Research On The Method Of Fault Feature Extraction Of DC-DC Circuit Under Variable Conditions

DC-DC converter is widely used,and its reliability is closely related to the safety and reliability of equipment operation.Once the failure occurs,it will lead to the failure of equipment function,or even the whole system will be paralyzed,resulting in economic losses.Fault feature extraction is the key technology of DC-DC circuit fault diagnosis and the premise of accurate fault diagnosis.Accurate diagnosis of DC-DC circuit fault can effectively reduce the loss caused by circuit fault and ensure the stable operation of equipment.When DC-DC circuit works,the input voltage and load will change on demand.In this paper,buck circuit and boost circuit.are taken as examples to study the feature extraction method of DC-DC circuit soft fault caused by the change of component parameters under variable conditions.Based on the analysis of the basic structure principle of Buck circuit and Boost circuit,this paper analyzes the types of faults caused by capacitance parameter changes and sets different working conditions of the circuit.According to the failure mode and working conditions,the Simulink simulation model of the open and closed loop circuit is built.The output voltage of each fault mode under variable working conditions is the research of the fault feature extraction method for the fault signal.Three methods are studied.The fault feature extraction method based on the optimal fractional wavelet.This paper optimizes the fractional order of the fractional wavelet transform based on the particle swarm algorithm Second,the relative wavelet energy of each component after the optimal fractional wavelet decomposition is used to construct the fault feature vector,and the SVM is used as the classifier for fault diagnosis.The results show that the method is more suitable for the fault feature extraction of closed-loop buck circuit and boost circuit than the optimal fractional wavelet;The fault feature extraction method based on VMD Shannon entropy,this paper combines VMD and Shannon entropy theory,and proposes a fault feature extraction method based on VMD Shannon entropy.Aiming at the difficulty of parameter setting in VMD decomposition,this paper is based on different decomposition scales.The center frequency of each modal component of VMD decomposition determines its optimal decomposition layer number,and finally,SVM is used for fault diagnosis to verify the effectiveness of the fault feature extraction method.The results show that compared to the feature extraction method of EMD Shannon entropy,VMD can effectively overcome the modal aliasing problem existing in EMD,with higher diagnostic accuracy and less diagnostic time;Based on the multi-scale CNN fault feature extraction method,deep learning is applied to the DC-DC fault diagnosis field,and a multi-scale CNN-based fault feature extraction method is proposed,which uses different sizes of convolution kernels to construct a multi-scale CNN model for the fault feature extraction of the DC-DC conversion circuit,the fusion of the fault feature information extracted by the convolution kernel of different scales,the extraction of the fully connected layer features to construct the fault feature vector,and the influence of the ratio of different training samples and test samples on the fault diagnosis results.SVM classifies the fault feature vector.The results show that this method can effectively extract the fault features of Buck circuit and Boost circuit.Figure[32]table[18]reference[76]...