优化FEEMD与相似度量的滚动轴承故障特征提取  

Fault feature extraction of rolling bearing based on optimized FEEMD and similarity measure

作　　者：马军[1,2] 李祥 秦娅 熊新[1,2] MA Jun;LI Xiang;QIN Ya;XIONG Xin(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Intelligent Control and Application,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650500 [2]昆明理工大学云南省智能控制与应用重点实验室,昆明650500

出　　处：《兵器装备工程学报》2025年第3期252-266,共15页Journal of Ordnance Equipment Engineering

基　　金：国家自然科学基金项目(62163020、62173168);云南省基础研究计划项目(202401AW070014)。

摘　　要：针对快速集合经验模态分解(fast ensemble empirical mode decomposition,FEEMD)方法信噪分离不准确的问题,提出一种优化FEEMD与相似度量的滚动轴承故障特征提取方法。该方法建立基于最小包络熵的目标优化函数,并利用北方苍鹰优化算法(northern goshawk optimization,NGO)确定FEEMD的模型参数后,利用优化后的FEEMD将滚动轴承振动信号分解为多个本征模态函数分量和残余项,融合形态波动一致性偏移距离(morphology fluctuation conformance deviation distance,MFCDD)指标筛选有效分量进行重构,最后对重构信号进行Hilbert包络解调,完成滚动轴承故障特征提取。试验结果表明,所提方法相比变分模态分解方法、峭度分量选取方法、改进的完备集合经验模态分解联合豪斯多夫距离与峭度值方法,信噪比分别平均提升了1.75、12.2639、2.0605 dB,均方根误差分别降低了0.0078、0.0430、0.0656,能够更加清晰、全面地提取出故障特征频率及其倍频。Aiming at the problem of inaccurate signal-to-noise separation in the Fast Ensemble Empirical Mode Decomposition(FEEMD)method,a rolling bearing fault feature extraction method based on optimized FEEMD and similarity measure is proposed.This method established an objective optimization function based on the minimum envelope entropy,and used the Northern Goshawk Optimization(NGO)algorithm to determine the model parameters of FEEMD.After that,the optimized FEEMD was used to decompose the rolling bearing vibration signal into multiple intrinsic mode function components and residual terms.The morphological fluctuation consistency deviation distance(MFCDD)index was used to screen the effective components for reconstruction.Finally,the reconstructed signal was demodulated by Hilbert envelope,and the fault feature extraction of rolling bearing was completed based on the demodulated envelope spectrum.The experimental results show that compared with the variational mode decomposition method,kurtosis component selection method and improved complete ensemble empirical mode decomposition combined with Hausdorff distance and kurtosis value,the proposed method improves the signal-to-noise ratio by 1.75 dB,12.2639 dB and 2.0605 dB respectively,and the root mean square error is reduced by 0.0078,0.0430 and 0.0656 respectively.It can extract the fault characteristic frequency and its frequency doubling more clearly and comprehensively.

关 键 词：滚动轴承 故障特征提取 集合经验模态分解 相似性 北方苍鹰算法 

分 类 号：TN911.7[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]