基于贝叶斯分类器的液压泵故障分级诊断方法  

Hierarchical Fault Diagnosis Method of Hydraulic Pump Based on Bayesian Classifier

作　　者：贾超广 肖海霞 JIA Chao-guang;XIAO Hai-xia(Information Engineering College of Zhengzhou Shengda University of Economics,Business&Management,Zhengzhou,Henan 451191;College of Science,Henan University of Engineering,Zhengzhou,Henan 451191)

机构地区：[1]郑州升达经贸管理学院信息工程学院,河南郑州451191 [2]河南工程学院理学院,河南郑州451191

出　　处：《液压与气动》2023年第6期181-188,共8页Chinese Hydraulics & Pneumatics

摘　　要：矿用液压泵故障种类较多,其分类过程一直存在无明确标签特征量化,导致故障诊断精准性较差的问题。对此提出应用贝叶斯分类器的矿用液压泵故障分级诊断方法,采用动态统计滤波计算矿用液压泵运行过程中的正常数据、噪声数据的特征性相关值,根据噪声和其他数据信息量不同的特点,离散噪声数据以达到过滤的目的。考虑到矿用液压泵故障数据集中存在无特征标签的问题,为避免混淆,划分无特征标签数据状态,计算二者属于同种类别的概率值,并采用损失函数对其进行约束。建立贝叶斯分类器,让已知种类和未知种类的矿用液压泵故障数据集处于同一分类空间,计算二者的先验概率和证据因子,通过概率值诊断待检测数据与已知数据间的属性差异,并利用贝叶斯分类器划分的故障诊断等级实现矿用液压泵故障分级诊断。实验结果表明:所提方法矿用液压泵故障诊断精准度高,不同类型故障的特征捕捉性能强,实用价值高。There are many types of mine hydraulic pump faults,and there has been no clear label feature quantification problem in its classification process,resulting in poor accuracy of fault diagnosis.A classification diagnosis method of mine hydraulic pump faults using Bayesian classifier is proposed.Dynamic statistical filtering is used to calculate the characteristic correlation value of normal data and noise data during the operation of mining hydraulic pump.According to the different characteristics of noise and other data information,the noise data is dispersed to achieve the purpose of filtering.Considering that there is no characteristic tag in the fault data set of mining hydraulic pump,in order to avoid confusion,the data state without characteristic tag is divided,the probability value of the two belonging to the same category is calculated,and the loss function is used to constrain them.A Bayesian classifier is established to make the known and unknown types of mining hydraulic pump fault data set in the same classification space,calculate the prior probability and evidence factor of the two,diagnose the attribute difference between the data to be tested and the known data through the probability value,and use the fault diagnosis level divided by the Bayesian classifier to realize the mine hydraulic pump fault classification diagnosis.The experimental results show that the proposed method has high accuracy in fault diagnosis of mine hydraulic pump,strong feature capture performance of different types of faults,and high practical value.

关 键 词：矿用液压泵 贝叶斯分类器 故障分级诊断 离散噪声 损失函数 

分 类 号：TH137[机械工程—机械制造及自动化] TH38[航空宇航科学与技术—飞行器设计] V444