A superstructure based approach to chemical reactor network synthesis

Abstract Frequently the product mix in a reacting system is influenced by both heat and mixing effects in the reactor. This paper presents a nonlinear programming (NLP) formulation for optimally generating reactor networks that would produce the desired effects, given a kinetic mechanism and expressions for the reaction rate. Using a generalized configuration of recycle reactors and heat exchangers (superstructure) as a basis, the NLP is able to extract the optimal subnetwork of units that would maximize an arbitrary objective function defined over a given kinetic scheme. Thus mixing and heat effects that affect reactions occurring in the homogeneous phase are taken into account. The resulting model and adjoint equations of this formulation form a two-point boundary value problem that interfaces with an efficient optimization strategy. Decisions representing network structure, reactor type and the amount of heat addition are made through continuous parameters in the model. The superstructure allows for serial and parallel connections involving reactor units, and is therefore fairly general. Since the method is equation based and since there are numerous ODE solvers it can be applied to large kinetic mechanisms with almost any objective function. Literature examples are presented and solved in order to demonstrate the effectiveness of this approach.