Parity space residuals

As (1.18) indicates, residual ri (r) is independent of the states a (r) and the unknown disturbances rf (f). It depends only on the faults / t) and if there are none, the residual vanishes. Besides w, further vectors can be chosen that satisfy (1.17) Hr such vectors constitute the rir y. q + l) o matrix W in (1.10). Residuals generated in this way are called parity space residuals [28, 37]. 

Equation (1.18) further shows that the computation of parity space residuals requires the derivatives of measured inputs and outputs up to the order q. As these measured signals carry noise, they need to be filtered which can be done by so-called... 

The parity space approach to FDI is based on a comparison of the behaviour of a real process with that of a model that describes the non-faulty process. Any discrepancies between the two are described by residuals. The development of parity relations for residuals using a state space model goes back to Chow and WiUsky [35] and has been presented in various publications, e.g. [8, 34, 36]. In the following, only the basic idea is outlined. 

Diagnostic observers consist in the definition of a set of observers from which it is possible to define residuals specific of only one failure [8]. Parity relations are relations derived from an input-output model or a state-space model [11] checking the consistency of process outputs and known process inputs. 

---