Kolloquium WiSe 2024/25

Prof. Dr. Philipp Neumann
Universität Hamburg, DESY
Wann: Mo, 18.11.2024, 17:15 Uhr (dieses Semester geänderte Zeit!)
Wo: Konrad-Zuse-Hörsaal (Raum B-201)
Thema
The Fast and the Curious: Molecular and Multiscale Flow Simulation in the Era of Data Science and Exascale Computing
Abstract
High performance computing (HPC) enables researchers to carry out compute- and data-intensive tasks, such as training of machine learning models or numerical simulations at unprecedented problem sizes and computational speed.
Molecular-continuum simulations in fluid dynamics, as subject of my talk, couple computational fluid dynamics (CFD) solvers and molecular dynamics (MD) simulations in a domain decomposition sense. This allows to invest into computationally intensive MD in small-sized local spots, where the molecular behavior requires to be resolved, and to rely on computationally cheap CFD everywhere else. Typically, the MD solver consumes most of the computational time in molecular-continuum simulations.
Although this multiscale approach itself renders respective flow simulations significantly cheaper compared to stand-alone MD systems, it can easily still require massive amounts of computational resources. Particular challenges arise from rapidly evolving exascale-enabling hardware technology and the vast amount of computational resources in exascale systems. Besides, machine learning and data science methods have evolved as additional scientific research paradigm, extending the computational approach of numerical simulations.
After introducing the field of HPC, I will discuss automated algorithm selection (auto-tuning) for MD algorithms at node level which allows to always select the fastest MD solver algorithm, independent from the underlying simulation problem or hardware. This feeds into efficient molecular-continuum simulations, which, leveraging efficient parallelization strategies, are executed on large-scale supercomputers.
I will further comment on different ways how to make use of the evolving data and machine learning approaches for molecular-continuum systems, such as modeling of artificial boundary forces, noise filtering or learning molecular flow behavior.
All developments have been implemented in mature software packages -- the MD software ls1 mardyn and the macro-micro-coupling tool MaMiCo for molecular-continuum simulations.
Bio
Philipp Neumann graduated in Engineering Mathematics (Technomathematik) from FAU Erlangen-Nürnberg in 2008. He moved to Technische Universität München afterwards where he received his PhD in 2013. In his postdoc phase, he took over coordination tasks for the DFG priority program "Software for Exascale Computing" (SPPEXA) and, after moving to the German Climate Computing Center/Universität Hamburg in 2016/2017, for the "Centre of Excellence in Simulation of Weather and Climate in Europe" (ESiWACE). In 2019, Philipp finished his habilitation and was appointed full professor for high performance computing at Helmut-Schmidt-Universität (HSU) Hamburg. Since May 2024, he is full professor at Universität Hamburg and lead scientist as well as head of the IT-department at DESY.