377340 Data-Driven Fault-Tolerant Control of Networked Distributed Processes

Monday, November 17, 2014
Galleria Exhibit Hall (Hilton Atlanta)
Zhiyuan Yao, Department of Chemical Engineering and Materials Science, University of California, Davis, Davis, CA and Nael H. El-Farra, Department of Chemical Engineering & Materials Science, University of California, Davis, Davis, CA

The identification and accommodation of faults in process systems are two of the fundamental problems at the intersection between process control and operations. A careful review of the extensive literature on fault detection and fault-tolerant control shows that while the majority of existing results have been developed for spatially homogeneous processes described by lumped parameter systems, methods for the diagnosis and handling of faults in spatially distributed systems described by Partial Differential Equations (PDEs), on the other hand, are few and limited by comparison. This realization, together with the prevalence of spatiotemporal dynamics in such systems, have motivated several efforts in recent years aimed at bridging this gap (see, for example, [1]-[4] and the references therein). In these studies, a framework for the integration of actuator fault detection, isolation and control system reconfiguration for highly-dissipative PDEs was developed on the basis of low-order models that are suitable for the design of practically implementable control and monitoring algorithms, and are amenable to rigorous closed-loop analysis.

Despite the continuing progress in this area, an examination of the above methods reveals that they are developed within the conventional feedback control framework where the sensor-controller communication is assumed to occur continuously and flawlessly. With the advent of networked control systems, however, and the increased reliance on advanced communication technologies in many industrial systems, the fundamental limitations imposed by communication constraints (e.g., in the form of resource constraints, real-time scheduling constraints, delays, etc.) on the closed-loop stability and performance properties can no longer be ignored and need to be explicitly accounted for. Efforts to address sensor-controller communication constraints in the context of spatially distributed control systems were undertaken in a number of recent studies (see, for example, [5]-[7]) which focused on highly-dissipative distributed processes and led to various finite-dimensional networked control and scheduling design methodologies that enforce closed-loop stability with minimal sensor-controller communication requirements. The resulting design methods, however, do not explicitly consider fault diagnosis or handling in the control system design. At this stage, the design of networked fault-tolerant control systems for spatially distributed processes remains an open problem that is in need of investigation and further development.

This work presents a methodology for the integrated identification and accommodation of control actuator faults in a class of spatially distributed systems controlled over a resource-limited communication network. The methodology brings together tools from model reduction, networked control and moving-horizon parameter estimation. Focusing on infinite-dimensional systems described by highly-dissipative parabolic PDEs, modal decomposition is initially used to obtain a finite-dimensional system that captures the slow dynamics of the infinite-dimensional system. The slow system is then used to design a model-based networked controller and explicitly characterize its stability region, in terms of the sensor-controller communication rate, the model uncertainty, the control actuator placement and the controller design parameters. Fault identification is then addressed using a moving-horizon least-squares parameter estimation scheme which is embedded in the sensors to estimate on-line the size of the fault using the sampled state and input data. Once the fault is identified and its magnitude estimated and communicated to the controller, a number of possible fault accommodation strategies are devised to maintain closed-loop stability. These include updating the post-fault control model,  adjusting the controller design parameters, or switching to an alternative control configuration. The selection of  the appropriate accommodation strategy is made on the basis of the estimated fault magnitude and the characterization of the networked closed-loop stability region. The implementation of the proposed fault identification and accommodation strategies on the infinite-dimensional system is also analyzed. Finally, the proposed methodology is illustrated using a diffusion-reaction process example.


[1] N. H. El-Farra and S. Ghantasala, "Actuator fault isolation and reconfiguration in transport-reaction processes,'' AIChE J., 53:1518-1537, 2007.

[2] S. Ghantasala and N. H. El-Farra, "Robust actuator fault isolation and management in constrained uncertain parabolic PDE systems,'' Automatica, 45:2368-2373, 2009.

[3] Z. Yao and N. H. El-Farra, "Robust fault detection and reconfigurable control of uncertain sampled-data distributed processes,'' in Proceedings of 50th IEEE Conference on Decision and Control, Orlando, FL, 2011, pp. 4925-4930.

[4] S. Ghantasala and N. H. El-Farra, "Active fault-tolerant control of sampled-data nonlinear distributed parameter systems,'' Inter. J. Rob. Nonlin. Contr., 22:24-42, 2012.

[5] Y. Sun, S. Ghantasala, and N. H. El-Farra, "Networked control of spatially distributed processes with sensor-controller communication constraints,'' in Proceedings of American Control Conference, St. Louis, MO, 2009, pp. 2489-2494.

[6] Z. Yao, Y. Sun, and N. H. El-Farra, "Resource-aware scheduled control of distributed process systems over wireless sensor networks,'' in Proceedings of American Control Conference, Baltimore, MD, 2010, pp. 4121-4126.

[7] Z. Yao and N. H. El-Farra, "Networked model predictive control of spatially distributed processes,'' in Proceedings of American Control Conference, Washington, DC, 2013, pp. 2062-2067.

Extended Abstract: File Not Uploaded
See more of this Session: Interactive Session: Systems and Process Control
See more of this Group/Topical: Computing and Systems Technology Division