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,更多相关内容在工程资料-天天文库。
1、基于遗传神经网络的汽车故障率预测上海交通大学硕士学位论文基于遗传神经网络的汽车故障率预测姓名:杨婷中请学位级别:硕士专业:控制理论与控制工程指导教师:杨根科20080101上海交通大学硕士学位论文i基丁遗传神经网络的汽车故障率预测摘要汽车在使用过程屮总会有故障出现,对于汽车生产商來说,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
2、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1上海交通大学硕士学位论文iF
3、ORECASTTHECARFAILUREBASEDONBPNEURALNETWORKABSTRACTTngeneral,theautomobilehasthebreakdownappearanceintheuseprocess.Regardingtheautomobileproducer,itwouldreducethecorrespondingtimecostandinventorycostafterknowingtheautomobilecondition.failurerateundercertaintravelAstothis,firstly,throughthea
4、utomobileoverallperformanceanalysis,thearticlefindsthepossiblecausestoreducetheautomobilebreakdown,suchastheautomobileeachinsiallmentbuckle,thecomponentsqualitycliffcrcnccs,theautomobileservicefrequency,theautomobilefueloil,theautomobiletravelenvironment,thedrivingtechnologyandthedrivingme
5、thod,thetravelcourseandsoonandetc.Allofthesehavedevelopedtheautomobilebreakdownmodelinthenewway,andanalysistheweighofthedifferentbreakdownfactorsinfluenee,thentheautomobilefailureratecontrolmodelhasbeengot.Next,thisarticlehasusedtheGA-BPneuralnetworktosolvethequestion.Onthefoundationofthet
6、raditionalBPnetworkbasicmodel,itintroducestheseveralcommonkindsofimprovementalgorithmintheBPnetwork:Additionalmomcntumandlearningrateauto-adaptcd■1上海交通大学硕士学位论文iiadjustmentimprovementBPalgorithmBPX、Levenberg-MarquardtoptimizedalgorithmLM,BayesstanclarclizcclBPalgorithm,andcarriesontheoptimi
7、zationontheneuralnetworkweightandbiasusingthegeneticalgorithm.IthasseparatelycarriedonthetrainingmodelbyMATLABonthebasisofthesealgorithms,andforecastthedatawhichweneedthroughregressionabilityusingtheneuralnetwork.KEYWORDS:Back-PropagationNeuralNetwork,GeneticA