Optimisation of pressure-swing distillation of a maximum-azeotropic mixture with heat integration

Separation of a maximum-boiling azeotropic mixture water(A)-ethylenediamine(B) by pressure-swing distillation is studied by simulation. Since the feed composition is between the azeotropic ones at the two pressures, it can be introduced either into the high- (HPC) or the low-pressure column (LPC), leading to two sequences: HPC-LPC and LPC-HPC. One of them will be superior in terms of energy demand or cost. The total annual cost (TAC) of both sequences without, with partial and full heat integration (HI) is minimised by a genetic algorithm (GA). Surrogate model-based optimisation (SMBO) is applied to reduce the computational intensity by calculating TAC with models fitted to simulation results. The environmental impact is evaluated by the CO 2 emissions. LPC-HPC is slightly more favourable than HPC-LPC whether HI is applied or not. By HI, TAC is reduced by ca. 30 %, while the CO 2 emission by more than 40 %. The TAC obtained by SMBO is lower than by GA in one case; the difference is always within 5 %. For the first time, surrogate models fitted without HI are used to optimise the heat-integrated process without additional simulations. The effect of increasing steam price was studied both without and with reoptimising the processes by SMBO. • Pressure-swing distillation with two different column sequences was studied. • Total annual cost (TAC) was minimised by two optimisation methods. • Surrogate model-based algorithm was faster but gave higher TAC than a genetic one. • Optimised heat integration led to a decrease in TAC by 31 % and 32 %. • The lowest CO 2 emission was reached by the fully heat-integrated LPC-HPC.