An approximation-concepts approach to shape optimal design (Q1057119)

In the past, much of the work done in structural optimization consisted in resizing the members of fixed configuration models. In that case, a powerful design procedure is now emerged, which is based on the coordinate use of explicit high-quality approximations of the behavior constraints and dual methods of mathematical programming. There is, however, a large class of problems for which the main degrees of freedom for the designer correspond to the shape of the structure itself. The main objectives of this paper are to recall briefly a convenient geometric representation, in which the boundaries of the structure are represented by Bezier or B-spline curves, and then to discuss the choice of optimization algorithm. It is shown that cost efficient methods for structural sizing may be advantageously extended to shape optimal design problems. Different approximation schemes are tested and a new general optimization algorithm is presented that combines mixed approximations and dual methods. Many large-scale applications are treated to demonstrate the generality and the efficiency of the new formulation. Finally, considerations are given about an integrated approach including CAD computer codes and finite element optimization software.