Research On Fault Prediction Method Of Press Lubrication System Based On DT

As an important manufacturing equipment of metal stamping,presses occupy an irreplaceable position in industrial production.Lubrication system can reduce friction and wear,and plays an indispensable role in mechanical press.Its healthy operation directly affects the stability and reliability of presses.During the operation of presses,once the unplanned shutdown is caused by the failure of lubrication system,the product will be scrapped,the production will be stagnated.Even more,the major property losses and casualties will happen.Therefore,it is particularly important to reduce unplanned downtime and improve fault-free operation time of lubrication system by scientific and intelligent maintenance means.Predictive maintenance can effectively ensure the reliability and stability of the system.It is an effective means to improve the fault-free operation time and reduce unplanned downtime.At present,there are mainly four methods for predictive maintenance,which are predictive maintenance based on historical statistical probability,predictive maintenance driven by sensor data,predictive maintenance based on physical model and fused predictive maintenance.However,these methods all have limitations and defects.Based on the concept of Digital Twin,this paper studies the predictive maintenance of press lubrication system.The main work is as follows:Firstly,based on the multi-domain unified modeling language Simscape,the main functional modules which includes power module and auxiliary modules are modeled.The power module modeling includes three-phase asynchronous motor modeling and gear pump modeling.The auxiliary module modeling includes the modeling of sensor module and fault simulation module.Finally,the construction of the digital twin model of the mechanical press lubrication system is completed by connecting above each modules and other components.Secondly,based on the socket communication method,the status data in the Siemens PLC is collected during the normal operation of the lubrication system.With the help of the Simulink Design Optimization toolbox of MATLAB,the parameters that cannot be directly obtained in the digital twin model are parameterized.In this way,the correction of the digital twin model of the press lubrication system is realized,which provides a theoretical basis for further fault diagnosis model construction and algorithm research.Finally,the fault data required to construct the fault diagnosis model and location of data collection are analyzed.Enough fault scenarios are designed according to the possible fault conditions.The configuration of fault scenarios and fault data collection are realized by MATLAB scripts.At the same time,the rapid restart function is added to the script to inject faults and accelerate the generation efficiency of fault data.The time domain and frequency domain features are extracted from the simulated fault data.The classification algorithm with the highest fault classification accuracy is selected with the help of the Classification Learner toolbox.Finally,the influence of the fault characteristics on the accuracy of the constructed fault prediction model is studied,the redundancy of the fault features is reduced,and the fault prediction models are verified.