In this page, we summarize our research on using set-theoretic methods to develop failure mode reconfiguration strategies with constraint admissible and recoverable sets.
There is a great demand for systematic approaches to Failure Mode and Effect Management (FMEM) system design for industrial applications. In fact, the software and algorithmic content of diagnostics and failure mode management is often larger than what is responsible for the nominal system function. For failure mode management, we need strategies to ensure that, in case of a single point of failure, the system operation can be reconfigured within a certain time period, so that in the new mode, another single point of failure cannot lead to safety hazards while the system availability can be maximized.
The distinguishing feature of our approach is guaranteeing that safety constraints are strictly enforced during the operation in the normal modes, during the operation in the failure mode and during reconfiguration. We additionally integrate failure mode detection and isolation into the overall framework.
Our approach is deterministic and is based on set-theoretic synthesis. Hence constraints are enforced strictly and not in a probabilistic case. Furthermore, failure mode detection and isolation can be performed in a finite time and not asymptotically.
We are addressing the issues of the overall design of FMEM system answering the question in what way should the system operation be (minimally) restricted in the preceding mode to ensure safe reconfiguration into possible subsequent failure modes and assure succesful failure mode detection and isolation.
A set of constant reference commands and initial states such that constraints are satisfied for all future time.
A set of initial states such that there exists a recovery sequence to steer the states into the state projection of the constraint admissible set within N_M steps, while satisfying safety constraints during the reconfiguration.
The state projection of the constraint admissible set of mode M' to be a subset of the recoverable set of the successor mode M.
The proposed strategy addresses multiple failure modes and failure paths/scenarios by combined use of constrained admissible and recoverable sets, specific mechanisms used to enforce the reconfigurability, and the use of reference governor for reference tracking.
H. Li, I. Kolmanovsky and A. Girard, "A Failure Mode Reconfiguration Strategy Based on Constraint Admissible and Recoverable Sets," Proceedings of 2021 American Control Conference (ACC), 2021, pp. 4771-4776, doi: 10.23919/ACC50511.2021.9482887.
H. Li, I. Kolmanovsky and A. Girard, "Set-Theoretic Failure Mode Reconfiguration for Stuck Actuators," in IEEE Control Systems Letters, vol. 6, pp. 1316-1321, 2022, doi: 10.1109/LCSYS.2021.3092953.
H. Li, I. Kolmanovsky and A. Girard, "Integrating Failure Detection and Isolation into a Reference Governor-Based Reconfiguration Strategy for Stuck Actuators," Proceedings of 2022 American Control Conference (ACC), pp. 4311-4316, 2022, doi: 10.23919/ACC53348.2022.9867778
Poster for the 2022 CPS Rising Stars Workshop
Presentation for the 2022 American Control Conference (ACC 2021)
Presentation for the 2021 Conference on Decision and Control (CDC 2021)
Presentation for the 2021 American Control Conference (ACC 2022)
This research is supported by the National Science Foundation under award number ECCS-1931738.
This page is maintained by Huayi Li and Ilya Kolmanovsky. Last updated 12/22/2022.
Huayi Li recieved her PhD from the University of Michigan, Ann Arbor in Feburary, 2022. She was co-advised by Professor Anouck Girard and Professor Ilya Kolmanovsky.
If you are interested in discussing about our work, you could reach out to us by emailing to huayil@umich.edu and ilya@umich.edu