Research On Analog Circuit Fault Diagnosis Method Based On Wavelet Analysis And Neural Network

Analog circuits are widely used in civil,military project,aerospace and other fields,Maintenance of electronic circuit is very important owning to the increasing complexity of modern electronic systems.The failure rate is high and the fault types are complex and diverse in a system because analog circuits are nonlinear and vulnerable to environmental interference.Fault diagnosis technology has been a difficult problem that restricts the development of modern electronic technology.With the development of analog circuits to multi functional and large-scale,traditional diagnostic methods have been difficult to apply,it is urgent to explore more intelligent automatic test and diagnosis methods.Good learning ability and approximation performance of neural network,along with good time-frequency localization characteristic of wavelet analysis provides a new way to solve many problems in the traditional methods of fault diagnosis of analog circuits.In this paper,the key theory and technology of analog circuit fault diagnosis based on wavelet analysis and neural network are studied deeply.Firstly,the research and development status at home and abroad is reviewed,and the basic principle and algorithm of neural network are introduced.A simulation example is given to verify the effectiveness of the neural network method.Then,the wavelet multi-resolution analysis and wavelet packet analysis are applied to the fault feature extraction,which effectively shorten the network training time and improve the accuracy of fault diagnosis.The wavelet function is used as the hidden layer transfer function of neural network to construct the compact wavelet neural network to improving the performance of neural network.Combine the theory of information fusion,a new fault diagnosis method based on D-S evidence theory is proposed as the fault diagnosis method based on single information source neural network has some problems such as low accuracy,poor reliability and lack of generalization ability of single neural network.The analog circuit output response frequency segments are independently diagnose and the obtained results are used as independent evidence bodies for fusion decision making.The experimental results show that the proposed method can effectively improve the accuracy and reliability of fault diagnosis.