Iterative Learning Control And Its Applications To Wafer Scanner Systems

This paper introduces iterative learning control in order to improve thedynamic tracking performance and productivity of wafer scanner systems, which arekey semiconductor manufacturing equipments. According to characteristics of waferscanner systems, robustness, convergent performance and tracking performance intime domain of iterative learning control are discussed to make the method appliedto the system. This thesis describes problems of the iterative learning controlmethod in actual practice, and analyzes robust convergence performance of themethod. Thus it presents an iterative learning control method based on μ-analysis,which combines μ-analysis with the optimal iterative learning control. This methodtakes advantage of the optimal weighting to adjust convergence performance of thesystem, and uses μ-analysis to ensure robustness of the method. Simulations areimplemented for validating the effectiveness of the proposed method.Then a segment iterative learning control strategy is presented according toexposure characteristics of wafer scanner systems. In order to improve the scanperformance, it enhances learning control action during the period from the dynamicstate to the steady state while attenuating learning control action during the periodfrom the steady state to the dynamic state. Segment iterative learning control isapplied to wafer scanner system, and simulations are implemented for validating theeffectiveness of the method. An engineering implementation method is presented forthe convenience of applications. This method employs experiments to achieve theimpulse response coefficients, which are used to calculate the learning coefficients.Then segment iterative learning control based on finite impulse response is formed.Finally experiments are implemented in wafer scanner test bench for validating theeffectiveness of the proposed method.This thesis also presents a design method of non-causal robust iterativelearning control law for the general uncertain systems to improve robustness andconvergent performance of the system. Robust convergence conditions are given fordifferent uncertain systems, and robust weightings are derived with regard touncertainties of the system. Thus robust iterative learning control for uncertainsystems is formed. Furthermore, conservatism and limitations of the robustconvergence conditions are analyzed by simulation results for two types of uncertainsystems. Then simulation results are used to validate advantages of the proposedmethod.Moreover, according to performance index of wafer scanner systems, aneffective non-causal iterative learning control method is proposed to improve learning effectiveness and adaptability of iterative learning control which is appliedto wafer scanner systems. It introduces time shift factors to compensate the phaselag of the closed loop system around bandwidth. The effect of time shift factors onconvergence of the system is analyzed, and the reference design method is given.Finite convergent domain of the learning control system is augmented by shapingconvergent characteristics of the learning control system. Low pass filters areemployed to make a second shaping for the system to ensure the robustness of thesystem while an effective learning function is utilized to improve the performance ofwafer scanner systems. Simulations are implemented for validating the effectivenessof the method. Then the method is applied to wafer scanner test bench, and thefeasibility of the method is validated by experiments. As an engineering designmethod, effective non-causal iterative learning control makes the designers get asatisfactory iterative learning control law by designing a finite amount ofparameters.Finally the thesis discusses macro-micro dual actuator control systems, whichare main modules of wafer scanner systems, and presents a dual loop iterativelearning control method to improve performance of the system. The force andelectromagnetism coupling effects between macro and micro motion stage andstroke limits of the micro motion stage are analyzed. Firstly master-slave and dualloop decoupling macro-micro control structure are presented according to themacro-micro motion characteristics of wafer scanner systems. Experiments areemployed to demonstrate tracking performance of the system. In order to improvetracking performance of micro motion stage and decrease the relative displacementof the macro-micro motion system, iterative learning control is applied to micromotion stage and macro motion stage respectively, which forms a dual loop iterativelearning control strategy based on dual loop decoupling control structure. Theconvergence of the method is analyzed, and experiments are implemented forvalidating the effectiveness of the proposed method.