-1- OBSERVABILITY AND REDUNDANCY DECOMPOSITION APPLICATION TO DIAGNOSIS José Ragot, Didier Maquin and Frédéric Kratz This paper describes different ways of generating analytical redundancy equations in the case of systems represented by either static or dynamic equations. A relation is called a redundancy equation when known variables only, i.e. measured variables or controlled inputs of the system, appear in its expression. Redundancies are a powerful tool for monitoring processes as they can be used to detect and isolate sensor and actuator faults. The classical methods for generating redundancy equations are presented first. A method using the decomposition of the process equations by the observability concept is also described. The general character of the methods of generation of redundancy equations is demonstrated by showing that the same formalism can be applied to static and dynamic systems. This general character also appears when redundancy equations are used for detecting faults. Indeed, the statistical tests, presented in this paper, which are used for detecting and localizing the faults, apply indifferently to static and dynamic representations. 1.1 INTRODUCTION The safety of processes can be greatly enhanced through the detection and isolation of the changes indicative of modifications in the process performances. If the models describing the process are accurate, the problem of fault detection may be solved by observer-type filters. These filters generate the so-called residuals computed from the inputs and the outputs of the process. This residual generation is the first stage in the problem of fault detection and identification (FDI). For them to be useful in FDI, the residuals must be insensitive to modelling errors and highly sensitive to the failures under consideration. In that regard, the residuals are designed so that the effects of possible failures are enhanced which in turn increases their detectability. The residuals must also respond