Extraordinariat - Current Research Topics

Current research focuses on creating advanced computer-aided models, methods and tools for product design and fabrication. This combines the areas of product development and design, fabrication, computer science and artificial intelligence. Emphasis is placed on supporting early design stages where there is less available technology support and the issues are more difficult and scientifically interesting.
Research is mostly developed generically and then applied to domains as opportunities for projects, both scientific and industrial, arise. Current applications are in the areas of automotive, e.g. hybrid powertrains, aerospace, e.g. cabin configuration, cognitive consumer products, e.g. a coffee service robot, manufacturing, e.g. CAD/CAPP/CAM and robotics.

Current research is split into five, inter-related areas:

• Model-based engineering and design libraries aims to create a new, formal approach to supporting concept design of complex, multi-disciplinary products. We are developing an integrated, formal product model for mechatronic systems, using SysML, to create a basis for generic, model-based design libraries. These include building blocks at the function, behavior and structure, or component, abstraction levels as well as formally defining their valid interconnections. Designers can use model-based libraries in concept stages to rapidly model designs and find alternative, feasible solutions as well as better understand relationships, e.g. among disciplines. Such models can be made “live” through integration of parametric modeling and solving as well as multi-engineering simulation. The libraries will support customization and extension so that they grow with use. This research is currently in an initial stage and receiving great interest from the academic and industrial community.

  Related projects: SFB 768 subproject B2, Computational Synthesis of Mechatronic Systems

• Computational design synthesis aims to support designers to discover new, creative solutions and better understand solution spaces through automatic generation of design alternatives. We are developing new methods and open-source software platforms for engineering design grammars, one using graph grammars for generative concept design of function-behavior-structure models, linked to the model-based libraries developed above, and the second using spatial grammars for generative shape design. One new area of research is using the graph grammar platform also to robustly support automated model transformations from concept models to simulation, e.g. Modelica, and CAD, to provide quantitative feedback during design generation. Key future areas include automated learning of engineering design grammar rules, robust integration of constraint solving and evaluation, e.g. FEM and multi-engineering simulation, and the integration of generalized optimization and search methods; see topic three below.

  Related projects: Design Automation Software for Gearboxes, SFB 768 subproject B2, Computational Synthesis of Mechatronic Systems

• Engineering design optimization extends on the research in computational design synthesis to be able to search vast spaces of solutions and find satisfactory or optimized designs considering multi-disciplinary objectives and constraints. Searching these spaces is difficult due to the strong coupling between the changing topology of a design, i.e. its configuration, and the parameters e.g. spatial layout and component sizes. Previous research involved developing tailored algorithms for specific application areas, e.g. structural topology optimization, building façade optimization or gearbox layout and optimization. We are now investigating new, more general multi-objective optimization and search methods for synthesis tasks that are robust, i.e. to enable a wide range of possible applications, and efficient, i.e. to integrate more lengthy simulation and user feedback. A further important topic in this area is improving support for optimization modeling and method selection to bridge the gap between research in optimization and effective industry use.

  Related projects: Design Automation Software for Gearboxes, Computational Synthesis of Mechatronic Systems

• Cognitive fabrication investigates how we can transfer understanding of cognitive capabilities in humans to technical systems, specifically focused on autonomous design-to-fabrication. Extending on the automatic design methods described above, this research addresses how to automatically fabricate customized, 3D parts. The aim is to create a fabrication machine that knows its own capabilities, can reason about these to plan actions, can re-design a part to meet current capabilities or provide feedback, can re-configure itself, can react to unexpected events and can learn new skills and improve performance. This research closes the loop between product design and fabrication. Through combination of all research topics, a long-term research goal is to create a machine that acts more like a human designer to autonomously create and fabricate a design, given basic requirements alone.

  Related project: Towards the Cognitive Machine Shop (CogMaSh)

• Cognitive products are the next-generation of mechatronic and smart products. They are defined as tangible and durable things with cognitive capabilities that consist of a physical carrier system with embodied mechanics, electronics, microprocessors and software. The surplus value is created through cognitive capabilities, which include reasoning from knowledge models, learning and planning, enabled by flexible control loops and cognitive algorithms. Customer needs are satisfied through the intelligent, flexible and robust behavior of cognitive products that meet and exceed user needs and desires. Research in this area links strongly with the topic model-based engineering and design libraries and currently focuses on creating a new functional taxonomy and integrated modeling approach for cognitive products to assist such complex interdisciplinary design involving mechanical, electrical and software disciplines. This topic is further explored through the interdisciplinary, project-based class, Innovation@CoTeSys, which has been run five times, twice in collaboration with industry.

  Related project: Innovation@CoTeSys