Implementation of topology optimization using openMDAO

Abstract

Topology optimization is one of the widely known branches among the structural optimization, and it distinguishes itself being able to generate extremely lightweight structures. Recently it has drawn particular interest from both industry and academia because of its natural applicability to additive manufacturing. However, its implementation is often a daunting task for engineers in practice. In particular there can potentially be a large programming effort required to modify the method, even from subtle tweaks in the problem definition or solution algorithm. Changes to the code can leave to corresponding updates to relevant derivative calculations, further compounding the problem. Implementation, therefore, is not only time-consuming but also repetitive and susceptible to human-induced errors. In this regard, topology optimization implementations stand to benefit from changes that result in more code modularity, ease of restructuring, and more automated derivative calculations. In this work we propose using OpenMDAO, a computational framework for multidisciplinary design optimization, as a generic platform for to built topology optimization implementations with in order to achieve these implementation improvements. Two widely used topology optimization techniques—density-based and level-set—are implemented as to serve as reference code designs. These techniques are implemented in a decomposed manner, with the aid of the modular architecture of OpenMDAO as well as state-of-the-art numerical methods. To demonstrate the flexibility of the new topology optimization architecture, two variations on the density-based topology optimization approach are shown.