基于Elman神经网络模型的IGBT寿命预测
|
摘要
建立了Elman神经网络模型来实现绝缘栅双极型晶体管(IGBT)的寿命预测。分析了IGBT的结构及其失效原因,结合NASA埃姆斯中心的加速热老化试验数据,确定了以集电极-发射极关断电压尖峰峰值作为失效预测依据。利用高斯滤波的方法对试验数据进行预处理,构建了单、多隐层Elman神经网络寿命预测模型,并构建了广义回归神经网络(GRNN)寿命预测模型作为对比模型。采用均方误差、平均绝对误差、最大相对误差作为各模型预测性能的评估指标。结果表明,提出的Elman神经网络模型比GRNN模型有更好的预测效果。二隐层的Elman神经网络模型均方误差为0.202 0%,平均绝对误差为0.387 6%,最大相对误差为3.023 0%,可以更好地实现IGBT寿命的预测。
Elman neural network model was established to realize life prediction of insulated gate bipolar transistors(IGBTs). The structure and failure causes of IGBTs were analyzed. Based on the accele-rated thermal aging experimental data of NASA Ames Center, the peak value of the collector-emitter switching off voltage was determined as the failure prediction basis. The experimental data were pre-processed by Gauss filtering method, life prediction models of the single and multi-hidden layer Elman neural network were constructed, and the life prediction model of generalized regression neural network(GRNN) was constructed as the contrast model. The mean square error, mean absolute error and maximum relative error were used as the evaluation indexes of model prediction performance. The results show that Elman neural network model has better prediction performance than GRNN model. The mean square error of Elman neural network model with two hidden layers is 0.202 0%, the average absolute error is 0.387 6%, and the maximum relative error is 3.023 0%. Then the life prediction of IGBTs can be realized better.
Elman neural network model was established to realize life prediction of insulated gate bipolar transistors(IGBTs). The structure and failure causes of IGBTs were analyzed. Based on the accele-rated thermal aging experimental data of NASA Ames Center, the peak value of the collector-emitter switching off voltage was determined as the failure prediction basis. The experimental data were pre-processed by Gauss filtering method, life prediction models of the single and multi-hidden layer Elman neural network were constructed, and the life prediction model of generalized regression neural network(GRNN) was constructed as the contrast model. The mean square error, mean absolute error and maximum relative error were used as the evaluation indexes of model prediction performance. The results show that Elman neural network model has better prediction performance than GRNN model. The mean square error of Elman neural network model with two hidden layers is 0.202 0%, the average absolute error is 0.387 6%, and the maximum relative error is 3.023 0%. Then the life prediction of IGBTs can be realized better.
引文
[1] JAYANT BALIGA B.IGBT器件—物理、设计与应用[M].韩雁,丁扣宝,张世锋,等译.北京:机械工业出版社,2018.
[2] 刘宾礼,刘德志,唐勇,等.基于加速寿命试验的IGBT模块寿命预测和失效分析[J].江苏大学学报(自然科学版),2013,34(5):556-563.LIU B L,LIU D Z,TANG Y,et al.Lifetime prediction and failure analysis of IGBT module based on accelerated lifetime test [J].Journal of Jiangsu University (Natural Science Edition),2013,34(5):556-563 (in Chinese).
[3] 李明,王跃,高远,等.采用动态电压上升控制的1 700 V大功率IGBT有源门极驱动技术[J].高电压技术,2014,40(8):2513-2519.LI M,WANG Y,GAO Y,et al.Active gate drive technology for 1 700 V high power IGBT with dynamic voltage rise control [J].High Voltage Engineering,2014,40(8):2513-2519 (in Chinese).
[4] 陈冰,鲁刚,房红征,等.基于退化数据和DBN算法的IGBT健康参数预测方法 [J].计算机测量与控制,2017,25(5):71-75.CHEN B,LU G,FANG H Z,et al.Prediction method of IGBT health parameters based-on degradation data and DBN algorithm [J].Computer Measurement & Control,2017,25(5):71-75 (in Chinese).
[5] 王宇琦.IGBT疲劳老化失效剩余使用寿命预测的研究[D].西安:西安电子科技大学,2017.
[6] 胡震.功率器件的故障诊断及疲劳寿命预测[D].天津:天津理工大学,2014.
[7] 方鑫,周雒维,姚丹,等.IGBT模块寿命预测模型综述 [J].电源学报,2014(3):14-21.FANG X,ZHOU L W,YAO D,et al.An overview of IGBT life prediction models [J].Journal of Power Supply,2014(3):14-21 (in Chinese).
[8] OH H,HAN B,MCCLUSKEY P.Physics-of-failure,condition monitoring and prognostics of insulated gate bipolar transistor modules:a review [J].IEEE Transactions on Power Electronics,2015,30(5):2413-2426.
[9] 赖伟,陈民铀,冉立,等.老化试验条件下的IGBT失效机理分析[J].中国电机工程学报,2015,35(20):5293-5300.LAI W,CHEN M Y,RAN L,et al.Analysis of IGBT failure mechanism based on ageing experiments [J].Proceedings of the CSEE,2015,35(20):5293-5300 (in Chinese).
[10] SUN X Y,GONG S J,HAN G,et al.Pruning Elman neural network and its application in bolt defects classification[J/OL].International Journal of Machine Learning and Cybernetics,2018:1-16[2018-12-12].http://doi.org/10.1007/s13042-018-0871-0.
[11] 王小川.MATLAB神经网络43个案例分析[M].北京:北京航天航空大学出版社,2013:197.
[2] 刘宾礼,刘德志,唐勇,等.基于加速寿命试验的IGBT模块寿命预测和失效分析[J].江苏大学学报(自然科学版),2013,34(5):556-563.LIU B L,LIU D Z,TANG Y,et al.Lifetime prediction and failure analysis of IGBT module based on accelerated lifetime test [J].Journal of Jiangsu University (Natural Science Edition),2013,34(5):556-563 (in Chinese).
[3] 李明,王跃,高远,等.采用动态电压上升控制的1 700 V大功率IGBT有源门极驱动技术[J].高电压技术,2014,40(8):2513-2519.LI M,WANG Y,GAO Y,et al.Active gate drive technology for 1 700 V high power IGBT with dynamic voltage rise control [J].High Voltage Engineering,2014,40(8):2513-2519 (in Chinese).
[4] 陈冰,鲁刚,房红征,等.基于退化数据和DBN算法的IGBT健康参数预测方法 [J].计算机测量与控制,2017,25(5):71-75.CHEN B,LU G,FANG H Z,et al.Prediction method of IGBT health parameters based-on degradation data and DBN algorithm [J].Computer Measurement & Control,2017,25(5):71-75 (in Chinese).
[5] 王宇琦.IGBT疲劳老化失效剩余使用寿命预测的研究[D].西安:西安电子科技大学,2017.
[6] 胡震.功率器件的故障诊断及疲劳寿命预测[D].天津:天津理工大学,2014.
[7] 方鑫,周雒维,姚丹,等.IGBT模块寿命预测模型综述 [J].电源学报,2014(3):14-21.FANG X,ZHOU L W,YAO D,et al.An overview of IGBT life prediction models [J].Journal of Power Supply,2014(3):14-21 (in Chinese).
[8] OH H,HAN B,MCCLUSKEY P.Physics-of-failure,condition monitoring and prognostics of insulated gate bipolar transistor modules:a review [J].IEEE Transactions on Power Electronics,2015,30(5):2413-2426.
[9] 赖伟,陈民铀,冉立,等.老化试验条件下的IGBT失效机理分析[J].中国电机工程学报,2015,35(20):5293-5300.LAI W,CHEN M Y,RAN L,et al.Analysis of IGBT failure mechanism based on ageing experiments [J].Proceedings of the CSEE,2015,35(20):5293-5300 (in Chinese).
[10] SUN X Y,GONG S J,HAN G,et al.Pruning Elman neural network and its application in bolt defects classification[J/OL].International Journal of Machine Learning and Cybernetics,2018:1-16[2018-12-12].http://doi.org/10.1007/s13042-018-0871-0.
[11] 王小川.MATLAB神经网络43个案例分析[M].北京:北京航天航空大学出版社,2013:197.