Study On Modal Parameter Estimation And On-line Identification Technology For Linear Time-varying Structures In Time-domain

With the continual expansion of applications in the field of aerospace structures, it isnecessary to research further on the time-varying structural dynamic characteristics in orderto meet the hard and limited requirement of structural analysis and design. Modal parameterestimation is useful for dynamic performance evaluating, load identification, modelupdating, structural heath monitoring, structural damage detection, vibration control and etc.Therefore, the research on modal parameter estimation of time-varying aerospacestructures recently has drawn a lot of attention, and it is of great research value.This thesis studies on modal parameter estimation method and on-line identificationtechnology for linear time-varying structures in time-domain. It researches into three coreproblems of time-domain models and their identification method, such as direction ofmodel prediction, selecting of model structure parameters, and on-line identification. Themain contents of this thesis are followed by:Time-varying characteristics of aerospace structures, the motivation of their modalparameter estimation, the current research situation and the methods of modal parameterestimation of time-varying structures are introduced to supply the primary support for thefurther research on the theory and experiment of modal parameter estimation oftime-varying structures. In addition, three core problems of time-dependent time seriesmodels are elaborated to give support to the research of this thesis.The basic theory involved in modal parameter estimation of time-varying structuresare summarized, including the introduction and discussion about the dynamic theory oflinear time-varying systems which presents the dynamic model and the definition andproperties of poles of linear time-varying systems, as well as the criterion of slowlytime-varying systems in the time-domain and frequency-domain, respectively, providing thebasis and assumptions for modal parameter estimation of time-varying structures.A forward-backward linear prediction method for modal parameter estimation oftime-varying structures is presented aiming at the direction problem of model prediction. Aforward-backward linear prediction model is established to estimate model parameter, expanding the identification model from simple output to multiple output. The simulationresults indicate that the proposed method not only has better anti-noise performance andovercoming the shortages of the one-step error in the low sample frequency case, but alsocould obtained the modal vector and antresonant frequencies.A mesh-free shape function modeling method for modal parameter estimation oftime-varying structural systems is proposed for the problem of selecting model structureparameters. According to the idea of the local approximation method using shape functionin the mesh-free technique, an identification model is built to be adaptable to time-varyingcharacteristics of structural responses, proposing it’s algorithm of model parameterestimation. On one hand, in order to solve the numerical conditions problem of theconditional moving least square shape function, a modified moving least square method formodal parameter estimation of time-varying structural systems is proposed based onweighted orthogonal basis function. The simulation results indicate that the proposedmethod has stable identification procedure and high precision without the numericalconditions problem. On the other hand, in order to improve the ability of tracking localtime-varying characteristics, a moving Kriging shape function modeling method for modalparameter estimation of time-varying structural systems is proposed. A model estimationscheme is presented based on optimization method. The scheme decomposes the probleminto two subproblems: estimating model parameters via two-stage least squares method andestimating shape function parameters via a discrete-continuous-variable hybridoptimization. The simulation results indicate that the proposed method has higher precisionof tracking local time-varying characteristics.A kernel adaptive filtering method for on-line modal parameter estimation oftime-varying structural systems is proposed for the problem of on-line identification. Inorder to solve the problem of on-line recursive using basis functions, beginning with thecompactly supported basis function, the method stems from the on-line recursive usingkernel functions, builds an identification model based on kernel function, and proposes it’son-line estimation algorithm insisting of sliding window kernel recursive least squaremethod and its modified counterpart. Compared with the traditional on-line recursivemethods, the proposed method is validated by the simulation case that it has higher precision and efficiency for the identification of multiple channel signals and it supplies theprimary support for the research on on-line identification.Based on a cantilever beam-like time-varying structure with a movable mass, a seriesof “configuration-frozen” reference experiments and a continuous-time structural dynamicexperiment for the time-varying structure are carried out. These experiments validate themethods presented in this thesis.