MINLP optimization of a heterogeneous azeotropic distillation process: Separation of ethanol and water with cyclohexane as an entrainer

• Rigorous optimization of a heterogeneous azeotropic distillation process.
• Modified GBD approach takes continuous and integer variables into account.
• Non-convexities are successfully tackled by decompose-and-discard strategy.
• Parameter studies indicate that approach is able to find global optimum.
• Method can be plugged easily via VBA into commercial flow sheet simulators.

Heterogeneous distillation processes are widely used in industry for the separation of azeotropic and close-boiling mixtures. This paper addresses the optimization of a heterogeneous distillation process for the separation of an azeotropic ethanol/water mixture using cyclohexane as an entrainer. Starting from a given process superstructure a MINLP problem is set up to consider continuous as well as discrete decision variables such as the feed locations and the number of stages of the distillation columns. A modified Generalized Benders Decomposition algorithm to account for non-convexities of the model equations solves the MINLP problem. The algorithm can be attached via Visual Basic for Applications (VBA) to any commercial process simulator with NLP and VBA capabilities. Various optimization runs show that the algorithm is easily applicable and returns solutions independent of the initial values.