Data-driven-based And Model-based Approaches For Structural Seismic Damage Detection

Great earthquakes in the world have caused huge economic damage, casualties andsocial implications in recent years. This is because that people on one hand, are lack ofknowledge of the characteristics of earthquake ground motion and on the other handpeople can not accurately understand the seismic damage mechanism of the realengineering structures under the strong ground motion. The structural health monitoring(SHM) systems, having been developed in recent30years, monitor the real responses ofthe prototype structures under the real environment and loads with placed sensors, andprovide real-time structural damage diagnosis and timely early warning alarm, thus theSHM is currently the most direct and effective measure to study and reveal the seismicresponses, damage mechanisms and failure mechanisms of engineering structures.Traditional structural damage detection methods based on vibration signals assume thestructure is linear before and after damage. Due to the non-stationarity of strong groundmotion, the engineering structures exhibit obvious nonlinear behavior, such as theopening and closing of cracks in concrete and the yielding and buckling of steels, etc.,which makes the traditional structural damage detection methods are not suitable fordetecting structural seismic damage directly. Therefore there is an urgent need to re-examine the non-stationary characteristics of the earthquake ground motion, and todevelopment structural seismic damage detection and assessment methods with the helpof nonlinear feature and nonlinear system identification approaches. To this end, thisthesis focuses on the time-frequency energy distribution of earthquake ground motion,data-driven based structural seismic damage detection method by extracting nonlinearfeature and model based structural seismic damage detection method by identifying thenonlinear system. The main contents are as follows:The method of generating earthquake ground motion based on matching pursuitdecomposition (MPD) and time-frequency energy distribution (TFED) is researched.Firstly the theory of MPD is introduced. Secondly the artificial earthquake groundmotions with same peak ground acceleration and Fourier amplitude spectrum aregenerated and the difference between the two TFEDs and their effects on the maximalelastic displacements of linear structures are studied. The amplitude, cyclingcharacteristics, phase lag angle and effective energy of time-frequency atoms aredefined, and the correlation between maximal elastic displacement and effective energyare analyzed. Finally the regular pattern of TEFDs’s effects on the responses of elastic perfectly plastic model (EPP) for the artificial earthquake ground motions are studied,and explaination of the dynamic ratcheting effect of EPP model are given by using theTFED of the real earthquake ground motion.The structural seismic damage detection method based on the fractal dimension(FD) of interstory displacement’s time-frequency feature (TFF) is proposed for shear-type structure. Firstly, the TFF is defined and the FD is used to quantify the TFF.Secondly, the FD of TFF at each degree of freedom (DOF) for linear system is derivedwith the help of modal superposition principle, and the FD of TFF for each substructurein nonlinear system is derived with the help of substructure method. Finally, thenumerical simulations of10-story shear-type structure with single or multiple damage,different damage extend, near field and far field ground motion and different noise levelare conducted, and the effects of damping ratios and using interstory acceleration on thedetection results are discussed.The structural seismic damage detection method based on the FD ofdisplacements’ TFF is investigated for moment-resist frame (MRF) structure. Firstly, theFD of TFF at each main DOF of linear MRF is derived with the help of staticcondensation method and modal superposition principle. Secondly, for nonlinear MRFwith plastic hinges at the ends of beams and columns, the transformation matrix fromthe rotations of plastic hinges to the inelastic displacements are derived based on theforce analogy method and the characteristics of the FD of TFF are studied; for thenonlinear MRF with added local nonlinear component such as the damper, the effects oflocal nonlinearity on the relative displacements are discussed and the characteristics ofthe FD of TFF are also studied. Thirdly, the numerical simulation of16-story MRF withsingle and multiple plastic hinges at the ends of beams and columns is studied. Finally,the shaking table test of a16-story steel MRF with added frictional dampers, whichsimulate local nonlinearities in the whole structure, is conducted to verify the proposedmethod.The decentralized identification method of structural seismic damage based on theextended Kalman filter (EKF) and unscented Kalman filter (UKF) is proposed. Firstly,the principles of classical KF, EKF and UKF are introduced. Secondly, the method thatidentifies the extended state spaces of the linear and nonlinear substructures using EKFand UKF respectively is investigated. Thirdly, a10-story shear-type structure withsingle or multiple damages is used for numerical simulation, and the identified resultsare compared with those using EKF or UKF for the complete structure identification.Finally, the quantitative evaluation of the structural seismic damage is obtained bycombining the estimated force-displacement curve and the modified Park-Ang damage index.