Multivariate Statistical Control Applied in a Marine Propulsion Engine.

Abstract

Changes in the correlation structure between the variables that define a process, can cause deviations which are difficult to detect with the only use of univariate monitoring techniques. This paper applies a methodology based on Multivariate Statistical Process Control (MSPC), which takes into account the correlation between variables.

The combustion process of a 2T marine diesel engine was monitored through Hotelling's T2 multivariate technique for specific ship conditions. The control chart identified sudden changes in the process, which would not have been possible to detect through univariate monitoring. The variables that generated the signals in the process were identified through the technique of decomposition MYT.