涡轴发动机自适应混合诊断模型高斯加权聚类方法  

Gaussian weighted sum clustering method of adaptive hybrid diagnostic model for turbo-shaft engine

作　　者：佘云峰[1] 黄金泉[1] 鲁峰[1] 

机构地区：[1]南京航空航天大学能源与动力学院,南京210016

出　　处：《航空动力学报》2011年第5期1178-1184,共7页Journal of Aerospace Power

摘　　要：针对整个飞行包线内涡轴发动机健康参数估计问题,提出基于高斯加权聚类的机载自适应混合模型建立方法.机载自适应混合模型由卡尔曼滤波器和神经网络组成,由于飞行数据样本庞大,采用高斯加权模型对涡轴发动机飞行数据进行实时聚类,利用聚类数据更新神经网络权重,并实现自适应混合模型工作范围的自动扩展.仿真结果表明了该方法的有效性,采用高斯加权聚类的自适应混合模型提高了全包线内性能跟踪的精度.A Gaussian weighted sum clustering method of on-board adaptive hybrid model was proposed for health parameter estimation of turbo-shaft engine in the full flight envelope. The on-board adaptive hybrid model was composed of Kalman filter and neural network. Gaussian weighted sum model was used to cluster a huge amount of flight data in real-time and update the neural network weighting. The working range of adaptive hybrid model was increased automatically with this method. The simulation results show the validity of the proposed method. Adaptive hybrid model based on Gaussian weighted sum clustering increases the accuracy for turbo-shaft engine performance tracking in the full flight en- velope.

关 键 词：涡轴发动机 高斯加权模型 机载自适应混合模型 性能参数估计 卡尔曼滤波 神经网络 

分 类 号：V233.7[航空宇航科学与技术—航空宇航推进理论与工程]