A RLS run-to-run control approach for semiconductor manufacturing process

Exponentially weighted moving average (EWMA) is a commonly used model-based algorithm for semiconductor manufacturing process Run-to-Run (R2R) control. However, it's very difficult to set up the mathematical modeling for the actual semiconductor manufacturing process. Therefore, data-based methods such as Recursive Least Squares(RLS) have received wide attention nowadays. This paper proposes a variable forgetting factor RLS R2R control approach for semiconductor manufacturing process with and without drift disturbance. The variable forgetting factor resolves the drift disturbance well, and this method is much superior to generic RLS algorithm in convergence speed and tracking effect. It has both a strong ability to track parameters, and a small convergence estimate error. Simulation results prove the feasibility and accuracy of the algorithm.