Research

Distributed Intelligent Design Systems

Finding design solutions in the sense of ideas is a central step in a design process. It's often mentioned in the literature of industrial designs that new designs include results from working steps similar to those of previous design projects. However, reuse of such design knowledge is normally bound to the size and knowledge culture of an organization. And in particular industrial design solutions that are customized to the demands of a customer's business (e.g. office designs, IT-solutions) show a special need for innovative design ideas, which have not been applied in previous projects. To support the reuse of design knowledge via information systems a formal knowledge representation is needed that can be used by design experts of an organization and design services.

Project RESPOND is dedicated to this field of research. From the view of informatics RESPOND's main aim is to design, develop, and evaluate following elements of an Distributed Intelligent Design System (DIDS):

Formal representations of design knowledge with practical use
Efficient inference mechanisms for design knowledge
Digital design spaces for interactive construction of design objects
Service-oriented architecture for design spaces
Functional evaluation services for formal representations of design objects

Technologies

For describing design solutions we use semantic representations, which can be used by means of logic and probabilistic inference mechanisms to suggest probable solutions, and to structure the design process. The required design knowledge is defined in form of semantic statements, which are connected to probability tables. The semantic representation of design knowledge is web-based (RDF), and can be be developed collaboratively via the decentral database management system S3DB (Almeida et al., 2010). To support a current design project the formalized design knowledge is reused for generating probabilistic FBS-BN representations of current design situations (Eichhoff & Maass, 2011). Explicit and implicit design decisions are then represented as changes in probability. Thereby it's possible to find design solutions even with problem descriptions of limited precision.

FBS-BN Bild Beispiel einer FBS-BN Repräsentation (cf. Eichhoff & Maass, 2011)