Abstract:
In this paper, an application of surrogate modeling for the analysis of structural problems is presented. The case study of this paper, a tension joint, is analyzed using a parametric non-linear Finite Element (FE) model in order to gather data for the generation of the surrogate model. The data points are defined using a design of experiments based on Optimal Latin Hypercube Sampling and the surrogate model is generated using a kriging function. The resulting surrogate model is validated using FE results from a second design of experiments and by comparison with an analytical approach. The comparison between the kriging method and the analytical approach shows respectively an average error of 1.2% and 7.6% over the entire design space. The maximum error is respectively of 4.0% and 20.1%. This approach shows great promise for structural analysis as it can greatly reduce computational cost for the analysis of recurring problems. It is also easy to implement and can be used for complex structural problems.