AI transforms simulation processes

What is already common practice for large companies is often still a major challenge for small and medium-sized enterprises (SMEs) because: There is often a lack of personnel with the necessary know-how or access to sufficiently powerful computers. The company Falquez, Pantle and Pritz from Karlsruhe shows how great success can be achieved with its nuberisim2ai project. With the support of the Smart Data Innovation Lab (SDIL), the small engineering firm managed to significantly increase the efficiency of 3D simulations and the resulting noise developments with the help of AI.

But right from the start: As a specialist in computational fluid dynamics (CFD) and computational aeroacoustics (CAA), Falquez, Pantle and Pritz supports and advises companies in the product development of flow-carrying devices - such as turbines, vehicles or pipelines. CAA is a technology that is particularly time and energy intensive. High-resolution and in 3D, flow-acoustic simulation technology typically takes up to 14 days, despite parallelization on high-performance computers. The huge amounts of data and their complexity push even modern computers and special AI hardware (such as powerful graphics processors, GPUs) to their limits. Research is also predominantly "only" concerned with "smaller", two-dimensional cases.

The aim of Falquez, Pantle and Pritz was therefore to pioneer 3D simulations and shorten the simulation time there. AI jumped out at them as a promising technology, but as a small company, the engineering firm had neither the resources nor the knowledge to use it for the application and therefore applied for a tender from SDIL.

Research meets practice: the brains behind the project

The SDIL was founded in 2014 with funding from the Federal Ministry of Education and Research (BMBF) as a "data clean room" at the Karlsruhe Institute of Technology (KIT). The SDIL's practice partners from research and industry offer companies access to modern AI methodology and infrastructure, helping them to drive innovation faster. On the research side, the KIT, the German Research Center for Artificial Intelligence (DFKI), the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS and Forschungszentrum Jülich are available to companies as sparring partners for projects. On the industry side, IBM Germany, SAP and Software AG are also available as providers, while Sicos BW GmbH acts as a link between research and industry. For selected projects, the SDIL regularly advertised so-called micro-projects, meaning that companies could apply for these and benefit from collaboration free of charge. In the case of Falquez, Pantle and Pritz, the KIT provided support with experts as well as physical and AI-based models to update the time series of a 3D flow simulation in high quality.

Reduce computing time with AI forecasts

Time series of the 3D simulation of a generic vacuum nozzle of varying quality served as the use case. In the course of the project, it ultimately became clear that an AI has different requirements for simulation data than engineers. Together with the KIT, Falquez, Pantle and Pritz therefore developed a strategy to prepare the data from simulations in such a way that it can be better used by the AI.