Concept FE models of the vehicle structure are often used to optimize it in terms of static and dynamic stiffness, as they are parametric and computationally inexpensive. On the other hand they introduce modeling errors with respect to their detailed FE equivalents due to the simplifications made. Even worse, the link between the concept and the detailed FE model can be sometimes lost after optimization.
The aim of this paper is to present and validate an alternative optimization approach that uses the detailed FE model of the vehicle body-in-white instead of its concept representation. Structural modifications of this model were applied in two different ways — by local joint modifications and by using mesh morphing techniques. The first choice was motivated by the strong influence of the structural joints on the global vehicle performance. For this type of modification the plate thicknesses of the most influent car body joints were changed. In the second case the overall car dimensions were modified.
The drawback of using detailed FE models of the vehicle body is that they can be times bigger than their concept counterparts and can thus require considerably more time for structural analysis. To make the approach proposed in this work a feasible alternative for optimization in the concept phase response surface models were introduced. With them the global static and dynamic performance of the body-in-white was represented by means of approximating polynomials. Optimization on such mathematical models is fast, so the choice of the optimization algorithm is not limited only among local-search strategies.
In the current study Genetic Algorithm was used to increase the chances for finding better design alternatives. Two different optimization problems were defined and solved. Their final solutions were presented and compared in terms of structural modifications and resulting responses. The approach in this paper can be successfully used in the concept phase as it is fast and reliable and at the same time it avoids the problems typical for concept models.