航空学报 > 2009, Vol. 30 Issue (10): 1795-1800

流体力学、飞行力学与发动机

航空发动机部件性能参数融合预测

鲁峰,黄金泉   

  1. 南京航空航天大学 能源与动力学院
  • 收稿日期:2008-07-21 修回日期:2009-01-16 出版日期:2009-10-25 发布日期:2009-10-25
  • 通讯作者: 鲁峰

Engine Component Performance Prognostics Based on Decision Fusion

Lu Feng, Huang Jinquan   

  1. College of Energy & Power Engineering, Nanjing University of Aeronautics and  Astronautics
  • Received:2008-07-21 Revised:2009-01-16 Online:2009-10-25 Published:2009-10-25
  • Contact: Lu Feng

摘要: 为了改善目前单独采用基于模型和数据驱动的部件健康参数预测精度,提高数据驱动方法的故障诊断的泛化能力,提出一种自调整决策融合机制,对航空发动机部件性能蜕化在连续蜕化空间进行融合诊断。传感器测量值同时输入到机载自适应模型和数据驱动的诊断模块中,分别利用卡尔曼滤波算法和自适应遗传算法优化的支持向量回归机(AGA-SVR)对主要部件性能进行预测,再利用自调整决策权重的量子粒子群寻优(QPSO)进行决策级融合诊断。以某型涡扇发动机为对象进行气路部件蜕化的仿真研究表明,与单独使用基于模型和数据驱动的诊断方法相比,采用决策融合机制有效地提高了部件故障诊断精度。

关键词: 航空发动机, 故障诊断, 自适应模型, 数据驱动, 卡尔曼滤波, 自适应遗传算法, 支持向量回归机, 量子粒子群寻优

Abstract: In order to improve the accuracy of aeroengine component fault diagnosis and the generalization ability which is either based on a model or data driven, a method of engine component fault diagnosis based on decision fusion of self tuning weight is proposed. The sensor output is fed to two different prognostic modules to acquire component performance parameters simultaneously. Estimated performance parameters are obtained by Kalman filtering and adaptive genetic algorthms support vector regression (AGA-SVR) respectively. Then the component performance parameters are fused by quantum particle swarm optimization (QPSO) of self tuning weight for fault decision. Simulation on a turbofan engine shows that compared with either of the model-based or data-driven module, this method improves significantly the accuracy of fault diagnosis.

Key words: aeroengine, fault diagnosis, self tuning model, data driven, Kalman filtering, adaptive genetic algorithms, support vector regression, quantum particle swarm optimization

中图分类号: