A further increase in the performance of supercomputers is expected over the next few years. So-called exascale computers will then be able to deliver more accurate simulation results. Fraunhofer SCAI is developing efficient data analysis methods for the much larger amounts of data generated in this way, which will also provide the engineer with detailed insights into the complex technical interrelationships.
Simulations on supercomputers answer important industrial questions, such as how air flows behave in air conditioning systems, on rotor blades or entire aircraft. New approaches to hardware design will make computers with a capacity of one trillion computing operations per second available. So-called exascale computers will have a computing speed a thousand times higher than previous supercomputers and can therefore cope with complex problems that are currently considered unsolvable. In addition, they will deliver more accurate results and thus more economical solutions for industrial tasks. However, exascale simulations also generate considerably larger amounts of data. This creates new challenges in data analysis.
The scientists of the Fraunhofer Institute for Algorithms and Scientific Computing SCAI have addressed this problem. They are developing methods that perform data analysis already during simulation. This minimizes memory requirements and computing time. In this way, even huge amounts of data can be analyzed efficiently. The SCAI scientists concentrate on methods for the comparative analysis of simulation data. By automatically comparing the data, engineers can see at a glance how, for example, a modified component shape affects the flow behavior and thus the volume of an air-conditioning system. This helps in the development of new products.
Fraunhofer SCAI works together with 13 partners from seven European countries in the newly founded "Competence Center of the European Union for Engineering Applications" EXCELLERAT. The aim is to make the best possible use of the new possibilities offered by supercomputing and thus strengthen Europe's competitiveness in this field. Over the next three years, the members of the competence center will deal with six selected program codes and associated workflows, such as data analysis, data management and data visualization. In addition to the efficient handling of large amounts of data, the project focuses on the scalability of the codes. Scalability here means that the programs work efficiently on both laptops and exascale computers. Since EXCELLERAT will also be used to expand the use of supercomputing in industrial environments, the participants will optimize the codes on the basis of twelve application examples from industry. These include the simulation of flows on rotor blades and in car air-conditioning systems as well as the calculation of combustion chamber emissions. In addition, a common knowledge platform is to be created within the framework of the cooperation, which will make existing knowledge more accessible.
Project duration: 12/2018 – 02/2022