摘要
为加快振动传感器故障诊断速度、提高诊断的准确率,提出了基于支持向量机(SVM)和主元分析法(PCA)特征信息提取相结合的故障诊断方法。首先进行诊断方法的调查,将振动传感器输出的时序电信号通过小波包分解得到频域领域的原始数据;然后经过PCA的特征信息提取得到原始数据的特征向量,加强振动传感器工作状态的可分性;最后进行二叉树与SVM结合的多分类算法,实现振动传感器运行故障的诊断。为提高分类速度,引入最小二支持向量机(LS_SVM)算法,并应用到多分类器中。仿真试验表明,改进后的方法提高了诊断准确率、加快了故障分类速度,优于单一方法进行故障诊断的情况,为其他种类传感器(如温度、瓦斯等)的故障诊断提供了参考。对传感器故障诊断方法的研究,为传感器的正常运行提供了保证,降低了因传感器故障而造成应用设备的损失。
In order to speed up the diagnosis of vibration sensor faults and improve its accuracy,a fault diagnosis method founded on support vector machine(SVM)and principal component analysis(PCA)feature information extraction is proposed.First,the time-series electrical signal output by the vibration sensor is decomposed by the wavelet packet to obtain the raw data in the frequency domain;then,the feature vector of the raw data is extracted by the feature information of the PCA to enhance the separability of the working state of the vibration sensor;finally,a multi-classification algorithm combining binary tree and SVM is used to implement vibration sensor fault diagnosis.In order to improve the classification speed,the least squares support vector machine(LS_SVM)algorithm is introduced to the multi-classification.Simulation experiments show that the improved method enhance the diagnostic accuracy and speed up the fault classification,which is preferable to a single fault diagnosis method.It provides reference for troubleshooting other types of sensors(e.g.,temperature,gas).The research on sensor fault diagnosis method provides a priori condition for the normal operation of the sensor,which reduces the loss of applied equipment caused by sensor faults.
作者
李翼飞
吴春平
涂煊
LI Yifei;WU Chunping;TU Xuan(Institute of Optoelectronic Information and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Shanghai Institute of Process Automation and Instrumentation,Shanghai 200233,China)
出处
《自动化仪表》
CAS
2019年第10期48-52,共5页
Process Automation Instrumentation
关键词
振动传感器
故障诊断
小波包分解
主元分析
支持向量机
二叉树分类
最小二乘支持向量机
Vibration sensor
Fault diagnosis
Wavelet decomposition
Principal component analysis(PLA)
Support vector machine(SVM)
Binary tree classification
Least square support vector machine(LS_SVM)
