基于ASMVMD和MOMEDA的齿轮特征提取方法  

Gear fault feature extraction method based on ASMVMD and MOMEDA

作　　者：唐贵基[1,2] 曾鹏飞 朱爽 TANG Guiji;ZENG Pengfei;ZHU Shuang(Department of Mechanical Engineering,North China Electric Power University,Baoding 071003,China;Hebei Key Laboratory of Health Maintenance and Failure Prevention of Electric Machinery Equipment,Baoding 071003,China)

机构地区：[1]华北电力大学机械工程系,河北保定071003 [2]河北省电力机械装备健康维护与失效预防重点实验室,河北保定071003

出　　处：《机电工程》2024年第12期2174-2184,共11页Journal of Mechanical & Electrical Engineering

基　　金：河北省自然科学基金资助项目(E2020502031)。

摘　　要：针对齿轮信号易被强噪声干扰,导致损伤特征难以提取的问题,提出了一种基于自适应逐次多元变分模态分解(ASMVMD)和多点最优最小熵解卷积(MOMEDA)的齿轮故障特征提取方法。首先,采用加权黑猩猩优化算法对SMVMD分解参数进行了自适应寻优,以SMVMD分解后各个通道的所有分量的平均包络谱峰值因子(Ec)之和的相反数作为寻优的适应度函数,确定了最大惩罚因子α和最大分解模态数k的最优值;然后,采用ASMVMD方法对齿轮多通道故障数据进行了自适应分解,根据Ec指标提取了各通道特定分量,并将这些分量相加,进行了信号重构;最后,采用MOMEDA解卷积处理了重构信号,进一步强化了齿轮故障的冲击特性,并利用包络谱分析解卷积信号,提取了齿轮的故障特征频率。研究结果表明:通过仿真信号和模拟实验信号的分析,可知利用ASMVMD-MOMEDA相结合的方法处理得到的信号降噪效果显著,能有效抑制无关干扰成分的影响,从包络谱中可以清晰地看到故障频率的前几阶倍频;与多元经验模态分解(MEMD)-MOMEDA相结合的方法进行对比,发现采用ASMVMD-MOMEDA方法得到的包络谱较MEMD-MOMEDA方法的谱线更加干净,各阶倍频更加明显,进一步证明ASMVMD-MOMEDA方法可以准确提取齿轮故障特征。Aiming to address the challenge of difficulty in extracting damage features from gear signals due to significant noise interference,a method combining adaptive successive multivariate variational mode decomposition(ASMVMD)and multipoint optimal minimum entropy deconvolution adjusted(MOMEDA)was proposed.Firstly,the optimization of SMVMD decomposition parameters was adaptively conducted using the weighted chimpanzee optimization algorithm.The optimal values for the maximum penalty factorαand the maximum decomposition mode number k were determined by using the negative sum of the average crest factor of envelope spectrum(Ec)of each channel component after successive multivariate variational mode decomposition as the fitness function for optimization.Then,the ASMVMD method was used to adaptively decompose the multi-channel fault data of gears.Specific components were then extracted from each channel based on the Ec index,and these components were summed for signal reconstruction.Finally,the signal was reconstructed using MOMEDA deconvolution processing to further enhance the impact characteristics of gear faults,and envelope spectrum analysis was used to deconvolve the signal and extract the characteristic frequency of gear faults.The research results show that through the analysis of simulated signals and experimental signals,it can be concluded that the combined method of ASMVMD-MOMEDA has a significant denoising effect on the signal.This effectively suppresses the impact of irrelevant interference components,allowing the fault frequencies and their first few harmonics to be clearly observed in the envelope spectrum.When comparing with the method that combines multivariate empirical mode decomposition(MEMD)-MOMEDA,the envelope spectrum obtained by the ASMVMD-MOMEDA method is cleaner and its harmonics are more distinct.This further demonstrates that the ASMVMD-MOMEDA method can accurately extract gear fault characteristics.

关 键 词：齿轮损伤特征 故障特征提取 自适应逐次多元变分模态分解 多点最优最小熵解卷积 多通道 解卷积 包络谱峰值因子 信号重构 

分 类 号：TH132.41[机械工程—机械制造及自动化] TH165.3