Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources.
Technological advancements in materials, architecture, and computer-aided design (CAD) have helped ensure that each new generation of buildings are safer than prior generations. However, intense wind gusts can still cause chaos, exposing flaws in buildings by stressing certain areas, ultimately ripping off roofs of buildings or causing other structural damage. Lightweight, quick-to-build structures are especially at risk.
Researchers investigating how turbulent wind gusts impact building integrity can design experiments to validate their predictions, constructing model buildings for wind tunnel tests and subjecting them to different wind loads. In recent years, though, high-performance computing (HPC) has allowed scientists to design virtual buildings in high-fidelity simulations to better understand fluid-structure interactions (FSI) between strong winds or wind gusts and buildings.
Recently, researchers from the Helmut-Schmidt University in Hamburg (HSU) have been using a combination of experimental investigations in the HSU lab and numerical simulations on Gauss Centre for Supercomputing (GCS) resources in an effort to design safer buildings.
“In the last few years, there have been a lot of structures built with lightweight materials based on membranous structures, such as stadiums that want to implement covered areas for fans while still keeping parts open,” said Prof. Dr.-Ing. Michael Breuer, principal investigator of the project. “If you have heavy, massive structures, the interaction between the fluid flow, such as wind, and the structure is not so critical in most cases. However, thin, lightweight structures can be easily deformed and ultimately destroyed by the wind or especially wind gusts, and we want to address this issue in our project.”
To that end, the team used the SuperMUC supercomputer at the Leibniz Supercomputing Centre (LRZ), one of the three centres making up GCS, in order to study the complex dynamics between turbulent flows and flexible structures.
Researchers across a variety of scientific disciplines struggle to accurately model one of the last unsolved mysteries of physics—the chaotic motion of turbulent fluid flows are notoriously difficult to describe in a realistic way.
In an effort to setup realistic simulations that are still capable of running on supercomputers, many researchers studying turbulence use large-eddy simulations (LES), which allow realistic predictions of turbulence to a certain level of accuracy. Turbulent fluid motions are comprised of a spectrum of “eddies,” which are essentially swirling fluids that create currents moving in different directions. These eddies cascade down to extremely small scales, where they eventually dissipate into heat. LES predictions focus primarily on the larger eddies that are dominating turbulence in a fluid flow. Nevertheless, the effect of the smallest eddies has to be taken into account by appropriate models.
“If you try to simulate turbulence on HPC systems, you are already running into challenges to understand it down to the smallest scales,” Breuer said. “If you are trying to take the interaction with a flexible structure into account, it generates even more length and time scales you have to account for in your simulation. All of this together adds to the complexity and time needed as well as the need for large computing resources.”
The computational challenge lies in the multi-physics and multi-scale nature of the underlying physical problem—that is, researchers must design a model capable of both simulating very small turbulent motions happening very quickly while also taking large length and time scales into account. Furthermore, the simulations have to run over a long enough period of time to understand the impact the strongly varying fluid flow can have on the building as a whole.
To address the full complexity, the researchers in Hamburg used a step-by-step approach to increasing complexity and detail in their simulations. First, they developed a high-fidelity, three-dimensional computational fluid dynamics (CFD) simulation methodology and coupled that to a finite-element solver for the structural part within a partitioned, but nevertheless strongly coupled solution approach. This tool was extensively validated based on specific FSI test cases, which were experimentally investigated in the wind and water tunnel at the lab. The team carried out these coupled FSI simulations for turbulent flows, but did not take wind gusts into account.
Second, the researchers performed flow simulations for the turbulent flow around static structures that took wind gusts into account in order to understand their effect on the fluid flow and the rigid structure.
The path towards structural safety solutions
In the near future, the team aims to incorporate the added challenge of modeling turbulent wind gusts in the context of high-resolution coupled FSI simulations. That will decisively advance the capabilities to better understand the fluid-structure interaction between turbulent wind loads and flexible structures in order to support the design process of such modern civil engineering buildings.
The team looks forward to using next-generation HPC resources, such as LRZ’s SuperMUC-NG (inaugurated this year, SuperMUC-NG provides a six-fold increase in peak performance over the prior generation SuperMUC machine) in order to run FSI simulations with higher resolutions that can better resolve the complex physical phenomena relevant for the interaction between the fluid flow and the structure.
“By modeling these lightweight structures before building them, we have a better understanding what the influence of different loads on structures can be, which helps to decide how thick the membrane of the stadium roof or awning needs to be, for example,” Breuer said. “This ultimately results in designing safer buildings, and we are trying to contribute to this process.”