摘要
为了直接对内燃机振动谱图像进行诊断识别,提出一种基于改进变分模态分解(VMD)、伪魏格纳时频分析(PWVD)与局部非负矩阵分解(LNMF)的内燃机振动谱图像识别诊断方法。该方法首先针对VMD分解过程中的层数选取问题,提出了一种中心频率筛选的VMD分解层数改进方法(KVMD),然后将内燃机振动信号利用KVMD分解成一组单分量模态信号,并对生成的各个单分量信号进行伪魏格纳分析处理后表征成振动谱图像;在此基础上,对生成的内燃机KVMD-PWVD振动谱图像分别采用非负矩阵分解(NMF)和LNMF形成编码矩阵,并采用最近邻分类器、朴素贝叶斯分类器和支持向量机对上述编码矩阵直接进行模式识别,以实现内燃机振动谱图像的自动诊断。最后,将该方法应用在内燃机故障诊断实例中,结果表明:该方法改进了传统图像模式识别中的特征参数方法,能有效诊断出内燃机气门间隙故障,三种分类器识别精度均大于93%,其中支持向量机的分类精度最高,达到99.8%,且采用LNMF形成的编码矩阵识别精度整体高于NMF,为内燃机振动诊断探索了一条新途径。
In order to directly diagnose and recognize IC engine vibration spectrum images,based on the improved variational mode decomposition( VMD),pseudo Wigner-Ville time-frequency analysis( PWVD) and local non-negative matrix factorization( LNMF),an IC engine vibration spectrum image recognition and diagnosis method was proposed.Aiming at the VMD layers selection during the decomposition process,a center frequency selected VMD decomposition method( KVMD) was proposed,then the vibration signal of IC engine was decomposed into a set of single component modal signals by KVMD,and each single component of the signal,by using PWVD,was characterized as a vibration spectral image. On this basis,to get a code matrix,the non-negative matrix factorization( NMF) and local non-negative matrix factorization( LNMF) were used to the IC engine KVMD-PWVD vibration spectral image,and the KNNC,NBC and SVM were applied to the code matrix for pattern recognition in order to realize the automatic diagnosis of vibration spectrum images. The method has been used in practical IC engine fault diagnosis and the results show that the method improves the traditional characteristic parameters of image pattern recognition,it can effectively diagnose the IC engine valve clearance fault,the recognition accuracy of the three classifiers is all not less than 93%,the SVM has the highest classification accuracy which reaches 99. 8%,and the code matrix using the LNMF has higher accuracy than the NMF.The method explores a new way for the IC engine vibration diagnosis.
出处
《振动与冲击》
EI
CSCD
北大核心
2017年第2期45-51,94,共8页
Journal of Vibration and Shock
基金
国家自然科学基金青年基金项目(51405498)
陕西自然科学基金项目(2013JQ8023)
中国博士后基金资助项目(2015M582642)
关键词
内燃机
故障诊断
时频分布
特征提取
局部非负矩阵
internal combustion(IC) engine
fault diagnosis
time-frequency distribution
feature extraction
local non-negative matrix factorization(LNMF)
