Proportional hazard‐based robust monitoring schemes using logistic distribution

Abstract

AbstractIn the manufacturing industry, process surveillance plays an important role in improving product reliability. Many monitoring procedures have been devised to improve the reliability of a product in the literature. Due to cascading nature of multistage processes, the quality variable of the final stage can be influenced by the quality variable in the previous stages. Furthermore, the ordinary least squares method produces biased estimates if there are outliers in the historical data. The existence of cascade property in the multistage process and the presence of influential observations (outliers) in the historical data make the analysis of control schemes more challenging. Therefore, it is essential to take into account the impact of effective covariates and outliers in the historical dataset. In this paper, a cumulative sum (CUSUM) and two exponentially weighted moving average (EWMA) charts have been developed to monitor a two‐stage‐dependent process. The proportional hazard (PH) model has been applied to model the relationship among the incoming and outgoing variables of the two stage process. To deal with the deleterious effects of outliers on the results, a robust regression technique known as the M‐estimation has been implemented and the performance of the proposed monitoring procedures has been analyzed through Monte Carlo simulations. Lastly, two real‐life applications of the proposed control schemes are presented.