Application of expert network for material selection in engineering design

Document Type

Article

Publication Date

9-30-1996

Abstract

The issue of representing knowledge in a manner so that the updating is easy and transparent to the user is very important in Artificial Intelligence. In recent years, an expert network approach for representing the knowledge has been used, which involves the development of an intelligent hybrid system. In this approach, the knowledge is stored in a neural network in form of weights associated with links. An expert system is used for validating the output of the neural network, and taking care of situations where the neural network cannot provide a unique solution. The expert system can also provide an explanation for the decisions made by the neural network. In this research, an expert network approach for material selection was used for the task of material selection in engineering design. Use of a rule based system alone makes it difficult to select a material in cases when there is no material available in the database whose properties are similar to the desired material properties. The neural network proved to be useful in such cases as it could select the best alternative from the database. Similarly, the neural network alone was not sufficient, as in some cases it may not be able to provide a unique solution. The expert system may evaluate the various materials selected by the neural network on the basis of price, availability in stock and closeness to desired properties and provide a unique solution.

Publication Source (Journal or Book title)

Computers in Industry

First Page

87

Last Page

101

This document is currently not available here.

Share

COinS