机械科学与技术2018，Vol.37Issue(6)：828-833,6.DOI:10.13433/j.cnki.1003-8728.2018.0602
柴油机振动谱图纹理增强的NMF特征提取方法
Feature Extraction Method of Diesel Engine Vibration Signal with Texture Magnified Non-negative Matrix Decomposition
摘要
Abstract
In view of the difficulty in diesel engine vibration acceleration signal feature extraction,a kind of vibration signal time-frequency characteristics extraction method,which joined the local texture feature information of nonnegative matrix decomposition method,is put forward.Firstly,decompose the diesel engine cylinder head vibration acceleration signal by the matching pursuit (MP) algorithm,and obtain the 2-d time-frequency distribution diagrams by superimposing the decomposed signal Wigner-Ville distribution.Then,use NMF to calculate the corresponding low dimensional time-frequency characteristic parameters of the time-frequency distribution image.Next,introduce local binary pattern (LBP) operator pair recoding,frequency of the image gray level matrix to improve the effect of feature extraction of NMF.By the analysis of four different states of diesel engine vibration acceleration signal,which is corresponding to the valve clearance is normal,the valve clearance is too small or too big,and valve leakage,the results showed that the combination of LBP and NMF can effectively express the time-frequency characteristics of diesel engine cylinder head vibration acceleration signal,which can be used for the diagnosis of diesel engine valve fault accurately.关键词
柴油机/时频特征/匹配追踪/局部二值模式/非负矩阵分解Key words
diesel engine/time-frequency characteristics/local binary pattern/non-negative matrix decomposition/matching pursuit分类
能源科技引用本文复制引用
王旭,岳应娟,蔡艳平,郭雅云..柴油机振动谱图纹理增强的NMF特征提取方法[J].机械科学与技术,2018,37(6):828-833,6.基金项目
国家自然科学基金项目(51405498)与中国博士后基金项目(2015M582642)资助 （51405498）
