Three-dimensional shape optimization using fully automatic mesh generation

Abstract

Introduction T HE technique for associating design variables with mesh data is the most crucial factor in three-dimensional shape optimization. Previously, work in three-dimensional shape optimization involved specifying design variables by associating parameters directly with grid points on an existing mesh. For realistic problems this can be a very tedious (and errorprone) process. In the past, shape optimization capabilities have been developed based on a variety of design/analysis capabilities ranging from associating parameters with a mesh created manually to associating parameters with control points of a mapped mesh generator. Special techniques have also been developed to properly move internal grid points during sensitivity calculations. More recently, a capability based on constructive solid geometry (CSG) has been developed, but CSG representations are not particularly suitable for design optimization. A new optimization approach is demonstrated in this Note that uses a new, fully automatic mesh generation capability. The design model is developed based on design-oriented geometric primitives that represent recognizable features of a part and can be assembled into complete solid models that are defined in terms of a small set of design parameters.