Disturbance decoupled residuals for a chemical process

The paper presents some results concerning fault diagnosis for dynamic processes using dynamic system identification and disturbance de-coupling techniques. The first step of the considered tpproach cor sists of exploiting inputoutput descriptions of the monitored system. In particular, the disturbance term of that model can be used to take into account unknown inputs affecting the system. The next step of the scheme leads to define a set of relations insensitive to the disturbance term, and therefore can be used as residual signals. The proposed fault diagnosis scheme has been tested on a real industrial chemical process in the presence of sensor, actuator and component faults. The results and concluding remarks have been finally reported.