资源描述:
《基于遗传神经网络的汽车故障率预测》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、上海交通大学硕士学位论文i基于遗传神经网络的汽车故障率预测摘要汽车在使用过程中总会有故障出现,对于汽车生产商来说,了解到汽车在一定行驶条件下的故障率,可以减少相应的时间成本和库存成本。有鉴于此,本文首先通过对汽车整体性能的分析,找出可能引起整体性能指标降低而导致汽车故障的因素:汽车不同零件的易损度不同,零件本身质量差异,汽车维修频率,汽车消耗品种类,汽车行驶环境,驾驶因素,行驶距离等,开拓了汽车故障模型建模的新途径;并对每个不同的故障因素影响导致汽车故障的权重进行了分析评估,得出汽车故障率的控制模型。其次,本文采用了GA-BP神经网络处理问
2、题。在描述了传统BP网络的基本模型的基础上,介绍了用于BP网络中的常见的几种改进算法:附加动量和学习率自适应调整的改进BP算法(BPX)、Levenberg-Marquardt优化方法(LM)、Bayes规范化BP算法。并利用遗传算法对神经网络的权值和阈值进行优化,在MATLAB上分别对这几种算法对模型进行了训练,利用神经网络的泛化能力,得到我们所需要的预测数据。关键字:BP神经网络、遗传算法、优化、汽车故障率、预测i上海交通大学硕士学位论文iFORECASTTHECARFAILUREBASEDONBPNEURALNETWORKABSTRA
3、CTIngeneral,theautomobilehasthebreakdownappearanceintheuseprocess.Regardingtheautomobileproducer,itwouldreducethecorrespondingtimecostandinventorycostafterknowingtheautomobilefailurerateundercertaintravelcondition.Astothis,firstly,throughtheautomobileoverallperformanceanal
4、ysis,thearticlefindsthepossiblecausestoreducetheautomobilebreakdown,suchastheautomobileeachinstallmentbuckle,thecomponentsqualitydifferences,theautomobileservicefrequency,theautomobilefueloil,theautomobiletravelenvironment,thedrivingtechnologyandthedrivingmethod,thetravelc
5、ourseandsoonandetc.Allofthesehavedevelopedtheautomobilebreakdownmodelinthenewway,andanalysistheweighofthedifferentbreakdownfactorsinfluence,thentheautomobilefailureratecontrolmodelhasbeengot.Next,thisarticlehasusedtheGA-BPneuralnetworktosolvethequestion.Onthefoundationofth
6、etraditionalBPnetworkbasicmodel,itintroducestheseveralcommonkindsofimprovementalgorithmintheBPnetwork:Additionalmomentumandlearningrateauto-adaptedi上海交通大学硕士学位论文iiadjustmentimprovementBPalgorithm(BPX),Levenberg-Marquardtoptimizedalgorithm(LM),BayesstandardizedBPalgorithm,an
7、dcarriesontheoptimizationontheneuralnetworkweightandbiasusingthegeneticalgorithm.IthasseparatelycarriedonthetrainingmodelbyMATLABonthebasisofthesealgorithms,andforecastthedatawhichweneedthroughregressionabilityusingtheneuralnetwork.KEYWORDS:Back-PropagationNeuralNetwork,Ge
8、neticAlgorithm,CarFailure,Forecastii上海交通大学硕士学位论文63上海交通大学学位论文原创性声明本人郑重声明:所呈交的学位论文,是本人在导师的指