| In the present research, distinguishing features of mechanical engineering are investigated first. Based upon this, a general decomposition schema is proposed, in which any machine can be represented functionally in terms of units of power, transmission, work, control, support, tribology and auxiliaries. Based upon this representation, mechanical engineering design problems are classified, and the conceptual design of mechanical assemblies is selected as the major focus for further research toward design automation.;In this, actual designs are investigated by means of case studies detailed herein. Based upon this analysis and upon cognitive theory, an "impulse matching" model is proposed. According to this model, design knowledge is stored in engineers' long-term memory as "chunks" in the abstract form of Function-device Token (F-dT) pairs. Moreover, design configurations are stored in a similar manner. During the conceptual design process, the specified needs are searched and compared against functional "slots" of various device tokens. When a match is found, the related device token is returned as a solution, and thus a design alternative is generated. Consequently, an engineer with more chunks stored in memory is capable of generating more acceptable alternatives, and thereby, improved designs.;A prototypical expert system is built in accordance with the impulse matching model. In this, overhead cranes are used as a vehicle to demonstrate the concept. The result reflects the cognitive process of conceptual mechanical design and is shown to be suitable for computer automation. |
