In this paper, a neural network-based event-triggered fault detection scheme is addressed within the finite-frequency domain for a class of nonlinear Markov jump system. Initially, an approximation model based on multilayer neural network to alternate the nonlinear Markov jump system is constructed. For the purpose of saving the communication network bandwidth, a transmission mechanism based on the event-triggered strategy is subsequently applied in which each signal is transmitted depending on the designed condition rather than the sampling period. Further, two theorems with considering the signal frequency and the applied event-triggered mechanism are derived which guarantee the fault sensitivity as well as disturbance attenuation for the augment systems in certain frequency ranges. Then, the desired filters can be synthesized by the linear solvable conditions that are derived with the aid of the previous theorems and some novel decoupling techniques. Eventually, the proposed algorithm’s efficiency is shown by a presented computational example.

本文针对一类非线性马尔可夫跳跃系统，在有限频域内提出了一种基于神经网络的事件触发故障检测方案。最初，建立了一个基于多层神经网络的非线性马尔可夫跳跃系统交替逼近模型。为了节省通信网络带宽，随后应用了基于事件触发策略的传输机制，其中，根据设计条件而不是采样周期来传输每个信号。此外，推导了两个考虑信号频率和所应用的事件触发机制的定理，这些定理在一定的频率范围内保证了增强系统的故障敏感性以及干扰衰减。然后，可以通过线性可解条件来合成所需的滤波器，这些条件是借助先前的定理和一些新颖的去耦技术得出的。最终，该算法的效率通过给出的计算示例得到了证明。