作　　者： (李娟娟）; (孟国营）; (谢广明）; (贾一凡）;

机构地区： 中国矿业大学（北京）机电信息与工程学院,北京100083

出　　处： 《传感技术学报》 2017年第7期1035-1039,共5页

摘　　要： 研究传感器实时故障诊断问题。首先采用MATLAB2015仿真得到传感器各种典型工作状态下的运行数据样本;其次将这些故障样本作3层小波包分解,分别求出第3层小波包基对应的各频率段的能量,利用这些能量值与正常工作时各频段的能量值之比构造出传感器故障诊断的特征向量;最后构建基于3×3的SOM神经网络的传感器故障诊断算法。测试证明了所提算法的有效性和准确性。 Sensor fault diagnosis is investigated. First, sensor＇s fault samples under various operation conditions are obtained by simulation with the MATLAB2015. Then fault samples are decomposed by a three-layerwavelet packet decomposition technology ,and the energies of cor^espbnding frequency in third layer are obtained ,respect[velyi Ei~[- envector of sensor fault diagnosis is extracted by using the ratio between fault energies and faultlessenergies on vari- ous frequencies. Finally, a 3 by 3 SOM neural network based algorithm isproposed for pattern classification and fault diagnosis. The effectiveness and accuracy of the proposed method are illustrated by diagnostic results.

关 键 词： 传感器 小波包变换 神经网络 故障样本 故障诊断