Shape Optimization of Potentially Slender Structures

Abstract

Shape optimization lies at the heart of modern engineering design. Through shape optimization, computers can, in theory, ‘synthesize’ engineering artifacts in a fully automated fashion. However, a serious limitation today is that the evolving geometry (during optimization) may become slender, i.e., beam or plate-like. Under such circumstances, modern 3-D computational methods, such as finite element analysis (FEA), will fail miserably, and so will the shape optimization process. Indeed, the recommended method for analyzing slender artifacts is to replace them with 1-D beams/ 2-D plates, prior to discretization and computational analysis, a process referred to as geometric dimensional reduction. Unfortunately explicit geometric reduction is impractical and hard to automate during optimization since one cannot predict a priori when an artifact will become slender. In this paper, we develop an implicit dimensional reduction method where the reduction is achieved through an algebraic process. The proposed method of reduction is computationally equivalent to explicit geometric reduction for comparable computational cost. However, the proposed method can be easily automated and integrated within a shape optimization process, and standard off-the-shelf 3-D finite element packages can be used to implement the proposed methodology.