DECOUPLER DESIGN AND CONTROL SYSTEM TUNING BY INA FOR DISTILLATION COMPOSITION CONTROL

• Published in: Chemical engineering communications 1985-01, Vol.35 (1-6), p.149-174
• Main Authors: WALLER, K.V., WIKMAN, K.E., GUSTAFSSON, S.E.
• Format: Article
• Language: English
• Subjects: Applied sciences; Chemical engineering; decoupling; Distillation; Distillation control; Exact sciences and technology; Inverse Nyquist array
• ISSN: 0098-6445; 1563-5201
• DOI: 10.1080/00986448508911224

Six distillation column models have been used in studies of the usefulness of the Inverse Nyquist Array (INA) method for design of dual composition control for distillation. Five of the column models are experimental, one describing an industrial column, the others four pilot plants. One column model is obtained by modeling from first principles. The control strategies investigated and compared are multiloop SISO, 2-way decoupling and l-way decoupling. The control variables are the standard ones, i.e. reflux flow and boilup. In most of the cases INA has been found to be a useful tool for design of the loops and for comparison of the different control approaches. INA has also been found to be useful for simultaneous tuning of the decouplers and the feedback controllers. For 2-way decoupling the criterion to minimize interaction at the critical frequency for each primary feedback loop has been found useful in most, but not all, cases. The scheme can be designed by pure gains in the decouplers and there is no reason to introduce dynamics into the decouplers-the systems studied are already sufficiently rich in dynamics. INA has been found to be a suitable vehicle for the choice between the two possible l-way decoupling schemes. With the design approach taken, l-way decoupling has been found to provide considerably better control quality than 2-way decoupling in two of the six systems, the differences in control quality being small in the other cases. The effect of model mismatching on the results, caused e.g. by process nonlinearities, is discussed and simple rules to decrease the parameter sensitivity for 2-way decoupling schemes are given. When model mismatching is considered, the advantage of l-way decoupling over 2-way decoupling seems to increase. One deficiency of the INA design as used in this paper is that it does not single out the design which is to be preferred on the grounds of robustness.