Title:	Large-scale Simulations in Complex Flows
DNr:	NAISS 2024/2-16
Project Type:	NAISS Large Storage
Principal Investigator:	Outi Tammisola <outi@mech.kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2024-07-01 – 2025-07-01
Classification:	20306 10105
Homepage:	http://www.mech.kth.se/
Keywords:

Abstract

We present a large-level request for storage allocation on high-performance computing (HPC) resources within NAISS. In particular, we request for 2,000,000 GiB storage on Klemming PDC (200 million files), 400,000 GiB on Centre Storage NSC (15 million files) and 300 TiB on dCache. We request more as we frequently run out of storage especially on Klemming, despite our efforts. The research is conducted by groups in Micro and Complex Flows, at FLOW, Dept. of Engineering Mechanics, KTH Mechanics. The group, with Assoc. Prof. Tammisola as PI and Profs. Brandt, Bagheri, Dahlkild, Lundell, Mihaescu and Prahl Wittberg as Co-Investigators consists of a total of 53 senior researchers, Postdocs and PhD students, at least 12 to be hired during the year. There are currently 24 ongoing research projects (see proposal for computational time) that rely entirely on compute and storage allocations of NAISS, grouped in six focal areas as below: Complex fluids - Non-Newtonian flows in laminar and turbulent regimes and with a microstructure (particles, fibers, bubbles, droplets). Flow at interfaces - Over porous, elastic, poro-elastic and micro-structured interfaces, and wetting. Bio-physical flows - Within the cardiovascular and respiratory systems, and cell transport. Multiphase and free-surface flows with phase change - Complex physical phenomena such as phase change, absorption/desorption, buoyancy-induced convection to study bubble/droplet spreading, breakup during evaporation and boiling process. Unsteady flows for clean vehicles - Compressible flows, with heat- and mass-transfer associated with energy conversion and propulsion systems Machine learning - Improved modelling, prediction, postprocessing and characterisation of complex and multiphase flows by a creative application of machine learning methods. In the associated activity report, we show that resources previously allocated were used efficiently and led to a significant number of relevant scientific publications. In the 2013 VR evaluation of the area of Engineering Mechanics, the fluid mechanics research at KTH was defined outstanding (the only group in Sweden), with findings considered as milestones in our research field. This consortium has been awarded a number of prestigious grants. Profs. Tammisola, Bagheri, Prahl Wittberg and Brandt have been awarded recent ERC grants whereof 3 are ongoing, Bagheri chosen the SSF Future research leader and Wallenberg Academy Fellow in 2017. We are a large part of the environment for multiscale modelling, INTERFACE, funded by VR (24 MSEK). Many EU and VR projects have been awarded to team members and consequently, our activities are now expanding significantly in new directions with many new people about to start. Therefore, the need for storage is also growing. Given the large amounts of data produced in the simulations of multi-physics phenomena as those proposed here, post-processing becomes more demanding and data need to be stored for longer times. Also for the new area of machine learning, data fields need to be stored often as training cannot always be performed on the fly. To exploit novel simulations and methods, it is therefore essential to keep the data for a time sufficient for a thorough analysis.