Motor Fault Diagnosis Based On D-S Evidence Theory Model Fusion

The generation of the motor liberated a large number of labor force,in production,life occupies an important position.If the motor in operation fails,it will cause more economic loss and social harm than the motor itself if it is not switched in time.Therefore,it is necessary to monitor and diagnose the motor.If the fault can be detected in the early stage of the fault,many risks can be avoided.The main research work of this paper is as follows:(1)The short circuit between turns of motor stator,broken rotor bar,air gap eccentricity and bearing faults are explained simply.The feature extraction method of motor fault data is introduced in this paper.The kurtosis and root mean square are taken as the feature values in the time domain,and the first four eigenmode components are used to extract the signal features by the method of empirical mode decomposition,and a6-dimensional feature vector group is formed.(2)For the faults existing in the motor,the whale algorithm is used to optimize the model of support vector machine to classify and identify the faults of the motor bearing.In terms of convergence factor and population size,the whale algorithm is improved.The improved whale algorithm,the annealing whale algorithm,the chaos whale algorithm and the original whale algorithm were combined with the support vector machine respectively to compare the accuracy of the four fault diagnosis models and the length of the operation time.It is found that the optimized host vector machine based on the improved whale algorithm in this paper has high efficiency and short computation time in motor bearing fault diagnosis.(3)Based on evidence theory,the credibility and probability distribution functions of evidence fusion are improved to build a model.Compare the diagnosis results using current data and vibration in the evidence model,and then improve the model and optimize the diagnosis results.Combined with support vector machine and evidence theory,the current data and vibration data were analyzed simultaneously,and the decision level fusion was compared with the analysis of the two separately.(4)Set up an experimental platform for motor fault diagnosis.Do destructive experiments on the motor,respectively,simulation machine stator interturn short circuit,rotor broken bar,air gap eccentricity,bearing fault types.In the experiment,multiple measurement channels are selected for the same motor running state to collect information at the same time,to ensure that the advantages of the signals are complementary,to obtain a more accurate description of the object,and to ensure that the original signal can be completely preserved.In order to ensure the applicability of the established model,the signals collected under different loads were substituted into the model to verify the accuracy,and the accuracy of the model under different loads was compared.