基于机器学习的SiC MOSFETS寿命预测 通信电气及其自动化工程专业VIP免费

基于机器学习的SiCMOSFETS寿命预测摘要以碳化硅材料、氮化镓材料为代表的第三代半导体材料，具有较大的禁带宽度。其中SiCMOSFET在高功率器件中脱颖而出，具有低损耗、导通压降较小、阻断电压高、耐受温度较高等高性能指标，在高压领域具有广阔的应用前景。在实际的工业应用以及相关开发的过程里，SiCMOSFETS器件的寿命分析对系统的效率有很大影响，但对于其可靠性特征的相关研究还处于起步阶段。本课题针对SiCMOSFETS器件的工作特点，以导通电阻为特征量进行器件的退化分析，以固定温度下多次功率循环后的电阻变化阈值为器件失效的判定标准，对器件进行失效建模。应用机器学习中的长短时记忆神经网络进行模型搭建，用以识别和预测器件寿命。利用开源的SiCMOSFETS加速退化试验数据对预测模型进行训练，输出拟合结果，达到了预定的性能指标。关键词：碳化硅功率管，机器学习，寿命预测，长短时记忆网络SAMPLESAMPLEBasedonSAMPLESAMPLESAMPLEAbstractThethirdgenerationsemiconductormaterials,suchassiliconcarbideandgalliumnitride,havelargebandgap.Amongthem,SiCMOSFETstandsoutinhigh-powerdevices,withlowloss,smallon-offvoltagedrop,highblockingvoltage,hightemperaturetoleranceandotherhighperformanceindicators,whichhasbroadapplicationprospectsinhigh-voltagefield.Intheactualindustrialapplicationandrelateddevelopmentprocess,thelifeanalysisofSiCMOSFETdeviceshasagreatimpactontheefficiencyofthesystem,buttheresearchonitsreliabilitycharacteristicsisstillinitsinfancyAccordingtotheworkingcharacteristicsofSiCMOSFETs,thispapertakesonresistanceasthecharacteristicquantitytoanalyzethedegradationofthedevice,andtakestheresistancechangethresholdaftermultiplepowercyclesatafixedtemperatureasthecriteriontodeterminethedevicefailure,soastomodelthedevicefailure.Thelong-termandshort-termmemoryneuralnetworkofmachinelearningisusedtobuildthemodeltoidentifyandpredictthedevicelife.ThepredictionmodelistrainedbyusingtheopensourceaccelerateddegradationtestdataofSiCMOSFETs,andthefittingresultsareoutputtoachievethepredeterminedperformanceindex.KeyWords:Siliconcarbidepowertransistor,machinelearning,lifeprediction,long-termandshort-termmemorynetwork目录1绪论...................................................................11.1研究背景..........................................................11.2研究意义..........................................................21.3国内外研究现状....................................................31.4研究目标与研究内容................................................61.4.1研究目标............................................................................................................61.4.2研究内容............................................................................................................61.5论文章节安排......................................................72SiCMOSFETS疲劳失效建模...............................................82.1SiCMOSFET结构....................................................82.2MOSFET主要参数...................................................102.3建立失效模型.....................................................112.4本章小结.........................................................113数据库及特征量提取....................................................123.1NASA预测数据存储库...............................................123.2滑动平均滤波.....................................................133.3特征量提取.......................................................143.4本章小结....................................................