Structural Optimization Using Approximations Based on Simplified Numerical Models

Abstract

The paper investigates the effectiveness of approximations that use simplified (low-fidelity) and detailed (high-fidelity) numerical models. Three types of such approximations are considered. The simplified models are used in the formulation of the approximation functions, which are to be fit in the numerical results obtained with the high-fidelity models. The approximations belong to a mid-range class and have been implemented in the MARS (Multipoint Approximations based on response Surface fitting) method, which is briefly described in the paper. Some numerical results of application of these approximations to optimal design of a railway track are given. The approximations have been used for stress constraints. The high-fidelity model is the 3-D finite element model of the railway structure while the low-fidelity model has been obtained by simplifying the finite element model (2-D FE model). The results are presented and discussed.