Design Assistants in Mechanics and Dynamics are the topic of a new priority program (PP) of the German Research Foundation (DFG) that aims to bring more artificial intelligence (AI) into the design process of technical systems. Coordinated by Prof. Peter Eberhard from the Institute of Engineering and Computational Mechanics at the University of Stuttgart, around 20 groups from mechanical engineering, mathematics, and computer science from across Germany will collaborate in an innovative research project. Their goal is to use numerical strategies to optimize and accelerate the industrial design process.
“Stuttgart’s mechanical engineering is pleased with this recognition. This is also based on the work in the Cluster of Excellence ‘Data-Integrated Simulation Science (SimTech)’ at the University of Stuttgart,” says Eberhard. “The PP will combine cutting-edge optimization and AI research methods with the latest analysis methods in engineering dynamics in order to develop design assistance systems that are at the frontier between engineering, mathematics, and computer science.”
The design of technical systems is one of the most demanding creative services in engineering. Time-consuming parameter studies are carried out in often interactive development processes. Simulation results are then evaluated by eye in order to manually make promising changes to the current design based on the company’s own expert knowledge. In many cases, often because of a lack of time, a result is declared as optimal based on subjective evaluations instead of mathematical criteria. In addition, humans are usually unable to recognize correlations in higher-dimensional design spaces, and the expertise of design engineers is usually limited to a few sub-disciplines.
Highly automated design processes as a goal
In order to be able to comply with customer requirements, new regulations, and social responsibility make the most of the opportunities offered by new types of systems with components based on AI, the system design of the future must be multidisciplinary. Only highly automated design processes can do this. In these processes, numerical optimization strategies take over the search for the best compromise solutions. For system evaluation, analysis methods from different disciplines are integrated; system behavior is evaluated and design decisions are made by AI-supported algorithms.
The goal of the PP is a sustainable paradigm shift in the industrial design process that replaces the current analysis-centric process with a criteria-driven design. This should result in design assistance systems that can be combined in modules, thereby equipping engineers with an artificial intuition that complements their own specialist knowledge and allowing them to deal with structural change, shorter development cycles, and increasingly interdisciplinary challenges. Thanks to design assistants, criteria that had previously been considered only late in design processes can now be taken into account in the early phases, thereby contributing to better results and shorter development times.
Thanks to the analytical methods that have matured in recent decades, the enormous increase in computing power, and the impressive advances in research (e.g., machine learning), now is the ideal time to develop design strategies that were unthinkable only a few years ago. “Mechanics and dynamics as centuries-old links between engineering and mathematics are the ideal topics to give mechanical engineering a new research direction through this PP,” says Eberhard. “Optimization of the design process and AI research are opening up challenging new application areas with untapped potential, particularly in the use of creative intelligence in design decisions.”
The DFG approved 13 new PP, providing around €82 million over three years. The PP „Daring More Intelligence - Design Assistants in Mechanics and Dynamics“ (PP 2353) will start in 2022.
Expert Contact:
Prof. Dr. Peter Eberhard, University of Stuttgart, Institute of Engineering and Computational Mechanics, Tel: +49 (0)711/685 66388, E-Mail