Optimal Process Synthesis in a Modular Simulator Environment: New Formulation of the Mixed-Integer Nonlinear Programming Problem

This paper deals with the formulation and the solution of a class of mixed-integer nonlinear programming (MINLP) problems, applied to optimal process synthesis, in a modular simulator environment. What is the first set out is the general solution strategy in a modular simulator environment. The proposed strategy leads to a reduced optimization problem and permits rigorous modeling of both logical and unit nodes of the superstructure. However, it involves implicit relations between the interest variables calculated by the simulator and both continuous decision and binary topological variables. In order to handle these implicit relations, a new formulation of the MINLP optimization problem is proposed. This formulation is based on the introduction of a new set of optimization variables and constraints : the pseudovariables and the pseudotorn streams. The proposed formulation leads to a reduced set of additional variables. The solution method is based on the outer approximation/equation relaxation (OA/ER) algorithm. The master problem (MILP) is constructed by partial application of the modeling/decomposition (M/D) strategy. Three examples illustrate the capabilities of the process synthesizer, one of which is the toluene hydrodealkylation (HDA) process.