基于PCA-RBF神经网络的WSN数据融合轴承故障诊断  

Fault diagnosis of bearing based on WSN and data fusion algorithm of PCA-RBF

作　　者：徐桂云[1,2] 蒋恒深[1] 李辉[1] 阮殿旭[1] 

机构地区：[1]中国矿业大学机电工程学院,江苏徐州221116 [2]西安电子科技大学综合业务网理论国家重点实验室,陕西西安710071

出　　处：《中国矿业大学学报》2012年第6期964-970,977,共8页Journal of China University of Mining & Technology

基　　金：综合业务网理论及关键技术国家重点实验室开放课题项目(ISN10-10);中国博士后科学基金项目(20060390277)

摘　　要：为了减少基于无线传感器网络(WSN)的轴承故障诊断系统数据传输总量和网络负载同时提高故障诊断准确性,提出一种采用主元分析(PCA)与径向基(RBF)神经网络结合轴承数据的融合与故障诊断算法.首先建立基于LEACH协议的3层融合模型,然后簇首节点采用PCA对大量多传感器数据降维,最后Sink节点采用RBF对数据进行决策级融合.仿真结果表明:该算法3个成员节点各上传10个数据包,簇头节点融合后剩余4个,融合率为86.7%,每组故障识别准确率大于85%.该算法具有很好的识别率和高压缩率,能够很好应用于煤矿设备故障监测.In order to reduce data quantity and wireless sensor network（WSN） load of bearing fault diagnosis system used by wireless sensor and increase accuracy of fault diagnosis, this research proposes a WSN devices based on monitor system, the principal component analysis （PCA） and radial basis （RBF） artificial neural networks as a data fusion diagnosis algorithm. Firstly a 3-layer data fusion model based on LEACH is established, and then dimension reduc- tion of sensor data is operated by cluster header node, and lastly sink node accomplish decision- level fusion of data. Simulation results show that 3 member transmits 10 data packets respec- tively, while only 4 packets are remained after data fusion by sink node, so data fusion ratio is 86.7%, and accuracy of fault diagnosis is 85 %. The algorithm with a good recognition rate and high compression ratio can be well applied in fault monitoring of the equipment for coal mine.

关 键 词：无线传感器网络 机械故障诊断 数据融合 主元分析 RBF神经网络 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]