Study On Dynamic Force Identification Method In Time Domain For Linear Elastic Structure

The safety and durability of engineering structures are highly associated with loadings applied on structures.As a result,it is necessary to learn about loading information applied on structures for analyizing structural vibration behavior scientifically and evaluating the"healthy" status of structural condition accurately.Due to limited sensing technology,as well as load characteristics of randomness,diversity and complexity and so on,it is extraordinary difficult to measure load directly.However,the vibration testing technique for structural dynamic responses are comparatively mature.Therefore,it is an important research orientation that force information are identified with structural vibration responses.This dissertation focuses on the critical problems and deeply investigates basic theoretical methods of dynamic force identification for linear elastic structures.The main research contents are listed as follows:The high precision force identification model is built after theoretical analysis.In order to avoid propagating and accumulating of force identification error,the mapping from input sequence to output sequence are established in whole time domain after structural state-space equation is discretized by classic with Runge-Kutta method.Then,the high-precision identification model is constructed in the whole time domain based on Topelitz matrix.The stability condition of identification model is also analyzed.In addition,a comparative analysis are carried out with respect to conventional state-space method and weak galerkin method.The numerical simulation results show that proposed method has the highest accuracy under noise free condition and have good performance in low noisy environment.Besides,the presented method circumvents the inverse of stiffness matrix,resulting in improvement of stability of force identification results.The dynamic loads identification method on multi-scale is presented in the framework of wavelet multiresolution analysis.In order to make force identification efficiency and stability significantly improved,the basic function expansion method with scaling function of Db wavelet is theoretically derived.Then,force identification problem is transformed into the solution of linear equations in wavelet subspace.Not only can one view identification results on various scales,but also the dimension of transmission matrix is highly reduced.Case studies investigated impulse and sinusoidal load identification in numerical simulation.In addition,the effects of noise level and sensor configuration are also analyzed.The modified sequential deconvolution dynamic force identification method are presented and parametric study are carried out subsequently.The dimension of transmission matrix increases with increasing of number of sampling points when load reconstruction problem is converted into the soluction of linear equations.This not only make ill-posed problem even worse,but also leads to computational inefficiency.As a result,sequential deconvolution method are investigated and modified.The relationship between load identification error at arbitrary moment and independent noise sequences are formulated and a new index of normalized standard deviation is presented.In addition,parametric studies are conducted and analyzed.The numerical simulation demonstrates the validity of modified method.The nonstationary random ground motion identification is investigated in the framework of augmented Kalman filtering.After band-limited white noise passed through Kanai-Tajimi filter,the stationary filtered white noise is generated,which is used to imitate nonstaionary randomness of gound motion with amplitude modulation operation.The simulation procedure of nonstationary random ground motion could be considered as ground motion model,which is combined with,structural state-space equation leading to an augmented state-space equation.The ground motion could be identified with structural dynamic responses under the framework of Kalman filtering.In addition,the effects of noise level and sensor placements are also investigated.The impulse load identification method with sparse regularization is presented.Signal reconstruction problem is investigated with sparse regularization(l1-norm regularization)for impulse loading characterized by sparsity in time domain.Impulse load identification is transformed to the optimization problem based on l1-norm regularization method,which is solved by fast iterative shrinkage-threshold algorithm with advantage of simplicity and fast convergence.The time history and location of impulse loading are simultaneously identified.In addition,basic principle is interpreted that l1-norm regularization results in sparsity of impulse load identification in perspective of mathematical point.