چكيده به لاتين
In this PhD thesis, a recursive Fault Detection and Isolation Kalman filter (FDIK) is extended based on communication control methods (deterministic and stochastic event-triggered scheduling) for Networked Control Systems (NCS). Since noise and unknown inputs are inevitable in real systems, the system in this thesis is considered linear consisting of deterministic/nonzero mean stochastic unknown input and colored measurement noise. First, a combined method is proposed to deal with the effect of colored measurement noise in designing FDI filter and to prevent white Gaussian measurement noise with zero covariance. Then, a stable subsystem of the main system is achieved which attenuates disturbance effects as much as possible and is independent of disturbance effects in ideal case.
Moreover, in order to control the communications between the FDI filter and sensor nodes, first the mentioned filter is extended recursively based on deterministic event-triggered scheduling. Because of the main challenge in control of communications based on deterministic scheduling (i.e. the impossibility of online tracking of the higher order systems due to complicated calculations), the FDI filter is extended based on three types of stochastic conditions in communication control. In the first type, the stochastic event-triggered condition is just based on measured data. In other words, the data-sending condition is considered open-loop. In the second type, the stochastic event-triggered condition is based on the difference between current measured data and the prediction of last sent data. In other words, the data-sending condition is considered closed-loop. In this case, the sensor needs not only the measured data, but also their estimation. Therefore, in such type of communication, the sensor needs the local copy of the estimator or the feedback from FDI filter to sensor whose unneeded information is deleted. In the third type, the stochastic event-triggered condition is based on the difference between the last measured sample and a coefficient of rough estimation of last sent sample. Therefore, in this type of communication, in contrast to closed-loop communication type, the sensor does not need feedback from the estimator, and in contrast to open-loop communication type, the sensor has got the rough estimation of last sent sample. Although there is no feedback loop from the FDI filter to sensor, this type of communication performs more like closed-loop type. It is notable that in order to make the design of constant parameters convenient, the design parameters are given based on LMI convex optimization methods in this PhD thesis. Finally, the proposed FDI method is evaluated to detect and isolate stator inter-turn short circuit and broken rotor bars faults with unbalanced voltage as disturbance in three-phase induction motors.
