Automatic Tuning of the Smith Predictor Controller

摘要

In this paper, we present a new automatic tuning technique for the Smith Predictor Controller which overcomes the disadvantages associated with current auto-tuners. The proposed new auto-tuner uses a single tuning experiment to estimate, with a high degree of accuracy, two points on the Nyquist curve of the open-loop process. Then, by employing an identification technique which uses these two points, a more accurate process model is obtained resulting in an auto-tuner which yields improved closed-loop responses. In addition, as only one tuning experiment is required, a shorter tuning time is achieved. In the paper, analysis is provided to show that the auto-tuned Smith Predictor Controller is robust to small modelling errors, and the paper also includes realtime experimental results in applying the auto-tuning technique for level control in a coupled-tanks system.