基于HRCMFDE、LS、BA-SVM的行星齿轮箱故障诊断  

Fault diagnosis of planetary gearboxes based on HRCMFDE、LS and BA-SVM

作　　者：庄敏[1] 李革[2] 范智军 孔德成 ZHUANG Min;LI Ge;FAN Zhi-jun;KONG De-cheng(Intelligent Manufacturing College,Hangzhou Polytechnic,Hangzhou 311402,China;School of Mechanical Engineering and Automation,Zhejiang Sci-Tech University,Hangzhou 310018,China;College of Mechanical and Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China;Zhengzhou Machinery Research Institute Co.,Ltd.,Zhengzhou 450052,China)

机构地区：[1]杭州科技职业技术学院智能制造学院,浙江杭州311402 [2]浙江理工大学机械与自动控制学院,浙江杭州310018 [3]河南工业大学机电工程学院,河南郑州450001 [4]郑州机械研究所有限公司,河南郑州450052

出　　处：《机电工程》2022年第11期1535-1543,共9页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金资助项目(51775516,51375459);浙江大学访问学者资助项目(FX2018140)。

摘　　要：针对行星齿轮箱的特征提取以及故障识别问题,提出了一种基于混合精细复合多尺度波动散布熵(HRCMFDE)特征提取、拉普拉斯分数(LS)特征降维优化和蝙蝠算法优化支持向量机(BA-SVM)故障识别的行星齿轮箱故障诊断方法。首先,提出了一种新的时间序列复杂度测量方法—HRCMFDE(其由5种不同粗粒化方式的RCMFDE组成,具备更全面和可靠的特征提取性能),用于从振动信号中挖掘出反映行星齿轮箱状态的故障信息,构成初始的混合故障特征;然后,考虑到由HRCMFDE组成的故障特征具有较高的维数和冗余,利用LS对初始特征进行了优化,生成了低维的敏感特征;最后,利用基于蝙蝠算法优化的支持向量机,对行星齿轮系不同故障特征向量进行了训练和分类,利用真实故障数据集对基于HRCMFDE、LS、BA-SVM的方法进行了验证。研究结果表明:利用行星齿轮箱数据集对该方案进行的有效性实验,能够准确地识别出齿轮箱的不同故障,其单次分类的准确率达到了98.13%,多次分类的平均准确率也优于对比方法;该结果验证了基于混合精细复合多尺度波动散布熵特征提取的有效性,采用该方法能够对行星齿轮箱的故障进行诊断。Aiming at the question of characteristics extraction and fault diagnosis of planetary gearbox,a fault diagnosis solution based on hybrid refined composite multi-scale fluctuation dispersion entropy(HRCMFDE),Laplacian score(LS)and bat algorithm optimized support vector machine(BA-SVM)was raised.Firstly,HRCMFDE,a new time series complexity measurement method,was presented.It was composed of five RCMFDEs with different coarse grained methods,which had more comprehensive and reliable feature extraction performance,and was used to mine the fault information reflecting the state of planetary gearbox from the vibration signal to form the initial mixed fault feature.Then,considering that the fault features composed of HRCMFDE had high dimension and redundancy,Laplacian score was used to optimize the initial features to generate low-dimensional sensitive features.Finally,bat algorithm optimized support vector machine(BA-SVM)was used to train and classify different fault feature vectors of the planetary gear train,and the methods based on HRCMFDE,LS,and BA-SVM were verified by using the real fault data set.The results show that the effectiveness experiment of the proposed scheme using the planetary gearbox data set can accurately identify the different faults of the gearbox,the accuracy of single classification is 98.13%,the average accuracy of multiple classification is also better than the comparison method.The results verify the effectiveness of the hybrid refined composite multi-scale fluctuation dispersion entropy feature extraction,and can provide a supplementary method for fault diagnosis of planetary gearbox.

关 键 词：特征提取 特征降维优化 故障分类识别 混合精细复合多尺度波动散布熵 拉普拉斯分数 蝙蝠算法优化支持向量机 

分 类 号：TH132.425[机械工程—机械制造及自动化] TH17