Frequency Domain Identification And Position/Force Motion Control System Design Based On Voice Coil Motor

With the development of the chip industry,the size and package size of the chip itself are becoming smaller and smaller,which makes the gap between pins and the diameter of pad decrease,and puts forward higher and higher requirements for semiconductor packaging equipment.The semiconductor packaging equipment is a highly integrated industrial automation equipment.The excellent motion control system is the core to achieve high precision,high acceleration,ultra-fine,high reliability and other performance of semiconductor packaging equipment.The voice coil motor is widely used in semiconductor packaging equipment because of its simple structure,small size,high frequency response,high acceleration,smooth force characteristics and high reliability.This paper focuses on the motion control system design of voice coil motor in semiconductor packaging equipment.Accurate and reliable system mathematical model is the basis for the design of motor motion control system.In this project,two frequency domain identification methods are used to identify the voice coil motor system.Mechatronic system identification belongs to "grey box" identification.It is necessary to understand the general structure of the system.For this reason,the mechanism models of two commonly used direct and swing voice coil motors in semiconductor packaging equipment are established.The entire voice coil motor motion control system is designed to be deployed around position control,force control,and position/force switching control.Firstly,starting with the structure and working principle of voice coil motor,two kinds of mechanism models of direct and swing voice coil motor are obtained.In terms of driving strategy,three driving schemes are analyzed.Finally,single-phase current driver is chosen as the driving scheme of voice coil motor.In the choice of the controller,an embedded motion controller scheme is adopted,which can obtain the DSP resources of the motion control card,and is convenient to freely develop the planning algorithm and the servo control algorithm.The excitation signal is very important for system identification,which needs to satisfy the identifiability condition of the system.Pseudo-random binary sequence(PRBS)and sine sweep signal satisfy this condition.In this paper,the generation mechanism and properties of these two signals are analyzed respectively.Using FFTW computing engine,the time domain signals of the two swept input and output are transformed into frequency domain.Based on the theoretical mathematical model of voice coil motor,the frequency domain identification of the voice coil motor system is carried out by recursive least squares method,and the identification results of PRBS and sine sweep frequency signals are compared and analyzed.In terms of fitting accuracy,iteration times,mean square error and final prediction error,the identification result under sine sweep signal is better than that under PRBS signal,which is reflected in higher identification accuracy and faster calculation iteration speed.However,from the experimental data collection scale,the sine sweep signal is much larger than the PRBS signal,and the latter has an advantage in data acquisition efficiency.The position control of voice coil motor adopts two-degree-of-freedom control strategy of feedforward control and feedback control,and the feedback control adopts three-loop nested control structure of current loop,speed loop and position loop from inside to outside.The position loop feedback controller adopts the PID controller,and the speed loop feedback controller adopts the IP controller.Two control strategies are adopted to implement feedforward control,which is based on direct feedforward compensation control and feedforward control based on iterative learning.Experiments show that the effect of direct feedforward compensation strategy is better than the feedforward compensation strategy of iterative learning control.However,the controller design of the former is complicated,the calculation amount is huge,and the generality of the controller is not good.Although the control performance of the latter is not superior to the former.However,iterative learning control does not require complex calculations,and iterative convergence is reliable and versatile.Open-loop force control is adopted as the force control strategy of voice coil motor,and trapezoidal force planning is used as the planning algorithm of open-loop force control combined with the process characteristics of semiconductor packaging equipment.In the selection of the position-force switching condition,the switching conditions such as the position feature,the position error feature,and the velocity feature are compared.The speed feature are better than the former two,the fault tolerance is better,the response is more rapid,and the switching efficiency is higher.Finally,the speed feature was selected as the switching condition of the position-force control mode.The above strategy was comprehensively verified by single experiment and multiple cycle experiments,and the experimental results were in line with expectations.