SENSOR FAULT DIAGNOSIS IN DYNAMIC PROCESSES

Fault detection and diagnosis are important technologies for the safe and efficient operation of a chemical plant. This paper describes a sensor fault identification approach using variable reconstruction for dynamic systems. The proposed methodology extends sensor fault reconstruction for Canonical Variate Analysis based process performance monitoring which admits process dynamic behaviour in a natural way, and evaluates its capabilities compared to a dynamic PCA approach using a mathematical benchmark problem and a simulation of a closed loop controlled CSTR previously used for studying both simple and complex faults.