基于离散小波变换和Kalman滤波的直升机主减智能状态预测  

Intelligent condition prognostic for a helicopter main gearbox using discrete wavelet and Kalman filtering

作　　者：刘立生[1,2] 杨宇航[2] 

机构地区：[1]南京理工大学自动化学院,南京210094 [2]总参谋部陆航研究所可靠性室,北京101121

出　　处：《振动与冲击》2012年第17期159-164,共6页Journal of Vibration and Shock

基　　金：军内重点科研项目

摘　　要：主减速器(简称"主减")是直升机传动系统的关键部件,它常处于高转速高负荷的恶劣环境下,对其运行状态进行预测,于直升机的安全性来说至关重要。鉴于此,提出了一种离散小波变换(DWT)、Kalman滤波以及Elman神经网络相结合的直升机主减智能状态预测系统:DWT使用"db44"母小波对振动信号进行分解提取特征向量,Kalman滤波对未来各时刻的特征向量进行预测,Elman神经网络对预测值进行故障辨识和分类。在Kalman滤波算法中,提出了一种新的预测算法,并用实验对该算法组成的系统进行验证,结果表明:该Kalman滤波算法预测效果好,更适用于对主减的特征向量进行预测;离散小波变换(DWT)、Kalman滤波以及Elman神经网络相结合组成的智能状态预测系统是可行的,它能很好地对主减的未来状态进行预测。Main gearbox （MGB） is a key component of a helicopter transmission system, it often runs under atrocious conditions of high rotating speed and high burden, it is very important to perform condition prognostic for safety of a helicopter. An intelligent condition prognostic system for helicopter MGB was presented here using discrete wavelet transformation （DWT）, Kalman filtering and Elman neural network. The mother wavelet of Daubechies 44 （db44） was selected to extract feature vectors in the process of DWT. Kalman filtering was used for feature vector prognostic, and Elman neural network was taken for fault discrimination and classification. In the algorithm of Kalman filtering, a new prognostic method was proposed, and it was verified with tests. The results indicated that the prognostic outcome of Kalman filtering with this method is better, it is more applicable for prognostic of feature vectors, and the intelligent condition prognostic system composed of DWT, Kalman filtering and Elman neural network is feasible, it can predict the future condition of a helicopter MGB accurately.

关 键 词：主减速器 离散小波变换(DWT) Parseval定理 KALMAN滤波 ELMAN神经网络 

分 类 号：V240.2[航空宇航科学与技术—飞行器设计] TH132[机械工程—机械制造及自动化]