(476c) Monitoring and Reconfiguration of Sampled-Data Nonlinear Hybrid Process Systems with Actuator Faults

Compared with the substantial and growing body of research work on fault diagnosis and fault-tolerant control of continuous process systems, results for hybrid process systems that combine continuous dynamics and discrete events remain limited at present. Recently, we developed in [1] an integrated approach for fault detection and monitoring of a class of nonlinear hybrid processes with control actuator faults, uncertain continuous dynamics and uncertain mode transitions. A key idea was the design of a bank of dedicated mode observers using unknown input observer theory to identify the active mode at any given time and distinguish between faults, mode transitions and uncertainties. Beyond mode identification, measurement sampling due to the inherent limitations on the sensing devices or sensor-controller communication medium represents an important issue that needs to be accounted for in the design of the monitoring and control systems, since the discrete availability of measurements limits our ability to accurately monitor and precisely control the process. Another important aspect that requires attention is the design of the control system reconfiguration logic following fault detection. Unlike continuous processes, hybrid systems involve switching between multiple modes with different dynamics. In general, the different operating modes will have different fall-back control configurations, and this may limit the availability of certain control configurations for use as backup in the event of faults. Furthermore, the activation of a fall-back configuration in a hybrid system introduces new dynamics within the operating mode that should be taken into consideration to maintain the stability of the overall hybrid system.

In this work, we present a model-based framework of fault detection and control system reconfiguration for nonlinear hybrid process systems under output feedback control with measurement sampling rate constraints and control actuator faults. A dynamic model for each mode of the hybrid system is initially embedded within each controller so that an estimate of the process output between the sampling instants can be generated and used to calculate the control action; while at the sampling instants, the estimate is updated using the actual measurements from the sensors. The stability properties of each mode of the hybrid system are then analyzed, leading to an explicit characterization of the fault-free behavior and the maximum allowable sampling period for each mode. In order to detect actuator faults within each mode, the fault-free behavior is used as the basis for deriving a time-varying residual alarm threshold which is used to obtain the fault detection rules. Once a fault is detected in a given mode, the supervisor needs to determine which fall-back actuator configuration to activate. There are two key considerations for designing the reconfiguration logic. The first is the fact that the activation of a fall-back actuator configuration (that is not identical to the faulty one) alters the dynamics of the current operating mode and thus introduces a new “sub-mode” that may have a different stabilizing sampling period. Instability may ensue if the sampling period that is currently in use is larger than the maximum allowable sampling period for the new “sub-mode”. The second consideration in designing the reconfiguration logic comes from the fact that a given actuator configuration that is suitable for use as backup in the current operating mode may not be available to future modes, and this requires that we take into account the availability of fall-back actuator configurations to future operating modes as well when performing the control system reconfiguration in any given mode. To address these two problems, one must identify all fall-back actuator configurations for each mode and characterize their maximum allowable sampling periods prior to process operation. Based on this information, the reconfiguration logic can then be implemented during process operation to ensure that the chosen fall-back configuration guarantees stability of the current mode as well as all future modes. Finally, the developed monitoring and reconfiguration strategies are illustrated using a simulated model of a hybrid chemical process.

References

[1] Y. Hu and N. H. El-Farra, “Robust fault detection and monitoring of hybrid process systems with uncertain mode transitions”, AIChE J., in press.