Abstract:

In this paper, we address the data-driven design of observers for the process monitoring and fault diagnosis purposes. Instead of identifying a standard state space model, our design schemes are based on the identification of the so-called parity space. Applying a well-established mapping between the parity vector and the solution of the Luenberger equations, two design schemes are developed, which allow a direct design of (a) observer-based fault detection systems (b) a full order observer for the estimation of the process variables. The achieved results are illustrated by an academic example.