严苛工作环境下绝缘栅双极型晶体管(IGBT)的可靠性及寿命预测研究

Insulated Gate Bipolar Transistor (IGBT) reliability relates to the performance of the entire mechanical and electrical products. At the extreme application environments, such as high temperature, high pressure, high current, high moisture salt spray, high mechanical vibration intensity etc, the IGBT's performance is very different to conventional environment. Some conventional analysis methods are no longer valid. This project researches IGBT's failure mechanism, running real-time monitoring and fault diagnosis, reliability metrics and screening, the remaining life prediction at the extreme application environments. (1) From the perspective of bonding wire shedding, solder layer fatigue, silicon chip technology, material defects etc, explore IGBT failure mechanisms, research fail criteria and treatment measures. (2) Proposed an IGBT operation status reverse monitoring method based on surface temperature and other parameters of external measurements, to establish fault sample database, using kernel principal component analysis (kernel PCA) to extract fault feature, using on line support vector machine diagnose the fault type. (3) Proposed an IGBT reliability metrics method based on generalized reliability computational model. Integrate probability reliability and fuzzy reliability to deal with the uncertainty of generalized stress and strength. (4) Proposed an IGBT life prediction model based on variable scale Gaussian process (VSGP) and use the phase trajectory recursive diagram to determine life curves' predictability. The proposed model can accurately express the non-stationary life curve. It owns high prediction accuracy and outputs prediction confidence.

绝缘栅双极型晶体管(IGBT)的可靠性关乎整个机电产品性能。在高气温、高气压、高电流、高潮湿盐雾、高机械振动强度等严苛工作环境下，IGBT的性状与常规环境迥然不同，一些常规分析手段不再有效。本项目研究该环境下的IGBT失效机理、运行状态实时监测及故障诊断、可靠性度量、寿命预测问题。（1）从键合线脱落、焊接层疲劳、硅芯片工艺等角度探索IGBT失效机理，研究失效评判标准和处理对策。（2）提出一种基于表面温度等外部参数的IGBT模块运行状态逆向监测法，建立故障样本数据库，用核主成分分析提取故障特征，用在线支持向量机诊断故障类型。（3）提出一种基于广义可靠度计算模型的IGBT可靠性度量方法，整合了概率可靠度和模糊可靠度，以处理广义应力和强度的不确定性。（4）提出一种基于可变尺度高斯过程的IGBT寿命预测模型，并用相轨迹递归图来判断寿命曲线的可预测性。能准确刻画寿命曲线的非平稳性，能输出预测置信度。

绝缘栅双极型晶体管(IGBT)广泛应用于工业生产的方方面面，因此，提高IGBT模块在严苛工作环境下的运行可靠性和精确预测IGBT寿命是提高机电产品性能的关键问题。本项目针对这些关键问题，在实验基础上，运用各种数据处理方法、智能算法模型和可靠性度量方法等对“严苛工作环境下绝缘栅双极型晶体管(IGBT)的可靠性及寿命预测”展开了研究。主要的研究成果和获得的进展如下：(1) 首先研究和分析了IGBT模块的失效机理，对键合引线失效和焊料层失效进行了研究，提出了相应的失效对策；(2) 模拟IGBT实际运行环境，针对实际工作环境中IGBT模块状态参数难以准确测量的问题，建立了IGBT模块结温、开关损耗和开关时间计算模型。其次在状态监测的基础上，建立了IGBT模块的数据库，基于饱和压降Vce建立模型实现了IGBT故障的快速诊断。(3) 建立了在应力、强度是随机变量、模糊变量四种组合中的任意一种时都适用的可靠性度量模型——广义可靠度计算模型(General Reliability Algorithm model, GRA模型)，建立基于该模型的可靠性方法与概率可靠性、模糊可靠性方法之间的联系，提高了其实用性。(4) 提出了一种具有最优尺度特点的高斯过程模型（OSGP）模型用于IGBT模块寿命预测。提出的高斯过程模型(OSGP)能够很好地处理时间序列预测中的随机性和非平稳性，分析结果表明OSGP的预测精度优于SVM、BP神经网络以及GP，是一种预测精度高、能适应较小规模训练样本的IGBT寿命预测模型，为IGBT寿命预测提供了重要的工具。