| In aviation,aerospace,power generation and other fields,all kinds of thinwalled frame parts are often used.These thin-walled parts have complex structure and are difficult to process,which can easily induce chatter in the process of machining.At present,how to suppress the occurrence of chatter is an internationally recognized problem,and chatter has become one of the main factors restricting the high-precision and efficient machining of thin-walled parts.Under this background,the milling stability prediction and online chatter monitoring and suppression technology are deeply studied in this paper.In the machining process of thin-walled parts,there are the characteristics of large workpiece deformation caused by milling force and obvious changes of dynamic parameters along the tool axis.To solve these problems,a multipoint contact dynamic model considering force-induced deformation is established in this paper.Firstly,the milling force is modeled,and the influence of milling cutter helix angle is considered.Secondly,an iterative extraction method of deformation is proposed.With the help of workbench software,the workpiece deformation caused by milling force at different axial cutting depths is extracted and fitted.Then,the modal analysis of the workpiece is carried out,the modal vibration mode of the workpiece is extracted and fitted.Finally,the tool-workpiece contact region is discretized along the tool axis,the vibration modes at each node and the workpiece deformation caused by milling force are calculated,and a multi-point contact dynamic model considering force-induced deformation is established.Finally,taking titanium alloy(Ti6Al4V)as the research object,the experimental comparison is made with the multi-point contact dynamic model without considering force-induced deformation,which proves the accuracy of the model proposed in this paper.On this basis,the effects of different tool wear states on milling stability are studied,and a threedimensional stability lobe diagram considering tool wear is established and verified by experiments.Aiming at the difficulty of chatter monitoring in milling process,a multiscale permutation entropy chatter feature extraction method is proposed based on adaptive signal decomposition technology.Firstly,the adaptive signal decomposition technology is studied.Aiming at the problem that the parameters of variational mode decomposition are difficult to choose,a method of automatic selection of decomposition parameters of variational mode decomposition is proposed,which takes the maximum crest factor of the envelope spectrum as fitness function and uses particle swarm optimization algorithm.Secondly,the multi-scale permutation entropy model of milling signals is studied,and its feasibility in milling chatter detection is analyzed.Then,the signals after adaptive decomposition were reconstructed based on the energy proportion,and the multi-scale permutation entropy of the reconstructed signals was extracted and applied to the monitoring of milling chatter.Finally,the multi-scale permutation entropy characteristics of signals processed by the optimized variational mode decomposition algorithm,the unoptimized variational mode decomposition algorithm and the empirical mode decomposition algorithm are compared and analyzed.Based on this,the optimal feature extraction method for online monitoring of chatter is selected.Aiming at the problem that the traditional chatter identification models is difficult to updated dynamically,resulting in poor accuracy and generalization performance,an online evolutionary chatter identification model based on multi-sensor fusion is proposed.Firstly,the performance of various sensors is analyzed,the best multi-sensor fusion scheme is selected,and different features are extracted according to the characteristics of each sensor.Secondly,aiming at the problem that the traditional K-Means model will expand infinitely with the accumulation of samples,and the clustering time will increase greatly,which can not meet the requirements of online clustering,an incremental sparse K-Means algorithm is proposed.Then,aiming at the problem that the online sequential extreme learning machine directly uses its predicted samples for learning without judging whether the prediction results are accurate,an online chatter evolution model combining unsupervised learning and supervised learning is proposed,and the model is designed.Finally,the multi-sensor fusion scheme and online evolution model proposed in this paper are experimentally verified and compared.In order to realize the online suppression of chatter,the application technology of chatter monitoring and suppression is studied.Firstly,the suppression method of cutting chatter is studied,focusing on the mechanism of discrete variable spindle speed to suppress chatter.Secondly,the software and hardware of the real-time acquisition platform are built,and the Windows+RTX technology is studied to solve the problem of weak real-time performance of Windows operating system.At the same time,the secondary development of FAGOR CNC system is carried out to open up the interaction between the external control module and the internal data of the machine tool.Then,combined with the chatter feature extraction method and chatter recognition method proposed above,the chatter intelligent monitoring and suppression module is designed,and the module is integrated into FAGOR open CNC system.Finally,on the basis of verifying the accuracy in side milling stability prediction of thin-walled frame parts,the initial cutting parameters of side milling chatter suppression experiment of thin-walled frame parts are selected through the stability lobe diagram,and the chatter in the machining process of no wear tools and wear tools is on-line suppressed by using the developed chatter intelligent monitoring and suppression module.The experimental results show that turning on the chatter intelligent monitoring and suppression module can significantly improve the workpiece surface quality.In the two groups of experiments,the surface roughness is reduced by 63.5% and 57.3%respectively. |
PDF Full Text Request