基于EEMD样本熵和SRC的自确认气体传感器故障诊断方法  

Self-validating gas sensor fault diagnosis method based on EEMD sample entropy and SRC

作　　者：陈寅生[1] 姜守达[1] 刘晓东[1] 杨京礼[1] 王祁[1] 

机构地区：[1]哈尔滨工业大学电气工程与自动化学院,黑龙江哈尔滨150001

出　　处：《系统工程与电子技术》2016年第5期1215-1220,共6页Systems Engineering and Electronics

基　　金：国家自然科学基金(61201306;61271094;61327804)资助课题

摘　　要：针对非线性、非平稳情况下自确认气体传感器的故障诊断问题,提出了对传感器不同故障模式信号进行特征提取和智能识别的在线故障诊断方法。首先,该方法根据传感器信号的变化进行集合经验模态分解(ensemble empirical mode decomposition,EEMD),自适应地获得一组固有模态函数(intrinsic mode functions,IMFs),对每个IMF及残余分量进行样本熵分析,提取传感器输出信号的完备特征;然后,利用稀疏表示分类(sparse representationbased classification,SRC)将各故障模式下训练样本的特征向量构成超完备字典。为了提高故障诊断方法的自适应能力,对SRC分类器进行在线更新。通过求解最小1范数约束问题,获得测试样本的稀疏表示系数,再由不同故障类型的重构误差确定测试样本归属,进行传感器故障类型识别。实验结果表明,与目前其他传感器故障诊断方法比较,本文提出的方法能够更显著地提取传感器故障信号特征,故障识别率提高4%以上,达到97.14%。Aiming at the fault diagnosis problem of self-validating gas sensor under the condition of non- linear and non-stationary, a sensor on-line fault diagnosis method is proposed to conduct the feature extraction and intelligent identification for the sensor signals in different fault modes. Firstly, the sensor output signal is a- daptively decomposed to a series of intrinsic mode functions （IMFs） by ensemble empirical mode decomposition （EEMD） according to sensor signal change, and each of IMFs and residue is conducted by the sample entropy a- nalysis to extract the complete features of the sensor output signal. Afterwards, the over complete dictionary is comprised of the feature vectors of training samples in different fault conditions by using sparse representation- based classification （SRC）. The SRC classifier is updated subsequently to improve the adaptivity of it for fault diagnosis. The minimum l - norm constraint problem is applied to obtain the sparse represent coefficient of tes- ting sample and sensor fault type identification is determined by reconstruction error minimum between test sample and its reconstructed signal in different fault conditions. The experimental results show that the pro- posed method can significantly extract more features of the sensor fault signal compared with the other fault diagno- sis methods and the fault diagnosis recognition rate increases over 4% and reaches 97.14%.

关 键 词：自确认气体传感器 故障诊断 集合经验模态分解 样本熵 稀疏表示分类 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]