On Topology optimization with elliptical masks and honeycomb tessellation with explicit length scale constraints

TL;DR

This paper introduces a topology optimization method using elliptical masks and honeycomb tessellation with explicit length scale constraints, employing a skeletonization algorithm and a Sequence for Length Scale (SLS) methodology to generate feasible, smooth, and near-optimal designs.

Contribution

It presents a novel skeletonization algorithm for hexagonal cell topologies and a SLS methodology that simplifies imposing explicit length scale constraints in topology optimization.

Findings

Solutions generally satisfy length scale constraints

Designs are close in performance to unconstrained topologies

Method can produce volume-distributed, uniform-thickness structures

Abstract

Topology optimization using gradient search with negative and positive elliptical masks and honeycomb tessellation is presented. Through a novel skeletonization algorithm for topologies defined using filled and void hexagonal cells/elements, explicit minimum and maximum length scales are imposed on solid states in the solutions. An analytical example is presented which suggests that for a skeletonized topology, optimal solutions may not always exist for any specified volume fraction, minimum and maximum length scales, and that there may exist implicit interdependence between them. A Sequence for Length Scale (SLS) methodology is proposed wherein solutions are sought by specifying only the minimum and maximum length scales with volume fraction getting determined systematically. Through four benchmark problems in small deformation topology optimization, it is demonstrated that solutions by-and-large satisfy the length scale constraints though the latter may get violated at certain local sites. The proposed approach seems promising, noting especially that solutions, if rendered perfectly {\it black and white} with minimum length scale explicitly imposed and boundaries smoothened, are quite close in performance compared to the parent topologies. Attaining {\it volume distributed} topologies, wherein members are more or less of the same thickness, may also be possible with the proposed approach.