Performance study of GPC for processes with time delays

The performance of the generalized predictive control (GPC) algorithm without parameter estimation is studied to investigate its robustness property. A thermal process which can be represented by a general first-order plus dead-time model is selected as the test plant. GPC is found to be robust to changes in the time constant, less robust to process gain, but rather sensitive to the dead time. It is suggested that the minimum expected time constant and the maximum process gain, together with dead time which is at midrange, should be adopted as the nominal values for the design of the controller. When compared to PI (proportional plus integral) control GPC exhibits faster responses, however GPC is less robust to parameter variations.<
>