Most work in knowledge acquisition and manipulation has focused on expert systems. Expert systems solve one kind of problem: heuristic classification. This thesis extends some advances in knowledge engineering to a broader class of problem: design.

Design is examined as a generic activity, found in many fields of professional practice. A theoretical framework is developed that supports the refinement of design from high-level concepts through implementation. This framework includes a computational model that is shown to be completely general (Turing-equivalent). Therefore, the theory and model are suitable for representing any design project. They are applied specifically to software development.

Practical support for software designers is offered in a prototype software design system. Existing work in automated knowledge acquisition is used to transfer knowledge about a design from the designer to the automated tool. Consistent support for refinement of design choices at any level of detail makes design a maintainable activity. This opens new possibilities for automated code generation, automated maintenance, and the nlore effective management of software at a higher-level design representation.