Research On The Key Technology Of Machine Tool Predictive Maintenance Based On Digital Twin

CNC machine tools are the mother machine of industrial production and the core basic equipment of the manufacturing industry.As CNC machine tools face high-speed,high-precision,and intelligent development,their functions are becoming more and more powerful and complex.How to ensure the safe,reliable and stable operation of CNC machine tools to meet the high automation/intelligence requirements of unmanned factories/smart factories directly affect the success of intelligent manufacturing implementation.However,the current domestic CNC machine tools,especially in terms of reliability and stability,still have a large gap with the advanced level of foreign countries.Unplanned shutdowns caused by failures occur from time to time,which seriously affects their use in key industries such as automobiles,national defense and militaryPredictive maintenance can effectively ensure the reliability and stability of the system,and is an effective means to improve the trouble-free running time of CNC machine tools and reduce unplanned downtime.At present.predictive maintenance mainly has three methods:historical statistical probability method,sensor data-driven method and physical model based method.However,the above single method has limitations and defects,such as poor model fidelity,low effective utilization of data,and prediction algorithm problems such as poor accuracy.The concept of Digital Twin's real-time objective mapping of virtual reality and multi-dimensional and multi-level time/space integration of virtual reality provides ideas for solving the above problems.Based on the concept and method of Digital Twin,this paper conducts the following research on the key technologies of predictive maintenance of CNC machine tools(1)The architecture of predictive maintenance for CNC machine tools based on Digital Twin is studied.Based on system engineering ideas,the functions and key technical issues of predictive maintenance of CNC machine tools based on Digital Twin are analyzed.An architecture including model construction,scene perception,and intelligent predictive maintenance of the CNC machine tool Digital Twin is designed.Then based on the analytic hierarchy process,the predictive maintenance plan of CNC machine tools was formulated from the system level,and the effectiveness evaluation mechanism of the plan was formulated based on the fuzzy evaluation method.(2)The construction method of Digital Twin model of CNC machine tool is studied.The object-oriented incremental numerical control machine tool Digital Twin multi-domain unified modeling method is studied.The mechanical model,electrical model.control model and hydraulic model of the CNC machine tool Digital Twin is conctructed,and coupled in multi-domain.The accuracy verification method and update mechanism of the model are designed to realize the high fidelity and consistency of the Digital Twin model of the CNC machine tool.(3)The digital twin scene perception method of CNC machine tools is researched The intelligent scene-aware software and hardware structure of distributed CNC machine tools is designed based on Hadoop,HBase and Map-Reduce.On this basis,algorithms such as data acquisition and storage,data preprocessing,feature extraction,and feature selection are realized,thereby reducing the data dimension,reducing the amount of machine tool perception data,and solving the low data usage efficiency caused by the large amount of data.(4)The hybrid predictive maintenance method of CNC machine tools based on Digital Twin data and model is studied.Based on particle filter algorithm and migration learning,the hyrbid method of Digital Twin model and data is studied,which overcomes the problems of poor consistency of model methods and poor adaptability of data-driven methods in traditional predictive maintenance,and the difficulty of data acquisition in predictive maintenance experiments.This method achieves more accurate prediction and diagnosis results than a single predictive maintenance method,and improves the feasibility of predictive maintenance(5)The predictive maintenance application and verification of CNC machine tools based on Digital Twin is carried out.A model simulation platform and a distributed storage and analysis platform for machine tool contex-awareness data were built on the model/data server;a data-driven fault diagnosis and life prediction algorithm was built on the high-performance computing server.Finally,hybrid predictive maintenance methods of model and data were realized based on particle filter algorithm and migration learning,then were applied to the life prediction of milling tools of CNC machine tools,the fault diagnosis of spindle system and the fault diagnosis of feed system.This verifies the feasibility of the method proposed in this paper.Through the above research,this paper solves the key issues in the predictive maintenance of CNC machine tools based on Digital Twin,such as system-level design and evaluation,multi-domain model construction,machine tools context-awareness,and hybrid methods of model and data.This paper provides effective predictive maintenance solutions based on Digital Twin for CNC machine tools and other complex electromechanical equipment.