Fault Diagnosis Method Of DC-DC Circuit Based On Multidimensional Features

The DC-DC circuit is used as the basic components of non-renewable energy systems,large computers and biomedical instruments,which plays a decisive role in the operation of equipment.Therefore,it is very important to monitor fault status and development trend of DC-DC circuit.A parameter fault diagnosis method for DC-DC circuit based on multidimensional features is proposed.The parameter fault diagnosis experiments of Buck and Boost circuits are carried out to study the parameter faults caused by capacitor degradation.The specific research contents are as follows:The basic principles of Buck and Boost circuits are briefly analyzed,and the fault types of capacitor parameter degradation are discussed.The Buck and Boost simulation circuit models and Boost physical circuit models are established.By setting the fault mode to collect the output voltage and current of each circuit as the characteristic signals,and three parameter fault diagnosis methods are explored.The DC-DC circuit fault diagnosis method based on variational modal decomposition and optimized LSSVM.The output voltage signal and output current signal of Buck simulation circuit were decomposed according to preset scale by using variational mode decomposition,and multiple IMF components were obtained.By analyzing the relative entropy of each IMF component,the IMF pseudo-component below the threshold was removed and the remaining signals were reconstructed to extract the time and frequency characteristics.Finally,the new feature vectors obtained by feature fusion are input into LSSVM for model training.To improve the classification capability of LSSVM,a new Political Optimizer(PO)is used to optimize the internal parameters of LSSVM.Experimental results show that this method can effectively solve the pseudocomponent problem of variational modal decomposition algorithm.In addition,the PO optimization algorithm can effectively improve the LSSVM classification ability,and has higher diagnostic accuracy than Particle Swarm Optimization(PSO)and Genetic Algorithm(GA).Fault diagnosis method of DC-DC circuits with convolutional neural network based on channel attention mechanism.Firstly,one-dimensional original voltage signal and load current signal of Boost simulation circuit are collected and preprocessed to obtain labeled data samples which are divided into training set and test set.Secondly,in order to improve the convolutional neural network model features the ability to learn,to channel attention module combined with convolution neural network,build convolution attention neural network model,the multi-dimensional features of the training set data are obtained through the convolution layer.Then,the attention module assigns different weights to the features of different dimensions of the data,so that the model pays more attention to the region with the highest degree of category differentiation,thus improving the feature learning ability of the model.Finally,the test set is used to verify the fault diagnosis performance of the model and the diagnosis effect of this method in Buck simulation circuit is discussed.Experimental results show that compared with other methods,this method can effectively enhance the feature learning ability of convolutional neural network model and improve the fault diagnosis rate.A DC-DC circuit fault diagnosis method based on hyperparametric automatic optimization of one-dimensional convolutional and long-short time memory networks.The method is used for pattern recognition of sequential signals by establishing a deep learning model.According to the advantages of one-dimensional Convolutional Neural Network(1DCNN)and Long short-term Memory(LSTM)in extracting the spatial and temporal features of signals,to construct the hybrid network model of 1DCNN and LSTM.Aiming at the network hyperparameters of deep learning model,the optimal hyperparameters of the network model are obtained by Improved Whale Optimization Algorithm(IWOA).Then,the model is used to train the data samples of one-dimensional original voltage signals of Boost physical circuits,and Softmax is used as the classifier for fault classification.In addition,the diagnostic ability of the proposed method in dealing with Boost physical circuit and Buck simulation circuit and fault diagnosis is verified by comparing with the algorithms in the previous two chapters.Experimental results show that compared with other methods,the proposed method is not only suitable for the fault diagnosis of Boost physical circuits but also for Buck simulation circuits,and the proposed method also has good performance in noise environment.