Subspace-aided data-driven robust fault observer design for a typical process system

Fault observer is an important mechanism for fault detection in industrial process, which plays as the premise in fault diagnosis and system safety monitoring. The fault alarm is raised if the generated residual signal exceeds the assigned fault detection threshold. With the unexpected but inevitable disturbances and noises in industrial process, it may generate increasing false alarm rate. Therefore, a robust fault observer design method is proposed based on the subspace-aided identification via system data in this paper. First the equivalent canonical form of the system is identified via the system data. Then the key parameters of the robust index is derived and the optimal parity vector is solved for constructing the robust fault observer. Finally, the performance of the proposed method is illustrated by the simulation of a typical three-tank flow process system, which indicates the improvement for its possible implementations.