Characteristics Analysis Of IGBT Module With Bond Wire Failure Based On APEN And Wavelet Transform

In recent years, new energy generation technologies represented by the wind powerhas achieved rapid and comprehensive development. As the proportion of wind powerin the grid is increasing year by year, in addition to water power, it has become a kind ofrenewable energy power generation of most mature and industrial-scale. However, theinformation returned from wind farms shows that the failure rate of wind powergeneration system is about three times that of traditional power generation systems, andit indicates that the overall reliability of wind power system still has the massiveinsufficiencies. The reliability of the power electronic converter, as IGBT for the corecomponents, has a direct impact on the reliability of the overall operation of the windpower device. In addition, IGBT has the both advantages of BJT and power MOSFET,making it applied in more and more areas. According to the latest survey, the powersemiconductor devices are the most vulnerable part of the power electronic converter,and IGBT also has become the most widely used among the power semiconductordevices.As bond wire fault is one of the most important reasons for the failure of the IGBTmodule, the further study was carried out in this paper. From the structural characteristicof the IGBT module, we know that the failure characteristics of bond wire andparalleled IGBT chip could be reflected through the gate turn-off voltage waveform andits trends were pointed out in the paper.By the destructive test on the bond wire of an unencapsulated IGBTmodule(2MBI150U4H-170), bond wire fault caused by bad working environment ornormal aging was simulated on purpose. Therefore, the chip fault result from variousreasons was simulated by cutting all the bond wires in the emitter of IGBT chip. Theinfluences of bond wire fault on working characteristic of IGBT module was compared.In order to extend the method on bond wire fault analysis to on-line monitoring, thispaper used the prevalent analysis methods based on wavelet transform for the unsteadysignal and approximate entropy analysis method based on rate of new informationgeneration to extract the fault characters of the gate voltage signal sequences. The usefulfault characters were extracted from huge amount of raw data successfully, and thevariation of lift-off number of bond wire to fault characters was compared. The resultsrevealed that the percentage variation of fault characters will change in different levelwhen the lift-off number of bond wire increased as well as a paralleled IGBT chip failure.