基于邻域粗糙集与支持向量极端学习机的瓦斯传感器故障诊断  

Gas Sensor Fault Diagnosis Based on Neighborhood Rough Set Combined with Support Vector Machine and Extreme Learning Machine

作　　者：单亚峰 汤月[1] 任仁[1] 谢鸿[1] 

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105

出　　处：《传感技术学报》2016年第9期1400-1404,共5页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(51274118);辽宁省科技攻关基金项目(2011229011);辽宁省教育厅基金项目(L2012119)

摘　　要：针对于瓦斯传感器故障诊断速度慢、诊断精度不高的问题,以常见的冲击型、漂移型、偏置型和周期型传感器输出故障作为研究对象,提出一种基于邻域粗糙集(NRS)和支持向量极端学习机(SVM-ELM)的故障诊断方法。首先对瓦斯传感器的特征属性值进行归一化处理,然后利用NRS信息约简理论降低属性维度,提取出影响瓦斯传感器的关键属性构成约简集。将约简集作为SVM-ELM的输入进行训练,利用训练好的SVM-ELM对测试样本进行模式识别。最后通过实验对比验证该方法具有训练速度快、分类精度高的特点,辨识正确率在95%以上,能够显著提高故障诊断的速度和准确性。In order to solve the problem that the gas sensor diagnosis speed is slow and the diagnosis accuracy is nothigh,this paper takes the common type gas sensor fault such as impact,drift,bias and periodic fault as research ob-ject and proposes a pattern classification and identification of the fault diagnosis of gas sensor method based onneighborhood rough set（NRS）combined with support vector machine and extreme learning machine（SVM-ELM）.First of all,normalize the feature attribute of the gas sensor,the reduction set is formed via reducing the attribute di-mension with NRS information reduction theory,including key attributes of the gas sensor. Train SVM-ELM takingthe reduction set for input data and recognize the fault patterns using test samples. Finally,through experiment con-trast analysis,this method has the features of fast training speed,high accuracy of classification,and the identifica-tion correct rate is more than 95%. It can significantly improve the effectiveness and accuracy of the fault diagnosis.

关 键 词：瓦斯传感器 邻域粗糙集(NRS) 支持向量极端学习机(SVM-ELM) 故障诊断 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] TP181[自动化与计算机技术—控制科学与工程]