资源描述:
《时间序列趋势分析及其在预防性维修中的应用》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、ADissertationSubmittedinPartialFulfillmentoftheRequirementsfortheDegreeofMasterinEngineeringTime-seriestrendanalysisanditsapplicationinpreventivemaintenanceMasterCandidate:WangPanMajor:CircuitsandSystemsSupervisor:LiuYiminWuhanUniversityofScienceandTechnologyWuhan,Hubei4300
2、81,P.R.ChinaMay24,2015摘要现代预防性维修大多是以RCM(以可靠性为中心的维修)为基础或者纯粹按维修周期进行计划检修。国内企业进行RCM分析时往往因很难得到设备故障概率而难以进行,通常使用国外的数据库,而计划检修容易造成维修过剩。因此,本文设计动态预防性维修管理系统,根据每一台设备的运行趋势来判断设备是否需要进行预防性维修,尽可能地减少维修不足与维修过剩。论文介绍了趋势分析常用算法以及预防性维修的国内外研究现状。在此基础上,采用一元线性回归算法、AR算法、Sen’s斜率估计与Mann-Kendall算法对设备运行数据进行分
3、析。通过对趋势分析的结果进行分析、比较,选出适用于本系统的算法。结果表明,Sen’s斜率估计与Mann-Kendall算法适用于本系统,Sen’s斜率估计计算设备劣化幅度变化率,Mann-Kendall算法实现设备运行数据的趋势分析、突变分析以及显著性分析。最后,开发了动态预防性维修管理系统,将Sen’s斜率估计与Mann-Kendall算法应用于本系统。实际应用表明,Sen’s斜率估计与Mann-Kendall算法在分析设备运行趋势时,具有较好的效果。关键词:趋势分析;预防性维修;Sen’s斜率估计;Mann-KendallIAbstrac
4、tMostlymodernpreventivemaintenancebasedonRCM(ReliabilityCenteredMaintenance)analysismethodorsimplycyclemaintenance.DomesticenterprisesareoftendifficulttogettheprobabilityofequipmentfailurewhenusingRCManalysismethod,theyusuallyuseoutsidedatabase.Andplanmaintenanceiseasygetti
5、ngexcessmaintenance.Therefore,thispaperdesigneddynamicpreventivemaintenancemanagementsystem,focusontheoperationtrendsofeachequipment,usinglogicaloperationswithotherconditions,todeterminewhethertheequipmentneedtopreventivemaintenance.Thepaperintroducedtheresearchstatusoftren
6、dsanalysisalgorithmsandpreventivemaintenance.Usingsimplelinearregressionalgorithm,ARmodel,Sen’sslopeestimationandMann-Kendalltoanalyzethetrendoftheequipmentoperation.TochoosewhichalgorithmissuitforthissystemTheresultisthattheSen’sslopeestimationandMann-Kendallalgorithmaresu
7、itforthesystem,Sen’sslopeestimationisusedtocalculatethechangerateofamplitude,Mann-Kendallalgorithmisusedtoanalyzetrend,toanalyzemutation,toanalyzesaliency.Finally,finishthedynamicpreventivemaintenancemanagementsystem,andapplySen’sslopeestimationandMann-Kendallalgorithmtothi
8、ssystem.PracticalapplicationshowsthatSen'sslopeestimationandMann-Kendallalgorithmh