Inverter Fault Diagnosis Research Based On Mathematical Morphology

With the development of power electronic device,variable frequency drive system which is power supplied by inverter is widely used in modern industry,aeronautics and astronautics,military and other fields.However,poor working conditions often make the system in abnormal condition even in failure state.If we can’t timely detection failure and take measures in time,whole variable frequency drive system will be paralyzed,which is lead to consequences,ranging from huge economic losses to the society to casualties.Therefore,it is of great significance to study the fault diagnosis of inverter and take measures in time.Under this background,this paper is taking inverter power transistors which is include three-phase voltage source inverter and AC-DC-AC variable frequency drive system as research object.This paper put forward to using basic mathematical morphology algorithm to deal with inverter output signals,distinguish the differences between all kinds of fault signals and realize fault diagnosis.Firstly,the method of inverter fault diagnosis and basic mathematical morphology algorithm is introduced.We build up the voltage source model and AC-DC-AC variable frequency drive system model in the MATLAB software.Using model are presented to simulate open-circuit fault diagnosis of inverter transistor by setting signals to zero at a fixed time in order to get fault signals.Then,this paper presents a new method to realize fault detection and diagnosis of inverter by combining with calculating mathematical morphology based generalized fractal dimensions.Analysis of the calculation results show that each kind of power transistor open circuit fault(single 6,double 15)the fractal dimension of the inverter output current calculation result are different.Using the calculation results combine with the BP neural network we can distinguish the differences between all kinds of fault signals and find the accurate position of fault transistor.Multi-scale morphology spectrum analysis is a popular method of signal processing and fault diagnosis in recent years.It has been applied in fault diagnosis of transmission line and rolling bearing.we can distinguish the five kinds of single current waveform accurately when single or double transistor open-circuit fault is occur by using multi-scale morphology spectrum analysis which has a ability to quantitatively analyze waveform.The three-phase current signal of the inverter can be regarded as the combination of the five single-phase fault signals which has been mentioned,then,we put forward effective fault diagnosis system,with the help of the Elman neural network to achieve the accurate positioning of the fault transistor.In view of this method,the article has given simulation results and verification experimental results show that the method is correct and workable.Finally,this article do some research on transistor fault diagnosis of AC-DC-AC variable frequency drive system.In view of this paper we can use the difference of the current of support capacitor to make a distinction between all kinds of fault when the single transistor failure,the same bridge arm under two transistor failure,cross two transistor failure in different bridge,the same side transistor failure in different bridge and normal condition is happen in AC-DC-AC variable frequency drive system.In the process of differentiation,we can use the concept of power spectrum entropy to analysis the current of support capacitor for the fault feature information which can achieve transistor fault diagnosis of inverter.