Study On Physical Parameter Identification Of Civil Engineering In Time Domain And Inversion Of Ground Motion

Structural physical parameter identification, parameter identification of horizontal layered soil and inversion of ground motion are studied in time domain in this paper with six aspects: (1) Classical Least-Square-based structural parameter identification with incomplete dynamic responses and with known input excitation; (2) Hybrid-inversion- based structural parameter identification with incomplete output and without input; (3) Structural parameter estimation by using extended Kalman filter with memory fading under incomplete output and known input; (4) Generalized-hybrid-inversion- based structural parameter estimation with limited output and without input; (5) Structural parameter identification and inversion of ground motion by using experimental data of shaking table; (6) Shear modulus identification of horizontal layered soil and inversion of bedrock ground motion with incomplete output and unknown input by using hybrid inversion method, and with limited output and without input information by using generalized hybrid inversion method, respectively. The concrete researches and conclusions are listed the following:(1) Accurate reconstruction methods of rotation responses are given in this paper. A pseudo-linearization (PL) method on damping is proposed for big difference on magnitude between stiffness and damping coefficients and nonlinear parameter system. The modified Levenberg-Marquardt (mLM) method is introduced. And since the mLM method is sensitive to the initial values of parameter, a SVD-mLM method is proposed with combination of singular value decomposition (SVD). A generation method of Gauss white noise with (0,1) distribution is presented by using Matlab software. Parameter identifications of shear-type structure and bending-type structure are simulated with incomplete output and known input by using classical Least-Square method. The results show that the parameter identification has good accuracy, the methods are robust on noise and the accuracy of stiffness parameters is higher than that of damping coefficients because of the big difference on magnitude between them.(2) Common used auxiliary conditions (or prior knowledge) are presented. The calculate equations of hybrid inversion method are modified with total compensation method (TCM) and statistical average algorithm (SAA). A Least-Square-based fast iteration method of structural parameter identification is proposed. The rectangle window method is introduced for parameter identification in time domain. Three examples of shear-type structure or bending-type structure for parameter identification are done with incomplete output and without input by using hybrid inversion method. The results show that the methods have good adaptive to the noise and are not sensitive to the initial values of parameters, and the identified parameters have good accuracy; For nonlinear parameter system, both the proposed PL-SVD method and SVD-mLM method can get good results on parameter identification, and the latter has better accuracy but reduces the computation efficiency; The maximum identification errors of structural parameter is reduced but computation time is increased by using the rectangle window method; The time history curves of inversed ground motion are consistent with ones of the real ground motion, but the inversed ground motions have higher noise component with the signal-to-noise ration increasing.(3) Memory fading (MF) filter technique is introduced in this paper, and an extended Kalman filter method with memory fading and with weighted global iteration (MFEKF-WGI) is constructed. The fourth-order Runge-Kutta Method or Gill method is used for sate prediction in MFEKF-WGI method. Simulation results show that the good results of parameter estimation can be get with noise polluted dynamic responses, and the parameter estimation for storey damping are get with a little bigger errors because of the number of state components is increased; But the time history curves of estimated velocity are coincide with the analyzed ones; The filter divergence is effectively prevented and the convergence of state estimation is sped up by using the memory fading filter.(4) The concept of generalized hybrid inversion method is given. Based on substructure technique, hybrid inversion method, MFEKF-WGI algorithm and fourth-order Rugge-Kutta method or Gill method, the equations of generalized hybrid inversion method is constructed. The results of shear-type structure model show that: step 1, the identified parameters of substructure have higher accuracy and the inversed ground motion is consistent with the real one by using the substructure technique and hybrid inversion method; step 2, based the identified parameters and inversed ground motion, the higher accurate parameters are estimated and the estimated dynamic responses are consistent with the analyzed ones.(5) Experiments of 5-layers steel frame model on shaking table are conducted. Modal analysis is done with the preprocessed experimental data. The modal frequencies can be get by using peak picking method and the first two modal shapes can be get by using the amplitude attenuation of dynamic responses. And modal analysis results show that the model is an approximate shear-type structure. Based on the preprocessed experimental data, the parameters of the model and the ground motion of the shaking table are estimated by using the hybrid inversion method or the generalized hybrid inversion method. Comparing the simulation results with the modal analysis ones and the measured ground motion, the identified parameters and the inversed ground motion under small amplitude and steady amplitude changing are worth of belief.(6) Based on the hybrid inversion method and the generalized hybrid inversion method, physical parameters identification of horizontal layered soil and ground motion inversion of bedrock are studied in this paper. A real engineering site is used as the example, the simulation estimation are conducted with incomplete output and unknown input by using the hybrid inversion method, and are done with limited observation and without input information by using the generalized hybrid inversion method. The results show that the estimated parameters and inversed ground motion have small errors; But since soil parameters of difference layers change more complex than structural parameters, the errors of the former are slightly bigger than ones of the latter; the convergence of parameters estimation are sped up and the phase of rough estimation is shortened by the reasonable selection of memory fading factor.The theoretical researches on parameter identification of structure and of horizontal layered soil and experimental investigation on estimation of parameters and ground motion of steel frame model on shaking table are listed above. The methods, such as the hybrid inversion method, the generalized hybrid inversion method, MFEKF-WGI technique, the proposed and the introduced methods, are proved to be effective and useful. In those researches, the inversion analysis of horizontal layered soil will provide a new technique for determining the physical parameters of soil and ground motion of bedrock in the real engineering.