Modeling of a Hierarchical Supervisory Controlled Industrial Process for Fault Diagnosis

Modern industrial processes are hierarchically controlled in two levels. A continuous time control logic, such as Proportional integral derivative (PID) control, is implemented at the lower level to assure that the system behavior attains specifications of transient and permanent regimes, while at the higher level, a supervisory control ensures safety specifications for routine and non-routine operations. In general, both control strategies are projected without considering that the system components can suffer faults that can modify their expected behavior. Thus, a fault diagnosis strategy is fundamental to quickly indicate a fault event occurrence that can cause a deviation from the expected system behavior. In this paper, a discrete event diagnosis strategy is exploited for a two-level controlled tank liquid level system. To do so, we consider that a fault event can occur in the valve, which causes the valve to be stuck closed. We propose a modeling strategy for the valve that considers the fault consequences and its interaction with the remaining components. We show that this approach does not interfere with the control project, which allows the diagnoser synthesis to be achieved and implemented in an already-running controlled system.