融合无量纲指标与信息熵的不同转速下旋转机械故障诊断  

Rotating machinery fault diagnosis under different rotating speeds based on fusion of non-dimensional index and information entropy

作　　者：陈仁祥 吴昊年 韩彦峰[2] 赵玲 吴志元 陈里里[1] CHEN Renxiang;WU Haonian;HAN Yanfeng;ZHAO Ling;WU Zhiyuan;CHEN Lili(School of Mechatronic and Vehicle Engineering, Chongqing Jiaotong University, Chongqing 400074, China;State Key Lab of Mechanical Transmission, Chongqing University, Chongqing 400030, China)

机构地区：[1]重庆交通大学机电与车辆工程学院,重庆400074 [2]重庆大学机械传动国家重点实验室,重庆400030

出　　处：《振动与冲击》2019年第11期219-227,共9页Journal of Vibration and Shock

基　　金：机械传动国家重点实验室开放基金(SKLMT-KFKT-201710);国家自然科学基金(51305471; 51605053);中国博士后科学基金(2014M560719);重庆市基础与前沿研究计划资助项目(cstc2015jcyjA70010,cstc2016jcyjA0526);重庆市留学人员回国创业创新支持计划创新项目(CX2018116);重庆市技术创新与应用示范项目(cstc2018jscx-msybX0012);城市轨道交通车辆系统集成与控制重庆市重点实验室开放基金(CKLURTSIC-KFKT-201809);交通工程应用机器人重庆市工程实验室开放基金(CELTEAR-KFKT-201803)

摘　　要：针对不同转速下旋转机械故障的特征同尺度表征与诊断问题,提出了融合无量纲指标与信息爛的旋转机械故障诊断方法。无量纲指标、信息爛等值与振动能量无关,取决于信号的分散程度与组分比率,对转速敏感性低,故利用无量纲指标与信息爛构建故障特征集,实现不同转速工况下故障特征同尺度定量表征;设计出基于核函数概率估计的故障敏感性指标算法,从所建立的故障特征集中选择对故障敏感性好的特征量构成表征能力更强的故障敏感特征集,并采用线性局部切空间排列(LLTSA)对其进行非线性降维与融合,获得分类特性好、受转速影响小的低维故障敏感特征集;最后,应用鲁棒性好的加权最近邻分类器(WKNNC)实现不同故障类型的诊断。对不同转速下齿轮箱故障进行诊断,结果证明了所提方法的可行性和有效性。Aiming at characterization at the same scale and diagnosis problems for rotating machinery fault features under different rotating speeds, a rotating machinery fault diagnosis method under different rotating speeds based on fusion of non-dimensional index and information entropy was proposed. It was shown that non-dimensional index and information entropy are not related to vibration energy, and they depend on vibration signal,s dispersion level to component ratio, they are less sensitive to rotating speed, so non-dimensional index and information entropy are used to construct fault feature set, and realize fault features 9 quantitative characterization at the same scale under different rotating speeds. The calculation method for fault sensitivity index was designed based on core function probability estimation to select features with better sensitivity to faults from the constructed fault feature set, and build a fault sensitive feature set with stronger characterization ability. The linear local tangent space arrangement ( LLTSA) was adopted to do nonlinear dimensional reduction and fusion for the fault sensitive feature set. Finally, different fault types were recognized using the weighted K- nearest neighbor classifier ( WKNNC ) with good robustness. This method was applied to diagnose gearbox faults under different rotating speeds. The results verified the feasibility and validity of the proposed method.

关 键 词：不同转速 无量纲指标 信息爛 故障诊断 

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]