Research On Feedforward Compensation Method For Precision Linear Motor Force Ripple

Integrated circuit(IC)manufacturing is the foundation of the Internet and information age.Photolithography is the most complex and critical process in the IC manufacturing.The industrial lithographic scanners,which contains extremely high technical contents,is the core equipment of the photolithography process.Therefore,the industrial lithographic scanners is known as the jewel in the crown of the semiconductor manufacturing.With the background of the wafer stage,the X-direction linear motor is regarded as the research object of this subject.The suppression and compensation of the cogging force of the linear motor is accomplished by modeling,identification and feedforward compensation.Firstly,the cogging force is suppressed by designing feedback controllers.The current loop transfer function is established according to the current loop sweep result,and the PI controller is designed to ensure the high bandwidth of the current loop.On this basis,the sweep test of the linear motor position loop is carried out.The transfer function of the motor is obtained by least squares fitting and the position loop feedback controller is designed.Secondly,the model identification problem of cogging force is transformed into the parameter estimation problem of multi-sinusoidal signal by modeling analysis.In order to solve this problem,the robust adaptive parameter estimation method,particle swarm optimization(PSO)algorithm and competitive swarm optimizer(CSO)are separately applied to complete the identification of the cogging force.Consequently,the cogging force model identified by the CSO algorithm is the most accurate.Direct compensation of the cogging force is accomplished by converting the cogging force model into a position-triggered feedforward control.Thirdly,in order to overcome the problem that the direct compensation method of the cogging force,which requires the compensation table must match the resolution of scale to avoid the compensation effect gets deterioration.A model-free iterative learning control algorithm applied to the linear motor system is introduced.The algorithm replaces the inverse model of the system by the ratio of the input and output signals in the frequency domain.Meanwhile,this algorithm can avoid the error caused by the system modeling process and update the model once every iteration.In theory,one-step convergence can be achieved.Furthermore,the relationship between the convergence error and the signal-to-noise ratio of the model-free iterative learning control algorithm under the noise and interference of input randomness is proved by theoretical analysis.Finally,aiming at the poor applicability of the model-free iterative learning algorithm,an iterative learning controller engineering design method for linear motor,which contain the Ptype ILC and the inverse model ILC with similarly design flow to the classical feedback controller design,is proposed.The performance of each parameter in the algorithm is compared by simulation,and the designed feedforward controller is experimentally verified on the workpiece table X to the linear motor.The results show that the feedforward controller can effectively reduce the tracking error of the motor and finally pass iterations make the error meet the requirements of the indicator.