Research On Dynamic Load Identification Method Based On Bayesian Method

For the accurate design of structural dynamics,it is necessary to know the dynamic loads on the structure.Traditional dynamic load identification not only needs to measure the response information of the structure,but also needs to obtain the dynamic characteristic information of the structure.How to obtain the external load of the structure with only the dynamic response information of the structure becomes a challenging problem.Bayesian method is a method of probability and statistics,which itself acknowledges the existence of statistical errors.Kalman filter based on linear minimum variance unbiased estimation criterion is one of the most widely used Bayesian methods in engineering applications.In the dynamic load identification method based on Bayesian principle proposed in this paper,the system is also unknown,only the response information of the structure under random excitation.The main research contents are as follows:(1)The Bayesian fast Fourier transform method for modal parameter identification under ambient excitation is studied.The posterior probability density distribution function and covariance of modal parameters are obtained by Fourier transform of environmental data,and then the modal parameters of the structure are obtained based on unconstrained optimization.If all modes are separated completely,it can be simplified to single mode analysis.Taking Bernoulli-Euler beam as the test object,the feasibility of this method is proved.The identified modal parameters are used for further dynamic load identification.(2)For linear syst MSE with unknown excitation,a Kalman filter-like method for identifying excitation in modal space is proposed,which identifies state and unknown excitation through system information and observation information.Taking the multi-degree-of-freedom system and the cantilever beam as examples,it is proved that the load information can be well recognized when there is noise pollution.The Kalman filter load identification method without observing the response of excitation points is further proposed and improved.The influence of acceleration sensor distribution on load identification accuracy is discussed.The results show that when the sensor is symmetrically distributed on both sides of the excitation point,the identification accuracy is the highest.(3)For discrete model errors and unknown excitations,a load identification algorithm based on extended Kalman filter is proposed,which modifies model parameters and reconstructs unknown excitations.This algorithm is introduced into modal space,and an extended Kalman filter algorithm in modal space is proposed.Taking Bernoulli-Euler beam as the test object,the modal parameters with discrete errors are identified based on Bayesian fast FFT transform method,and the time-domain information of excitation is identified by combining the extended Kalman filter load identification algorithm in modal space,and the modal parameters in the model are modified.