Research On Structural Dynamic Load Identification Technology Based On Different Regularization

In practical engineering problems,the information of dynamic load is of great significance to parameter identification,fault diagnosis and fatigue life prediction of the system.However,it is difficult or even impossible to measure load directly in many practical cases due to the constraints of the actual environment or economic conditions.Therefore,it is very important to study the indirect acquisition of dynamic loads based on system response and system characteristics.In this paper,different identification methods are proposed for periodic dynamic loads and impact loads,and corresponding regularization techniques are used to deal with the ill-posedness in the identification process.The research work of this paper mainly is included the following aspects:?1?The periodic dynamic load is described by a series of impulse response functions in the time domain,and the dynamic response of the structure is discretized by convolution relation.The identification equation of periodic dynamic load based on Green function is established.Tikhonov and generalized truncated singular value decomposition?TSVD?regularization methods are used to deal with the ill-conditioned characteristics of identification problems.The identification of single source dynamic load and multi-source dynamic load is studied by numerical comparison,and the identified results under different noise levels and response types are analyzed.Then,the identification of sinusoidal loads on cantilever beams is verified by experiments,and the finite element model is modified by multi-island genetic optimization method to obtain a more accurate Green function matrix.?2?For the sparse characteristics of impact load identification,the generalized sparse identification equation of the impact load is constructed.And the equation is based on the transfer matrix using the convolution integral relationship between the impact load and the response.Then,the ill-conditioned equation is solved based on the l₁ norm regularization technique of truncated Newton interior point method.Meanwhile,the search direction of l₁norm regularization is calculated by preconditioned conjugate gradient algorithm.The accuracy of the identification method is studied by the impact load identification test of plate and scaled car body structure,and the identified results are compared with the traditional Tikhonov regularization.?3?Combined with the substructure technology,the Green function method proposed is applied to the three-dimensional scaled car body structure,and the ill-posedness of the recognition problem is processed by the TSVD regularization technique.The numerical simulation analysis is carried out by using the established periodic dynamic load identification model,and the effectiveness of the identification method under different cyclic dynamic loads is studied.The results are shown that the acceleration signal is selected as the input response for the identification of periodic dynamic loads.And the identified results of TSVD regularization technology are always better than Tikhonov regularization.The overall identification error is lower and the correlation coefficient is higher.For the identification of impact loads,the proposed l₁ norm regularization has stronger noise immunity and robustness when identifying impact loads than Tikhonov regularization.