Statistical properties of exponentially weighted moving average algorithm for change detection

In this paper, the statistical properties of a change detection algorithm are considered. More specifically, we considered the exponentially weighted moving average (EWMA) algorithm. Analytical expressions for the probability distribution of detection delay and the time between false alarms are proposed, and the results are validated by simulations. The results can be used in examining the abrupt changes in a signal, and also in the design of active fault tolerant control systems.