Identification techniques for chemical process fault diagnosis

The paper presents the application results concerning the fault diagnosis of a dynamic process using dynamic system identification and model-based residual generation techniques. The first step of the considered approach consists of identifying different families of models for the monitored system. In particular, it is selected the most accurate identified model able to describe in the best way the dynamic behaviour of the considered process. The next step of the fault diagnosis scheme requires the design of output estimators e.g., dynamic observers or Kalman filters which are used as residual generators. The proposed fault diagnosis and identification scheme has been tested on a real chemical process in the presence of both sensor, actuator, component faults and disturbance. The results and concluding remarks have been finally reported.