Anomaly detection for high precision foundries

Microshrinkages are known as probably the most difficult defects to avoid in high-precision foundry. This failure renders the casting invalid, with the subsequent cost increment. Modelling the foundry process as an expert knowledge cloud allows machine learning algorithms to foresee the value of a certain variable, in this case, the probability that a microshrin-kage appears within a casting. However, this approach needs to label every instance for generating the model that can classify the castings. In this paper, we present a new approach for detecting faulty castings inspired on anomaly detection methods. This approach represents correct castings as feature vectors of information extracted from the foundry process. Thereby, a casting is classified as correct or not correct by measuring its deviation to the representation of normality (i.e., correct castings). We show that this method achieves good accuracy rates to reduce the cost and testing time in foundry production.