適應性干擾觀測器於半導體製程干擾之抑制

Abstract

常見的批次控制器，如EWMA和double EWMA等，在系統存有模型誤差時，常導致系統輸出會有震動，因此降低系統性能，且控制器參數的穩定區間也會隨著模型誤差的存在而受影響。除此之外，當我們根據某些性能指標去尋找一組最佳的批次控制器參數時，往往須先知道干擾的型態及系統模型誤差的大小，使得控制器最佳參數的選取在實際情況時難以達成。此篇論文的主要目的是提出一個架構，其結合干擾觀測器(DOB)及一組使用投影法修正過的參數調整律，並且取名為適應性干擾觀測器。而此架構的優點有，消除干擾、減少系統的模型誤差，並且使得系統輸出達到期望的性能。而論文中分別推導了在單輸入單輸出，及多輸入多輸出的情形。並且在論文最後，以化學機械研磨的製程的做驗證。透過模擬，其結果較傳統固定參數的干擾觀測器改善了約22%。

It is well known that the common RtR controllers like EWMA and double EWMA suffer the decrease in stability region for weights and the oscillatory or long transient performance when there is a model mismatch. Besides, the optimum weights of RtR control obtained from some kind of performance index are always function of the a priori knowledge of the disturbance and model mismatch, which makes the optimum weights difficult to choose in the realistic situation. The purpose of this paper aims to present a design of Run-to-Run (RtR) controller by using the Output Disturbance Observer (ODOB) structure and adaptive algorithm with projection, which is called Adaptive Disturbance Observer (ADOB). The method has the advantage of rejecting process disturbance, reducing model mismatch and thus achieving expected performance. The cases of SISO and MIMO are demonstrated. At last, a relevant application to Chemical Mechanical Polishing (CMP) process is presented to illustrate the proposed control structure. Through the simulation study, the results show that the performance is improved about 22% as compared to the fixed-parameter ODOB structure.