A General Intelligent Method for Degradation Sensitive Parameters Extraction and Life Prediction of Bipolar Transistor

In recent years, key instrumentation and control devices in the nuclear power industry have entered the life depletion period. Due to diverse failure mechanisms and scatter orbits of degradation, the life of core component Bipolar Transistor is difficult to be determined. A migration learning based life prediction method is proposed to determine the sensitive degradation mechanism parameters. According to the degradation mechanism, the degradation principle of transistor is given, and degradation path of the R
BE
, R
BC
, and
$\beta$
parameters are proposed. A Bayesian classifier algorithm is developed to classify the triode to be tested into one of the pre-defined types according to the measured parameter values. Finally, the proposed three-parameter competitive degradation model is used to obtain the final triode lifetime distribution with reference to the parameter degradation curve corresponding to the type. The application case demonstrate that the method can intelligently generalize the degradation characteristic of different kind of transistor to get the accurate prediction of lifetime through empirical data, reduce the test costs significantly.