关于多种模态分解方法的分离效果的差别探讨

发布时间：2019-03-26 10:14

【摘要】：经验模态分解(Empirical Mode Decomposition,EMD)是一种自适应信号分解方法,将数据从高频到低频分解成一系列的本征模式函数(Intrinsic Mode Functions,IMF)和一个余量。局域均值分解(Local Mean Decomposition,LMD)一定程度上解决了EMD方法的端点效应问题,但仍不容忽视。变模态分解(variational mode decomposition,VMD)解决了EMD方法在噪声恶劣背景下,IMF淹没在噪声背景中,导致不能得到信号特征分量的问题。多分别奇异值分解算法(Multi-resolution singular value decomposition,MRSVD)利用矩阵二分递推结构原理和SVD方法相结合,能够很好地把信号中微弱的细节信号和主体信号多层次体现出来,从而提取到其中隐含的信号特征。在此主要讨论EMD、LMD、VMD和MRSVD处理含噪信号时的效果差异,并对四种处理方法在滚动轴承故障振动信号的实际应用中出现的问题进行探讨。
[Abstract]:Empirical Mode decomposition (Empirical Mode Decomposition,EMD) is an adaptive signal decomposition method, which decomposes data from high frequency to low frequency into a series of intrinsic mode functions (Intrinsic Mode Functions,IMF) and a margin. Local mean decomposition (Local Mean Decomposition,LMD) solves the endpoint effect of EMD to a certain extent, but it can not be ignored. VMD (variable Mode decomposition) (variational mode decomposition,VMD) solves the problem that the IMF is submerged in the noisy background and the EMD method can not obtain the characteristic components of the signal in the background of bad noise. The Multiseparate singular value decomposition (Multi-resolution singular value decomposition,MRSVD) algorithm combines the principle of matrix dichotomy recursive structure with the SVD method, which can well represent the weak detail signal and the main signal in the signal. Thus, the implicit signal features are extracted. This paper mainly discusses the difference between EMD,LMD,VMD and MRSVD in processing noisy signals, and discusses the problems of four processing methods in the practical application of rolling bearing fault vibration signals.
【作者单位】：昆明理工大学信息工程与自动化学院;
【基金】：国家自然科学基金面上项目(60872157;61271007)
【分类号】：TH133.33

【参考文献】