.dc / Dc Switching Power Supply Fault Diagnosis

The switching power supply is a typical equipment. It is used widely. At the same time, it is difficult to solve its fault diagnosis problem by using general or traditional fault diagnosis theory and method due to such factors as lack of uniform fault model, existing component tolerance, non-linearity and so on. Neural network representing the intelligent technology which provides an effective approach, receives more academic attention nowadays.The topologies and working principle of the switching converter, which is the core of the switching power supply, are mentioned in this dissertation firstly, and Full-bridge converter is selected as the study object. In succession, the converter's working principle is introduced in detail. Based on analysis of its working principle, a new method of making the output voltage of the converter as the fault character parameters is presented. The operation situations of different parts of the converter in open-circuited fault are simulated by using Simulink toolbox, and output voltage waves are produced. Then the feature extraction of the waveform is got through wavelet packet transform. Character vectors are constructed by obtaining the power spectrum of decomposed coefficients.In this paper, a new efficient intelligent fault detection method based on the wavelet packet transform and the artificial neural network is also presented. BP Neural network is applied to Fault pattern recognition, design method of neural network learning samples is discussed, and the simulation of intelligent fault diagnosis of the converter is made. Such method is proved effectively by the research. It can improve the performance of fault diagnosis and decrease much more workload than expert system which needs a great deal of data analysis.Finally, the design scheme for an intelligent fault diagnosis system of the converter including hardware structures, software configuration and program framework is introduced briefly.