Title: A fuzzy-rough case-based learning approach for intelligent die design

Authors: Chi Zhou, Feng Ruan, QingXiang Xia, ZhenYuan Huang

Addresses: School of Mechanical Engineering, South China University of Technology, Guangzhou 510640, PR China. ' School of Mechanical Engineering, South China University of Technology, Guangzhou 510640, PR China. ' School of Mechanical Engineering, South China University of Technology, Guangzhou 510640, PR China. ' School of Mechanical Engineering, South China University of Technology, Guangzhou 510640, PR China

Abstract: Tacit knowledge plays an important role in stamping die design, but it is hard to be mined and reused. This paper presents a fuzzy-rough approach of mining rules from existing successful die designs, which improves learning capability of intelligent stamping die design systems. The core of the learning mechanism includes: (1) a feature-based case representation, (2) fuzzification of feature attributes, (3) a fuzzy classification method to partition the cases into clusters based on similarities, (4) a rough set theory-based approach to compute attribute reduct and mine rules. A case base containing 53 bending features was used to test the proposed approach. 38 If-Then rules were mined after computing, and these rules could preserve the basic properties of the origin decision table. The results show that this approach can speed up the retrieval process of the case-based systems and reduce the difficulties in acquiring and updating rules for the rule-based systems.

Keywords: intelligent stamping; intelligent die design; rough sets; fuzzy classification; tacit knowledge; mining rules; learning capability; feature-based case representation; bending features.

DOI: 10.1504/IJCAT.2009.026584

International Journal of Computer Applications in Technology, 2009 Vol.35 No.2/3/4, pp.76 - 83

Published online: 20 Jun 2009 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article