| One important characteristic and an essential element of design is that it is a creative act, as the purpose of design is to improve the existing solutions to old problems using new technologies and find solutions to new problems that have not been encountered before. Computers have proven to be powerful tools in removing from the designers' shoulders the burdens of processing and managing large amounts of design information. However, their potential in assisting designers in creative problem solving has not been adequately explored.; In this document, a study of the potential of using computers to support human designers' creativity is reported. A model for creative engineering design based on a network-based design prototype representation and a mixed mutation and analogy technique has been developed. The model distinguishes between a representation of domain-dependent design knowledge and domain independent mutation operators. These mutation operators manipulate the representation based on syntactic information to produce or suggest new design solutions that are not implied by the existing design knowledge. This model extends the representation of design prototypes beyond a parametric expression of function, performance, behavior, and structure to include an explicit representation of the relationships between parameters within a prototype. The categorization of design attributes and the explicit representation of the relations between design attributes are the two most important aspects of the network-based prototype representation and can be used to support creative design.; The model is illustrated using two particular mutation operators called the SCALE operator and the COMBINE operator. In order to produce new design solutions, the SCALE operator mutates the basic structure of a given failed solution by introducing a radical change to an attribute's value. The COMBINE operator utilizes both mutation and analogical reasoning to produce creative design ideas. A failed design solution is repaired by introducing into the failed solution the structure, thus the behaviors and functions, of another prototype. It uses a partial matching technique to identify the useful prototypes based on matching criteria represented as graphs. New solutions are generated by graph manipulation, where the graphs represent parts of a dependency network of a prototype. Both mutation operators produce design solutions using only the information of the relations, functional categories and syntactic description of attributes of prototypes and therefore are domain independent operators. The model has been tested in the domain of structural system design for buildings but has the potential to be extended to other engineering design domains. |
