Online Trimming of Feed Forward Trajectory by System Identification and Model Predictive Control

Abstract

Order to improve performances, feed forward control is widely applied to process control, especially for plant with nonlinearity and transport delay. However, the pre-set feed forward trajectory cannot adapt to uncertainty in plant due to its run-to-run discrepancy and material deteriorations. It needs to be adjusted online according to current plant dynamics. In this paper, online trimming is implemented by adding the output of a MPC (model predictive control) controller to feed forward trajectory. The model used in MPC controller is acquired through system identification. This scheme is applied to CZ crystal growth control for verification. The results show that it can effectively compensate the inaccuracy of temperature trajectory in CZ control system. The trimmed temperature trajectory matches plant dynamics quickly due to the prediction functionality in the MPC controller.