Gradient-based approximate design optimization

Gradient-based approximate design optimization

Vervenne, K.

Van Keulen, F. (promotor)

Aerospace Engineering

2005-11-01

The research presented in this thesis deals with gradient-enhanced approximate design optimization. This research has been carried out as part of the ADOPT project (Approximate Design OPTimization), a joint STW project with the Eindhoven University of Technology. In structural optimization, sensitivity analysis is often performed by means of the semi-analytical method. In the present thesis it has been shown that the accuracy of the design sensitivities can be improved by using Laplacian smoothing instead of a boundary node approach. In many applications of structural optimization, sensitivity information is available at low computational cost. Various methods to include this information in the response surface method have been investigated. It has been shown that including gradient information in the response surface may be considered as a multi-objective optimization problem, in which a trade-off between accuracy in function values and gradients has to be made explicitly. In addition, the use of fast reanalysis is considered in order to reduce the computational time required to carry out the function evaluations. As an application, the optimization of composite laminates has been considered. These types of problems lead to a non-smooth response. It has been investigated whether an optimizer based on a sequential approximate method can be used to solve these types of problems. The proposed approach has been applied to the optimization of composite tubes.

gradient
response surface
design optimization

http://resolver.tudelft.nl/uuid:e592b6f1-7f0f-499b-a985-2705a7a2ecba

Delft University Press

90-407-2608-6

Institutional Repository

doctoral thesis

(c) 2005 K. Vervenne