Gamma generalized linear model-based control charts for high-purity processes

Statistical process monitoring of high-purity manufacturing processes becomes challenging if the defect rate depends on the fluctuations of a set of covariates (e.g., inspected weight, volume, temperature). This paper applies the generalized linear model framework to statistical process control for detecting contextual anomalies in high-purity processes. Different types of predictive residuals (i.e., Pearson, deviance, and quantile) and recursive residuals are considered, and the performance of these schemes is compared via a simulation study.