Smith Predictor-based Controllers for Temperature Process with Time Delay

Abstract

Smith Predictor (SP) control structure is usually developed around a proportional-integral (PI) controller to improve the performance of processes with time delay. However, this technique has shortcomings as it depends on the exact representation of the process model, does not guarantee robustness and it is sensitive to process delay variation. A temperature control experiment is conducted on a shell-tube heat exchanger due to its time delay presence characteristic. A process model is obtained through a statistical modelling identification. Using Matlab/Simulink to simulate the control of the temperature process loop, a PI and Smith Predictor-PI (SP-PI) controllers are developed and compared to prove the effectiveness of the latter. Later both controllers are compared to a Filtered Smith Predictor (FSP) to offer benchmark comparison. Investigation is conducted on the ability of the SP-PI and FSP control structures to robustly control a temperature feedback loop when the process time delay is varied and results show that they are both efficient. Moreover, limits to reducing the effects caused by time delay are dependent on the choice of robust element value and sampling time choice for the SP-PI and FSP structures respectively. The control performance of both structures are compared using indicators such as Integral Absolute Error (IAE), Integral Squared Error (ISE), settling time, rise time and overshoot.