaneous eigenfrequency, so the real-time damage detection can be performed in workplace. When and how extensive the structural instantaneous characters changes reflect the time and extent of structural nonlinear state respectively. Time realization of aforesaide method need time domain data analysis to obtain structural nonlinear information.Frequency spectrum analysis plays an important role in earthquake engineering field, because ground motions is a complex non-stationary stochastic process and structure vibration represents obvious nonlinear properties when damaged under the action of earthquake waves. Hereby, the traditional spectral analysis method based on FFT can hardly exert effects.Various of signal processing technology come into being and develop rapidly coupled with technical development of modern information in the past 20 years. Some time frequency analysis methods in wide application such as short-time Fourier transform analysis, Wigner-Ville distribution and Wavelet analysis are presented briefly and compared with. Whereas aforesaide methods can not describe the time frequency characteristic of nonlinear and non-stationary signal primely.After calculating and analyzing the marginal spectrum and Fourier frequency spectrum respectively of three typical ground motion records of EI Centro, Loma prieta and Northbridge and contrasting the results, the thesis points out the marginal spectrum has obvious physical meaning, which can reflect the time-frequency characters ofground motion records more exactly. This thesis introduces a new Hilbert-Huang Transform (HHT) which raises empirical mode decomposition (EMD) and defines instantaneous frequency owning clear physical concepts. Signal's time-frequency spectrum--Hilbert spectrum, marginal spectra, energy spectra and instantaneous energy spectra are all derived from EMD and instantaneous frequency.In this thesis, marginal spectra and Fourier spectrum of three typical earthquake waves, EI Centro, Loma Prieta and Northbridge were calculated and analysed. Comparing analysis result, it is to be pointed out that marginal spectra possesses obvious physical meaning and reflects frequency domain characteristics more exactly, which reveales the advantage of HHT.Based on the reflection of Fourier Frequency Response Function, the author defines HHT Frequency Response Function (FRF) and HHT Time-Frequency Response Spectrum (TFRS) to the ratio of system dynamic responses output and input. The impact due to instantaneous frequency calculating error is eliminated through using a method of signal processing -- "Mobile Frequency Bands Weighting" to the marginal spectrum curve. Numerical simulation to a three-dimension trilinear stiffness degradation model suffering from different magnitude earthquake motion is given by taking HHT Frequency Response Function (FRF) and HHT Time-Frequency Response Spectrum (TFRS) methods. In elastic state, the structure vibration state is identified and the preceding two order eigenfrequency is obtained by calculating "Center Frequency" under different threshold. When exposed to strong earthquake, the structure is undergoing the state from health to crack and finally to yield, gradually exhibiting its non-linear vibration behavior. In this condition, TFRS can identify the structure damage time accurately. From the change of first-order eigenfrequency in time and extent, the damage time and extent were fixed on respectively.The thesis makes use of HHT Frequency Response Function and HHT Time-Frequency Response Spectrum to analyse the data of one-story frame structure under the shaking table test experimented by X. T. Zhou at the Hong Kong Polytechnic University. The natural vibration frequency identified from theearthquake response records by HHT Frequency Response Function is less than the result identified from the ambient vibration test .From the analysis of 400s-long really measured output and input date of structure artificial earthquake wave prompting, the time and extent of structure damage are identified preferably.Real-time damage detection method lodged by the author is proved to be effective through the numerical simulations and real measured data analysis.