Research On Fault Analysis Of Rolling Bearing Based On Machine Learning And Virtual Instrument Technology

Rolling bearing is a widely used core supporting component in rotating machinery.Its failure will affect the safe and stable operation of the whole equipment;At the same time,for the manufactured bearing,it is necessary to measure the vibration according to the national standard to judge whether it meets the quality requirements.Therefore,the research on vibration measurement and fault diagnosis technology of rolling bearing has important theoretical significance and practical application value.Taking rolling bearing as the object,the paper mainly focuses on bearing vibration signal processing and fault diagnosis technology including the following three aspects: vibration signal denoising technology,machine learning technology based on artificial feature extraction and deep learning technology based on automatic feature extraction.A wavelet neighborhood threshold denoising algorithm,random forest based on joint features and bearing fault diagnosis algorithm model based on MSKACNN model are proposed.Finally,a rolling bearing vibration measurement and intelligent fault diagnosis system is designed based on virtual instrument technology.The main research contents are as follows:1)A wavelet neighborhood threshold denoising algorithm is proposed.Due to the measurement system and environment,the bearing vibration signal is often accompanied by useless information and noise.To solve this problem,a new neighborhood threshold function noise reduction method is proposed based on wavelet threshold noise reduction algorithm.Through simulation analysis and comparative experiments,compared with other common noise reduction functions,this method has better noise reduction effect,higher signal-to-noise ratio and avoid the defect of excessive noise reduction.2)A random forest fault diagnosis algorithm based on joint features is proposed.In the process of machine learning fault diagnosis based on artificial features,feature extraction is an essential step.The extracted features are directly related to the diagnosis results.In the paper,the energy entropy feature of modal component of empirical mode decomposition of vibration signal is introduced to form a feature set with time-domain feature and frequency-domain feature,and the random forest algorithm is applied to fault diagnosis.Through the comparative analysis with other common machine learning algorithms,the algorithm obtains higher diagnosis accuracy and generalization ability,and the effectiveness of this method is verified.3)A bearing fault diagnosis model based on MSKACNN is proposed.In the neighborhood of pattern recognition,convolutional neural network is widely used,because of its superior automatic feature extraction ability,to replace the cumbersome manual feature extraction.Based on the one-dimensional convolutional neural network model,a multi-dimensional kernel convolution structure is proposed in the paper.Compared with the unmodified model WDCNN,through experimental analysis and comparison,the model has higher diagnosis accuracy and neighborhood adaptive ability.In the recognition of mixed ball and normal type,it shows that MSKACNN has the advantage with higher recognition accuracy.4)A rolling bearing vibration measurement and intelligent diagnosis system based on VI(Virtual Instrument)& AI(Artificial Intelligent)is designed and completed.With the mixed development of Lab VIEW,MATLAB and Python,a complete set of rolling bearing vibration signal measurement and fault diagnosis system is designed according to relevant national standards,including signal measurement,storage and diagnosis functions.In the aspect of intelligent diagnosis,a method of embedding intelligent diagnosis algorithm into vibration signal acquisition system is developed to realize the real-time analysis of collected signal flow by machine learning algorithm model.The developed system can be used not only for fault bearing diagnosis,but also for bearing quality judgment,analyzing the problems existing in bearing processing,and providing a basis for bearing manufacturers to improve bearing processing quality.