Cyber-physical systems are getting more and more complex and thus are increasingly prone to faults. Since it is intractable to model all faults a-priori, online plant identification and reconfiguration is key to a successful fault handling strategy. This paper presents an approach to control reconfiguration via online identification of CPS to increase system dependability against actuator or plant faults. Using sparse regression (SINDYc), closed-loop system dynamics, including faults, are identified and used to reconfigure the control law, utilising plant redundancies. We will study the fault handling approach along a canonical control systems example, the inverted pendulum on a cart, which is inherently nonlinear and unstable. To disambiguate between input and system dynamics in the closed-loop systems, a perturbation signal is injected. The SINDYc algorithm is then applied to the measurement vectors of in-and output signals of the systems, rendering an up-to-date dynamic model, including possible faults. In the case of an actuator fault, the identified model will be used for control reconfiguration using the Pseudo-Inverse method, ensuring an optimal use of available redundancies. Both abrupt and incipient faults in the actuator dynamics are considered. In this work, we will limit our online identification to linear models and will reconfigure a full-state feedback controller, for which full observability of the system is assumed. In a parameter study we have shown the influence of perturbation signal power and measurement noise on the identifiability of the closed-loop system. We conclude that our approach of online control reconfiguration is performing satisfactorily for actuator faults in the studied use-case and can readily be extended to nonlinear model identification and consecutive reconfiguration of nonlinear controllers, such as MPC or INDI.     «

Cyber-physical systems are getting more and more complex and thus are increasingly prone to faults. Since it is intractable to model all faults a-priori, online plant identification and reconfiguration is key to a successful fault handling strategy. This paper presents an approach to control reconfiguration via online identification of CPS to increase system dependability against actuator or plant faults. Using sparse regression (SINDYc), closed-loop system dynamics, including faults, are identi...     »