Chatter Identification And Surface Topography Prediction In Milling Of Thin-walled Parts

Thin-walled parts are widely used in many fields,such as aviation,aerospace,energy and marine,to meet specific structural requirements and weight limitations.Thin-walled parts are prone to chatter during machining due to their low damping,poor stiffness and time-varying dynamic parameters,and vibration has a significant impact on the surface quality even in a stable machining condition.Chatter can seriously affect the machining efficiency,part accuracy and surface quality.Poor surface quality can affect the service performance of the parts,including wear resistance,sealing,fatigue strength,optical properties,corrosion resistance and mating properties.The digital twin can visualize and digitize objects,processes and phenomena related to manufacturing activities,contributing to the networking and intelligence of machine tools.To ensure the machining accuracy and surface quality of thin-walled parts,in this dissertation,the stability prediction method,online chatter identification method and 3D surface topography prediction method are studied.Based on this,a digital twin platform for thin-walled milling system is developed and integrated with open CNC system.The milling process is monitored and optimized,and the prediction of chatter and surface topography of aluminum thin-walled parts is realized.The dynamic model of a milling system is the key to predicting the stability of thin-walled parts milling,and the accuracy of the dynamics model is largely influenced by the accuracy of the modal parameters.In this dissertation,multi-sourcel least square complex exponential(LSCE)and multi-source autoregressive moving average(ARMA)methods are used to extract the modal parameters of thin-walled parts.Compared with the single-source method,the identification accuracy is improved and the modal omission is avoided.According to the characteristics of thin-walled part milling processing,the dynamic model and milling force model are established by considering multi-order vibration modal,multi-point contact,tool and workpiece flexibility.Based on this,a stability lobe diagram(SLD)with three dimensions is constructed.It provides a theoretical basis for optimizing the milling process,reasonably selecting the machining parameters and improving the machining efficiency.Due to the complexity of the machining process,it is difficult to completely avoid the occurrence of milling chatter even if the machining parameters are selected strictly according to the stable lobe diagram.A chatter identification model based on deep learning is proposed in this dissertation.For feature engineering,the time-domain,frequency-domain and time-frequency-domain features of the multi-channel vibration and force signals are extracted.The correlation among different features of different channels and chatter is evaluated using mutual information method.In terms of model construction,a temporal attention network is established,which contains self-attention mechanism,mutual information,channel attention mechanism and statistical attention pooling.The capture of feature correlations,feature weights and their variation with moment is achieved.In terms of model evaluation,to prevent the influence of unbalanced data on accuracy,the model is trained using signals with different machining conditions from the test set.The accuracy,precision,recall,F1 value and confusion matrix of the model in the test set are calculated.An accurate evaluation of the classification performance and generalization performance of the chatter discrimination model is achieved,and chatter can be accurately identified.The formation of the surface topography during the milling process is related to a variety of geometrical and physical factors,and the mechanism is complex.In this dissertation,according to the characteristics of milling of thin-walled parts,to reflect the material removal effect,the time-varying dynamic parameters of thin-walled parts are extracted using a structural dynamics modification method.According to the milling system dynamics model,the relative vibration displacement under different cutting areas is calculated.Based on the dynamic model of the milling system,a kinematic model of the cutting edge considering the material removal effect and vibration is established.The 3D surface topography of the thin-walled part is constructed based on the Z-map method.The surfaces of the machined thin-walled parts are inspected using white light interferometry.The accuracy of the simulation results is verified,and the effects of material removal effect and tool vibration on the surface morphology are analyzed.Based on the above research,this dissertation constructs a digital twin model of a thin-walled part milling system based on the milling chatter and surface topography prediction model.A digital twin platform for the milling process is developed.To realize bidirectional communication among different modules and between physical entities and virtual models,based on the idea of hardware abstraction layera,a data acquisition module and a digital twin module are built and integrated into the open CNC system.In the digital twin platform,the milling chatter and surface topography prediction algorithm is loaded.Based on fuzzy control theory,the spindle speed and feed rate are changed.Monitoring of machining state and surface topography as well as milling chatter suppression is achieved.The initial machining parameters are selected based on the 3D stability lobe diagram,and the effectiveness of the digital twin monitoring platform is verified by machining an aluminum curved thin-walled part.