Residual generator design for the FDI of linear multivariable dynamic systems

This paper addresses the problem of the detection and isolation of the input and output sensor faults for a linear multivariable dynamic system, in the presence of disturbance. In particular, this work proposes a polynomial approach for the design of residual generators in order to realise complete diagnosis schemes when additive faults are present. It is shown that the use of an input-output description for the linear dynamic model of the system under investigation allows to compute in a straightforward way the residual generators. The residual generator design is performed in order to maximise a suitable fault sensitivity function. Thus, the suggested design approach leads to dynamic filters that achieve both disturbance de- coupling and desired transient properties in terms of a fault to residual reference transfer function. The results obtained in the simulation of the faulty behaviour of a linearised aircraft model are finally reported. The fault diagnosis strategy performances are tested in the presence of measurement noise, disturbance and model uncertainty by means of extensive simulations.