An investigation of the feasibility of a CBR approach to the design of sand castings in the foundry industry

The research described in this thesis investigates the feasibility of automating a key aspect of the Methods Engineering process of designing sand castings in the foundry industry by a case-based reasoning (CBR) approach. The research answers four basic questions. These have been organised into a primary question and three subsidiary questions. The primary question asks: Is it possible to automate the retrieval processes that a Methods Engineer uses in recalling analogous design cases for a given target design case by using a component decomposition of a shape derived from knowledge acquisition? From this, the three subsidiary questions investigate the mechanism for component decomposition, the formulations of appropriate similarity metrics of components and finally, evaluating metric performance. Methods Engineering is introduced and the investigation is focussed on an existing process of shape componentisation of 2-D section slices into elementary components corresponding to the way Methods Engineers reason about solid objects. A novel contribution of this research is the identification, abstraction and formulation of this procedure, and its automation by computer software. A further contribution is made by providing a series of similarity metrics for shapes in the casting domain; these include both metrics based on feature-value pairs, and metrics based on graph matching. A set of performance measures appropriate to the casting problem was formulated and a prototype retrieval system was produced and implemented as a CBR system. A test case base representing a sub-domain of rotationally symmetric castings was constructed and performance figures show that the system comes close to expert retrieval performance. This thesis concludes with a section regarding new research questions that emerge from the research and an agenda for future work that these questions give rise to.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

University of Greenwich

Year dc:date.issued

2002

Author dc:creator

• Mileman, Tony

Advisors dc:contributor.advisor

• Knight, Brian
• Petridis, Miltiadis

• TA Engineering (General). Civil engineering (General)

Language dc:language

en