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)、Margenau-Hill(MHD)时频分析与双向二维主成分分析(Two-directional,Two-dimensional PCA,TD-2DPCA)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MHD处理后表征成振动谱图像;在此基础上,对生成的内燃机KVMD-MHD振动谱图像采用双向二维主成分分析形成编码矩阵,并采用最近邻分类器(KNNC)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,将该方法应用在气阀机构4种工况下的缸盖表面振动信号诊断实例中,结果表明:该方法不仅改进了传统图像模式识别中的特征参数提取方法,而且能很好地消除时频分布中的交叉干扰项,使各时频分量物理意义明确,能有效诊断出内燃机气阀机构故障,为内燃机振动诊断探索了一条新途径。
In order to direct diagnose and recognize of IC engine vibration spectrum image,based on improved variational mode decomposition（VMD）,Margenau-Hill（MHD）time-frequency analysis and Two-directional,Two-dimensional PCA（TD-2 DPCA）IC engine vibration spectrum image recognition and diagnosis method are proposed in this paper.Firstly,for VMD layers selection problem during the decomposition process,a center frequency of selected VMD decomposition method（KVMD）is proposed,then the vibration signal of IC engine is decomposed into a set of single component modal signals by KVMD,and for each single component of the signal using MHD then characterized as vibration spectral image;on this basis,to get code matrix,TD-2 DPCA are used to IC engine KVMD-MHD vibration spectral image,and using the KNNC to the code matrix for pattern recognition in order to realize the automatic diagnosis of vibration spectrum image.Finally,the method is used in IC engine fault diagnosis,the results showed that：the method not only improved the traditional pattern recognition characteristic parameters extraction methods,but also can eliminate cross terms in the time-frequency distribution,the clear physical meaning of frequency components,can effectively diagnose of valve mechanism of internal combustion engine fault,fault recognition accuracy up to 100%,explored a new way for the IC engine vibration diagnosis.
出处
《振动工程学报》
EI
CSCD
北大核心
2017年第4期688-696,共9页
Journal of Vibration Engineering
基金
国家自然科学基金资助项目(51405498)
中国博士后科学基金资助项目(2015M582642)
陕西省自然科学基金资助项目(2013JQ8023)
