National representative in the World Foundry Organization

Technical library ·

Innovations in machine learning and defect diagnostics

R. S. Ransing and M. R. Ransing

Abstract

Analysis of cause and effect relationships for diagnosis has always been a favorite application for the artificial intelligence community. The first paper on expert systems for casting defect analysis has appeared in mid eighties. However, over the last decade, the research has slowed down with very few new publications emerging in this area.

However, the problem is very relevant even today. With globalization and experts retiring from their jobs, the industry is already facing skills shortage. The future casting industry needs to move away from the traditional approach for defect analysis.

This paper will look back twenty years and review the technological advancement. It will also identify why the technology failed to meet foundry-man’s expectations.

We have also revealed the patented self-learning knowledge base technology of X1Recall and discussed how it has managed to overcome the limitations of existing expert systems, neural networks or statistical techniques for defect reduction problems. A case study has also been presented.