A Neural Network Based Approach to Knowledge Acquirement in Conceptual Design of Mechanical Derive

Selecting an appropriate drive type that meets design requirements depends on a large amount of experiential knowledge in conceptual design of mechanical drive, whereas the complexity of conceptual design results in a great difficulty existing in knowledge acquirement. To tackle it, a BP network based approach to knowledge acquirement is proposed, in which a binary coding method is used to structure learning samples to accumulate expert knowledge, and through training this constructed network using these samples, the knowledge related with mechanical drive is acquired and expressed with the trained weight and threshold matrix. This paper provides a promising approach to deal with the bottleneck problem in knowledge acquirement of intelligent design system. Under this approach, the automation of knowledge acquirement is effectively solved and in a sense, this trained network can be used as a knowledge base of expert system to facilitate the design process of mechanical drive.