| With the rapid development of power electronics technology and new energy technology,motor drive system is widely used in various fields.As the core equipment in the motor drive system,three-phase inverter plays an important role in electric vehicles,aerospace equipment,new energy and other fields.However,Due to the long-term influence of high voltage,high current and high-frequency switching,inverters are particularly prone to failure,which will degrade the performance of the equipment and affect personal safety in serious cases.Therefore,it has important research significance for fast and accurate fault diagnosis of inverters.In this thesis,the fault diagnosis method of three-phase two-level voltage source inverter based on signal processing is studied.The details of the work are as follows:Firstly,the working principle of the inverter is analyzed,and the model of the inverter motor drive system is established at the same time.By simulating the fault state of the power switch tube of the inverter.And the phase voltage of the faulty bridge arm is used as the input signal for fault diagnosis.The voltage fault waveform under different voltages and different loads.It provides the basis for the subsequent fault signal preprocessing and fault diagnosis.Secondly,to solve the problems of fault redundant information,a large amount of signal data and noise interference.A signal preprocessing method combining Compressed Sensing（CS）and Wavelet Packet Decomposition（WPD）is proposed.Reduce the size of fault redundant information and data volume through signal sparse representation and compressed measurement.To reduce the error of the reconstructed signal,a fusion of the Gram matrix and fast iterative systolic thresholding algorithm（FISTA）is proposed to optimize the measurement matrix in CS.The signal is decomposed into different scales by the WPD algorithm for denoising.Finally,the fault signal preprocessing method is verified on the simulation platform.The simulation results show that,compared with the CS algorithm with the Toeplitz matrix as the measurement matrix,the voltage signal reconstruction error of the preprocessing algorithm in this thesis is small,the amount of data is small,the noise interference is significantly reduced,and the important characteristic signals of the original data are included.Then,to solve the problems of low fault diagnosis accuracy and slow diagnosis speed.Based on the above preprocessing method,a fault diagnosis model based on RBFNN is established.To quickly determine the sample center of the hidden layer,the fault feature information is used to improve the K-means algorithm.To avoid the RBF neural network falling into local optimum,the Beetle Antennae Search（BAS）algorithm is used to optimize the weight threshold of RBFNN.The fault diagnosis method is verified by the simulation platform.And the simulation results show that,compared with the traditional RBF neural network fault diagnosis,the improved RBFNN algorithm has the advantages of fast convergence speed and high fault diagnosis rate.Finally,the experimental platform of a three-phase inverter fault diagnosis system is built.The inverter fault data acquisition system with a motor drive is designed.The above-mentioned fault signal preprocessing and fault diagnosis methods are verified on this experimental platform.The experimental results show that the algorithm proposed in this paper has the characteristics of fast diagnosis speed,high diagnosis rate,and strong anti-interference ability.The experimental results demonstrate the validity and practicability of the CS-WPD-RBF fault diagnosis method. |
