Observer-based Fault Detection and Isolation of Three-tank Benchmark System

In modern process industry such as chemical and petrochemical plants, it is important to detect and isolate faults in closed loop, while the plant is running. This paper investigates an observer-based process fault detection and isolation. Unknown input observer (UIO) is used for on-line detection and isolation of faults. Three-tank benchmark system is used as a good prototype of many industrial applications in process industry. Sensor and actuator faults are considered. UIOs are designed using the linearized and discretized model of the system around the operating point. It is assumed that the probability of occurrence of more than one fault at the same time is negligible. Residuals are evaluated using a cumulative weighted average of the estimation error, with exponential forgetting factor. Resulted fault detection system is tested with the continuous nonlinear model of the three-tank benchmark system. Different fault cases are studied. Simulation results are presented and discussed. Results show that the fault detection and isolation system has good performance.