Research On Fault Diagnosis Of Motor System Based On EEMD And Support Vector Machine Classifier

Motor is an important power equipment in industrial production and plays an important role in national security,industrial production and national life.In recent years,the motor because of the fault problem to the country and people bring huge economic losses of numerous cases.In order to reduce the loss caused by motor fault,ensure the stability of industrial power system,and achieve the economic purpose of motor maintenance,the research and development of motor fault diagnosis technology is of great engineering practical significance.Motor fault parts are numerous,common can be summed up as stator,rotor and bearing three categories,especially the bearing part of the frequent failure.Through the analysis of bearing structure and failure mechanism,the mathematical model is established to obtain the bearing failure frequency combined with vibration signal for analysis.The traditional spectrum analysis method is more effective for the original signal with stable characteristics,but it is difficult to extract the characteristic components of motor fault when applied to motor fault diagnosis.This paper starts from the analysis of vibration signal by hot spot signal processing theory--EMD decomposition,and finds that the mode aliasing problem occurs in the result.The improved algorithm is put forward overall average way of empirical mode decomposition(EEMD),and makes a detailed interpretation of approximate entropy theory,at last,by experimental simulation on bearing under normal condition and have unbalanced fault state,respectively,to extract the vibration signal of EEMD decomposition,found on different IMF analysis of approximate entropy,can lead to form the main component failure.Although the approximate entropy value can measure the complexity of the physical quantity represented by the component,it cannot accurately distinguish which fault mode the fault component belongs to.Therefore,this paper introduces support vector machine to recognize the fault mode.In this paper,the problem of model selection and performance test under the new method is analyzed and studied in detail.Aiming at the research object of this paper,it is decided that the support vector machine with polynomial kernel function is preferred for classification.Finally,the feasibility of the technical scheme is verified by on-site diagnosis of cnooc's fault diesel units.Decorate more groups of sensors sampling points,first of all to EEMD processing of the original vibration signal collected,the IMF component vibration fault feature extracting,the early acquisition of sample set under condition of the four patterns of the SVM classifier for training and testing,and then use SVM to identify the vibration fault features of approximate entropy,accurately identify the fault source,eventually achieved very good results.Therefore,the introduction of white noise EEMD method for fault signal processing can effectively eliminate mode aliasing and obtain clear vectors representing fault characteristics,which is a new and efficient signal processing method.Support vector machine(SVM)can accurately identify the approximate entropy of fault and diagnose the typical fault mode of motor.