A hybrid artificial intelligence approach for improving yield in precious stone manufacturing

Abstract

Reducing wastage from the unnecessary cutting of raw material is a key issue in the manufacture of diamonds and gemstones. The accuracy with which stones are graded prior to their being processed through the various manufacturing stages of cutting and finishing is a key determinant of yield and so profit. This presently manual activity requires a skilled craftsman to assess the grade and spot opportunities for upgrading through the judicious cutting away of imperfections in the raw material. There is however a balance to be struck between raising quality and lowering wastage. This paper describes iGem, an artificial intelligence tool that integrates rule-based knowledge representation, fuzzy logic and genetic algorithms to produce a system for automating, and introducing consistency into, the grading of diamonds and gemstones. In this paper we show how iGem derives its knowledge from repeated examples of previously correctly graded stones and can improve its performance by learning from experience. The industrial benefit of iGem extends beyond simply improving grading but also to the introduction of consistency and so greater control into the overall manufacturing process. We believe the approach described has application in other situations where overall yield and manufacturing efficiency depends on trade-off decisions between removal of imperfections and loss of material as well as consistency in quality assessment. A further noteworthy aspect of the iGem project is its development of an objective quality assessment technique out of a hitherto substantially subjective one.