摘要
为满足模式识别故障诊断算法的鲁棒性要求,在小波包分解提取特征向量的基础上,提出了有监督模式分类与无监督模式分类相结合的故障诊断方法.利用小波包分解提取各个频带的能量作为特征向量;采用LVQ神经网络作为有监督的模式分类器进行故障诊断;运用无监督的减法聚类方法对新型故障模式进行辨识.最后,通过动力系统管路流量传感器数据对算法进行检验,验证了所提出方法的实用性和有效性.
To meet the robustness of the fault diagnosis algorithm, a method is proposed, which combines the supervised classification and unsupervised classification based on the feature extraction with wavelet package decomposition. As the pattern vector, the energy in different frequency with the wavelet package decomposition is calculated. Then, learning vector quantity neural network is employed as the supervised classification for fault diagnosis. As the supervised classification, subtractive clustering is applied to identify the novel fault pattern. Finally, the applicability and effectiveness of the proposed methodology are illustrated by flow sensor data of the dynamical system.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第7期783-786,共4页
Control and Decision
基金
国家自然科学基金项目(60572010).
关键词
模式识别
小波包
LVQ神经网络
减法聚类
传感器故障诊断
Pattern recognition
Wavelet package
LVQ neural network
Subtractive clustering
Sensor fault diagnosis