Analog Circuit Fault Diagnosis Methods Based On LMD And Neural Network

In the information society,electronic equipment has become a necessary tool for human production and life.With the rise of the electronic industry,people put forward higher demand for the repair and maintenance of electronic products.Because of the nonlinear factors of analog circuit itself and the tolerance of electronic components,it is difficult to diagnosis the analog circuit fault.The traditional analog circuit fault diagnosis technology can not be applied to high complexity,high integration circuit,therefore,it is urgent to explore new diagnostic methods.With the rapid development of artificial intelligence,neural network,wavelet analysis,genetic algorithm,support vector machine and other artificial intelligence algorithm are widely applied to analog circuit fault diagnosis,which opened up a new direction for the field of fault diagnosis.In fault diagnosis,feature vector extraction and pattern recognition are two important steps in fault diagnosis,therefore,how to construct feature vector and classifier is the research direction of analog circuit fault detection technology.This paper focuses on the extraction of fault feature vector,and proposes a fault diagnosis method for analog circuits based on LMD and neural network.The main contents are as follows:(1)In this paper,the wavelet packet theory is used to extract the fault features of analog circuits,and BP neural network is used to classify the faults.Simulation results show the effectiveness of this method.(2)LMD(local mean decomposition)algorithm is researched,and a fault feature vector extraction method based on LMD energy feature is proposed.The feature vector extraction method decomposes the complex multi-component FM signal into a single component frequency modulated signal,and the result of the decomposition maintains the amplitude and frequency of the original signal,therefore it is suitable for the circuit fault diagnosis.(3)On the basis of LMD,combined with multiscale entropy theory,a new method of fault feature extraction based on LMD and multiscale entropy is proposed.Multiscale entropy reflects the self similarity and complexity of time series under different scale factors,and has strong anti noise and anti interference ability,Can effectively extract fault features,can effectively extract fault features.Simulation results demonstrate the effectiveness of the proposed method.(4)In this paper,the principle and application of neural network are introduced,and the structure model of BP neural network with its learning algorithm are studied.Aiming at the defects of slow training speed and easy to fall into the local optimum value of BP neural network,the application of extreme learning machine in analog circuit fault diagnosis is studied.Without adjusting the weights and thresholds,the algorithm can get the unique optimal solution only by setting the number of neurons in the hidden layer.This method has fast learning speed and good generalization performance,which greatly reduces the time of fault diagnosis.