(654a) Data-Based Monitoring and Reconfiguration of Distributed Model Predictive Control Systems

Authors: 
Chilin, D., Univ. of California, Los Angeles
Liu, J., University of California, Los Angeles
Muñoz de la Peña, D., University of California, Los Angeles
Davis, J. F., University of California - Los Angeles


MPC is a natural control framework to deal with the design of cooperative, distributed control systems because of its ability to handle input and state constraints, and also because it can compensate for the actions of other actuators in computing the control action of a given set of control inputs in real-time. With respect to available results in this direction, several distributed MPC (DMPC) methods have been proposed in the literature that deal with the coordination of separate MPC controllers that communicate in order to obtain optimal input trajectories in a distributed manner. In our previous work [1], we proposed a DMPC architecture with one-directional communication for nonlinear process systems. In this architecture, two separate MPC algorithms designed via Lyapunov-based MPC (LMPC) were considered, in which one LMPC was used to guarantee the stability of the closed-loop system and the other LMPC was used to improve the closed-loop performance. In [2], we also considered the design of DMPC architectures for systems with asynchronous and delayed measurements. In a recent work [3], we extended the DMPC architecture developed in [1] to include multiple distributed controllers and relaxed the requirement that one of the distributed controllers should be able to stabilize the closed-loop system. In the new proposed DMPC architecture in [3], there are several distributed controllers, where individually they can not stabilize the closed-loop system, but cooperatively can achieve closed-loop stability and a desired level of closed-loop performance. The above results deal with the design of DMPC systems and do not address the problems of monitoring and reconfiguration of DMPC in the event of actuator faults.

On the other hand, the occurrence of faults in chemical processes poses a number of challenges in process monitoring and fault-tolerant control (FTC). Over the last three years, we have initiated an effort on FTC of nonlinear processes trying to bring together the disconnected fields of process fault-diagnosis and nonlinear process control. We have looked at both actuator and sensor faults and their impact and handling in the context of chemical process control. Despite this progress, there are no results on monitoring and reconfiguration of cooperative, distributed control systems.

The focus of this work is on the development of FDI and FTC systems for the monitoring and reconfiguration of DMPC systems applied to general nonlinear processes in the presence of control actuator faults. Specifically, we consider a DMPC system in which two distributed LMPC controllers manipulate two different sets of control inputs and coordinate their actions to achieve closed-loop stability and performance specifications. We first consider that case that the faulty actuator can be isolated from the process. In this case, we design a model-based FDI system which effectively detects and isolates actuator faults; and then based on the assumption that there exists a backup control configuration which is able to stabilize the closed-loop system within the DMPC system, we develop FTC switching rules to handle faults in the actuators of the distributed control system to minimize closed-loop system performance degradation. Subsequently, we consider the case that the faulty actuator can not be isolated from the process. In this case, we design a data-based fault identification system which can identify the maganitude of the fault based on process state measurements. Based on the identified fault value, we develop FTC switching rules to handle faults. Sufficient conditions for the stabilizability of the FDI and FTC system are obtained based on a detailed mathematical analysis. The proposed design is applied to a chemical process example, consisting of two continuous stirred tank reactors (CSTRs) and a flash tank separator with a recycle stream operated at an unstable steady state, to demonstrate its applicability and effectiveness.

[1] J. Liu, D. Munoz de la Pena, and P. Christofides, ?Distributed model predictive control of nonlinear process systems,? AIChE Journal, vol. 55, pp. 1171?1184, 2009.

[2] J. Liu, D. Munoz de la Pena, and P. D. Christofides, ?Distributed model predictive control of nonlinear systems subject to asynchronous and delayed measurements,? Automatica, vol. 46, pp. 52?61, 2010.

[3] J. Liu, X. Chen, D. Munoz de la Pena, and P. D. Christofides, ?Sequential and iterative architectures for distributed model predictive control of nonlinear process systems,? AIChE Journal, in press.