Residual Generator Functions for Linear Multivariable Process Fault Detection

This paper investigates the design of residual generators in order to perform the fault detection task for linear multivariable models with additive faults and disturbances. The use of input–output polynomial forms leads to characterise in a straightforward fashion the basis of the subspace described by all the possible residual generator functions. The minimality of the residual generator function can be obtained by considering canonical input–output polynomial descriptions. These tools show how the same mathematical description of these filters can be obtained also by following a black–box identification approach. A simulated example is finally reported in order to highlight the main features of the proposed fault detection strategy.