Neural Network-Based Model-Free Adaptive Fault-Tolerant Control for Discrete-Time Nonlinear Systems With Sensor Fault
针对非线性单输入单输出无模型系统，仅利用输入输出数据，提出一种基于神经网络的故障检测、估计与容错控制方法，通过动态线性化数据模型和时变残差阈值实现。
In this paper, the main focus is to cope with the fault detection and estimation (FDE) and fault-tolerant control (FTC) issues of nonlinear single input single output model-free system (MFS), while only the input/output data are utilized. First, in accordance with the pseudo-partial-derivative approach, the original system is transformed into a compact form dynamic linearization data model, in which only one parameter is employed. Second, an estimator is developed to detect the fault. A key highlight is the design of a time varying residual threshold. Moreover, an online neural network (NN) approximator is utilized to learn the unknown fault dynamics and an FTC strategy is reconstructed based on the optimality criterion. In contrast to the previous methods, the main features of the proposed method are as follows: 1) the fault related problem is solved for MFS; 2) the number of system parameters is largely reduced; and 3) NNs are utilized to establish a novel fault estimation scheme. Finally, a numerical simulation is provided to show the effectiveness of the proposed FDE and FTC strategy.