No abstract

A controllability index, which quantifies the cost associated with dynamic controllability, is proposed. The index depends quantitatively on the process alternative and control system alternative for each process alternative, and for each of these, the process dynamics, process constraints, product variability, disturbance characteristics, and controller/control structures. Its properties are discussed and its dependence on each of these factors is illustrated for the simplest possible case of a single unit (i.e., a distillation column).

The importance of selecting controlled variables appropriately for an entire process has long been recognized. However, because of its combinatorial nature, finding an optimal set of controlled variables requires solving a large number of nonconvex optimization problems, a computationally intensive task for any realistic chemical process. In this paper, we propose a shortcut method to eliminate poor choices and to generate and rank attractive alternatives without solving optimization problems. The method is based on scaling of all of the candidate controlled variables so that they have similar effects on the steady-state profit. The procedure is illustrated on a simplified butane alkylation process.

It is well-known that the most economical design from a steady-state economics perspective does not necessarily need to be easily controllable and might, in some extreme cases, even be uncontrollable. It is important to evaluate and quantify the dynamic controllability of a process at the design stage. Mahajanam (Ind. Eng. Chem. Res. 1999, 38, 999-1006) introduced a simple controllability index (ν) to quantify dynamic controllability in terms of economics. The dynamic controllability index is defined as the smallest additional surge capacity required to meet all of the control objectives and constraints dynamically for all of the expected disturbances. The index depends quantitatively on the process design and the control system alternative for each design. It has been found useful to compare plantwide control alternatives quantitatively (AIChE J. 1999(AIChE J. , 45, 1255(AIChE J. -1265. The main difficulty in applying the above method at the design stage is the need for a rigorous dynamic model, which is rarely available at the conceptual design stage. To resolve this difficulty, we propose that the controllability index (ν) be estimated with a steady-state model of the process. The proposed approach is based on several simplifying assumptions and is illustrated with a simple reactor-separator-recycle system.

No abstract