Finite element model updating based on multi-outputs support vector regression

In order to overcome the shortcomings in finite element model updating by means of traditional neural network and single-output support vector regression, a new method based on multiple-outputs support regression algorithm was proposed. The parameters of support vector regression were selected according to 5-fold cross validation, the uniform design method was used to construct the samples, the static and dynamic response data were taken as the inputs and a number of design parameters as the outputs, the support vector regression machine was applied to approximate the mapping relationship between the inputs and outputs, and then the generalization ability of the support vector regression machine was utilized to get the target values of the design parameters. The result of numerical application to a space grid structure shows that: the method proposed can update a number of design parameters with high precision in the case of a small number of samples. It provides a new exploration for finite element model updating.