The single-piece large-scale repetitive processing represented by 3C products has become an urgent need.In the current processing,the single-piece repetitive processing is still used.In order to make full use of historical processing data to guide the current processing to avoid repetitive interference and noise with the adverse effects,we introduce the idea of iterative learning to improve the machining accuracy.Based on the servo system of CNC machine tool,the model analysis and parameter identification experiment are carried out.The feedforward controller and feedback controller are analyzed and designed.The accuracy of motion control depends to a large extent on the design of the controller.In this paper,from the perspective of the control layer,aiming at the problem that the traditional feedback control is difficult to achieve high precision in the face of complex trajectories,the iterative learning control method is introduced.With iterative learning controller and the position loop feedback controller together to control,it can achieve fast and good convergence.After confirming the use of modified control-type open-loop iterative structure,the control effect of iterative learning control algorithm was tested on the virtual simulation platform.In order to reduce the influence of contingent error,the idea of ladder-type iterative learning is proposed.Each iteration is run repeatedly multiple times.The error of the softening process is extracted,and then added to the next iteration process.In order to reduce the amount of calculation,the iteration mechanism can be closed when the accuracy converges to the allowable range.The current iteration control amount is stored,and the constant value compensation is performed.Moreover,repetitive noise and repetitive interference during system operation can be suppressed by the iterative learning algorithm.However,it has been found in simulation and experiment that the system tracking error will be locally divergent after iterative learning of the controller.In order to analyze this problem,we start from the time domain and the frequency domain,and mainly analyze the velocity and acceleration features of the input trajectory in the time domain.The reason of the frequency domain analysis is about that high frequency noise is introduced.In response to the above problems,we introduce filters to effectively filter out high-frequency components to solve the above problems.Because of the phase lag caused by the on-line low pass filter,it is difficult to achieve the effect,and we finally choose the offline low pass filter,which motivates us to choose the open loop iterative learning structure.In order to further improve its convergence accuracy,for the effective high-frequency information in a particular section into the learning,based on the analysis of the input trajectory frequency characteristics,proposed a switched filter.Through SIMULINK simulation and d SPACE experimental verification,the effectiveness of the modified control-type iterative learning controller and the filter-based iterative learning control method is proved. |