Shape optimization is very important for many industrial applications. The
typical problem is to find the optimal shape of a machine such that it minimizes
a certain cost functional while satisfying given constraints.
The practical problems in this thesis are given by the ACCM (Austrian
Center of Competence in Mechatronics). The task is to determine the design
of an electric drive such that a cost functional is minimized. Due to the
fact that the analytic definition of the cost functional is not available but
all function values can be computed, this thesis considers optimization by
black box simulations.
The model problem in this thesis is firstly considered with two design
parameters and it is described in detail. The mathematical background
for the optimization is presented. It is mainly based on the book
Practical Optimization
by Philip E. Gill, Walter Murray and Margaret H. Wright.
The optimization programs are written in Matlab.
This thesis presents different numerical results by using a certain
Matlab routine. Moreover, results are
shown with the availability of parallel processors and it is presented how this
would reduce the number of function evaluations in each optimization step.
A proposal is made at the end of the consideration of the problem with two
design parameters, how the whole procedure of optimizing the motor via
black box simulations can be automatized.
Finally, the gained knowledge is applied on real application problems
with more design parameters given by the ACCM and the results of the
optimization are presented. These results are compared to the results of a
different optimization method, a genetic algorithm.
