Active Fault-tolerant Control Method Of A Classified Dynamic Non-linear System

In this paper, an active fault-tolerant control algorithm mentioned for a nonlinear systems with bounded uncertainties will be made further simulation researches. For MIMO system, a method of learning a multidimensional nonlinear function using many modified low dimensional cerebellar model articulation controller (CMAC) is presented. The fault-tolerant control law can be realized by utilizing the effective control law reconfiguration strategy based on sliding model control algorithm, using identification algorithm of the improved CMAC based on balanced learning estimates faults function. In the past simulation researchs, only for single failure such as single abrupt filure or single incipient failure had been researched, but for multi failures such as consecutive abrupt faults, consecutive incipient faults or mixed incipient and abrupt failures cases haven't been discussed deeply. The purpose of successful fault tolerance is in real applications. Thinking about the complexity of real applications, it is necessary for a series of research works that is devoted to more general failure cases is reported.For fault-tolerant control, it is necessary to think about two important problems, one is on-line fault diagnosis and fault detection strategy. The other is how to reconfigure the control law. This paper discusses factors of CMF situations likelihood and the approach and algorithm of MVDS. In this paper, using improved CMAC neural network online learning approach improves the on-line learning speed and accuracy of fault diagnose of nonlinear system, The credit assignment CMAC uses the learned times of the addressed hypercubes as credibility. The more times the hypercube has been trained, the more accurate the weight is. The updating algorithm improved the learning speed and accuracy of the neural network without considering the balance of"learned knowledge"and"learning knowledge". The improved credit assignment CMAC neural network, reserved the idea of the credit assignment weight and introduced into the concept of"learning balance", improve on the credit assignment regulations. The values of the correcting errors are proportional to the inversion of the pth power of learned times of addressed hypercubes, and the learning speed and accuracy can be indeed improved when p is at the optimal value. The fault's estimator is designed by using this algorithm. When the couple fault can be expressed combination of the fault noise, this method can be used on the decoupling of the couple fault.In a word, on the basis of single fault theory for nonlinear systems, we proposed ICA-CMAC as the measure of on-line fault diagnosis to design on-line fault estimator in this paper. Under various multi failures scenarios, we can study unknown faults fast and truly, and the discrete-time sliding model control algorithm is applied to reconfigure the control law to on-line fault-tolerant on the dynamic nonlinear system.