| Conceptual design is a higher level design phase requiring more creativity, imagination, intuition, and knowledge than the detail design stage. In the conceptual stage, designers usually begin the design process by referencing design cases that they feel are, based upon experience or existing documentation, similar to the case on which they are currently working. Through associative references, designers modify the design cases that they retrieve into a target design that satisfies all geometric and technological requirements of their current case.;Although several computer-aided design retrieval systems have been developed for assisting the conceptual design process, several limitations and drawbacks make them impractical. For instance, Group Technology (GT) based approaches simultaneously conduct indexing and classification so that the provided query must be accurate in order to find a fixed number of similar designs. Similarly, image bitmap and skeleton based approaches exclude technological attributes and are difficult to query in terms of case context.;To address some of the difficulties found in current systems, this research develops a flexible feature-based design retrieval system. First, each design in the developed system is modeled by feature-based representation that consists of its geometric and technological information. With the help of a pre-defined feature library and CSG indexing scheme, designers can easily index the geometric shape of design cases and query into vector formats. Meanwhile, technological attributes of design cases and query can be stored as quantitative and qualitative data.;Based upon the proposed system, when input preparation is complete, design retrieval tasks can be accomplished by three proposed modules. The first module, fuzzy feature association procedure, is for solving the problems of imperfect query input through data composition. The second module, Fuzzy ART neural network, is developed as a searching mechanism for finding reference designs based on overall geometric similarity. By adjusting a similarity parameter of the Fuzzy ART neural network, designers can control similarity of retrieved reference designs. The third module, fuzzy evaluation procedure, is developed for finding reference designs based on overall technological attributes. This procedure is capable of dealing with qualitative and quantitative data in technological attributes. By controlling a parameter of this procedure, a desired number of reference designs ranked by orders can be received. Implementation examples show the robustness of the proposed system in that it can flexibly retrieve reference designs based on geometric features and technological attributes even when incomplete query is provided. |
