Health Evaluation And Residual Life Prediction Of Packing Machine Rolling Bearings Based On Deep Learning

Packaging machine usually needs to quantitative packaging of a variety of materials,the increase of demand,the long-term stability of the packaging machine,the speed of higher requirements,urgent need to grasp the real-time operation and health of the moving parts,timely targeted maintenance of the packaging machine in advance,in order to reduce sudden shutdown and production,affect the normal production of enterprises.Rolling bearing is the core component of packaging machine,which plays the role of supporting high-speed operation in the transmission system of packaging machine,and is very important to the stable operation of packaging machine.This thesis takes the rolling bearings of the horizontal seal transmission system of the single-line powder packed packing machine as the research object,studies the health index construction method and the remaining life prediction method of the rolling bearings,develops the rolling bearing health status evaluation and remaining life prediction system,realizes the remote real-time monitoring of the rolling bearing status of the packing machine,and provides an effective reference for enterprises to develop maintenance plans.It is of great value and significance to ensure the stable operation of packaging machine and improve the production efficiency and benefit of enterprises.The main research content and results of this thesis are as follows:(1)Vibration signal acquisition and pretreatment of rolling bearing of packaging machine.The operating characteristics,motion characteristics and laws of the transmission system of packaging machine were analyzed,and the changing laws of the cycle working cycle and motor running speed of packaging machine were obtained.In view of the noise and impact interference of the vibration signal of the rolling bearing in the transmission system of packaging machine,the relatively stable movement of the rolling bearing was extracted after the impact signal.Combined with the operation rules of the servo motor,the software interpolation isometric resampling method was adopted to study the Angle domain reconstruction method of the extracted vibration signal data,and the unsteady signals in the time domain were converted into steady-state signals in the Angle domain.Based on the reconstructed steady-state data,the maintenance cycle data set of the rolling bearing of the packaging machine was constructed for the subsequent research.(2)Study and establish health index construction and health state evaluation method of rolling bearings based on convolutional neural network(Convolutional Neural Networks,CNN)and selfattention(Transformer)model.In view of the characteristics of large volume of vibration signal data and serious multi-noise interference of rolling bearing of packaging machine,convolutional neural network was firstly used to further de-noise vibration signals,extract effective data and screen key health information.Then Transformer network is used to process vibration timing signals.Based on vibration signal data with health labels,Transformer network is established through learning and training.By learning and processing vibration signals,the health information in the data is fully mined to represent the health state,and the mapping between the rolling bearing vibration signal data and health indicators is established.After that,the health stage,degradation stage and failure stage of rolling bearings were divided according to the health indicators to complete the health status assessment of rolling bearings.Finally,IEEE PHM 2012 data set was used to test the construction of health indicators and the effect of health status evaluation.The results showed that the proposed method could effectively extract the health indicators of rolling bearings,and the health indicators were better than cyclic neural networks and convolutional neural networks in monotonicity,correlation and robustness.Based on the constructed health indicators,the health status can be evaluated.(3)The remaining life prediction method of rolling bearings was proposed based on Sparrow Search optimization(Sparrow Search Algorithm,SSA)and BI-LSTM(Bi-LSTM).Aiming at the problem that the residual life prediction results based on deep learning greatly depend on the parameters of the prediction model due to the long acquisition span of vibration signals of rolling bearings of packaging machines,the sparrow search optimization algorithm was used to find out the best parameters of the residual life prediction model based on two-way long and short term memory network,so as to improve the accuracy of the model prediction results.Through IEEE PHM 2012 data set,the residual life prediction effect test experiment,compared with the residual life prediction model based on long and short term memory network and the residual life prediction model based on bidirectional long and short term memory network,the prediction accuracy of this method is higher,the error is smaller,and the effect is better.(4)Design and application test of the rolling bearing health status evaluation and residual life prediction system of packaging machine.Using Labview,Python,SQL and Matlab programming languages,based on Html and Web API interface technology,the health state evaluation and remaining life prediction system of rolling bearing of packaging machine was designed and developed,and the system was successfully deployed on packaging machine.Through the actual data acquisition,health status assessment and residual life prediction application test,real-time vibration data of rolling bearings can be obtained in real time,and the current health status can be analyzed in time,and the estimated residual life can be calculated.The test results show that the constructed health index of packing machine rolling bearings can well fit the health decline trend,and the remaining life prediction is in line with expectations,which verifies the effectiveness and stability of the system.The method and system of health status assessment and remaining life prediction of rolling bearing of packaging machine studied in this thesis are helpful to improve the production efficiency of enterprises and have certain application value when applied to the actual running of packaging machine.