SELF-TUNING CONTROL STRATEGIES FOR DISTILLATION COLUMNS (ADOPTIVE, TIME DELAY SYSTEMS)

A multivariable self-tuning controller is developed and its performance evaluated via simulation studies of two distillation columns. This self-tuning controller is based on Koivo's approach, however, modified so as to apply to systems which contain multiple time delays and/or multi-rate sampling. Simulation results are used to study the effects of important design parameters on the controlled system.This new approach has also been extended to incorporate decoupling for multivariable control problems. The same algorithm is shown to result from two very different derivations. The first approach employs multiple single-input/single-output STC-TDC controllers but with a classical decoupling scheme incorporated. The second approach utilizes a novel selection of design parameters in a multivarible STC-TDC. Simulation studies utilizing two distillation column models show that the controller can provide excellent control even for highly-nonlinear and highly-interacting processes.Finally, the STC-TDC, both with and without decoupling, have been evaluated using a pilot-scale distillation column. Experimental results comparing these two methods with a well-tuned dual PI control system show that these two new methods perform much better than standard control for both set-point and load changes.In a separate study, single variable and multivariable self-tuning controllers with Smith-predictor type time delay compensation (STC-TDC) are developed to handle open-loop stable processes with unknown or varying time delays. Several numerical examples and a nonlinear simulated air heater model show that the new approach performs well for both set-point and load changes when process time delay changes occur.