作者：韩中合, 赵文波, 朱霄珣, 李震涛
作者单位：华北电力大学动力工程系，河北 保定 071003
关键词：风机传动系统振动信号;虚假分量;互补集合平均经验模态分解;KL散度
摘要：
互补集合平均经验模态分解（complementary ensemble empirical model decomposition，CEEMD）作为一种时频特征分析方法，可以较好地提取复杂非线性非平稳信号的故障特征，但其存在虚假分量，很大程度限制诊断过程中的准确性。针对该问题，提出一种基于KL散度(Kullback-Leibler divergence, KLD)的CEEMD虚假分量识别方法（KL-CEEMD）。该方法在原有CEEMD方法基础之上，进一步计算各分量IMF与原信号之间的KL散度值，从而量化各分量与原信号之间的相关性。最后通过对各个IMF的KL散度值进行聚类分析，找出虚假分量和真实分量，最终解决CEEMD的虚假分量问题。为验证KL-CEEMD的有效性，研究搭建风力机传动系统振动试验台，基于该方法对实验台实验数据以及仿真数据进行验证性研究，最终证明所提方法可以很好改善CEEMD的虚假分量问题，能够有效提取出故障信号的真实特性。
Research on fault diagnosis method of fan drive system based on KL-CEEMD
HAN Zhonghe, ZHAO Wenbo, ZHU Xiaoxun, LI Zhentao
Department of Power Engineering, North China Electric Power University, Baoding 071003, China
Abstract: As a time-frequency feature analysis method, complementary ensemble empirical model decomposition (CEEMD) distinguishes the fault features of complex nonlinear non-stationary signals effectively. However, the problem of false components limit the accuracy of diagnosis process. In response to it, a CEEMD false component recognition method (KL-CEEMD) based on KL divergence is proposed, which precisely calculates the KL divergence value between the IMF and the original signal based on the original CEEMD method. Finally the the problem of false component of CEEMD is solved through clustering analysis of the KL divergence values of each IMF and distinguishing the false and true component. In order to verify the effectiveness of KL-CEEMD, a wind turbine transmission system vibration test bench is built. Based on the KL-CEEMD method, the experimental data and simulation data of the experimental platform are verified and researched. It is finally proved that the KL-CEEMD method improves the false component problem of CEEMD and effectively extracts the true characteristics of fault signal.
Keywords: fan drive system vibration signal;false component;complementary ensemble empirical model decomposition;Kullback-Leibler divergence
2022, 48(5):88-95,101 收稿日期: 2021-04-13;收到修改稿日期: 2021-05-31
基金项目: 河北省自然科学基金（E2019502080）
作者简介: 韩中合（1964-），男，河北衡水市人，教授，博士，研究方向为设备状态监测与故障诊断和两相流计算与测量
参考文献
[1] 蔡艳平, 范宇, 陈万, 等. 改进时频分析和特征融合在内燃机故障诊断中的应用[J]. 中国机械工程, 2020, 31(16): 1901-1911
[2] 林水泉. 基于旋转机械滚动轴承的时域故障诊断方法[J]. 自动化技术与应用, 2020, 39(8): 1-5, 35
[3] XIE Y X, YAN Y J, LI G F, et al. Scintillation detector fault diagnosis based on wavelet packet analysis and multi[J]. Journal of Instrumentation, 2020, 15(3): 1-13
[4] HUANG N E, SHEN Z. The empirical mode decomposition and the Hilbert spectrum for non-linear and non-stationary time series analysis[J]. Proceedings of The Royal Society A Mathematical Physical and Engineering Sciences, 1998, 454: 903-995
[5] HUANG N E. A new view of non-linear waves: the Hilbert spectrum[J]. Annual Review of Fluid Mechanics, 1999, 31(5): 417-457
[6] 周颖涛, 周绍骑, 姚远航. 减少模态混叠的改进EEMD算法[J]. 重庆理工大学学报(自然科学), 2015, 29(1): 111-114, 130
[7] 王志坚, 韩振南, 宁少慧, 等. 基于 CMF-EEMD 的风电齿轮箱多故障特征提取[J]. 电机与控制学报, 2016, 20(2): 104-111
[8] 胡君林, 赵炎堃. 基于改进HVD和包络谱的轴承故障诊断方法[J]. 机械, 2020, 47(1): 30-34
[9] YEH J R, SHIEH J S, HUANG N E. Complementary ensemble empirical mode decomposition: a novel noise enhanced data analysis method[J]. Advances in Adaptive Data Analysis, 2010, 2(2): 135-156
[10] 戴婷, 张榆锋, 章克信, 等. 经验模态分解及其模态混叠消除的研究进展[J]. 电子技术应用, 2019, 45(3): 7-12
[11] 徐统, 王红军, 宋智勇, 等. 基于K-L散度的VMD瞬时能量与PNN的滚动轴承故障诊断[J]. 电子测量与仪器学报, 2019, 33(8): 117-123
[12] BOUNOUA W, BENKARA A B, KOUADRI A, et al. Online monitoring scheme using principal component analysis through Kullback-Leibler divergence analysis technique for fault detection[J]. Transactions of the Institute of Measurement and Control, 2020, 42(6): 1225-1238
[13] JONES M, MARRON J, SHEATHER S. A brief suevey of bandwidth selecrion for density estimation[J]. Joumal of the American Statistical Association, 1996(91): 401-407
[14] 王志杰. 基于K-L散度的滚动轴承故障诊断及状态监测方法研究[D]. 北京: 北京交通大学, 2019.
[15] 朱霄珣, 周沛, 苑一鸣, 等. 基于KL-HVD的转子振动故障诊断方法究[J]. 振动与冲击, 2018, 37(16): 249-255
[16] 章永来, 周耀鉴. 聚类算法综述[J]. 计算机应用, 2019, 39(7): 1869-1882
[17] 张玉学, 潘宏侠, 安邦. 基于EEMD与FCM聚类的自动机故障诊断[J]. 中国测试, 2017, 43(3): 106-110
