绝缘栅双极型晶体管（Insulated Gate Bipolar Transistor，IGBT）因其驱动功率小、通流容量大和易于并联等优点，逐渐成为主流功率器件并广泛应用于新能源、智能电网、电机驱动和工业变流等领域。IGBT作为电力电子装置的核心部件，其可靠性是提升整个电力电子装置性能的基础，因此，针对IGBT可靠性研究中晚期失效的两大难点展开研究：IGBT疲劳失效机理和IGBT寿命预测。

  针对目前IGBT多物理场耦合分析采用静态功率损耗导致有限元仿真精度低、边界条件理想化导致未能结合实际工况以及现有键合线疲劳失效研究不完整—未考虑二极管芯片键合线等问题，以BPJ5-630/1140矿用变频器中逆变器部分IGBT为研究对象，提出了一种考虑其应用工况，结合IGBT特征模型来完整分析其多物理场耦合下键合线疲劳失效机理的方法。首先，建立FZ800R33KF2C型IGBT三维模型，再通过仿真与实验方法验证ANSYS Simplorer搭建的IGBT特征模型动态特性，得到矿用逆变器额定工况下IGBT芯片和二极管芯片瞬时功率损耗，并根据矿用逆变器温升实验确定隔爆腔内环境温度参数。最后，在ANSYS Workbench平台下进行热-电-力多物理场耦合有限元仿真，完整分析了不同芯片下键合线脱落对IGBT的影响。分析表明：在大功率IGBT键合线多物理场耦合分析中不能忽略二极管芯片及其键合线，IGBT芯片键合线脱落占总数62.5%及以上或二极管芯片键合线脱落数占总数33%及以上时，IGBT结温超过最大工作温度，且温升和等效应力迅速增大。在二者共同脱落的情况下，二极管芯片键合线脱落对IGBT影响更大。

  针对传统寿命预测模型预测精度不足、数据特征挖掘不充分等问题，提出了结合互补集合经验模态分解（Complementary Ensemble Empirical Mode Decomposition，CEEMD）和相空间重构（Phase Space Reconstruction，PSR）的Tent混沌映射改进鹈鹕算法（Tent Pelican Optimization Algorithm，TentPOA）优化长短期记忆网络（Long Short-term Memory，LSTM）的寿命预测模型。首先，分析IGBT寿命表征参数及老化数据样本，考虑训练数据的完整性和真实性，确定选用NASA PCoE实验室公开IGBT老化数据中集-射极关断峰值电压V_CE-P为IGBT寿命表征参数并作为原始数据集。再通过CEEMD分解和PSR挖掘数据潜在信息和扩充数据样本，并引入过零率优化数据处理流程，实现高效构建高质量训练数据集。最后，在Tent混沌映射改进POA算法优化LSTM预测模型下进行IGBT寿命预测，并与其他预测模型进行对比分析。分析表明：以LSTM模型预测结果为基准，所构建的TentPOA-LSTM寿命预测模型，其预测结果对比POA-LSTM预测模型评价指标MAE相对下降18.21%、RMSE相对下降10.71%、MAPE相对下降18.22%、R² 相对提升178.8%，验证所提出模型的优越性。

  所提出的疲劳失效分析方法完善了IGBT键合线失效机理分析，为进一步研究与实际工况相结合的IGBT多物理场耦合有限元仿真提供参考。所构建的IGBT寿命预测模型表现出良好的预测精度和稳定性，具有一定的工程应用价值。

  Insulated Gate Bipolar Transistor (IGBT) has gradually become the mainstream power device and is widely used in new energy, smart grid, motor drive and industrial converter due to its advantages of low driving power, large current capacity and easy parallel connection. IGBT is the core component of power electronic devices, and its reliability is the basis for improving the performance of the entire power electronic device. Therefore, two difficulties in the middle and late failure of IGBT reliability research are studied: IGBT fatigue failure mechanism and IGBT life prediction.

  Aiming at the problems of the current IGBT multi-physical field coupling analysis, the low accuracy of finite element simulation due to static power loss, the failure to combine the actual working conditions due to the ideal setting of boundary conditions, and incomplete research on fatigue failure of existing bonding wire without considering the bonding wire of diode chip, taking the IGBT of the inverter in the BPJ5-630 / 1140 mine inverter as the research object, a method of considering its application conditions and combining the IGBT characteristic model to completely analyze the fatigue failure mechanism of the bonding wire under multi-physical field coupling is proposed. Firstly, the three-dimensional model of FZ800R33KF2C IGBT is established, and then the dynamic characteristics of IGBT characteristic model built by ANSYS Simplorer are verified by simulation and experiment. The instantaneous power loss of IGBT chip and diode chip under rated working condition of mine inverter is obtained, and the ambient temperature parameters in flameproof cavity are determined according to the temperature rise experiment of mine inverter. Finally, the thermal-electric-mechanical multi-physics coupling finite element simulation is carried out under the ANSYS Workbench platform, and the influence of bonding wire shedding on IGBT under different chips is completely analyzed. The analysis shows that the diode chip and its bonding wire cannot be ignored in the multi-physical field coupling analysis of high-power IGBT bonding wire. When the IGBT chip bonding wire shedding accounts for 62.5 % or more of the total or the diode chip bonding wire shedding accounts for 33 % or more of the total, the IGBT junction temperature exceeds the maximum operating temperature, and the temperature rise and equivalent stress increase rapidly. In the case of both shedding, the diode chip bonding wire shedding has a greater impact on IGBT.

  Aiming at the problems of insufficient prediction accuracy and insufficient data feature mining of traditional life prediction models. A life prediction model of Long Short-term Memory (LSTM) optimized by Tent Pelican Optimization Algorithm (TentPOA) combined with Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Phase Space Reconstruction (PSR) is proposed. Firstly, the IGBT life characterization parameters and aging data samples are analyzed. Considering the integrity and authenticity of the training data, the V_CE-P  in the IGBT aging data published by NASA PCoE laboratory is selected as the IGBT life characterization parameter and used as the original data set. Through CEEMD decomposition and PSR mining data potential information and expanding data samples, and introducing zero-crossing rate to optimize data processing flow, high-quality training data sets can be efficiently constructed. Finally, IGBT life prediction is carried out under the LSTM prediction model optimized by Tent chaotic mapping improved POA algorithm, and compared with other prediction models. The analysis shows that based on the prediction results of the LSTM model, the TentPOA-LSTM life prediction model is constructed. Compared with the evaluation index MAE of the POA-LSTM prediction model, the prediction results are relatively reduced by 18.21%, RMSE is relatively reduced by 10.71%, MAPE is relatively reduced by 18.22%, and R²  is relatively increased by 178.8%, which verifies the superiority of the proposed model.

The proposed fatigue failure analysis method improves the failure mechanism analysis of IGBT bonding wire, and provides a reference for further research on IGBT multi-physics coupling finite element simulation combined with actual working conditions. The constructed IGBT life prediction model shows good prediction accuracy and stability, and has certain engineering application value.

［1］ 王晨苑.基于温敏电参数的高压大功率IGBT模块可靠性评估方法[D].武汉:武汉大学,2021.

［2］ 周文鹏,曾嵘,赵彪,等.大容量全控型压接式IGBT和IGCT器件对比分析：原理、结构、特性和应用[J].中国电机工程学报,2022,42(08):2940-2957.

［3］ Yang Shao yong, Bryant A T, Mawby P A, et al. An industry-based survey of reliability in power electronic converters[J]. IEEE Transactions on Industry Applications, 2011, 47(3): 1441-1451.

［4］ 李辉,刘盛权,冉立,等.大功率并网风电机组变流器状态监测技术综述[J].电工技术学报,2016,31(08):1-10.

［5］ 李红岩,杨朝旭,荣相,等.矿用逆变器功率器件故障预测与健康管理技术现状及展望[J].工矿自动化,2022,48(05):15-20.

［6］ 黄思思.航天关键功率电子器件IGBT筛选技术研究[D].成都:电子科技大学,2018.

［7］ 刘泽浩,肖岚.模块化多电平变流器IGBT开路故障诊断与定位方法综述[J/OL].中国电机工程学报:1-16[2023-04-05].

［8］ 肖飞,刘宾礼,罗毅飞,等.IGBT疲劳失效机理及其健康状态检测[M].北京:机械工业出版社,2019.8.

［9］ 李亚萍,周雒维,孙鹏菊.IGBT功率模块加速老化方法综述[J].电源学报,2016,14 (06):122-135.

［10］ Schuler S, Scheuermann U. Impact of test control strategy on power cycling lifetime[C]. Proc. PCIM, 2010: 355-360.

［11］ Xiong Y, Cheng X, Shen Z J, et al. Prognostic and warning system for power-electronic modules in electric, hybrid electric, and fuel-cell vehicles[J]. IEEE Transactions on Industrial Electronics, 2008, 55(6): 2268-2276.

［12］ 曾东,孙林,周雒维,等.基于加速老化试验IGBT性能退化特征参量的可靠性评估[J].电工电能新技术,2019(7):20-28.

［13］ 姬煜轲,侯婷,何智鹏,等.一种柔直换流阀用压接型IGBT功率子模块加速老化试验方法[J].南方电网技术,2021,15(05):1-11.

［14］ 赵子轩,陈杰,邓二平,等.负载电流对IGBT器件中键合线的寿命影响和机理分析[J].电工技术学报,2022,37(01):244-253.

［15］ 曾东.基于加速老化试验IGBT性能退化特征参量的可靠性评估[D].重庆:重庆大学,2019.

［16］ Maogong Jiang, Guicui Fu, Martin Bendix Fogsgaard, et al. Wear-out evolution analysis of multiple-bond-wires power modules based on thermo-electro-mechanical FEM simulation[J]. Microelectronics Reliability, 2019, 100-101(C).

［17］ 王春雷,郑利兵,方化潮,等.键合线失效对于IGBT模块性能的影响分析[J].电工技术学报,2014,29(S1):184-191.

［18］ 郑利兵,韩立,刘钧,等.基于三维热电耦合有限元模型的IGBT失效形式温度特性研究[J].电工技术学报,2011,26(07):242-246.

［19］ 何怡刚,周健波,李兵.NPC三电平中IGBT焊料层多物理场耦合失效分析[J].电子测量与仪器学报,2019,33(04):71-78.

［20］ 方佳怡,刘宪云,夏丽,等.基于ANSYS及MATLAB的温度场分布对半桥型IGBT器件的影响[J].固体电子学研究与进展,2018,38(03):178-183.

［21］ Jianbo Zhou, Bing Li, He Yigang, et al. Electro-Thermal-Mechanical Multiphysics Coupling Failure Analysis Based on Improved IGBT Dynamic Model[J]. IEEE Access, 2019, 7: 174155-174166.

［22］ 史明明,葛雪峰,任政燚,等.基于ANSYS的IGBT模块健康状态研究[J].电力电子技术,2021,55(12):65-68.

［23］ 喻磊,肖标,黄安迪,等.IGBT模块材料选型的仿真研究[J].南方电网技术,2020,14(05):22-27.

［24］ 张科峰,丁伟伟.基于有限元法的IGBT模块焊料层可靠性研究设计[J].电源世界,2017(1):36-38.

［25］ Y. Jia, F. Xiao, Y. Duan, et al. PSpice-COMSOL-Based 3-D Electrothermal–Mechanical Modeling of IGBT Power Module. IEEE Journal of Emerging and Selected Topics in Power Electronics, 2020, 4(8): 4173-4185.

［26］ 张钢,王运达,刘志刚,等.城轨牵引供电系统多尺度和多物理域建模仿真方法[J].电工技术学报,2022,37(12):3097-3107.

［27］ 骆小满,阮江军,邓永清,等.基于多物理场计算和模糊神经网络算法的变压器热点温度反演[J].高电压技术,2020,46(03):860-866.

［28］ 耿学锋,何赟泽,李孟川,等.IGBT多物理场建模技术与应用研究概述[J].中国电机工程学报,2022,42(01):271-290.

［29］ Liu H P, Wu W T, Wu H, et al. Lifetime prediction of a modified Y-source inverter in photo-voltaic application[J]. Microelectronics Reliability, 2018, 89-90: 1157-1163.

［30］ 徐丽宾.基于电热耦合方法的IGBT模块结温测量工程应用研究[D].长沙:湖南大学,2020.

［31］ 苗硕.一种改进的IGBT寿命预测方法[D].西安:西安理工大学,2021.

［32］ 吕高泰,雷万钧,赵佳琪,等.MMC子模块关键器件长时间尺度运行工况老化分析与可靠性评估[J].高电压技术,2020,46(10):3469-3476.

［33］ 王彦刚,Chamund Dinesh,李世平,等.功率IGBT模块的寿命预测[J].机车电传动,2013(02):13-17+27.

［34］ 帅双旭,熊炜,彭月,等.基于电热耦合模型和寿命预测的IGBT可靠性评估[J].电力科学与工程,2021,37(06):17-25.

［35］ 张经纬.基于有限元法的压接型IGBT器件单芯片子模组疲劳寿命预测[D].北京:华北电力大学(北京),2018.

［36］ 郭来小.IGBT基于故障物理的失效分析及电子器件加速寿命模型的研究[D].成都:电子科技大学,2018.

［37］ 蒲亚博,王艳艳,赵方超,等.典型制导弹药故障预示预警系统的设计[J].兵器装备工程学报,2021,42(08):39-45.

［38］ Liu Jiacheng, He Yigang, Zhou Jianbo, et al. IGBT junction temperature prediction method based on improved artificial bee colony algorithm for optimizing SVR[C]. IOP Conference Series: Materials Science and Engineering, 2019, 59(6): 1965- 1984.

［39］ 郭子庆,王学华.基于神经网络的IGBT模块剩余使用寿命预测模型[J].电测与仪表,2023,60(01):132-138.

［40］ 王新春,赵金金.基于小波神经网络的IGBT寿命研究[J].电工技术,2020(10):114-116.

［41］ 刘子英,朱琛磊.基于Elman神经网络模型的IGBT寿命预测[J].半导体技术,2019,44(05):395-400.

［42］ 史业照,郭斌,郑永军.基于LSTM网络的IGBT寿命预测研究[J/OL].中国测试:1-6[2021-12-08].

［43］ 韩恒贵.基于深度学习的IGBT故障预测研究[D].北京:北京交通大学,2019.

［44］ 孙梓涵.基于数据驱动的IGBT寿命预测研究[D].北京:北京交通大学,2020.

［45］ 葛建文,黄亦翔,陶智宇,等.基于Transformer模型的IGBT剩余寿命预测[J].半导体技术,2021,46(4):316-323.

［46］ 维诺德·库马尔·卡纳.IGBT理论与设计[M].杨兵,康玄武,王杨译,译.北京:机械工业出版社,2020．

［47］ 徐延明,赵成勇,周飞,等.适用于电路仿真的IGBT模块暂态模型研究[J].电源学报,2016,14(03):28-37.

［48］ 唐新灵,崔翔,张朋,等.感性负载条件下IGBT开通过程分析[J].华北电力大学学报(自然科学版),2017,44(02):33-41.

［49］ 安德烈亚斯·福尔克,麦克尔·郝康普.IGBT模块：技术、驱动和应用[M].韩金刚,译.北京:机械工业出版社,2016．

［50］ 杜泽晨,吴沛飞,任志军,等.15kV SiC IGBT芯片常温动态特性测试[J].半导体技术,2020,45(09):723-729.

［51］ 李永皓,陈敏.IGBT模块开关特性测试研究与驱动参数选择方法[J].电源学报,2020,18(01):162-167.

［52］ 陈树宗,李旭,彭文,等.基于数值积分与功率损耗测试的冷轧电机功率模型[J].东北大学学报(自然科学版),2017,38(03):361-365.

［53］ Elzbieta Ewa Lesniewska, Jan Olak. Improvement of the Insulation System of Unconventional Combined Instrument Transformer Using 3-D Electric-Field Analysis[J]. IEEE Transactions on Power Delivery, 2018, 33(6).

［54］ 王胜辉,许育帅,孙凯旋,等.多物理场耦合下不同短路与接地故障对变压器绕组状态影响的仿真研究[J/OL].高电压技术:1-12[2023-04-09].

［55］ 贺啸秋,熊永亮,彭泽瑞,等.旋转肥皂泡热对流能量耗散与边界层特性的数值模拟[J].物理学报,2022,71(20):174-192.

［56］ Goud B. Shankar, Srilatha Pudhari, Mahendar D., et al. Thermal radiation effect on thermostatically stratified MHD fluid flow through an accelerated vertical porous plate with viscous dissipation impact[J]. Partial Differential Equations in Applied Mathematics, 2023, 7.

［57］ Yueqiang Yu, Tingang Ma, Suling Wang, et al. Simulation Analysis of Stress Field of Walnut Shell Composite Powder in Laser Additive Manufacturing Forming[J]. Journal of Renewable Materials, 2023, 11(1).

［58］ Bowen Liang, Jefferson Cuadra, Kavan Hazeli, et al. Stress field analysis in a stony meteorite under thermal fatigue and mechanical loadings[J]. Icarus, 2020, 335(C).

［59］ Xie Fei, Zhu Wenhao, Sun Dongxu, et al. Finite element simulations of corrosion defect with different inclination angles in submarine pipelines using multi-physics field coupling and deformation geometry[J]. International Journal of Pressure Vessels and Piping, 2022, 200.

［60］ 杨鹏,胡业荣,王贵山.温度对半导体激光器退化的影响[J].国防科技大学学报,2020,42(01):45-50.

［61］ Daryl L. Logan.有限元方法基础教程[Ｍ].北京:电子工业出版社,2014．

［62］ 胡金宝.基于机器学习的IGBT多参数性能退化预测方法研究[D].西安:西安电子科技大学,2020.

［63］ 彭英舟.基于端口电特性的IGBT模块可靠性研究[D].重庆.重庆大学,2017.

［64］ M. Tounsi, A. Oukaour, B. Tala-Ighil. Characterization of high-voltage IGBT module degradations under PWM power cycling test at high ambient temperature[J]. Microelectronics Reliability. 2010, 50: 1810–1814.

［65］ Zhou Luowei, Zhou Shengqi, Xu Mingwei. Investigation of gate voltage oscillations in an IGBT module after partial bond wires lift-off. Microelectronics Reliability. 2013, 53: 282–287.

［66］ 汪波,胡安,陈明,等.IGBT关断瞬态电压尖峰影响及抑制[J].半导体技术,2011,000(007):501-504．

［67］ Wang Panjie, Wu Jiang, Wei Yuan, et al. CEEMD-MultiRocket: Integrating CEEMD with Improved MultiRocket for Time Series Classification[J]. Electronics, 2023, 12(5).

［68］ Jiaxun Liang, Kexin Yan. PM2.5 prediction based on CEEMD-SSA-KELM[J]. Academic Journal of Environment & Earth Science, 2021, 3.0(2.0).

［69］ Wang Shuai, Zhao Xiaodong, Liu Wenhang, et al. Impact-Type Sunflower Yield Sensor Signal Denoising Method Based on CEEMD-WTD[J]. Agriculture, 2023, 13(1).

［70］ Chan HsiaoLung, Chang HungWei, Hsu WenYen, et al. Convolutional Neural Network for Individual Identification Using Phase Space Reconstruction of Electrocardiogram[J]. Sensors, 2023, 23(6).

［71］ 黄飞江,陈演羽,李廷会,等.基于灰色模型和混沌时间序列的卫星钟差预测算法[J].电子学报,2019,47(07):1416-1424.

［72］ 崔峰,何仕凤,来兴平,等.基于相空间重构与深度学习的冲击地压矿井时间序列b值趋势研究[J/OL].煤炭学报:1-13[2023-02-04].

［73］ Guneet Bedi, Ganesh Kumar Venayagamoorthy. Development of an IoT-driven building environment for prediction of electric energy consumption[J]. IEEE Internet of Things Journal, 2020, 7(6): 4912-4921.

［74］ 刘超,张爱琳,李树刚,等.基于Pearson特征选择的LSTM工作面瓦斯浓度预测模型及应用[J/OL].煤炭科学技术:1-9[2023-02-04].

［75］ 马志侠,张林鍹,邱朝洁,等.基于CEEMD-SSA-LSTM的园区综合能源系统两阶段优化调度[J/OL].高电压技术:1-11[2023-02-04].

［76］ Trojovský P, Dehghani M. Pelican Optimization Algorithm: A Novel Nature-Inspired Algorithm for Engineering Applications. Sensors (Basel). 2022 Jan 23; 22(3): 855.

［77］ 滕志军,吕金玲,郭力文,等.一种基于Tent映射的混合灰狼优化的改进算法[J].哈尔滨工业大学学报,2018,50(11):40-49.

［78］ 商立群,李洪波,侯亚东,等.基于特征选择和优化极限学习机的短期电力负荷预测[J].西安交通大学学报,2022,56(04):165-175.

［79］ Strumberger I, Bacanin M, Tuba M, et al. Resource scheduling in cloud computing based on a hybridized whale optimization algorithm[J]. Applied Sciences, 2019, 9(22): 4893.

［80］ 段锦,姚安妮,王震,等.改进的麻雀搜索算法优化无线传感器网络覆盖[J/OL].吉林大学学报(工学版):1-11[2023-02-03].