多包传输机制下航空发动机分布式控制系统故障检测

考虑航空发动机分布式控制系统中,传感器与控制器间采用多包传输机制,传感器节点将测量数据封装为多个数据包发送给控制器,在每一采样时刻,同时产生的多个数据包只有一个能够发送,不能发送的数据包则被数据总线丢弃,针对这种情况,设计了神经网络预估器,用预估值来更新未被传输的数据包,使多包传输分布式控制系统"变为"单包传输分布式控制系统,建立了基于状态预估的多包传输分布式控制系统模型,构建了相应的故障观测器,将观测器误差系统看成一个离散周期系统,并应用周期系统稳定性理论推导了神经网络预估器和故障观测器能够稳定工作的条件。最后,给出了具体的故障检测步骤,仿真结果验证了所提出的故障检测算法的有效性。

A Fault Detection for Aero Engine in the Distributed Control System under Condition of Multiple Packet Transmission

Abstract:Aimed at the problems that the information between the sensors and the controller is transmitted in the form of multiple packet in aero engine distributed control system (DCS), the sensor nodes encapsulate the data measured into several data packets to transmit to the controller, at each sampling time, and a part of the data packets could be sent while the data packets failed to send would be discarded, in view of the mentioned above conditions, a neural network predictor is designed to update the packets failed to be transmitted by a predicted value, making the aero engine DCS with multiple packet transmission an aero engine DCS with single packet transmission. And then a multiple packet transmission DCS model based on state estimation is established, and an observer is constructed. The observer error system could be regarded as a discrete periodic system, and the periodic system stability theory is utilized for deriving the stable working conditions of the neural network predictor and fault observer. Finally, the specific fault detection steps are given, and the simulation results show that the proposed fault detection method is valid.