Reduction of calibration effort in FEM-based optimization via numerical and experimental data fusion

Abstract

In this paper a fusion metamodeling approach is suggested as a method for reducing the experimental and computational effort generally required for calibrating the parameters of FEM simulations models. The metamodel is used inside an optimization routine for linking data coming from two different sources: simulations and experiments. The method is applied to a real problem: the optimal design of a metal foam filled tube to be used as an anti-intrusion bar in vehicles. The model is hierarchical, in the sense that one set of data (the experiments) is considered to be more reliable and it is labeled as “high-fidelity” and the other set (the simulations) is labeled as “low-fidelity”. In the proposed approach, Gaussian models are used to describe results of computer experiments because they are flexible and they can easily interpolate data coming from deterministic simulations. Since the results of experiments are obviously fully accurate, but aleatory, a second stage (“linkage”) model is used, which adjusts the prediction provided by the first model to more accurately represent the real experimental data. In the paper, the modeling and prediction ability of the method is first demonstrated and explained by means of artificially generated data and then applied to the optimization of foam filled tubular structures. The fusion metamodel yields comparable predictions (and optimal solution) if built over calibrated simulations vs. non-calibrated FEM models.