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

Boost circuit,as an important component of power electronic circuit,is widely used in industrial production,biomedical,rail transit and national defense construction.If a fault occurs during operation,it will cause significant losses in production economy.Therefore,accurate diagnosis and timely elimination of circuit faults are particularly important.The specific research contents of this paper are as follows:On the basis of analyzing the basic structure of Boost circuit,this paper first builds the simulink simulation model of Boost circuit,sets the capacitor parameter degradation standard and fault types,and collects the output voltage signals of 10 different fault types;Similarly,the capacitance of 150 W Boost physical circuit is studied for parameter degradation fault.Through parameter failure experiment platform,the circuit capacitance is tested for parameter degradation and voltage signals of 4 measuring points under 16 types of capacitor parameter degradation fault are collected;The collected voltage signals under different fault types of Boost simulation circuit and 150 W Boost physical circuit are taken as the data basis of this method.Based on this,this paper analyzes three different aspects of fault diagnosis of boost circuit capacitance parameters,and researches three methods in total.Aiming at the problems of signal aliasing,poor feature discrimination and difficulty in distinguishing fault types when the capacitor parameters of Boost circuit degenerate to different degrees,a fault feature extraction method based on adaptive variational mode decomposition is studied.Firstly,the sample entropy is used to determine the level K of the fault signal through the variational mode decomposition;Secondly,the fault signal is decomposed into subsequences of layer K for Pearson correlation comparison with the original signal,and extremely weak correlation subsequences are eliminated;Finally,the remaining subsequences are reconstructed linearly,and the 8-dimensional time domain parameters of the reconstructed signal are extracted as the feature vectors of the fault signal.The experimental results show that the accuracy of fault diagnosis of this method is higher than that of many common fault feature extraction methods in various classifier verifications,which verifies the effectiveness of this method.Aiming at the problems of high randomness,low accuracy and difficult parameter setting of the classifier in fault diagnosis,a soft fault diagnosis method of boost circuit capacitance based on improved sparrow search algorithm and optimized limit learning machine(ISSA-ELM)is studied.Firstly,the advantages and disadvantages of sparrow search algorithm are analyzed.The original sparrow search algorithm is improved by three strategies: infinite mapping initialization population,introduction of inertia weight and Levy mutation,and the ISSA algorithm is tested on the benchmark function;Secondly,in order to improve the classification performance of ELM,the improved sparrow search algorithm is used to optimize the weight and threshold of ELM,establish the ISSA-ELM diagnostic model,and select five UCI classification datasets to test the classification performance of ISSA-ELM;Finally,the extracted fault features are input into ISSA-ELM for fault classification experiment.The experimental results show that ISSA optimization speed and accuracy are improved compared with SSA,and the diagnostic accuracy and stability of ISSA-ELM in UCI dataset and circuit capacitance fault diagnosis are higher than the original ELM,and the classification performance is better than that of ELM optimized by multiple swarm intelligence algorithms.Aiming at the problem that the combination of signal feature extraction and classifier has good effect but poor universality in fault diagnosis and requires high expert experience,a capacitor soft fault diagnosis method of Boost circuit based on improved one-dimensional convolutional neural network is studied.In order to realize the end-to-end fault diagnosis effect,the one-dimensional fault voltage signal is directly input and the fault result is output.Firstly,the depth of the network structure is deepened.Secondly,residual structure,Disout and dynamic learning rate are introduced to improve the network performance.Finally,Softmax is used as the classifier to achieve fault classification.The experimental results show that the method can be used to diagnose capacitor parameter faults of Boost circuit with excellent results.Figure [54] Table [41] Reference [90]...