ABSTRACT
This article not only shows that the CUSUM multi-chart which consists of several CUSUM charts, has the asymptotic optimal performance in jointly detecting and diagnosing the unknown change in a sequence of observations but also provides a design method of optimizing the CUSUM and EWMA multi-charts. The numerical comparisons illustrate that the optimized CUSUM multi-chart has better performance in jointly detecting and diagnosing the mean and variance shifts in
normal observations than that of the optimized EWMA multi-chart. A real example for engineering surveillance using the electric power generation data was used to demonstrate the practicality of the schemes.
Disclosure statement
No potential conflict of interest was reported by the author(s).