By designing the most effective way to inject fuel into an engine, setting up a machine in farmland water, or painting a car, humans rely on liquid spray for numerous industrial processes that enable and enrich everyday life.
To understand how to make fluid jet spray cleaner and more efficient, researchers need to focus on small things. To understand one of the great challenges of science, the turbulent movement of fluids,
Experiments are used as an important tool in understanding industrial spray processes, but there is a growing trend to rely on simulations to understand and model how to control the turbulent and turbulent movements that exist when fluids flow quickly.
Professor Markus Klein (Germany: Universität der Bundeswehr München) professor at Bundeswehr University Munich has to adopt high performance computing (HPC) to accurately and efficiently model the complexity of turbulence, which is the Leibniz Supercomputing Center (LRZ ) Have used the GCS (Supercomputing for GCS) resources to generate high-end flow simulations to better understand turbulent flow motion.
Dr. Josef Haßlberger, co-author of the Klein team, said: "Our goal is to develop simulation software that someone can apply commercially to real-world engineering problems. He works with Sebastian Ketterl on a computer project. Recently Journal of Fluid Dynamics.
It is a (multi) step.
When scientists and engineers talk about liquid spraying, there is a slightly subtle difference. Most sprays are actually polyphase, meaning that some combination of liquid, solid, and gas flows at the same time. In spraying, atomization or liquid fluids are typically disintegrated into water droplets and ligaments, forming vapor in some applications.
The researchers must explain this multistep mixing, which has enough detail to understand the fine and basic processes of turbulent motion, especially the formation of liquid droplets, bonding and cleavage, and surface tension dynamics between liquid and gas. It also captures enough space to see how these actions affect jet spray. The droplet is formed and influenced by the turbulent motion, but it affects the turbulent motion after molding, so very detailed and accurate numerical simulation is required.
When modeling fluid flow, researchers have several different ways they can be used. Among them, Direct Numerical Simulation (DNS) has the highest accuracy because it reproduces processes from "from scratch" numerically to the minimum level of turbulent motion, without starting with a physical approximation of how the fluid flows ("Kolmogorov-scale "resolution). Because of the high computational demands, DNS simulations can only be run on the world's most powerful supercomputers, such as SuperMUC from LRZ.
Large-Eddy Simulation (LES), another common approach to modeling fluid flow, assumes that the fluid will flow at its smallest size and focuses instead on simulating larger amounts of fluid over time All. However, for LES simulations to accurately model fluid flow, assumptions built into the model require DNS calculations because they depend on high-quality input data for these small families.
To simulate turbulent flow, the researchers created a three-dimensional grid with more than a billion individual small cells and solved equations for all forces acting on the fluid volume. Newton's second law causes fluid acceleration. As a result, the fluid velocity can be simulated both in space and in time. The difference between turbulence and laminar flow, or smooth flow, depends on how fast the fluid moves, not just the size of the flow structure, but also how thick or viscous it is. Then the researchers calculated the liquid properties from the moment they left the nozzle until it became a droplet, which caused the model to move.
Based on the team's DNS calculations, we began to develop new models for fine turbulence data that could be used to inform LES calculations, which ultimately helped to bring accurate jet spray simulations to a more commercial level. The LES computes the energy carrying large structures, but the smallest scale of the flow is modeled. This means that LES calculations can provide high accuracy for much less computational work.
It is flowing in the right direction.
The team has made progress in improving the LES model by understanding the fluid flow more fundamentally through DNS simulation, but there is still room for improvement. Although the team can simulate the current atomization process in detail, we would like to observe additional phenomena occurring on longer time scales such as evaporation or combustion processes.
Next-generation HPC resources help to reduce the gap between DNS and real-world lab and industrial applications in academic flows. This will provide a more realistic database for model development and provide detailed physical insight into phenomena that are difficult to experimentally observe.
The team also needs to do more to improve the LES model. The next challenge is to model droplets smaller than the actual grid size in a typical large eddy simulation, but they can still interact with turbulence and contribute to momentum exchange and evaporation.