Design And Implementation Of Health Monitoring System For Rolling Bearings Of CNC Machine Tools

CNC machine tools are the basic industry of the manufacturing industry as well as the strategic industry.With the advancement of modern technology,CNC machine tools are also developing on the road of high reliability,high precision,intelligence and networking.On this basis,monitoring the health status of CNC machine tools and fault diagnosis and early warning is an inevitable choice to achieve this goal.Rolling bearings are the key components of CNC machine tools that are prone to failure.The fault vibration signal usually has strong nonlinearity and non-stationaryity,and is often disturbed by noise due to the complexity of the working environment.Therefore,many scholars at home and abroad have conducted research.Various time-frequency methods and their improvement methods have been continuously proposed,and deep learning has been gradually applied to the diagnosis of rolling bearing faults,which makes the accuracy of diagnosis continuously improved and the intelligence is also continuously enhanced.However,with the advent of the industrial big data era,the monitoring of mechanical health has entered the era of big data,bringing new opportunities and challenges.In this paper,the numerical control machine rolling bearing is taken as the research object.Starting from the analysis of the causes and vibration characteristics of the rolling bearing fault,the various time-frequency analysis methods are compared,and the traditional time-frequency analysis method is combined with the deep convolutional neural network.A method for diagnosing the health of rolling bearings combined with EEMD and VMD with convolutional neural networks.The method can not only exploit the advantages of EEMD and VMD in processing nonlinear non-stationary signals,but also fully exploit the powerful ability of deep convolutional neural networks to extract fault characteristics,adapt to the needs of big data fault diagnosis,and avoid the choice of human characteristics.It enhances the intelligence of feature extraction and is a method for end-to-end diagnosis of the health of rolling bearings.The accuracy of this method for the CWRU rolling bearing data set reached 99.6%.The multi-class fault diagnosis methods were tested and compared with them,and the advantages of the method were obtained.Finally,the rolling bearing health monitoring system based on this method is completed,and the research results are summarized and forecasted.