Finite element model modification combining SVR response surface and PSO

A model updating method independent of model sensitivity is proposed in which the parameter of the structural finite element model can be updated with less computational effort. Surrogate model is firstly established to replace the structural finite element model, and the support vector regression can be trained to predict the response corresponding to the parameters via the calculation on a small number of sample points; Secondly, taking the residual of natural frequency as objective function, the particle swarm algorithm is used to achieve the global optimization, which leads to the updated parameters of the finite element model; Then, a plate with holes is taken as the experimental research object, and the effectiveness of the proposed method is verified based on the measurement data. The influence of different parameters and different number of sample points, etc. on the model updating accuracy are further discussed. Finally, a satellite structural model updating example is given to prove the advantages of the proposed method on computational efficiency over the method based on sensitivity analysis. This research work aims to provide technical support for the modification of structural models with complex parameter-response characteristics. © 2023 Chinese Vibration Engineering Society. All rights reserved.