Revisiting the origin to bridge a gap between topology and topography optimisation of fluid flow problems

TL;DR

This paper revisits the foundational assumptions of topology optimization in fluid flow, proposing a new model that allows for continuous variation in channel height, leading to more accurate and efficient designs.

Contribution

It introduces an augmented mass conservation model that accurately describes flow with variable channel heights, improving topology optimization results for fluid flow problems.

Findings

Traditional model is inaccurate for intermediate design values.

Proposed model reduces degrees-of-freedom while maintaining accuracy.

Better topological designs are achieved with the new model.

Abstract

This paper revisits the origin of topology optimisation for fluid flow problems, namely the Poiseuille-based frictional resistance term used to parametrise regions of solid and fluid. The traditional model only works for true topology optimisation, where it is used to approximate solid regions as areas with very small channel height and, thus, very high frictional resistance. It will be shown that if the channel height is allowed to vary continuously and/or the minimum channel height is relatively large and/or meaning is attributed to intermediate design field values, then the predictions of the traditional model are wrong. To remedy this problem, this work introduces an augmentation of the mass conservation equation to allow for continuously varying channel heights. The proposed planar model accurately describes fully-developed flow between two plates of varying channel height. It allows for a significant reduction in the number of degrees-of-freedom, while generally ensuring a high accuracy for low-to-moderate Reynolds numbers in the laminar regime. The accuracy and limitations of both the traditional and proposed models are explored using in-depth parametric studies. The proposed model is used to optimise the height of the fluid channel between two parallel plates and, thus, the topography of the plates for a flow distribution problem. Lastly, it is observed that the proposed model actually produces better topological designs than the traditional model when applied to the topology optimisation of a flow manifold.