发动机数控系统智能容错技术的半物理仿真研究

A Matter in the Loop Simulation of Intelligent Fault Tolerance Technology in Digital Control System of Engine

提出基于自联想神经网络的发动机数控系统的智能容错技术。该方法不依赖系统模型 ,只需要发动机的测量值来离线训练网络。若发生性能蜕化 ,该方法能自动切入故障诊断与最优估计的综合逻辑。文中开展了半物理仿真实验 ,实验结果充分证明了该技术能适应实验现场多变的环境 ,能正确诊断发动机传感器故障并提供解析余度 。

An intelligent fault tolerance technology based on autoassociative neural network is presented, which doesn′t depend on model, and needs only engine measurement to train the network, then it can work on line. If there is performance degeneration, a compensation algorithm will be automatically used. A matter in the loop simulation by using engine model and real control system is developed. The simulational results show that the proposed intelligent fault tolerance technology has good robust performance, and can detect sensor failures on time and present good accommodation to replace the failed sensors, even there is a lot of noise and performance degeneration