Learning from cases : retrieving cases in CAD systems based on design criteria

Abstract

Architects usually regard case studies as an important way to study architecture. Cases stand for existing answers to real problems and provide resources for free interpretation. This thesis investigates encoding design knowledge into architectural design projects through formal computation, and retrieval of cases through satisfying criteria provided by designers. Knowledge in architectural design has been distinguished into domain knowledge and design knowledge [Schmitt et al 90]. Domain knowledge investigates precise rules in design, and design knowledge deals with more flexible or personal rules in aesthetics or spatial arrangement. The Retrieving Cases Based on Design Criteria (RCBDC) program proposed in this thesis is a prototype for a CAD system with Case-Based Reasoning to deal with design knowledge. This thesis concentrates on the issues of retrieval within CAD systems using Case-Based Reasoning. Two kinds of retrieval have been developed: Critical Retrieval retrieves cases that meet the assigned single design criteria. Designers can filter among these retrieved cases by adding more and more design criteria. General Retrieval calculates all the possible permutations among repeated elements, and retrieves cases that have spatial relationships that are most similar to the current project.