Eigenstructure Assignment-based And Neural Network Adaptive Fault-tolerant Control Technology

The technique of fault tolerant control has been more researched recent years, and it plays the role of irreplaceable to improve the reliability and maintainability of the modern complex system. This paper makes a study for the fault tolerant control based on the eigenstructure assignment, direct adaptive and neural network adaptive control method respectively.First of all, specific to the system of has obtained the knowledge of the failure, we give a designing method of active fault tolerant control based on the eigenstructure assignment. So that, we can get the algorithm of reconfigurable through this method to make the eigenvectors of the closed-loop system all same as normal system and the eigenvalues of the closed-loop system get the maximum recovery. The results of simulation show the effectiveness and feasibility of the method.Secondly, according to the system of unknown parameter and failure, in the case of actuator failure, we give a method of direct adaptive fault tolerant control based on state feedback. The condition of plat-model output matching is studied, and gives the sufficient and necessary condition for the matching. According to the direct adaptive control method, this paper gives the direct adaptive failure compensation control strategy based on proportional-actuation scheme. So that in the case of unknown priori knowledge of system failure, the output of the system can asymptotically tracks the output of the reference model. The results of simulation show the effectiveness and feasibility of the method.Finally, specific to the MIMO nonlinear system, we give a fault tolerant control method of neural network adaptive based on dynamic inversion. Using the method of dynamic inversion, the nonlinear system will converts to pseudo linear system. For the model inversion error caused by failure, designing a direct adaptive control signal based on RBF neural network to eliminate the error partly or all. In order to reduce the dynamic inversion error rapidly, we usually use neural network compensator of high learning rate, which may saturate the control authority and influence the stability of system. To address this issue, we design an indirect adaptive controller based on RBF neural network to perform an on-line estimation of the plant dynamics due to failure, so that the learning rate can be reduced, stability of the system is guaranteed. The results of simulation show the effectiveness and feasibility of the method.