航空发动机自适应传感器故障重构方法研究  

Research on Adaptive Sensor Fault Reconstruction Method for Aero-engine

作　　者：姚凯翔 彭靖波[1] 谢寿生[1] 王立国[2] 

机构地区：[1]空军工程大学航空航天工程学院 [2]中国人民解放军95507

出　　处：《计算机仿真》2016年第9期137-141,共5页Computer Simulation

基　　金：国家自然科学基金(51506221);陕西省自然科学基础研究计划青年人才项目(2015JQ5179)

摘　　要：当航空发动机控制系统传感器故障时,为保证发动机正常工作,需进行故障诊断与故障隔离,并实现重构控制。为了提高传感器故障重构控制的准确性,提出一种采用自适应模拟退火遗传神经网络(ASAGANN)的传感器重构方法。为能准确、及时发现智能传感器故障,采用序贯概率比阈值判别法对故障进行检测;利用改进的神经网络对发动机试车过程进行辨识,建立高精度的传感器正常工作模型,提高了故障重构的准确性;通过三种传感器不同故障的数值仿真表明,上述方法可以实现对传感器故障的准确检测,并完成对传感器的故障重构控制,证明故障重构方法有效。When faults occur in the sensors of aero-engine control system, the fault diagnosis and isolation progress should be conducted to ensure the stable operation of the closed-loop system, and the reconstruction control is desired to be achieved finally. A sensor fault reconstruction method based on adaptive simulated annealing genetic algorithm neural network （ASAGANN） was proposed to promote the accuracy of the fault reconstruction control. To monitor the sensor faults accurately and timely, a sequential probability ratio diseriminance method was utilized for fault detection. The improved neural network was employed to identify the hot-firing test progress, and the high precision normal working model was established, which can boost the accuracy of fault reconstruction. The numerical simulation was carried out with three types of sensors and with different faults. The results show that the proposed method can detect the sensor faults accurately. And the ideal reconstruction control effect can be achieved, which verifies the effectiveness of the proposed method.

关 键 词：传感器 故障重构 神经网络 序贯概率比 模型辨识 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP391.9[自动化与计算机技术—计算机应用技术]