Abstract IGBT has multiple degradation mechanisms. The existing methods using a single characterization parameter cannot characterize the comprehensive health status of the device, but multiple characterization parameters cannot quantify and intuitively reflect the comprehensive health level. Therefore, this paper proposes a characterization method of IGBT comprehensive health index (CHI) based on online status data of aging life experiment. Different parameters contain different degradation information. By extracting and fusing effective information and eliminating overlapping and invalid information from multiple parameters, the obtained CHI is used to characterize the health level. The main steps are as follows: (1) Determining the sensitive degradation parameters; (2) Using the physical model method for feature selection; (3) Using the data-driven method for feature extraction; (4) Performing feature fusion and obtaining the CHI. The online state data of the IGBT aging life experiment provided by NASA were used to verify the algorithm, and the results proved that the method of feature extraction and fusion can more accurately characterize the comprehensive health status of the device. Based on this work, it is proved that extracting and fusing effective information from features is a valuable technique.

中文翻译：

---

基于在线状态数据的IGBT综合健康指标表征方法

摘要 IGBT 具有多种退化机制。现有的使用单一表征参数的方法无法表征设备的综合健康状况，而多个表征参数无法量化和直观地反映综合健康水平。因此，本文提出了一种基于老化寿命实验在线状态数据的IGBT综合健康指数(CHI)表征方法。不同的参数包含不同的退化信息。通过从多个参数中提取和融合有效信息，消除重叠和无效信息，得到的CHI用于表征健康水平。主要步骤如下： (1)确定敏感退化参数；(2) 使用物理模型方法进行特征选择；(3) 使用数据驱动的方法进行特征提取；(4)进行特征融合，得到CHI。利用NASA提供的IGBT老化寿命实验在线状态数据对算法进行验证，结果证明特征提取融合的方法能够更准确地表征器件的综合健康状态。基于这项工作，证明从特征中提取和融合有效信息是一项有价值的技术。