摘要
针对传感器故障探测和诊断,提出了一种基于稀疏分解残差的氢气传感器故障探测和辨识方法。基于信号稀疏分解理论,对采集的传感器正常信号数据集,利用K奇异值分解(K-SVD)学习算法得到一超完备字典D;在字典上对非正常(故障)信号进行分解,根据稀疏分解的残差大小和范围完成对传感器故障的探测及辨识。实验结果表明:对氢气传感器的故障探测率和总辨识率分别达到98.75%和97.25%,可以有效地解决氢气传感器的故障探测和辨识。
Aiming at detection and diagnosis of sensor,a fault detection and identification method for hydrogen sensor based on residual of spares decomposition is proposed. The method bases on theory of signal spares representation,collects the normal signal data of hydrogen sensor to learn an over-complete dictionary D by K-SVD learning algorithm firstly,then uses the dictionary D to decompose abnormal（ fault） signals and get the decomposed residuals. Finally,according to size and range of the residuals,the sensor faults can be detected and identified. The experiment results show that for hydrogen sensor,the detection and total recognition rate of the proposed method reachs to 98. 75 % and 97. 25 % respectively,which can be applied to detect and identify the fault of hydrogen sensor effectively.
出处
《传感器与微系统》
CSCD
2017年第8期32-34,38,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61663013)
江西省自然科学基金资助项目(20161BAB212051)
江西省重点研发计划项目(20161BBE50076)
江西省教育厅科学技术项目(GJJ160491)
关键词
氢气传感器
故障探测
故障辨识
稀疏分解
K奇异值分解
hydrogen sensor
fault detection
fault identification
spares decomposition
K-singular value decomposition（K-SVD）
作者简介
韦宝泉（1978-），男，硕士，副教授，从事信号处理与故障诊断技术研究工作。
