Research On Fault Diagnosis Method Of Main Circuit Of Converter Based On Bayesian Network

With the rapid development of frequency conversion technology and microelectronics technology,frequency converter has been widely used in the field of mine hoisting because of its excellent control ability and energy saving ability.Frequency converter is the key equipment in the mine hoisting system,which plays an important role in the transformation of electrical energy.The structure of the converter’s main circuit is complex,and the components usually work in the high voltage and high frequency environment,which are prone to failure and can not work order.The normal operation of the elevator frequency converter directly affects the safety of equipment and personnel,and also relates to the production income of the mine.Therefore,the research on the fault diagnosis of inverter main circuit has very important application value.The paper studies the fault diagnosis method of converter’s main circuit based on Bayesian network.Firstly,this paper takes ACS800 inverter of ABB company as the research object,and divides the main circuit of the converter into three parts: rectifier link,DC link and inverter link.In the study of the structure and working principle of these three links,the common faults of each link are analyzed.According to the experimental platform,the simulation models of the three links are established,the output waveforms of various faults are analyzed,and the corresponding fault features are extracted.Secondly,the Bayesian fault diagnosis model is established respectly for the three parts of the study.The data-driven method is used to build the rectifier link’s Bayesian fault diagnosis model.According to the extracted fault features,the Bayesian diagnosis model structure is established,and the parameters of the network are learned by EM algorithm.Considering a lot of data is needed for parameter learning and validation of the model,different fault modes are set in the simulation model to obtain sample data.The model is verified by test data,the results show that the model has a high accuracy.In this paper,a Bayesian model building method under the condition of small sample is proposed for DC link model building.For the structure of Bayesian model of DC link,the prior probability of root node is obtained by the historical operation data correction basic probability method,and the conditional probability of non root node is obtained by the prior beta distribution method.The proposed method can fully combine the prior knowledge of experts and the existing sample data.The test data is used to verify the established model.The results show that the model has a good effect.In order to solve the problem of attribute redundancy in fault diagnosis of frequency converter,a method of fault diagnosis model is proposed,which combines rough set and Bayesian network.Rough set method is used to reduce the complexity of attributes,because of the reduction results are not unique,a model performance evaluation algorithm based on accuracy and time is proposed to select and establish the optimal model.Test data is used to verify the optimization model,and the results show that the model has higher diagnosis accuracy than the fuzzy neural network model,which shows that the method is effective.Finally,the design of the fault diagnosis system of the converter’s main circuit is completed based on the GUI module in MATLAB,the fault diagnosis and location of the three links are realized.The fault diagnosis system has good human-computer interaction ability and strong data processing ability.It can realize waveform data preprocessing,fault diagnosis,model self-learning and so on.The function of each module of the fault diagnosis system is verified by examples.