期刊名称：International Journal of Future Generation Communication and Networking
印刷版ISSN：2233-7857
出版年度：2016
卷号：9
期号：9
页码：1-14
出版社：SERSC
摘要：Health parameters prediction of radar transmitter needs to complete the electronic components failure characteristics analysis and fault symptom parameters prediction. The article looks the Insulated Gate Bipolar Transistor (IGBT) as the research object, combines the accelerated life experimental data of NASA-ARC, and determines the turn-off voltage spike peak value of collector-emitter as the basis of failure prediction, carries out IGBT health prediction method research based on process neural network. In view of the slow convergence speed and easily trapped in local minimum of back-propagation learning algorithm in the traditional process neural network, develop a kind of learning algorithm based on orthogonal basis function of Levenberg-Marquardt(LM) optimization. The experimental results show that the process neural network algorithm based on LM optimization can better predict the performance degradation trend of IGBT; it has high accuracy and achieves the short-term prediction of IGBT health state.
