An approach for casting defect analysis employing finite element design optimisation, media axis transformation and neural networks

Abstract

Our experience of using an approach based on the defect-metacause-rootcause relationship coupled with other novel computational techniques for casting defect analysis, has been discussed in this paper. A rootcause is a design, process or material parameter, which may be controlled to minimise the occurrence of defective castings. A metacause is defined as a scientific rationale that governs the occurrence of defects and is influenced by controlling one or more rootcauses. The defect-metacause-rootcause relationship has been linked to Finite Element modelling, as in most circumstances the objective of any modelling exercise is to understand the influence of rootcauses on defects by numerically modelling one or two metacauses. It has been shown that an optimal specification of rootcauses can be predicted by coupling the numerical modelling techniques with optimisation techniques. Use of the medial axis transformation has been explored for defining objective functions within an optimisation analysis. The influence of process, design or material parameters on the occurrence of defects has been studied using the “learning from examples” strategy of neural networks. Current limitations of these techniques have also been highlighted.