Modular learning structure for knowledge-based CAD systems

CAD system with the ability of learning to improve the design strategies from its own experience is not only desirable, but also necessary and essential. This paper investigates the related issues of learning in a knowledge-based CAD system, and explores a modular structure for incorporating learning mechanism into that system. CAD system can be developed that initially implements whatever expert knowledge is available, learning structure then allows the system to improve, expand, and transfer knowledge from one problem to another in the system running practice. The learning agents in the structure act as the learning experts to carry out the domain-specific learning activities. A pattern matching method in learning process is discussed. An interactive refutation approach for knowledge induction is described.