Research On Circuit Fault Diagnosis Method Based On Deep Learning Fusion Technology

Electronic circuits play a major role in the automotive,aerospace,and guidance fields,and are the basic structures that make up various types of electronic products.As the complexity of electronic products increases,the electronic circuit structure becomes more and more complex,which brings a great challenge to circuit fault diagnosis.The traditional circuit fault diagnosis technology mainly relies on manual identification,which requires a lot of empirical knowledge and low efficiency of real-time diagnosis.In order to simplify the circuit fault diagnosis process,reduce the workload,and improve the efficiency of real-time electronic circuit fault diagnosis,this paper proposes a circuit fault diagnosis method based on long and short term memory networks and applies it to electronic circuit fault diagnosis.In order to improve the model fault feature extraction capability,a circuit fault diagnosis model based on CNN-LSTM fusion technology is built by introducing convolutional neural network on this basis.The main work of this paper is as follows:(1)The construction method of the circuit fault sample data set is given,the circuit components are analyzed by the sensitivity analysis method,and the two cases of single fault and compound fault are considered respectively.In each case,eight types of circuit faults are selected as the research object,and then reused PSpice software performs Monte Carlo simulation to complete the production of the original sample data set of circuit faults.In order to facilitate the observation of the circuit fault data set information,the principal component analysis(PCA)method is used to visualize it.(2)To reduce the complexity of the circuit fault diagnosis method and improve the efficiency of circuit fault diagnosis,the LSTM,which is good at processing sequence data,and the SOFTMAX classifier are combined to perform fault diagnosis of circuits.By analyzing the parameters of the LSTM model,an LSTM-SOFTMAX fault diagnosis model is built on the Keras deep learning library with Tensorflow as the backend.(3)To further improve the circuit fault diagnosis accuracy and enhance the model generalization ability,the LSTM-SOFTMAX model is optimized,on which CNN is introduced to process the data set and small convolutional kernel is used to extract features from the data set,and a circuit fault diagnosis method based on CNN-LSTM fusion technology is proposed,which can solve the LSTM-SOFTMAX The problem of inadequate feature learning of the model is solved.The experimental results show that the circuit fault diagnosis method based on CNN-LSTM fusion technology proposed in this paper can obtain a higher accuracy rate of circuit fault diagnosis.This method does not require too much manual intervention and does not require researchers to have deep electronic circuit expert experience,The circuit fault diagnosis model built by this method has strong generalization ability and practicability.