Inferential Self-Optimizing Control of a Reactor–Separator–Recycle Process

Abstract

Self-optimizing control of a reactor−separator−recycle process with A + B → C reaction chemistry is evaluated for mode I (minimize recycle rate or column boilup at given production) and mode II (maximize throughput with maximum recycle or boilup as the bottleneck). In both modes, one unconstrained degree of freedom (DOF) remains after accounting for the active constraints. Through an analysis of the overall plant material balance, it is shown that the reactor A to B ratio is a good self-optimizing variable for the unconstrained DOF. The separator distillation column top temperature is directly correlated to this ratio and is a good inferential self-optimizing controlled variable (SOCV) that avoids cumbersome reactor composition measurements. A plant-wide control system with column boilup as the throughput manipulator (TPM) is synthesized with two candidate CVs for stoichiometric feed balancing, namely, reactant composition in the reactor (CS1) and separator top temperature (CS2). Steady state and closed loop dynamic results demonstrate that both CS1 and CS2 provide effective process regulation in the face of principal disturbances with CS2 achieving significant economic benefit up to 7% and 0.8% in mode I and mode II, respectively. Holding the reactant ratio constant is also shown to be selfoptimizing for two variants of the A + B → C + D process and the realistic cumene process with recycle-to-extinction of the side product. The work emphasizes the substantial economic impact of the CV corresponding to an unconstrained DOF.