Research On Failure Prediction Of IGBT Modules

Over the last decade, IGBT has broad development prospects for its gate voltage control characteristics like MOSFET and the low on-resistance characteristics like BJT. However, there are some shortcomings in IGBT such as low tolerance of overvoltage, overheating, shock and interference that affect its stability. Economic loss due to IGBT failure is considerable. Therefore it is necessary to achieve Prognostics and Health Management Technology of IGBT modules by monitoring critical parameters, predicting failure to a certain extent, taking timely measure to realize on-condition maintenance. That can not only reduce routine maintenance cost, but also prevent downtime due to IGBT failures. Based on the perspectives above, following jobs have been carried out in this thesis centering on the IGBT modules failure prediction technology.1. Analyze IGBT failure factors, failure modes and failure mechanism by deeply understanding IGBT structure, characteristics and operating principle. On this basis, condition monitoring technologies of IGBT modules are analyzed and compared.2. In order to expedite IGBT failure procedure, accelerated life test platform is designed. Apply over-stress to accelerate the aging of IGBT modules while obtaining performance degradation parameters for IGBT failure prediction.3. Random filtering method combines the recurrence relation obtained from a prior probability and the current observations to estimate the system status. It is suitable for real-time failure diagnosis and remaining useful life prediction. Therefore, this paper presents an IGBT RUL prediction method based on particle filter theory. By preprocessing the raw data from accelerated life test, a bond between particle filter theory and exponential fitting model is used for failure prediction of IGBT modules.