Research On Fault Diagnose Method Of Motor Based On Data-driven

Motor fault diagnosis has always been a research hotspot in the field of equipment diagnosis.Due to the complex structure of the motor and the harsh operating environment,the fault features are subject to many interference factors.Based on the traditional method,the fault characteristics cannot be accurately obtained from the original data,resulting in fault diagnosis results less accurate.With the maturity of signal processing technology and the rapid development of artificial intelligence methods,fault information that was difficult to obtain can be effectively extracted using advanced methods.In this paper,based on the characteristics of motor fault data,based on signal processing and intelligent methods,the fault diagnosis method is studied.The key research contents are as follows:First,the motor fault is modeled,and the mechanical and electrical fault characteristics are analyzed through the mechanism.For mechanical faults,the vibration signal mainly generates harmonic components,and the electrical fault mainly causes the current signal to change.When the running state of the motor is stable,the fault characteristics can be effectively identified by using the mechanism characteristics;however,the motor is often in an unstable state of the speed fluctuation,and the fault frequency characteristic cannot be accurately extracted by the conventional Fourier transform.In this paper,the equal-angle resampling method is studied for this problem.The experimental results show that the method can effectively extract the fault characteristic frequency under the condition of speed fluctuation,and then accurately identify the motor fault type.Then the variational mode decomposition method is studied.For the rotor breaking fault of the motor,the characteristic frequency and the fundamental frequency of the power supply are very close.Based on Fourier transform and empirical mode decomposition,the two frequencies cannot be effectively separated.This paper uses the variational mode.Decomposing the original signal can accurately separate the two.Due to the choice of different variational modal parameters,the decomposition results are very different.Therefore,the influence of modal parameters on the decomposition results is analyzed by simulation.Further,in order to reduce the manual adjustment of the variational modal parameters,the genetic algorithm is used for optimization,and the optimized parameters are substituted into the model to obtain better decomposition results.Finally,the motor rotor broken bar fault is simulated to verify that the method can accurately separate the fault characteristic frequency and the fundamental frequency of the power supply,and improve the accuracy of the diagnosis results.Secondly,the correlation vector machine is studied.The correlation vector machine is used to classify multiple faults of the motor,and the influence of kernel function on classification performance is studied.Due to the complex characteristics of motor faults,the meanings of different fault characteristics are different.In order to select the appropriate features as the input of the classifier,the experimental results show that the frequency characteristics and wavelet energy characteristics are the most accurate as the classifier input.In the initial stage,the number of fault samples is small,and the correlation vector machine and the support vector machine classification model are applicable to small sample data,but the latter cannot directly output probability values,and the kernel function is limited by Mercer.The correlation vector machine overcomes this shortcoming.Finally,the classification effect is compared by experiments.The results show that the correlation vector machine has high classification accuracy and good sparsity.Finally,the fault diagnosis system platform was designed and the software was developed using MFC.