基于LE-DBN故障诊断模型的滚动轴承振动信号特征提取  

Feature Extraction and Diagnosis of Bearing Fault Signals Based on LE and DBN Algorithms

作　　者：韩春红[1] 伊洪彬 薛涛[1] 刘玉芳[3] HAN Chunhong;YIN Hongbin;XUE Tao;LIU Yufang(School of Information Engineering,Jiaozuo Normal College,Jiaozuo Henan 454000,China;Intelligent Manufacturing College,Kaifeng Technician College,Kaifeng Henan 475004,China;College of Resources and Environment,Henan Polytechnic University,Jiaozuo Henan 454000,China)

机构地区：[1]焦作师范高等专科学校信息工程学院,河南焦作454000 [2]开封技师学院智能制造学院,河南开封475004 [3]河南理工大学资源环境学院,河南焦作454000

出　　处：《机械设计与研究》2023年第3期131-134,共4页Machine Design And Research

基　　金：河南省高等学校重点科研资助项目(18B170004)。

摘　　要：为了提高滚动轴承的高效率运行,设计了一种基于拉斯特征映射-深度置信网络(laplacian eigenmap-deep belief network,LE-DBN)故障诊断模型的滚动轴承振动信号特征提取方法。利用LE算法从高维振动数据中提取获得流形参数,分别测试包含少量有标签与大量无标签样本的DBN网络训练结果,再对各类故障实施分类。研究结果表明:采用训练集识别时准确率在99.8%附近,表明该模型可以对训练数据发挥理想拟合性能。LE算法比PCA、KPCA算法都达到了更优特征提取效果,选择合理参数可以使准确率达到99.8%。采用多传感器实施特征融合时相对单个传感器的诊断性能更优。有标签样本个数在60~120之间时,采用DBN网络可以获得比CNN网络更理想的分类结果。该研究可以达到可靠性标准,更能适用于其它的机械传动设备。In order to improve the operation efficiency of rolling bearings,a method of vibration signal feature extraction of rolling bearings based on LE-DBN fault diagnosis model is designed.Manifold parameters are extracted from high-dimensional vibration data using the LE algorithm.The training results of the DBN networks containing a small number of labeled and a large number of unlabeled samples are tested respectively,and then various faults are classified.The results show that the accuracy rate is around 99.8%when the training set is used,which indicates that the proposed model can provide an ideal fitting performance for the training data.Compared with PCA and KPCA,the LE algorithm achieves better feature extraction effect,and the accuracy rate can reach 99.8%by selecting reasonable parameters.The diagnostic performance using multiple sensors in feature fusion is better than that with a single sensor.When the number of labeled samples is between 60 and 120,the DBN network can obtain better classification results than the CNN network.The research can meet the reliability standard and can be applied to other mechanical transmission equipment.

关 键 词：滚动轴承 振动信号 特征提取 拉普拉斯特征映射 深度置信网络 

分 类 号：TH133[机械工程—机械制造及自动化]