Title:	Topology Optimization considering high-resolution 3D Multiphysics Fluid-Thermal-Structural models
DNr:	SNIC 2020/13-47
Project Type:	SNIC Small Compute
Principal Investigator:	Jonas Lundgren <jonas.lundgren@liu.se>
Affiliation:	Linköpings universitet
Duration:	2020-06-03 – 2021-09-01
Classification:	20301
Keywords:

Abstract

Topology Optimization (TO) is a powerful design tool that can be used to facilitate generation of design concepts by an efficient exploration of the design space. TO becomes especially powerful in combination with Additive Manufacturing (AM), a manufacturing technique on the rise, which enables very complex geometries to be realized. To fully exploit the natural combination between TO and AM, there is a need for high-resolution models in 3D. One possible application is the design of hot components in gas turbines, where the material is exposed to temperatures of around 1000 degrees Celsius. An optimal layout of the cooling channels inside these components could mean a higher efficiency of the turbine, but more interesting, it also means that renewable energy sources, such as hydrogen, could be used to fuel the gas turbines. This possibility is introduced due to that the coolant in the most optimal way keeps the temperature below the critical point, because of the optimal design. For this to be a competitive tool, it has to be able to take care of conflicting interests from different physical domains. For example, the design for optimal cooling of a structure might not be compatible with an optimal design for fatigue properties or the layout for a minimal pressure drop across the domain, which is highly linked with the performance of the turbine. Because of this, a simultaneous approach is needed, where these conflicting goals are evaluated and accounted for continuously during the optimization process. This kind of model is believed to have a very high potential to contribute in the design of more efficient and environmental friendly turbine components, as well as other similar applications. But for this to happen, there is a need for high performance computing resources to be able to simulate and compute these optimal shapes efficiently. The computation time grows rapidly with the size of the models, and to achieve a resolution that utilizes the precision of todays AM printers, desktop computers are not enough. Dedicated resources at a supercomputing center would be essential for this project. This would mean that customized, highly parallell codes could be tested out and evaluated substantially faster than today.