Process fault detection in the presence of a Markov noise

The method of cumulative sums with a reflecting screen is employed to detect variations of the mean for a normal process against the backgroud of a Markov noise. The likelihood relation of the observed process can be computed if the Markov noise is nonlinearly filtered. A computing the a posteriori probability density of the Markov noise one of the parameters of the normal process has to be reassigned om tje points where the cumulative sum algorithms is canceled.