Motor Bearing Fault Diagnosis Based On Data Drive

In today’s era,sustainable development is increasingly emphasized,and the safety maintenance of equipment is an important way to extend the service life of equipment,ensure safe production,and achieve sustainable development.Therefore,equipment inspection and diagnostic technology,as the core technology of equipment safety maintenance,is of great significance and deserves our high attention and attention.In modern industry,the application of electric motors has penetrated into various fields,including machinery,automobiles,electronics,aviation,energy,etc.,and almost all production equipment relies on electric motor drive.Normal operation of electric motors is essential for improving production efficiency,product quality,and extending equipment service life.Therefore,timely and accurate diagnosis of electric motor faults can not only avoid downtime and maintenance costs caused by faults but also improve production efficiency and equipment reliability,ensuring production safety and equipment lifespan.Traditional motor fault diagnosis methods mainly rely on accurate mathematical modeling of the motor system,and simulate the response of the system to different inputs and conditions,so as to compare the actual behavior of the system and the behavior predicted by the model,and finally achieve the effect of fault identification and diagnosis.Traditional motor fault diagnosis methods mainly rely on accurate mathematical modeling of the motor system,and simulate the response of the system to different inputs and conditions,so as to compare the actual behavior of the system and the behavior predicted by the model,and finally achieve the effect of fault identification and diagnosis.However,the traditional model-based fault diagnosis method requires a detailed mathematical model,and has some defects,such as heavy calculation and difficult application in complex motor systems.In contrast,the method of motor fault diagnosis based on data drive has strong real-time performance,high precision and easy implementation.And motor bearing vibration data is easy to obtain,high resolution,can reflect the running state of the motor.Therefore,this paper takes the motor bearing as the research object and adopts the data-driven method for motor fault diagnosis.However,the traditional model-based fault diagnosis method requires a detailed mathematical model,and has some defects,such as heavy calculation and difficult application in complex motor systems.In contrast,the method of motor fault diagnosis based on data drive has strong real-time performance,high precision and easy implementation.And motor bearing vibration data is easy to obtain,high resolution,can reflect the running state of the motor.Therefore,this paper takes the motor bearing as the research object and adopts the data-driven method for motor fault diagnosis.(1)Select i-ALERT pressure sensor according to experimental conditions and install it on different motors to obtain vibration signal data of motor bearings.The collected data will be uploaded to the cloud to provide data basis for the following research.(2)For different faults,the frequency amplitude is different,the signal-to-noise ratio is low,and the discrimination rate of single feature or several features is low.Based on the collected vibration signals of HC1 bearing of horizontal lathe and its envelope signals,the corresponding data sets are classified according to different faults,and then the features are extracted according to the original vibration signals and envelope signals of motor bearing.Then,the Principal Component Analysis(PCA)method is used to reduce the dimensionality of these high-dimensional eigenvalues from the highdimensional space to the two-dimensional space.Finally,machine learning algorithm is used to classify.3)Machine learning fault diagnosis algorithm needs to extract motor fault features and relies on expert experience and other defects.In this paper,the Hilbert transform and envelope spectrum are used to extract the fault frequency of motor bearings,and the fault center frequency is set according to the fault frequency,and the continuous wavelet transform is used to generate the two-dimensional image of each bearing state on the fault frequency band.Finally,Convolutional Neural Network(CNN)is used to learn the discriminant features of two-dimensional images and classify the bearing faults.Experiments show that the deep learning method using convolutional neural network has a higher accuracy than the machine learning method.(4)Aiming at the problems such as large parameters and slow running speed of the traditional convolutional neural network model,this paper proposes to add 1×1convolutional before flattening to improve the CNN deep learning model,reduce the input dimension of the full connection layer and make the model lightweight.And compared its performance with the unimproved CNN deep learning model.Experiments show that the improved CNN deep learning model has fewer parameters and faster speed.