基于能量特征和RBF神经网络的柱塞泵状态识别方法-论文.pdf

《基于能量特征和RBF神经网络的柱塞泵状态识别方法-论文.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、JournalofMechanicalStrength机械度2013，35(2)：127—132基于能量特征和RBF神经网络的柱塞泵状态识别方法STATERECoGNITIoNMETHoDoFPISToNPUMPBASEDONENERGYFEATUREANDRADIALBASISFUNCTIoNNEURALNETWoRK何庆飞”陈桂明陈小虎姚春江杨庆张宪宇(第二炮兵工程学院装备管理工程系，西安710025)HEQingFeiCHENGuiMingCHENXiaoHuYAOChunJiangYANGQingZHANGXianYu(DepartmentofEquipmentManagementE

2、ngineering．eSecondArtilleryEngineeringCollege，Xian710025，China)摘要柱塞泵状态发生改变时，其振动信号各频带的能量发生相应变化，利用各频带的能量特征可识别柱塞泵状态。基于此提出一种基于能量特征和径向基函数(radialbasisfunction，RBF)神经网络的柱塞泵状态识别方法。首先利用改进阈值的小波包降噪方法对原信号进行降噪处理，然后对降噪后的信号进行经验模态分解(empiricalmodedecomposition，EMD)，选取含主要状态信息的IMF(intrinsicmodefunctions)分量，提取能量特征作为神经

3、网络的输入参数，识别柱塞泵的状态。同时利用小波包分析技术提取能量特征，并运用神经网络进行状态识别。应用实例表明，以EMD提取各频带能量作为特征参数的RBF神经网络状态识别方法比小波包分析提取各频带能量特征的方法具有更高的识别率，能有效地识别柱塞泵的状态。关键词柱塞泵能量特征经验模态分解小波包分析RBF神经网络状态识别中图分类号TH322TP277AbstractTheenergyofvibrationsignalwouldvaryindiferentfrequencybandswhenpistonpumpfaultoccurs．Thereforeenergyfeatureparametere

4、xtractedfromvibrationsignalcouldbeusedtoidentifythestateofpistonpump．Forthispurpose，staterecognitionmethodofpistonpumpbasedonenergyfeatureandradialbasisfunction(RBF)neuralnetworkisputforward．Firstly，thevibrationsignalisde—noisedbyimprovedwaveletpocketthresholdde—noisingmethod．Thende—noisedsignalsar

5、edecomposedintoafinitenumberofstationaryintrinsicmodefunctions(IMFs)byempiricalmodedecomposition(EMD)，andIMFscontainingmainstateinformationareselectedforfurtheranalysis．EnergyfeatureparameterextractedfromIMFscouldbeservedasinputparameterofRBFneuralnetworkstoidentifydiferentstatesofpistonpump．Theapp

6、licationexamplesshowthattheapproachofRBFneuralnetworkstaterecognitionbasedonEMDextractingenergyparameterofdiferentfrequencybandsasfeatureissuperiortothatbasedonwaveletpacketanalysisandcanidentifypistonstateefectively．KeywordsPistonpump；Energyfeature；Empiricalmodedecomposition(EMD)；Waveletpacketanal

7、ysis；Radialbasisfunction(RBF)neuralnetwork；StaterecognitionCorrespondingauthor：HEQingFei，E—mail：qingfeihe@yahoo．COrn．cn，Tel：／Fax：+86—29—84741370TheprojectsupportedbytheNationalDefenseResearchFoundationofChi