Analog Circuit Fault Diagnosis Based On Deep Belief Network And Neural Network

Since the 1960 s,relevant researchers and scholars have devoted their energy to the research of analog circuit fault diagnosis methods.After several generations of efforts,the field has achieved remarkable research theories and technical achievements.However,the characteristics of the analog circuit itself,that is,the tolerance of its component parameters and the failure modes of the circuit are complex and diverse,which leads to errors in the diagnosis method in practical applications and affects the accuracy of the results.In this paper,it is proposed to apply deep belief network to extract fault features,and establish a BP neural network fault classification model.At the same time,the two are combined with intelligent optimization algorithms to achieve better analog circuit fault diagnosis.Based on the existing fault diagnosis theories and methods,this paper first proposes a method of using deep belief networks to solve the problem of obtaining deep feature vectors in fault diagnosis.As a tool for fault feature extraction,deep belief network can discover deeper rules in fault signals,obtain the optimal feature vector,and then use it as the input of a trained BP neural network to classify faults.Deep belief networks overcome the problems that other feature extraction methods tend to ignore useful information in the signal and that the extracted features are prone to overlap,thereby extracting depth and essential features.The simulation example proves the feasibility of the implementation steps of the method and the effectiveness of the diagnosis effect.Aiming at the difficulty of fault pattern recognition for analog circuits under the conditions of existing theoretical methods and corresponding technologies,this paper proposes an analog circuit diagnosis method that uses quantum particle swarm optimization to optimize deep belief networks and BP neural networks.The method first uses quantum particle swarm optimization to optimize the structural parameters of the deep belief network,and then uses it to extract the fault characteristics of the circuit,and then uses quantum particle swarm optimization to optimize the structural parameters of the BP neural network to determine the best network structure and obtain the obtained network structure.The optimal feature set is used as the input of the optimized BP neural network to obtain the optimal classification result.Quantum particle swarm optimization has a strong optimization ability,and can quickly improve the generalization ability of the network.This method eliminates the blindness inselecting network structure parameters when building a fault diagnosis model,thereby greatly improving the convergence speed of the network.A simulation example verifies the accuracy of this method for fault diagnosis.