Consistency-based failure detection and reconfigurable control
Final Report Abstract
Fault-tolerant control (FTC) aims at increasing the availability of technical processes. In the active FTC scheme, a diagnosis unit has to detect and to identify the fault and a reconfiguration unit has to adapt the controller to the faulty plant. However, in literature the diagnosis task and the reconfiguration task are usually considered separately. The aim of the project was to develop a unified FTC framework for sensor and actuator failures. A structural representation of the system was used to identify the redundancy in sensor and model information, which allows fault diagnosis, as well as the redundancy in sensor information and actuator effect required for control reconfiguration. Based on the result of the structural diagnosability analysis, a method for consistency-based sensor or actuator failure detection and isolation was elaborated. A structured bank of observers for the residual generation was combined with dynamical, observer-based residual evaluation methods, which provide diagnosis results that are accurate enough to be used for control reconfiguration. The structured bank of observers was combined with the concepts of the virtual sensor and of the virtual actuator for control reconfiguration. These concepts were extended in a way that the result from the structural reconfigurability analysis was integrated in order to allow a more precise reconfiguration of the control loop with selected components. The concept of simultaneous control reconfiguration extends the concepts of the virtual sensor and the virtual actuator towards systems with non-isolable sensor or actuator faults, which ensures that the diagnosis result is sufficiently accurate for control reconfiguration. Furthermore, the virtual sensor and the virtual actuator have been considered in the context of interconnected systems with a distributed or decentralized controller architecture. Since the observer used for sensor fault isolation is the same as the observer used as a virtual sensor for control reconfiguration after a sensor fault, a second approach to combine diagnosis and control reconfiguration aimed at residual generation with virtual sensors. This method is applicable, whenever the system state is measurable, which is the case in mechanical systems like multirotor unmanned vehicles. The elaborated methods were implemented in MATLAB / Simulink. In a case study, several multirotor vehicles subject to the complete loss of a rotor have been analysed with respect to diagnosability and reconfigurability. The active FTC scheme was evaluated in simulations and experiments with hexrotors, which demonstrated the robustness and the on-line, real-time applicability of the proposed active FTC scheme without human interaction.
Publications
-
Control reconfiguration of fullstate linearizable systems by a virtual actuator. 9th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, pp. 339–344, Paris 2015
D. Vey, J. Lunze, E. Žáčeková, M. Pčolka, and M. Šebek
-
Control reconfiguration of physically interconnected systems by decentralized virtual actuators. 9th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, pp. 360–367, Paris 2015
D. Vey, S. Hügging, S. Bodenburg, and J. Lunze
-
Experimental evaluation of an active fault-tolerant control scheme for multirotor UAVs. 3rd International Conference on Control and Fault-Tolerant Systems, pp. 119–126, Barcelona 2016
D. Vey and J. Lunze
-
From fault diagnosis to reconfigurable control: a unified concept. Plenary paper at 3rd International Conference on Control and Fault-Tolerant Systems, pp. 413–421, Barcelona 2016
J. Lunze
-
Simultaneous control reconfiguration of systems with non-isolable actuator failures. American Control Conference, pp. 7541–7548, Boston 2016
D. Vey, K. Schenk, and J. Lunze
-
Towards a reduced virtual actuator: A graph-theoretic approach. European Control Conference, pp. 1915–1921, Aalborg, Denmark, 2016
D. Vey and J. Lunze
