Research On Finite Element Model Updating Method Based On Surrogate Model And FRF Eigenvalues

At present,most of the finite element model updating methods are based on modal parameters,but these kinds of methods depend on the accuracy of modal identification.The model updating method based on frequency response function(FRF)does not need modal analysis,and FRF can provide more data.However,considering that the frequency points need to be selected when directly using FRF for model updating.Therefore,the constant Q transformation on the frequency response function is performed to extract the first layer of coefficients as the response features in this thesis.In the process of model updating,by constructing a surrogate model instead of the finite element model,the computational efficiency of model updating can be greatly improved,which is of great significance for solving complex structural optimization problems.At the same time,due to uncertain factorsmeasurement error,environmental noise and model simplification,the data obtained by the experiment has a large error.Based on the above background,the deterministic and stochastic model updating methods are studied respectively in the thesis.The main research contents are as follows:The model updating method based on Gaussian process model(GP)and frequency response function is studied.Firstly,Latin hypercube sampling is used to extract the sample set of parameters to be updated as the input of the model,the frequency response function is calculated,and the constant Q transformation is used to extract the first layer coefficient as the output of the model.The Jellyfish Search algorithm(JS)is used to optimize the parameters of GP model,and the JS-GP model is constructed.Finally,the minimum difference between the response of the test model and the predicted response of the JS-GP model is taken as the objective function,and the Jellyfish Search algorithm is used to update the structural parameters.Numerical example of space truss and experimental example of simply supported beam show that the proposed method can effectively update the structural parameters and has good updating effect.The method of updating stochastic model based on JS-BP model and JS divergence is studied.Firstly,assuming that the parameters to be updated and the characteristic quantities of the frequency response function obey normal distributions,the uncertainty model updating problem is transformed into the updating problem of the mean and standard deviation;then the Latin hypercube sampling is used to select the parameter sample set to be updated as the model input,and calculate its corresponding frequency response function to perform constant Q transformation to extract the first layer of coefficients as the model output,optimize the weights and thresholds of the BP neural network through the Jellyfish Search algorithm(JS),build the JS-BP neural network model.Finally,minimize the JS divergence as the objective function to realize the simultaneous updating of the mean and standard deviation of the parameters to be updated.The results of the space truss show that the proposed method can effectively update the mean value and standard deviation of the structural parameters,and it can get good effect when the standard deviation of test data is different.