Building Design Support by Hierarchical Expert Networks

For building design, computational intelligence systems use a knowledge base formed by means of neural network and fuzzy logic (neuro-fuzzy) techniques, from a building design database. The application of such a system to a building design task was preliminarily demonstrated earlier. The present research describes a systematic neural fuzzy modelling of data that forms a knowledge base in a hierarchical form. Each sub-knowledge base represents a local expert i.e., level-one expert, and the association of local experts forms a more comprehensive expert, i.e., a global domain expert. The association of the experts is accomplished by means of fuzzy-logic-driven gating network that performs the information handling as required. Although, the present paper describes hierarchical experts as local and global, the associations can be done in more subtle form, i.e., in a more than two steps so that the level of experts can be categorized in multi-level form. In such more complex structures, multi-level experts require a related gating network that could similarly be designed. The building design support system with the expert network developed, as a whole, is generic enough for decision-makings with a novel systematic approach concept using an appropriate database. Accordingly, the research deals with a particular architectural building design with efficiency and consistency features using the hierarchical expert network system described.