Abstract
In structural optimization, the computational cost increases rapidly as the number of design variables increases. In this paper, the multi-level optimization is proposed, which decreases computational cost. Moreover, the multi-level optimization is more natural way of carrying out optimal structural design in the distributed design practice in which different engineers design different parts of the structures.
The optimal designs of ship structure that minimize weight subjected to strength constraints are derived. The equivalent thicknesses of the stiffened plates are taken as global design variables that are used in overall optimization, and the detail sizes of stiffened plates are taken as local design variables that appear in single subsystem. The transformation of the constraints into objective function is carried out to include local design variables in the objective function, which is necessary for multi-level optimization.
By comparing the single level optimal design and multi level optimal design, it was shown that in the multi level optimization the number of iteration to reach optimal design decreases dramatically, and the computational cost also becomes quite small.
