Approach to a reliable solution strategy for performing phase equilibrium calculations using MINLP optimization

摘要

The objective of this contribution is to propose a reliable strategy to solver the problem of phase equilibrium calculations for non-ideal systems, usingthe Gibbs free energy minimization. This type of problem, using the Gibbs free energy minimization, is usually formulated as a Mixed Integer Non-LinearProgramming (MINLP) Optimization. This optimization problem allows the compositions to be associated with continuous variables, and the presence of phases inthe equilibrium to be associated with the integer variables. The solution strategy proposes a bi-level approach. The first level combines a stochastic (SimulatedAnnealing - SA) and a local deterministic algorithm (Sequential Quadratic Programming - SQP), and solves a Non Linear Programming Problem (NLP). Thecontinuous variables are considered at this level. The second level considers the integer variables. The advantage of this bi-level strategy lies in its easyimplementation and in its proven efficiency to locate global optima with acceptable computational load. This article includes the study of the Water-Ethanol-Cyclohexane and Water-Ethanol-Glycerin systems. A comparative analysis was conducted using experimental data reported in published works and theoreticalcalculations by means of the Gamma-Phi classic method.