非负矩阵分解在发动机故障特征提取中的应用  

Feature extraction for engine fault diagnosis by utilizing adaptive multi-scale morphological gradient and non-negative matrix factorization

作　　者：张培林[1] 王怀光[1] 张磊[2] 王卫国[2] 李兵[3] 

机构地区：[1]军械工程学院七系,河北石家庄050003 [2]军械工程学院科研部,河北石家庄050003 [3]军械工程学院四系,河北石家庄050003

出　　处：《振动工程学报》2013年第6期944-950,共7页Journal of Vibration Engineering

基　　金：国家自然科学基金资助项目(51205405)

摘　　要：信号处理与特征参数提取是发动机故障诊断的核心和关键。提出了采用自适应多尺度形态梯度算法对信号进行处理,综合利用小尺度下能保留信号细节和大尺度下抑制噪声能力强的优点,能够在强噪声背景下有效地提取振动信号中能够反映发动机工作状态的有用分量;在此基础上提出采用非负矩阵分解的特征提取方法对信号进行压缩,计算用于发动机故障诊断的特征参量。结果表明:与传统的信号处理与特征参量提取方法相比,所提的方法具有更高的分类精度,为准确判断发动机故障状态提供了一种行之有效的新方法。Signal processing and feature extraction are key steps for engine fault diagnosis. An adaptive multi-scale morphologi cal gradient （AMMG） algorithm, which has an edge of keeping the details of the original signal in small scale structure ele- ments and suppressing noises in large scale ones, is employed to extract the useful signal components hiding in the original vi- bration signals blurred by strong noises. Furthermore, the non-negative matrix factorization technique is utilized to calculate the features of the signal which is pre-processed by AMMG for engine fault diagnosis. The effect of application in practical en- gine fault diagnosis demonstrates that AMMG and the proposed feature extraction scheme have comparatively higher precision of fault classification, therefore providing an effective method for engine fault diagnosis.

关 键 词：故障诊断 发动机 特征提取 自适应多尺度形态梯度 非负矩阵分解 

分 类 号：TH165.3[机械工程—机械制造及自动化]