Modeling And Characteristic Analysis Of Machine Tool Feed System Driven By Mechanism-Data Mixture

The manufacturing industry’s "workhorses" are CNC machine tools,not only essential for intelligent modernization of the equipment manufacturing sector,but also a gauge of a nation’s industry’s progress.To enhance machining accuracy of CNC machine tools and reduce their development and production preparation times,amulti-domain digital models and virtual debugging are effective means of implementation.This thesis aims to improve the machining precision of CNC machine tools.Firstly,dynamic measurement technique is utilized,and in conjunction with the error similarity theory,the positioning error model of the CNC machine tool feed system is constructed and compensated,significantly reducing the positioning error of the feed system.Secondly,the feed system itself needs improvement.By using a hybrid mechanism and data-driven approach,we can construct a model of CNC machine tools with great accuracy.This,in turn,enhances the predictive accuracy of the feed system model.The specific research contents aimed at prediction accuracy are as follows:Firstly,the error data measured dynamically at different speeds are combined with error similarity theory to derive the error data under static conditions as the positioning error of the single-axis feed system by linear fitting,and the data is used to construct the positioning error model and to compensate for the positioning error verification.The positioning error model derived from error similarity has enhanced the accuracy of the CNC machine tool feed system’s positioning error measurement and modeling,when compared to traditional methods.Subsequently,a mathematical representation of each connection of the CNC machine tools’ feed system is executed,and the AC servo system and mechanical drive system are then modeled mathematically are built respectively.The key parameters in the feed system are identified by the traditional parameter identification and CARLA reinforcement learning parameter identification methods,and the mechanism model of the feed system is constructed.Through testing various curves and position commands and comparing the actual operating position of the feed system with the simulated position,we have determined that the proposed model’s prediction error achieves millimeter-level accuracy under different test curves.This indicates that the model is capable of effectively simulating the actual dynamic response of the machine tool feed system.To enhance the accuracy of the mechanism model’s mismeasurement,we suggest a data model that is based on CNN-LSTM.This model enables us to predict the mechanism model’s residuals with greater precision,and fuse the mechanism model with the data model to build a hybrid mechanism-data driven feed system model,which can not only reflect the influence of physical.This model takes into account the impact of physical parameters on the feeding system and significantly enhances the accuracy of the model by predicting the residuals through the data model.This thesis provides technical assurance for compensating for positioning errors in CNC machine tools and modeling the machine tool feed system.Through this,it promotes the improvement of CNC machine tool machining precision and the accurate construction of machine tool feed system models.The establishment of the positioning error model and machine tool feed system model is a valuable contribution to the field.