ABSTRACT
This research presents a run-to-run (R2R) multiple input-multiple output (MIMO) controller, termed Genetic Algorithm Quality Controller (GAQC), for semiconductor manufacturing processes. A recursive least-squares algorithm is used to estimate the parameters of the Hammerstein MIMO model. A real-valued genetic algorithm (RVGA) is applied to obtain the near optimal solution for the next run. The simulation is conducted for Chemical Mechanical Planarization (CMP) processes based on real models provided by SEMATECH. Experimental results show that GAQC outperforms Optimizing Adaptive Quality Controller (OAQC) and Multivariate Adaptive Controller (MAC) under the presence of noise and drift disturbances of different magnitudes; meanwhile, GAQC maintains adequate removal rate and nonuniformity under severely nonlinear input/output transfer functions.