Prognostics for electronics components of avionics - NASA IGBT accelerated ageing case study

摘要

Insulate gate bipolar transistors (IGBTs) are widely used in electric vehicles,railway locomotive and new generation aircrafts, due to the IGBTs haveadvantages in small conduction resistance and small drive current. Hence, thereliability of IGBTs directly affect the reliability and performance of these vehiclesystems. In recent years, a series of research works about IGBT reliability,failure mode and ageing analysis have been carried out widely, and a suitableprognostics method for IGBT and an efficient algorithm for predicting the IGBTRemaining Useful Life (RUL) become increasingly important.In recent years, despite the research works on IGBT reliability, failure mode andeffect analysis are carried out widely, the prognostics and prediction of the IGBTremaining useful life are still the bottleneck in the research work to developIGBT health management system.In this thesis, a framework and algorithm of IGBT prognostics and RULprediction were developed. The research was based on the IGBT acceleratedageing experiments. The IGBT ageing data were processed and analysed, andthe mechanism of the IGBT degeneration was studied. Additionally, the IGBTdegradation models were built to simulate the IGBT degeneration processwhich were utilised to develop the prognostics algorithm for predicting the IGBTRUL.Gamma distribution model, Exponential distribution model, Poisson distributionmodel and the combining distribution model were established, and Monte Carlosimulation was utilised in the algorithm to compute the IGBT remaining usefullife. The collector emitter voltage (VCE) was used as precursor parameter inprognosis to predict the RUL. Seven IGBTs were experimented in thisprognostics research.The RUL prediction results were analysed and compared, and the prognosticsalgorithm was developed and summarised. The accuracy of the RUL predictionwas presented, and the root mean square error was utilised to analyse andcompare the efficiency and applicability of different models. The study of theIGBT prognostics and algorithm development were summarised anddemonstrated.