首页|基于边缘计算的轴向柱塞泵磨损状态辨识方法研究

基于边缘计算的轴向柱塞泵磨损状态辨识方法研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
判断轴向柱塞泵的磨损状态对维护轴向柱塞泵正常运行具有重要意义.然而,现有的轴向柱塞泵故障诊断方法大多为离线式和基于云计算的在线式,存在延时长和数据量大的问题,无法满足轴向柱塞泵磨损状态辨识的实时性需求.为了减少延迟时间和传输数据量,提出一种基于边缘计算的轴向柱塞泵磨损状态辨识方法.构造出一个集成信号采集、信号预处理、特征提取和磨损状态分类的边缘节点,用于正确和实时地辨识磨损状态.设置4种轴向柱塞泵滑靴副磨损状态作为故障源,并构建相应的磨损故障数据集.为了减少边缘节点计算量,在上位机中利用随机森林包外误差选择敏感特征值.为了实现磨损状态辨识,在上位机中训练磨损状态分类的特征选择人工神经网络模型,并将信号预处理和特征提取功能算法以及模型参数嵌入边缘节点.通过与其他方法的比较和在线磨损状态辨识试验证明所提方法的正确性和实时性.
Wear State Identification Method for Axial Piston Pumps Based on Edge Computing
Wear state identification is of great significance to maintain the normal operation of axial piston pump.However,most existing fault diagnosis methods for axial piston pump are offline while those online are based on cloud computing,producing a long delay or a large amount of data,which cannot meet the real-time requirements of wear state identification.To cut down latency and data amount,a wear state identification method for axial piston pump based on edge computing is proposed.An edge node integrating the functions of data acquisition,data pre-processing,feature extraction and wear state classification is constructed to identify the wear state accurately and timely.Four kinds of wear states of axial piston pump are set as fault sources,and the corresponding wear fault dataset is established.To reduce the computational load of the edge node,the out-of-bag error of random forest is employed to select the sensitive features at the host computer.To classify the wear state,a feature-selected artificial neural network for wear state classification is trained at the host computer,and the data pre-processing and feature extraction algorithm together with model parameters are embedded into the edge node.The accuracy and real-time performance of the proposed method are demonstrated through comparisons with other methods and the online wear state identification experiment.

axial piston pumpedge computingwear state identificationfeature sensitivityartificial neural network

王丹丹、黄伟迪、张军辉、赵守军、于斌、刘施镐、吕飞、苏琦、徐兵

展开 >

浙江大学流体动力基础件与机电系统全国重点实验室 杭州 310027

北京精密机电控制设备研究所 北京 100076

航天伺服驱动与传动技术实验室 北京 100076

轴向柱塞泵 边缘计算 磨损状态辨识 特征敏感程度 人工神经网络

国家重点研发计划国家自然科学基金国家自然科学基金浙江省自然科学基金航天伺服驱动与传动技术实验室开放基金

2019YFB20045045183500952105075LQ21E050022LASAT-20210104

2024

10.3901/JME.2024.04.189

机械工程学报
中国机械工程学会

机械工程学报

CSTPCD北大核心
影响因子：1.362
ISSN：0577-6686
年,卷(期)：2024.60(4)
  • 27