Effect of phase-equilibrium uncertainties on the process design of selected C2 and C3 alcohol separation systems: a Monte Carlo approach

[摘要] ENGLISH SUMMARY: Process design uncertainties can significantly influence the safety and reliability of separation processes. It is broadly accepted that the process engineer must not only specify the best available thermodynamic model to obtain reasonable results, but also quantify the effects of uncertainties in thermodynamic models and data on the final process design.Phase equilibrium correlations are reported as the most significant source of property uncertainties. Thus, how can one best account for the thermodynamic model parametric input uncertainty and the propagation of said uncertainty through the process simulation model? In this work, the effect of phase equilibrium uncertainties on the process design of the dehydration of C2 and C3 alcohols using extractive and azeotropic distillation was investigated. The extractive distillation of diisopropyl ether (DIPE) and isopropanol (IPA) with 2methoxyethanol as the solvent, as well as the heterogeneous azeotropic distillation of ethanol and water using DIPE as entrainer, were considered.Firstly, a systematic evaluation of thermodynamic models was performed. The objective was to identify the model that offered the closest prediction of experimental data of the underlying system (What is the best model?). The performance of the NRTL activity coefficient model in predicting the phase equilibria of the DIPE / IPA / 2-methoxyethanol extractive distillation system was of a high degree of accuracy. The prediction of the azeotrope temperature and composition were improved, although marginally, with the Hayden O'Connell and Nothnagel equations of state. However, this benefit was not extended to the binary vapour-liquid equilibrium (VLE) correlation ability of the model, thus the NRTL model was used. For the DIPE / ethanol / water azeotropic distillation system, the evaluation process revealed that the NRTL activity coefficient model offered largely excellent results, with a high degree of accuracy apparent in the azeotrope and phase envelope predictions. The inclusion of liquidliquid equilibrium (LLE) data provided a meaningful improvement of the model's ability to predict experimentally measured equilibrium data, confirming the usefulness of the NRTL model for this system. Secondly, a combined computer-based approach of stochastic models and process simulation was used to assess the effect of phase equilibrium uncertainties on the sizing of key process equipment. The Monte Carlo simulation technique generated a set of random input variables that represent the range of parametric uncertainty. For each system a process model was developed and the simulation solved for each unique set of input parameters with Aspen Plus® v8.8. The results were subsequently combined to develop cumulative distribution functions (CDF) for each design output of interest e.g. reboiler duty, heat exchanger surface area or column diameter and thus used to estimate the confidence level of the design.For the DIPE / IPA / 2-methoxyethanol system, it was observed that the extraction column uncertainty was predominantly in the bottom section of the column and was mainly related to the reboil rate. The design confidence could be improved to an acceptable level through a marginal increase of the reboil rate. The investigation further determined that the recovery column uncertainty was also limited to the bottom section of the column and only reboil ratio was of concern. The recovery column condenser and reboiler design confidence were 45% and 82%, but a small increase in duty restored the design confidence to the required levels. It was therefore concluded that the design of the extractive distillation process for the separation of the diisopropyl ether + isopropanol azeotrope with 2-methoxyethanol is acceptable and the identified risk areas can easily be resolved. For the DIPE / ethanol / water system, it was observed that the azeotropic distillation column geometry was not significantly impacted by the phase equilibria uncertainty, but that reboil ratio, bottoms flow rate and condenser surface area were. It was further noted that the dilute component uncertainties were high in the decanter, but did not appear to effect the overall performance of the decanter as the ratio of organic to aqueous liquid phase was high. Lastly, in the top section of the recovery column the key design output variables sensitive to phase equilibrium uncertainty were those related to condenser thermal requirements and that the effect of the phase equilibria uncertainty on the column geometry was negligible.In this work, it was thus shown that a systematic uncertainty quantification process based on a Monte Carlo approach reveals the effect of phase equilibrium uncertainty on the process design of C2 and C3 low molecular weight alcohol separation systems. The approach presented can be used to facilitate decision making in fields related to safety factor selection.

[发布机构] Stellenbosch University