Start-up and control of a distillation column using adaptive dynamic matrix control: An experimental study

Manual control is conventionally exercised during the start-up of a distillation column since most linear controllers fail to perform satisfactorily during this highly sensitive transient period. This paper discusses the application of an adaptive dynamic matrix control (DMC) scheme to start-up and control of a pilot scale distillation column. A closed-loop on-line identification technique was used to update the DMC controller model to accommodate process-model mismatch. Experimental results showed that performance comparable to that obtained with a non-linear model predictive control can be achieved with the present scheme, during both start-up and subsequent single composition control for servo as well as regulatory problems, The adaptive dynamic matrix control (ADMC) algorithm was also extended to multi-input, multi-output systems. For a 2 x 2 system, only two of the four partitioned matrices of the system dynamic matrix could be updated at a time. For the dual composition control of a distillation column, the reflux flow was found to have a more profound influence on the column performance as compared with the steam pressure to the reboiler. Hence, only the models corresponding to the former were updated. Experimentation on a pilot scale column showed that adaptive DMC performed better both for regulatory and servo problems, when compared with conventional DMC.