Research On Fault Feature Parameter Extraction Techniques For Power Electronic Circuits

The prognostics and health management(PHM)technology is of critical importance to enhance system reliability and maintainability.As the wider use in various fields and severe situations,power electronic equipment faces the challenge to develop PHM technology.On the one hand,PHM methods can now be divided into three categories(fault diagnoisis,condition monitoring,and fault prediction)and fault feature detection is an essential part for all three.On the other hand,power switch and electrolytic capacitor are the devices with the highest failure rate in power electronic equipment.Therefore,power switch and electrolytic capacitor are selected as the subjects for study,and this thesis is focused on the fault precursor parameters and the detection methods of these parameters.Firstly,the research status is reviewed.Fault precursor paramters of power switch and electrolytic capacitor are introduced,and the corresponding detection method are classified.The emphases of fault feature extraction technique are summarized,which lays the foundation for the follow-up studies.To have knowledge of failure mechanisms,accelerated aging experiments are carried out.For power switch,thermal cycling experiment,power cycling experiment and gate aging experiment are designed.For electrolytic capacitor,storage experiment,direct current(DC)aging experiment and alternating current(AC)aging experiment are designed.To find corresponding precursor parameters,several parameters are tested after different level aging.Threshold voltage,on-state voltage,diode forward voltage,leakage current of power switch and capacitance,loss agle(tan?),leakage current of electrolytic capacitor are chosen for tests.With the analysis and discussion of these measured parameters,on-state resistance(voltage),ringing frequencies,and threshold voltage are chosen as fault precursor parameters for power switch,and equivalent series resistance(ESR)is indicated to be the best fault precursor paramter for electrolytic capacitors.Then,the fault feature parameter extraction methods for MOSFETs are studied.For on-state voltage extraction,a voltage clamp circuit based on voltage source and Zener diode is proposed,which effectively suppresses voltage spikes and reduces RC delay during switching transients.For ringing frequency extraction,a method based on the voltage across stray inductance is proposed first.To simultaneously otain the information about input voltage,a method based on empirical mode decomposition(EMD)and fast Fourier transform(FFT)is also proposed.For threshold voltage extraction,a threshold voltage monitoring method based on stray inductance is proposed,which avoides the use of current sensor.Simulation and experimental results are both provided to verify the effectiveness of these proposed methods.Besides the power switch,the fault feature parameter extraction methods for electrolytic capacitor are also studied.The ripple voltage and ripple current of the output capacitor are analyzed and their internal relationship is revealed.On this basis,a tan? and ESR extraction method based on inductor ripple current and output ripple voltage is proposed for Buck converter,and a ESR extraction method based on the output voltage jump is proposed for Boost converter.Both the proposed methods apply to CCM and DCM resistive-load conditons and converter-load condition.The validity of analysis and the effectiveness of the proposed methods are verified by simulation and experimental results.Finally,parameter identification methods based on hybrid system are studied.The fundamental principle of parameter identification method based on hybrid system model is introduced and conventional hybrid system models of Buck and Boost converters are presented.A deep analysis of conventional hybrid system models is conducted and the drawbacks of them are pointed out.For Buck converter,a parameter identification method based on local linear model is proposed,which eliminates the negative effect of nonideal factors on identification precision.For Boost convereter,a corretion term for the suddenly-change in the diode current caused by switching actions is introduced to modify the primary model,which achieves high identification accuracy.By comparing the simulation identification results of the proposed model with the primary model,the effectivenss of the modified model is confirmed.Experimantal results are provided to verify the effectiveness of the proposed methods.The work in this dissertation provides technical support and theoretical basis for PHM technology of power devices in power electronic converters,which has certain value for promoting power electronic system PHM technology.