KurVMDPgram:一种用于旋转机械故障诊断的信号分解算法  

KurVMDPgram:Signal Decomposition Algorithm for Fault Diagnosis of Rotating Machinery

作　　者：李华[1,2] 王天杨 张飞斌[2] 褚福磊 LI Hua;WANG Tianyang;ZHANG Feibin;CHU Fulei(State Key Laboratory of Public Big Data,Guizhou University,Guiyang 550025;Department of Mechanical Engineering,Tsinghua University,Beijing 100084)

机构地区：[1]贵州大学省部共建公共大数据国家重点实验室,贵阳550025 [2]清华大学机械工程系,北京100084

出　　处：《机械工程学报》2025年第4期11-23,共13页Journal of Mechanical Engineering

基　　金：国家自然科学基金青年科学基金(52205092);中-波合作交流(52161135101);中国博士后科学基金面上(2023M731939);贵州大学校级人才((2021)27,[2020]25)资助项目。

摘　　要：变分模态分解(Variational mode decomposition,VMD)是一种在旋转机械故障诊断领域广泛应用的信号分解方法。然而,VMD的若干影响参数尤其是模态数和惩罚因子对其分解性能影响很大,但需要事先确定,使得其参数优化成为研究热点。首先,结合小波包变换(Wavelet packet transform,WPT)的二元分解模式和VMD的维纳滤波特性构造一种新的信号分解方法,即变分模态分解包(VMD packet,VMDP)。然后,构造平铺结构的VMDP,提出VMDPgram方法。VMDPgram的提出是为了利用VMD实现信号的WPT模式的分解。在VMDPgram中,原始信号被分解为一定层次的若干子分量(Sub-component,SC),其中每个VMD生成两个SC,由此可将多参数问题简化为单参数问题。利用峭度指标对VMDPgram中的各VMD的惩罚因子进行优化,进而提出KurVMDPgram方法。随后,提出累积峭度指标的策略选取KurVMDPgram的最佳SC,可以解决共振频带分解到不同SC中的情形而不需要预先对其进行判定。最后,用仿真案例和实际轴承故障案例证明了KurVMDPgram的有效性和优越性。Variational mode decomposition(VMD)is a signal decomposition method which is widely used in the field of rotating machinery fault diagnosis.However,several influencing parameters of VMD,especially the mode number and penalty factor,need to be determined in advance,which makes the parameter optimizations of VMD become a research hotspot.A new signal decomposition method,namely variational mode packet decomposition(VMD Packet,VMPD),is constructed by combining the binary decomposition mode of wavelet packet transform(WPT)and the Wiener filtering characteristics of VMD.Then,a tiled VMDP is constructed,and the VMDPgram method is proposed.The VMDPgram is proposed to use VMD to realize the decomposition of the WPT mode of the signal.In VMDPgram,the original signal is decomposed into a certain level of sub-components(SCs),where each VMD generates two SCs,thereby simplifying the multi-parameter problem into a single-parameter problem.The penalty factor of each VMD in the VMDPgram is optimized by using the kurtosis index,and then the KurVMDPgram method is proposed.Then,the strategy of cumulative kurtosis index is proposed to select the optimal SC of KurVMDPgram,which can solve the situation that the resonant frequency band is decomposed into different SCs without pre-judging them.Finally,the validity and superiority of KurVMDPgram are proved by simulation case and actual bearing fault cases.

关 键 词：变分模态分解 小波包变换 VMDPgram KurVMDPgram 累积峭度指标 故障诊断 

分 类 号：TH165[机械工程—机械制造及自动化] TH17