摘要
针对行星齿轮箱振动信号非线性、非平稳性特点及故障特征难以有效提取的问题,提出基于改进的集成经验模态分解(MEEMD)多特征融合和最小二乘支持向量机(LS-SVM)的行星齿轮箱故障诊断方法。首先,利用MEEMD分解不同工况下的齿轮振动信号,得到一系列固有模态分量。其次,根据相关系数筛选出3阶敏感模态分量并计算对应的样本熵和能量,将二者融合组成高维特征向量,最后,将融合特征向量作为最小支持向量机(LSSVM)的输入,对齿轮进行故障分类。在行星齿轮箱实验台上开展实验,与基于单特征构成的特征向量进行对比,并与概率神经网络(PNN)分类算法进行对比,结果验证该方法的有效性和优越性。
In order to solve the problem that the vibration signal of the planetary gearbox is nonlinear,nonstationary and difficult to be extracted effectively,a planetary gearbox fault diagnosis method based on modified ensemble empirical mode decomposition(MEEMD)multi-feature fusion and least square support vector machine(LS-SVM)is proposed.Firstly,with the MEEMD method,gear vibration signals under different working conditions are decomposed into a series of intrinsic mode functions(IMF).Secondly,the third-order sensitive modal components are screened out according to the correlation coefficient and the corresponding sample entropy and energy are calculated.The two are fused to form a high-dimensional eigenvector.Lastly,the fused eigenvector is taken as the input of the minimum support vector machine(LSSVM)to classify gear faults.Experiments are conducted on the planetary gearbox test bed,and compared with the feature vector based on single feature composition,and with the probabilistic neural network(PNN)classification algorithm.The results verify the effectiveness and superiority of the proposed method.
作者
蔡波
黄晋英
杜金波
马健程
王智超
CAI Bo;HUANG Jinying;DU Jinbo;MA Jiancheng;WANG Zhichao(School of Mechanical Engineering,North University of China,Taiyuan 030051,China;Beijing North Vehicle Group Corporation,Beijing 100072,China;School of Data Science and Technology,North University of China,Taiyuan 030051,China)
出处
《中国测试》
CAS
北大核心
2021年第9期126-132,共7页
China Measurement & Test
基金
山西省重点研发计划(国际科技合作方面)(201903D421008)
山西省自然科学基金项目(201901D111157)。