柱塞泵空化振动信号提取及其极限学习机检测  

Extraction of Cavitation Vibration Signal of Piston Pump and Detection by Extreme Learning Machine

作　　者：张保生[1] 王强东 李明 ZHANG Bao-sheng;WANG Qiang-dong;LI Ming(School of Mechatronicsl and Automotive Engineering,Xuchang Vocational and Technical College,He’nan Xuchang 461000,China;School of Mechanical Engineering,He’nan University of Science and Technology,He’nan Zhengzhou 446000,China;China Power Technology Information Industry Co.,Ltd.,He’nan Xinxiang 453006,China)

机构地区：[1]许昌职业技术学院机电与汽车工程学院,河南许昌461000 [2]河南科技大学机械工程学院,河南郑州446000 [3]中电科信息产业有限公司,河南新乡453006

出　　处：《机械设计与制造》2023年第6期101-104,共4页Machinery Design & Manufacture

基　　金：中国高校产学研创新基金(2019ITA02018)。

摘　　要：随真空度的持续提高,泵运行过程中开始形成空化并且空化程度逐渐变得明显,由此引起泵流量减小。为实现对柱塞泵的实际振动信号精确分析,选择小波基函数对采集信号实施三层小波包分解处理,经过阈值去噪以及小波包重构方法完成信号去噪过程。采极限学习机用EMD自适应分解方法处理振动信号,从中提取可以显著反馈信号复杂度的样本熵,并根据该参数判断空化情况。当样本熵值提高后形成更复杂的样本序列。开展了通过ELM实现的柱塞泵空化状态检测,结果表明:本设计特征提取方法都能够准确检测各工况的空化状态,与BP神经网络以及随机森林(FR)相比,ELM达到了更高检测准确度,并且ELM模型时间也显著缩短。该研究对提高柱塞泵空化振动信号提取及故障诊断效率分析具有很好的实际指导价值。With the continuous improvement of vacuum,the pump began to form cavitation in the process of operation and the degree of cavitation gradually becomes obvious,resulting in a decrease in pump flow.In order to accurately analyze the actual vibration signal of piston pump,the wavelet basis function is selected to implement the three-layer wavelet packet decomposition processing of collected signal,and the signal denoising process is completed by the threshold de-noising and the wavelet packet reconstruction method.Using EMD adaptive decomposition method,the extreme learning machine is used to process vibration signals,and sample entropy which can feedback signal complexity is extracted,and the cavitation situation is judged according to this parameter.When the sample entropy is increased,a more complex sample sequence is formed.The piston pump cavitation status detection based on ELM was carried out.The results show that:the designed feature extraction method can accurately detect the cavitation status of each condition.Compared with BP neural network and random forest(FR),ELM achieves higher detection accuracy,and the ELM model time is also significantly shortened.The research has a good practical guiding value for improving the signal extraction and fault diagnosis efficiency analysis of piston pump cavitation vibration.

关 键 词：极限学习机 柱塞泵 空化 振动信号 特征提取 检测 

分 类 号：TH16[机械工程—机械制造及自动化] TH32