Using neural networks for control of multicomponent distillation



Journal Title

Journal ISSN

Volume Title


Texas Tech University


Distillation colunms comprise a large part of the chemical industry's processing infrastructure. Currently, distillation is the most widely used separation process in the industry. The behavior of these processes, however, can be highly nonideal and nonlinear. Additionally, the variables that contiol the columns can be highly coupled. Thus, contiol of distillation columns is often a difficult, yet important process contiol challenge.

Three-product distillation is a process where, in addition to the overhead and bottoms products, an additional vapor or liquid product is removed from the side of the column. This distillation technique is applied to process cmde stieams with multiple components. The side stieam draw is used to recover a product whose boiling point lies between those of other components in the feed. Normally, a vapor draw is removed below the feed stage, or a liquid draw is removed above it. This additional product provides an extia dimension of complexity over ordinary distillation in the control challenge by adding another manipulated and controlled variable.