In the paper, we propose a new nonlinear self-tuning run-to-run (R2R) controller for the semiconductor manufacturing process. Self-tuning R2R controllers evolved from linear to nonlinear through former studies. Although some nonlinear controllers have been introduced, they were not fully supported in the areas of nonlinear update and optimization. For that reason, the proposed controller has three characteristics. First, a simplified model which reduces complexity in estimating the state of the actual process is applied to describe the process. Second, the nonlinear recursive estimation algorithm using the extended Kalman filter process is applied to update the nonlinear model. Third, nonlinear optimization using a modified Levenberg-Marquardt algorithm is applied to calculate recipe or input vectors. The test results show that the proposed controller can bring quality characteristics closer to the target than other controllers.
- Constrained nonlinear optimization
- Nonlinear recursive least square estimation
- Nonlinear self-tuning R2R controller
ASJC Scopus subject areas
- Computer Science(all)
- Health(social science)
- Environmental Science(all)