Design and Additive Manufacturing of Manifolds for Navier-Stokes Flow: A Topology Optimisation Approach

Abstract

The Netherlands Aerospace Centre is increasingly improving Selective Laser Melting (SLM) capabilities. It is working with academics and industry on demonstrators to prove the current state of the art of SLM. Recently
they started with additive manufacturing (AM) of internal manifolds. This thesis aims to provide a novel design approach by including topology optimisation for Navier-Stokes flow with the focus on SLM. This approach intents to bridge the gap between academics and industry regarding fluid topology optimisation by
practically implementing the methodology. As a result, this thesis places emphasis on 3D topology optimisation instead of the very common 2D approach in academics. Fluid topology optimisation was chosen as the primary design step because it is a relatively new and unexplored method in academics and especially new in the industry. Moreover, it is a highly versatile design
approach as it can come with any shape that yields the best result for the given objective and boundary conditions. Due to the complex nature of topology optimisation and its numerical implementation, it was chosen to divide this topology optimisation approach into multiple steps. Hence the goal of this thesis is to combine AM and topology optimisation of Navier-Stokes flow and to incorporate it into a design approach.
The flow optimisation step aims to improve the performance, that is, minimising the power dissipation of the problem at hand. Besides incorporating fluid optimisation, this method also includes a weight reduction step by introducing constraints (volume and perimeter constraints), a specific AM post processing step to create self-supporting cross sections and a mass minimising structural topology optimisation step. Combining these methods lay the foundation for this unique design process for additivemanufacturing.
The volume constraint is not directly proportional to the weight of the manifold, but it is an intuitive constraint that penalises the amount of fluid domain that can be used. The second constraint that is implemented
is the perimeter constraint (including a greyness constraint). This constraint penalises the perimeter of amanifold, which is directly proportional to the weight of the manifold. The novelty of the perimeter constraint is that in fluid flow optimisation it serves as a direct mass control of the structural part of our manifold.
Including this constraint will result in rather straight manifolds as opposed to the more organic shapes arising from the volume constraint.
A three-dimensionalmanifold is used as the case study for thesis and is compared to a manifold that would be manufactured conventionally. The performance objective is power dissipation minimization, and it is shown
to reduce the power dissipation by 47% as compared to the conventional tee, almost doubling the performance.
The optimised manifolds are not always manufacturable due to the limitations of the AM process. In particular downward facing surfaces perpendicular to the build direction (overhang) cannot be printed without support. Adding sacrificial support material is not an option in manifolds, as it often cannot be removed.
Therefore, a post-processing step is implemented to adapt the geometries to remove any internal overhanging regions. This post process step results in self-supporting manifold cross sections that are shaped like a droplet i.e. a 45-degree roof on top of a circular cross section. This approach yields manufacturable design, but it comes at the cost of performance. The x, y, and z build direction adjustments decreased the power dissipation by 41.5%, 27.9%, and 38.4%, respectively. Each design adjustment reduces the performance of the
tee; nonetheless, they performsignificantly better than the conventional tee.
Subsequently, amethod to optimise the local wall thickness is applied. For this purpose, the internal pressure field can be usedwhich is found with the computational fluid dynamics (CFD) simulation . This pressure field
acts as the load condition in the structural topology optimisation. The result from this approach is a progressive wall thickness such that it is thick where pressures are high, and thin where pressures are low.
The resulting geometries were manufactured after executing the completed design approach. It is shown that this method works well and can be practically implemented. A pressure measurement test showed a difference of 2.4% to 11.8% as compared to the simulations for the different designs, which is reasonable considering various assumptions such as the frozen turbulence, the turbulence model, mesh size and wall functions. Another test is carried out to look at the flow distribution of the tee and showed a larger difference
of 2.8%-24% in flow distribution ratios. This larger difference is likely the effect of transient effects and possible internal roughness in the topology optimised manifold.