This article is more than 1 year old
Scientists model turbulence to boost space propulsion
Running direct numerical simulation on 'an industrial scale'
Scientists at the Technical University of Munich (TUM) say they have developed a method to run suites of simulations to better understand turbulence in fluid flows and so deliver more efficient combustion for improved space propulsion systems.
Turbulence is notoriously difficult to model yet is an important aspect of many parts of the physical world, including aircraft aerodynamics, fuel efficiency in combustion engines, and basically any system where there is an interaction between fluids (which here refers to both liquids and gases) and surfaces.
Space propulsion in particular requires high-pressure environments, which help to efficiently turn the energy generated by burning fuel into thrust for the engine. To make it more efficient, TUM has been running massive simulations focused on understanding turbulent interactions of gas in high-pressure environments.
According to the Gauss Centre for Supercomputing, direct numerical simulations (DNS) are the most accurate way to model the complex interactions involved in turbulent fluid flows. It is one of several methods that simplify the task by modeling fluid flow on a grid that breaks the system into many smaller cells that can be calculated separately.
DNS makes no assumptions for how fluids will behave in the simulation, but also requires immense computing power, such that most DNS is limited to only modeling small systems over short periods of time. To model larger systems, researchers perform large-eddy simulations, which make assumptions about how the smallest eddies behave, then extrapolate that across the whole system.
Andrej Sternin, lead researcher on the project at TUM, said his team has been running "quasi-DNS" simulations in order to more fully understand these interactions for space propulsion systems. Quasi-DNS refers to a simulation that is performed with a grid that is too coarse to capture all the detail at the smallest scales. This allows the researchers to reduce the computational costs, but still simulate the smallest eddies and their influence on the larger system, according to Sternin.
The TUM scientists have been working with computational experts at the Leibniz Supercomputing Centre (LRZ), part of the Gauss Centre for Supercomputing, and making use of the high-performance computing (HPC) resources there for their simulations.
- Saudi Arabia turns to HPE for AMD-powered Cray supercomputer
- Nvidia claims 'record performance' for Hopper MLPerf debut
- Polaris supercomputer boots up, paves way for Aurora exascale system
- Japanese space agency to put massive HPC cloud to the test
With the help of LRZ staff, Sternin and other researchers have developed a process to run suites of these simulations, which he describes as being "like an industrial process."
By running many high-accuracy DNS iterations, the team said it aims to provide insights into fluid behavior to generate data that can be used to improve the inputs for further simulations that will take more assumptions into account.
In future, the team said it also hopes to try an alternative computational method called smoothed-particle hydrodynamics (SPH) that has recently been applied to modeling turbulence. This apparently allows researchers more geometrical flexibility in their simulations by treating the system as a collection of particles.
"SPH-based multi-physics gives us a better flow-structure interaction as well as the possibility to change mesh all the time without spending a lot of CPU power on that process," Sternin said. This method allows the researchers to adapt the resolution locally, he claimed, enabling them to add more aspects of chemistry or radiation, and the like into the simulations. ®