摘要
针对航空发动机多源信息冗余的健康参数估计问题,提出基于信息熵融合的特征提取方法,将其用于涡轴发动机气路分析中。分别采用近似熵和互信息熵2种方法分析不同故障模式下传感器参数,对特征信息进行融合提取,根据2种信息熵不同特点将每种故障模式下传感器参数分成强、弱2类,利用弱特征信号构建虚拟传感器,最后通过简约强特征信息和虚拟传感器信息解决最小二乘支持向量回归机的样本稀疏性问题,实现健康参数蜕化估计。仿真结果表明采用信息熵融合的特征提取方法有效地减少了输入参数维数,简约了特征样本,从而提高了发动机健康估计能力。
Aiming at the problem of health parameter estimation of aero-engine with multi-source information, a hybrid diagnosis method based on feature extraction of information entropy is proposed. Because of the information redundancy in different types of measurements for fault diagnosis, the feature under each fault mode is extracted and divided into strong relative characteristic signal and weak one, and the weak one is used to set up virtual sensor; then sparse least squares support vector regression is used to estimate the degradation of health parameter according to the two classes of information. Simulation result on some turbo-shaft engines shows that the method can effectively reduce the characteristic parameters, decrease the input dimension and simplify the training samples for the health estimation.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2012年第1期13-19,共7页
Chinese Journal of Scientific Instrument
基金
南京航空航天大学基本科研业务费项目(V1042-021)资助
