Actually, we’re always faced with the same task in the field of simulation research. We develop or optimize procedures that enable us to solve new problems by using as few and as simultaneously executable arithmetical operations as possible. The motivation always comes from the application.
Prof. Miriam Mehl
Simulations currently complement or replace complex and expensive experiments in many areas, and optimize and accelerate processes. Or else they simply make things possible that would otherwise be inconceivable without supercomputers such as those at the High Performance Computing Center, Stuttgart (HLRS).
During the interview, Mehl, an expert in numerics and supercomputing cited examples from very different areas. One of the reasons for this is because she likes to leave her own professional comfort zone every now and again. “The really exciting thing is that we learn more from every new field of application”, says the 45-year-old, who, until October 2019, was Vice Dean at the Faculty of Computer Science, Electrical Engineering and Information Technology. Even when she was still at school, the daughter of a physicist knew that she would go into mathematics or science. Yet, having taken her Diplom in mathematics at the Technical University of Munich (TUM), Mehl decided to do a doctorate in computer science. the deciding factor having been some fascinating application examples, which she saw in a summer school on simulations in South Tyrol. As a deputy professor, she made another detour into mathematics at the TUM before finally becoming professor for the simulation of large systems at the University of Stuttgart in 2013.
Determining the starting points of tumors
Since then, she has been involved in a large number of projects simultaneously, working at an international level and in interdisciplinary settings, among other things as head of a project network at the Data Integrated Simulation Science (SimTech) Cluster of Excellence. Together with scientists from the University of Texas in Austin, her group has, for example, developed a method that enables us to determine the starting point of a brain tumor on the basis of a single MRI image, as well as the parameters that determine its growth, and all within a matter of seconds. Until now, such a review of the developmental history of a tumor has not been possible, but could provide valuable diagnostic and therapeutic information.
Miriam Mehl is head of research at the “SimTech” Cluster of Excellence in Stuttgart.
“All” that is required in this context are two systems based on complicated differential equations which have to be made to communicate with one another. “One for the real MRI image, the other for mapping a statistically healthy brain onto that of the specific patient”, Mehl explains. In contrast, more systems come into play in other applications. For example, optimizing the design of wind turbines, requires the calculation and coupling together as simultaneously as possible of three phenomena: the flow, structural deformations due to forces from the flow, and the resulting acoustics. Not only do these systems differ per se and have interactions in all directions, they also require different computational periods – which presents a challenge in terms of the efficient use of parallel computers. The field of geophysics provides further examples of such interactions when, for example, evaporation processes in porous rocks are calculated – as in the Collaborative Research Center (CRC) 1313: "Interface-Driven Multi-Field Processes in Porous Media – Flow, Transport and Deformation". In this context, it is necessary to harmonize the processes that take place within the cavities of such structures with the air flows on their surfaces.
Complex programming of the super computers
Not only do these examples demonstrate the range of possible applications, but also the increasingly complex requirements in the so-called field of “multi-physics”, in which simulations only work if one proceeds iteratively rather than linearly, i.e., if one approaches as accurate a result as possible in an step by step manner. This involves combining programs, causing completely different and unpredictable computing costs. However, the simulation as a whole should still use tens of thousands of computing cores within a supercomputer in as balanced a manner as possible. Obtaining additional information, for example about uncertainties within the results, requires thousands of simulations. In addition, programming supercomputers is becoming increasingly complicated because of the parallel calculations and the use of heterogeneous, i.e., different, computing components.
The really exciting thing is that we learn more from every new field of application.
Prof. Miriam Mehl
Sustainability in simulation science
However, in addition to such technical challenges, Mehl is also facing new overriding tasks, including the coordination and communications within heterogeneous research networks. She also wants to develop simulation programs that are not only suitable for a single doctoral thesis, but can be extended and used more widely for future, as yet unknown applications. In view of the enormous resources consumed by super computers, sometimes several megawatts which is as much energy as consumed by a small town, Mehl would also like to contribute towards increased sustainability in simulation science using new mathematical methods rather just by simulations.
So, the mother of two is unlikely to get bored in the future. She keeps herself fit for this multitasking between her family, teaching, research, supervision of doctoral candidates and committee work with lots of sport and is already looking forward to the next project: “I’m always happy when someone walks in the door with some new subject up their sleeve.”
Text: Jutta Witte
This text was published in the magazine “forschung leben”.