A Study For Fault Detection Of Inverter Based On The Neural Network

To solve some faults that appear in the inverter, this paper builds a fault diagnosis system base on Neural Networks(NN). Firstly, the varied-frequency varied-speed system which used MATLAB software was established to obtain the fault prototype, then combining it with Neural Networks(NN) and using a linear BP algorithm to analysis the fault information included in the fault prototype so that the fault type of the inverter can be judged. Through this process, the requirement of the fault analysis is satisfied, then the loss can be reduced in the production.Because of the high cost, the inverters are usually set on an important position, that whether it is safe or not indirectly the production. So the reliability of the inverter is critical. But there are too many fault type, it is necessary to analysis accurately which kind of fault it is and solve it in time, so as to prevent the fault from extending. Under this condition, in order to analysis the fault quickly and accurately, a kind of fault diagnosis method must be developed to deal with the fault information.Because of the limit of condition ,we cannot get the fault information on the condition of fault which can be used as the base of fault detection and diagnosis. At first, the paper build the simulation model of varied-frequency varied-speed system, and on this base, simulate some kinds of inverter faults that can be meet usually such as short and open circuit fault of switch etc, then extract the fault information-the three-phase output current of inverter. When the fault information are transformed by Fourier -Transform, we can get the amplitude and frequency information of the three-phase current on the condition of the fault which can be used as the information of the fault diagnosis. The information input a NN that has been optimized. The NN can diagnosis the place where the fault took place so we can reduce the faultmaintain time and provide the base of the next step analyse, the NN is first adopted in the fault diagnosis of the inverter in this paper, and at the same time the characteristics of the NN, such as auto-learning, non-linearity, memory, are utilized to make the run of the inverter more reliable.The NN algorithm, which has been used in this paper, has a integrate theory base and has been validated by the theory and experiment. Under the condition that there are many faults in the inverter, applying the MATLAB programme, we can judge precisely in which part and which component the faults were taking place.