基于MED-EEMD和ELM的轴向柱塞泵松靴故障诊断研究  

Loose slipper fault diagnosis of axial piston pump based on MED-EEMD and ELM

作　　者：刘生政 张琳 曾祥辉[1] 兰媛 王志坚 程珩[1,2] LIU Sheng-zheng;ZHANG Lin;ZENG Xiang-hui;LAN Yuan;WANG Zhi-jian;CHENG Hang(College of Mechanical and Vehicle Engineering,Taiyuan University of Technology,Taiyuan 030024,China;Key Laboratory of Ministry of Education in Advanced Transducers and Intelligent Control System,Taiyuan University of Technology,Taiyuan 030024,China;School of Mechanical Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]太原理工大学机械与运载工程学院,山西太原030024 [2]太原理工大学新型传感器与智能控制教育部(山西省)重点实验室,山西太原030024 [3]中北大学机械工程学院,山西太原030051

出　　处：《机电工程》2020年第3期241-246,252,共7页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金青年项目(51405327);国家自然科学基金面上项目(51675364);山西省科技重大专项项目(20181102016)。

摘　　要：针对轴向柱塞泵松靴故障在强噪声干扰下故障信号微弱、故障特征提取困难和故障诊断识别精度低等一系列问题,提出了基于最小熵反褶积、集合经验模态分解和超限学习机相结合的轴向柱塞泵松靴故障诊断的方法。首先采集了轴向柱塞泵在正常和松靴故障两种状态下的振动信号;然后对振动信号进行了最小熵反褶积降噪,排除了噪声干扰,增强了冲击特性;之后利用集合经验模态分解将去噪后的信号分解成了若干个本征模态函数分量,并通过奇异值分解获得了特征矩阵;最后将得到的特征矩阵输入超限学习机、反向传播神经网络和支持向量机等3类分类器,并将识别结果与集合经验模态分解特征提取方法的识别结果进行了对比。研究结果表明:最小熵反褶积和集合经验模态分解结合的方法弥补了最小熵反褶积在强背景噪声下提取特征的局限性,克服了经验模态分解对微弱故障特征不敏感的缺陷;以最小熵反褶积-集合经验模态分解特征提取方法为输入的超限学习机分类模型,在少量样本的情况下可以有效地对轴向柱塞泵松靴故障进行检测与诊断。Aiming at series of problems such as weak fault signal, difficulty in extracting fault features and low fault diagnosis accuracy under the strong noise interference for loose slipper fault diagnosis of axial piston pump, a method of fault diagnosis of loose slipper fault diagnosis of axial piston pump based on minimum entropy deconvolution, ensemble empirical mode decomposition and extreme learning machine was proposed. Firstly, the vibration signal of the axial piston pump under normal and loose slipper conditions was collected. Then the MED method was applied to the vibration signal to eliminate noise interference and enhance impact characteristics. Afterwards, the denoised signal was decomposed into several intrinsic mode functions(IMFs) by EEMD, and the feature matrix was prepared by singularity value decomposition(SVD) of the obtained IMFs. Finally, the obtained feature matrix was input into the extreme learning machine, back propagation(BP) and support vector machine(SVM)classifiers. The recognition results were compared with those of the feature extraction method of EMD. The results indicate that the combination of MED and EEMD compensates for the limitation of EEMD extraction feature under strong background noise, and overcomes the defect that EMD is not sensitive to weak fault characteristics;the MED-EEMD feature extraction method and ELM classifier can be used to detect and diagnose the axial piston pump loose slipper fault in the case of a small number of samples.

关 键 词：最小熵反褶积 集合经验模态分解 超限学习机 故障诊断 

分 类 号：TH322[机械工程—机械制造及自动化]