This thesis addresses the issues related to the acquisition and transformation of design knowledge during the conceptual design process and provides a framework for representing this using connectionist methods.; Conceptual design is represented as a three-task process: problem analysis, performance analysis, and conceptual synthesis. The decisions made during the conceptual design process is represented as an information flow (translations and transformations) from abstract requirements to final definition of design requirements. This is represented with a computational model using series of cascading neural networks. Here, the outputs and inputs are interlinked in a successive fashion representing the information translations and transformations that occur during the conceptual design of process machinery.; The design of process machinery offers many challenges. The product requirements are given in terms of final results expected from the material they process instead of design specifications for individual process equipment. Although there are many books on the subject, much of the knowledge is not formalized and relies on considerable expertise. This thesis codifies some of the knowledge essential to the design of process machinery.; Theoretical guidance is used to analyze, formulate, and structure the conceptual design problem while experimental implementation is used to verify the proposed concept. A new and unique way of describing the design environment is explored with the use of connectionist methods. These ideas are experimented with the domain of process machinery design. As a result, several real process machinery design problems are used to test the concepts and refine the overall approach.
展开▼