Improved Bleach Plant Control Using Internal Model Control with Smith Predictor

Abstract

This paper presents the improved control of a benchmark pulping process. Open-loop tests were conducted to obtain a multiple-input multiple-output process model in the form of transfer function matrix. Using this model, the best set of input-output pairings was selected by using relative gain array (RGA) and relative disturbance gain techniques, both in static and dynamic modes. These analyses confirmed the setting provided by the authors of the Benchmark, based on static RGA analysis. Controller settings for each control loop were calculated using different internal model control (IMC) tuning methodologies and the best set of controller parameters was chosen by evaluating the control system performance for set-point tracking and disturbance rejection in terms of the integral of absolute error, settling time, time constant and percentage of overshoot. PI controllers combined with Smith-predictors, and tuned with IMC, providing the set-points of the Kappa factor controllers related to the quality variables of the process and Pi-only controllers tuned using IMC controlling the secondary variables give the best control performance. Smith predictors allow the controller designer to provide to the process controllers larger controller gains and smaller reset times, making the controlled response faster. Their ability to provide estimations of the process measurements when the real measurements are not available was especially useful in this process due to its large time delays.