Detection of operating mode changes, without a priori model and in uncertain environments

Abstract

Starting from the general observation that a measurement delivered by a sensor is subject to an uncertainty, its use in a decision chain must take into account this imprecise character. It is thus advisable to propagate this imprecision in all the chain of treatment and use of the measurement in question. In what follows, this principle is applied to the diagnostic function, one of the components of system monitoring. More precisely, we propose to analyze the temporal data collected on a system in order to detect and locate possible changes in the behavior of this system.