卡曼濾波器與模型誤差自我調變法於批次控制之應用

Abstract

近年來，批次控制技術已被充分地發展並且應用於半導體製造過程中，以維持製程水準並改善產品的良率。在這些批次控制之中，EWMA控制器、double EWMA控制器和PCC控制器在線上批次估測中已是被常使用的方法。但是，製程輸出在選擇不適當的批次控制器參數下，反而會有負面的效果。特別是當系統存在模型誤差和環境上有擾動時，一般批次控制器可能無法有令人滿意的性能效果。 本文先討論已知製程干擾類型：DT、RWD、IMA(1,1)、ARMA(1,1)、ARIMA(1,1,1)，比較EWMA控制器、結合卡曼濾波器批次控制架構之控制效能。後針對未知製程干擾，本文提出一動態調變系統模型誤差的控制結構，在此控制結構裡主要有二種機制：系統模型誤差調變模組和使用卡曼濾波器估計干擾，將此控制架構命名為自我調變模型誤差卡曼濾波器，該控制架構具有排除製程干擾與減少系統模型誤差之能力，並使系統輸出達到所期望之效能。

In the last few years, Run-to-Run (RtR) control techniques have been developed and used to semiconductor manufacturing processes to maintain process targets and improve the yield of products. Among the RtR controllers, Exponentially Weighted Moving Average (EWMA), double-EWMA and Predicted Correct Control (PCC) are useful methods for online RtR estimation. However, incorrect selection of the RtR control parameters can have the opposite effect on the controlled process output. Conventional RtR controllers may fail in satisfying performance requirement especially when the system has model mismatch and the environmental perturbation. First, this thesis has discuss and analysis performance of EWMA controller and combine Kalman Filter in run-to-run control to deal with known disturbance : DT、RWD、IMA(1,1)、ARMA(1,1)、ARIMA(1,1,1). Then, a dynamic-tuning control structure having the capability of adjusting the system parameters dynamically that used to unknown disturbance is proposed in this thesis. There are two schemes in this control structure: a model mismatch self-tuning module and Kalman Filter used to estimate disturbance. The control structure, termed Self-Tuning model mismatch Kalman Filter controller (STKF), can reject process disturbance, reduce model mismatch and achieve expected performance.