Research On Key Issues About Condition Monitoring For Power Semiconductor Device

With power electronic systems in various areas of application range, they haveplay a principal role in maximizing energy efficiency, developing and utilizing newand renewable sources of energy. So, power electronics systems become growingimportance for economic and social development of different countries. However, thewider use of power electronic systems also results in some corresponding problemsand consequences, namely incalculable losses caused by the faults in powerelectronics systems. Related researches show that power semiconductor devices aremain factors affecting the reliability of power electronic systems. Thus, in order toenhance reliability of power electronic systems, it is inevitable to study the somemethods improving the reliability of power devices. Currently, the best solution toimprove reliability is condition monitoring for power electronic system. For conditionmonitoring, this paper mainly studies the thermal performance of power devices,semiconductor physics modeling for power devices and their bonding wires lift-off.The research contents are as follows:1. The thermal issues of power devices are one important reason for affectingreliability of high integration power electronic systems. The paper selects theinsulated gate bipolar transistor (IGBT) module as an example, and constructs thethermal network model using lumped parameter method based on known geometricdimensions, material properties, and thermal parameters of IGBT module. Basing onthis, the paper proposed two model parameters extraction strategies. One method isthat the unknown model parameters are extracted by using numerical simulationmodel of IGBT model. Another method is that model parameters are obtained bycalculating unsteady state heat conduction with inverse heat conduction problemmethod. Simulation and experimental results all show that the above model and itsparameters extraction methods are higher accuracy.2. Considering gradual failure of IGBT module, a temperature prediction methodof IGBT module based on autoregressive moving average (ARMA) model is putforward in this paper. The feasibility of this temperature prediction method is verifiedby temperature simulation and experiments of IGBT module. Combining thetemperature prediction method and thermal network model, the temperature in different layers of IGBT module can be all predicted.3. In order to obtain real-time health parameters of IGBT module for conditionmonitoring, this paper constructs the semiconductor physics model of IGBT chip, andthis model takes into account the different injection of carriers (electron and hole) inIGBT base. In addition, the extraction methods of unknown model parameters aregiven in the paper. Comparison between simulation results based on the physicsmodel and calculation results of Hefner model of IGBT chip or experimental datum ofIGBT module, the correctness of IGBT chip model can be checked.4. Bonding wires are one of the most vulnerable parts in IGBT module.Considering shorter response times in power electronic systems, the short period oftime is required for condition monitoring for power devices in system. So, the strategyof indirectly monitoring the bonding wires by measuring electric performancesoutside IGBT module is an appropriate way for the fault of bonding wires lift-off.Hence, in this paper, firstly, the principles of bonding wires lift-off affecting theterminal capacitances, parasitic inductances and equivalent resistances in IGBTmodule are studied. Secondly, influence of the terminal capacitances, parasiticinductances and equivalent resistances on the electrical parameters outside IGBTmodule. Finally, the fault of bonding wires lift-off can be monitored by measuring thevoltage or current of IGBT module. The experimental results prove the correctness oftheoretical analysis.