High-precision Motion Positioning Error Compensation Of Dicing Saw Based On Iterative Learning Control

As the key equipment of electronic component packaging,the dicing saw currently adopts the form of mechanism modeling for most error compensation methods.However,there are many factors that cause errors in the actual operation process,so it is very difficult to measure and model the error source.The motion form of dicing saw is a repeated process of mass production,which produces a large amount of data in the process of operation.However,previous methods all ignored the repeated motion process of mass production,and did not make full use of these data,which resulted in a waste of data.With the development of global intelligence,data and information technology,data will play a more important role,so the effective use of data to improve the production efficiency of manufacturing industry has become a development trend.In the cutting process of the dicing saw,the error factors may cause that the accurate cutting point can not be reached,which will affect the positioning accuracy of the cutting movement.However,due to such factors as the difficulty in modeling error compensation and the incomplete consideration of error sources,it is difficult to establish a model that includes all error sources.Based on this,error compensation is carried out through input and output data by using data-driven idea.In this thesis,based on the existing data driven control algorithm,the iterative learning error compensation method of neural network is proposed by combining the iterative learning control algorithm and neural network in data driven.This method that using the iterative learning control algorithm is suitable for large quantities of repeated movement in the process of online running form,and using the neural network is suitable for the ability of dealing with complex nonlinear systems,a nonlinear mapping and adaptive and generalization,and other functions,can guarantee the dicing saw movement stability and robustness of the control system.This method does not need to model the error source mechanism,but only uses the displacement data generated in the process of motion to compensate,which solves the difficulties and limitations of the traditional modeling methods.This thesis applies neural network iterative learning error compensation method in dicing saw operation control system,which the error compensation controller was designed,by using this method will ever produce data information through the control law is applied to the current movement process,error compensation on current motion process,improve the positioning accuracy of the current row cutting operation.In this thesis,the proposed method to make use of the Matlab simulation analysis,the first type of PD iterative learning error compensation method for the simulation analysis,the effect is not ideal,and model-free adaptive iterative learning error compensation method for the simulation analysis,through the comparison shows the method can make up for PD type iterative learning fixed learning law of error compensation method and cause lack of anti-interference and the disadvantage of compensation effect is insufficient.However,the error compensation method model-free adaptive iterative learning cannot guarantee continuous convergence,and the repeated positioning accuracy cannot meet the requirements of materials with high cutting accuracy.Aiming at the above problems,neural network iterative learning error compensation method is proposed,which can ensure the convergence and further improve the accuracy of repeated positioning.Compared with the previous two methods,the simulation results show that the proposed method has better compensation effect.Finally,the above method is analysed on the dicing saw platform to verify the feasibility and effectiveness of the method presented in this thesis.The method presented in this thesis does not rely on the traditional mechanism model,but only uses the input and output data,and the calculation amount is simple and easy to analyze.It is a good attempt to apply the iterative learning control algorithm to the motion positioning error compensation of the dicing saw.