
文章编号:10062971(2013)06000706
设计研究
基于RBF神经网络和多信息融合技术的往复压缩机状
态评估研究赵雨薇",马波",刘锦南
(1.北京化工大学诊断与自愈工程研究中心,北京100029;2.北京博华信智科技发展有限公司,北京100023)
摘要:为及时、准确地捕提设备运行的异常信息,提出基于径向基函数神经网络(RadialBasisFunetionNeuralNetwork,简移RBFNN)和多信息融合挂术的设务状态评估方法;利用往复压缩机不同测点特征值的趋势变化特点,研究神经网络分类技术,优选敏感特征参数,对不同测点特征值的趋势变化特点进行分类与组
合,建立往复压缩机状态评估流程,实现设备运行异常状态的预警评估。关键调:状态评估;趋势分类;神经网络;状态监测;故障诊断
中图分类号:TH457
文献标志码:A
TheStateAssessmentStudyof ReciprocatingCompressor Based onRBFNeuralNetwork
and Multi-informationFusionTechnology ZHAO Yu-wei',MA Bo', LIU Jin-nan?
(1.Diagnosis and Self-Recovery Engineering Research Center, Bejing University of Chemical Technology, Beijing100029, China;2.Beijing Bohua Xinzhi Science and Technology Development CO.,Lad.,Beijing 100023, China)
Abstract:ln order to capture the abnormal signals of the equipment timely and accurately,in this paper,the equipment state assess ment method based on RBF neural network and multiinformation fusion technology is proposed.Depending on the trend change fea-tures of characteristic indexes of different measuring points of reciprocating compressor,classification techniques based on neural net-work is studied,sensitive characteristic indexes are preferred and trend change features are classified and combined so that the state issessment process of reciprocating compressor is set up,which can achieve early prediction of abnormal running state of the equip. ment.
Key words state assessment;trend classification;neural network; condition monitoring;fault diagnosis
1引信言
往复压缩机是炼油、化工企业的关键设备,与离心压缩机相比,其存在往复、旋转2种周期性运动方式,结构复杂,运行部件受交变载荷的作用,受力复杂且不均勾,易损件较多,无法通过状态监测系统的单值报警限反映设备的运行状态是否异常,因此需要专业人员定期查看状态监
收稿日期:20130522
露2013年06期(总第242期)
测系统,分析并判断设备是否存在异常。若在设备运行出现早期故障征兆时就能及时、准确地捕提到设备异常动态信息,一方面可以提前采取相应措施,避免故障发生;另一方面,专业人员只需在得到准确报警信息的情况下再去查看监测系统,可以提高效率。因此,进行往复压缩机状态预警评估,及时、准确地捕提设备运行状态的异常信息具有十分重要的意义。
压机技术_107 sa