自适应频域字典的机车轮对轴承稀疏诊断方法  

Adaptive frequency domain dictionary for sparse diagnosis of locomotive wheelset bearing

作　　者：张龙 赵丽娟 王朝兵 刘杨远 ZHANG Long;ZHAO Lijuan;WANG Chaobing;LIU Yangyuan(Key Laboratory of Conveyance and Equipment of the Ministry of Education,East China Jiaotong University,Nanchang 330013,China;CRRC QISHUYAN Co.,Ltd.,Changzhou 213011,China)

机构地区：[1]华东交通大学载运工具与装备教育部重点实验室,江西南昌330013 [2]中车戚墅堰机车有限公司,江苏常州213011

出　　处：《铁道科学与工程学报》2023年第4期1456-1468,共13页Journal of Railway Science and Engineering

基　　金：江西省自然科学基金资助项目(20212BAB204007);江西省研究生创新资金资助项目(YC2020-S335,YC2021-S422)。

摘　　要：稀疏分解是强噪声环境下故障特征提取的一种有效方法,构造与故障振动信号匹配的字典对稀疏分解效果至关重要。小波参数字典因小波的灵活多变性和良好的局部时频特性而被广泛应用于轴承故障诊断领域。然而,现有小波参数字典多是通过时域相关滤波法(CFA)以小波与故障冲击间的相关系数为指标获取字典原子的最优中心频率、阻尼比等参数,时域滤波耗时长、抗噪性差,相关系数指标没有考虑冲击发生的周期性特点,导致字典匹配性欠佳。针对上述问题,提出一种自适应频域滤波进行参数字典设计的机车轮对轴承故障诊断方法。该方法以新提出的时频域指标——加窗包络谱峭度(WESK)和相似度指标——皮尔逊相关系数(PCC)作为字典原子参数选取依据,以粒子群优化算法(PSO)优化的Morlet小波带通滤波器确定轴承故障产生的系统共振频率作为字典原子的中心频率,按照PCC值最大原则选取最优阻尼比完成字典原子的构造,改变时移变量张成小波字典后,结合正交匹配追踪算法(OMP)稀疏重构原始信号,提取故障特征频率。自制试验台数据以及机车轮对轴承的工程实际应用均验证了所提方法和新指标(WESK)的有效性和稳定性,诊断效果优于现有时域相关滤波法(CFA)以及常用可调品质因子小波变换法(TQWT),具有一定的工程实用价值。Sparse decomposition is an effective method for fault feature extraction under strong noise conditions.It is of great significance to construct a dictionary that matches the fault vibration signal with the results of the sparse decomposition.The wavelet parameter dictionary is widely used in the field of bearing fault diagnosis due to the flexibility and good local time-frequency characteristics of the wavelet.However,most existing wavelet parameter dictionaries obtain optimal center frequency,damping ratios of dictionary atoms via the time domain correlation filtering algorithm(CFA)method,where the correlation coefficient between wavelets and fault impulses is taken as an indicator.The CFA method is time-consuming and anti-noise;and the correlation coefficient indicator does not consider the periodic character of impulses,resulting in poor dictionary matching.To address these drawbacks,a parametric dictionary design based on adaptive frequency-domain filtering for locomotive wheelset bearing was proposed.It employed the novel time-frequency domain index with window envelope spectral kurtosis(WESK)and the similarity index Pearson correlation coefficient(PCC)as the optimization index.Further,the central frequency of the dictionary wavelet atom was determined through system resonance frequency of a bearing fault using Morlet wavelet filter that Optimized by particle swarm optimization(PSO)algorithm.Then,the most optimal damping ratio was obtained according to the principle of maximum PCC to accomplish the construction of dictionary wavelet atoms.Ultimately,by forming a wavelet parameter dictionary under various time-shifted parameters,the original signal was sparsely reconstructed by incorporating the orthogonal matching pursuit(OMP)algorithm to extract the fault feature.The validity and stability of the proposed method and the new index(WESK)were confirmed by the self-made test bench data,the Oriental Institute data,and the locomotive wheelset bearing engineering practical applications data.The diagnostic effect o

关 键 词：频域滤波 加窗包络谱峭度 字典构造 故障诊断 特征提取 

分 类 号：TH133.3[机械工程—机械制造及自动化]