Sliding Mode Based Fault Reconstruction And Fault Tolerant Control

In recent years, fault detection and isolation (FDI) and fault tolerant control (FTC)have receved more attention than ever before due to the increasing demand for morereliable dynamic systems such as aircraft, nuclear power station and off-shore largescale wind turbine. The failure of actuators, sensors and other components can result insignificant performance degradation, severe damage of the physical systems or evendisaster with the possibility of the loss of human lives. To avoid these consequencescaused by the failure of the elements in the control systems, it is critical to make thecontrolled plant fault tolerant or to detect or identify any possible faults at the earlieststage.The main function of FDI scheme is to generate an alarm when a fault occurs, andthen to determine the location of the fault. There is a large variety of FDI approachsbased on model including observer, parity space and parameters estimation. Recently,sliding mode observer (SMO) based FDI has been extensively studied for the faultreconstruction and estimation ability when equivanlent output injection concept isempolyed, which has great advantage in the sense that it can not only detect and isolatethe fault, but sharp the size, location and dynamics behavior of the fault than residualgeneration. Besides, SMO has been recognized as a promising robust observer methodto confront uncertain or perturbed systems. For this reason, robust SMO design for faultreconstruction and FTC design for faulty systems are the focus of this paper.First, the theory of SMO-based fault reconstruction for a class of uncertain lineartime invariant systems is improved and further extended. Simultaneous actuator faultand uncertain reconstruction is first considered and obtained by using an auxiliaryobserver. New sufficient conditions for the existence of observer and asymptoticalstability are derived by using an LMI formalution and Lyapunov theory. Then, inspiredby conventional result, this paper also proposes the scheme to design robust SMO forsimultaneous actuator and sensor fault reconstruction using an augmented observermethod, i.e., making the sensor fault in the form of actuator fault, and actuator robustreconstruction method. Existence conditions and asymptotical reconstruction propertyare also detailedly addressed. Furthermore, numerical simulations are presented todemonstrate the effectiveness of proposed schemes.Then, this paper proposes a new adaptive SMO design approach for a class of nonlinear systems, particular for a class of so-called one-sided Lipschitz nonlinearsystems, to reconstruct the faults. For Lipschitz nonlinear systems, a lot of work hasbeen done, and many excellent findings have been obtained. It has to be pointed thatmost results only work when the Lipschitz constant is small and usually need to beknown in advance. So this paper expends the scope of applicable systems fromLipschitz to one-sided Lipschitz nonlinear systems. Moreover, adaptive mechanism isembedded into the observer design to cope with the one-sided Lipschitz constantunknown case where the value is often difficult to get in practical systems. A sufficientcondition using linear matrix inequality is given to gurantee the asymptotically stabilityof the estimation error dynamics and compute the observer gains.The last chapter concentrates on the design of a fault tolerant controller to handleactuator and sensor faults using sliding mode control techniques. A sliding mode faulttolerant control design method based on state feedback scheme is presented for a classof uncertain linear multivariable systems subject to actuator effectiveness decrease faultaccording to the fact that this kind of actuator fault can be expressed in the form ofmatched uncertainty while the classical sliding mode control has the intrinsic robustnessto this type of uncertainty. The gains associated with the nonlinear term are allowed toadaptively increase when the onset of fault is occurred. For sensor faulty systems, underthe assumption that all states are available for measurement, the idea of utilizing asliding mode observer to reconstruct the faulty signals and using these reconstructedsignals to correct the measured values before they are used in the control law isproposed as a means of achieving fault tolerance. The advantage over other methods lieswith the controller without the need to be reconfigured or redesigned. The results ofsimulations highlight the capabilities and benefits of SMO as a way of sensor faulttolerant control.