Fundamentals of automated system development for detection of casting defects in shaped castings based on neural networks and fuzzy logic

An integrated approach that combines traditional non-destructive testing methods for castings with elements of artificial intelligence based on convolutional neural networks (CNN), recurrent neural networks (RNN), and fuzzy logic is presented. The proposed methodology is intended to enable automated detection and diagnosis of a wide range of defects in shaped cast blanks. By combining conventional casting quality control techniques with advances in artificial intelligence, it becomes possible to automate defect detection, classification, and decision-making regarding either the remediation (healing) of identified defects or rejection of the casting.