The separation of detergent range alkanes and alcohol isomers with supercritical carbon dioxide

[摘要] ENGLISH ABSTRACT: Data on the process performance at different operating conditions are required to determine thefeasibility of a separation process. Such data can be experimentally measured, but due to the time andcosts associated with pilot plant scale experiments, the use of predictive process models are oftenpreferred. The main aim of this project is to establish a working process model in Aspen Plus® thatcan be used to predict the separation performance of a supercritical fluid fractionation process aimedat the separation of mixtures of detergent range alkanes and alcohol isomers where similar boilingpoints or low relative volatilities can occur.Currently, an azeotropic distillation process is employed for the separation of detergent rangealkanes and alcohols. Although this process shows good separation performance, some concernsregarding the operating conditions are raised: the preferred entrainer, diethylene glycol, is toxic tohumans; very low operating pressures of 0.016 – 0.031 MPa and high temperatures of 473 K arerequired; additional processing units and materials are required to remove the entrainer from theproduct streams. An alternative process, supercritical fluid fractionation, is proposed in this workafter previous studies have reported that this process have potential for the separation of alkanes andalcohols. The supercritical fluid fractionation process addresses the concerns of the azeotropicdistillation process in the following ways: a non-toxic solvent, CO2, is used as the separating agent;mild temperatures of 344 K is proposed, but at the cost of the low operating pressures of theazeotropic process; and a single process unit and no additional material is required to separate thesolvent from the product streams.A process model was developed in Aspen Plus® to evaluate the separation performance of thenewly proposed supercritical fluid fractionation process and compare it to the current azeotropicdistillation process. The development of the process model included the development of an accuratethermodynamic model in Aspen Plus®. After thorough evaluation of a number of cubic equations ofstate, the RK-ASPEN model was found to be superior in its representation and prediction of phasetransition pressures for multi-component mixtures of detergent range alkanes and alcohols in thetemperature range 318 – 348 K. Phase transition pressures could be predicted with an error of lessthan 6 % with the inclusion of regressed polar parameters and binary solute-solvent interactionparameters for two multi-component mixtures: CO2 + (20 % n-dodecane + 70 % 1-decanol + 10 % 3,7-dimethyl-1-octanol) and CO2 + (25 % n-decane + 25 % 1-decanol + 25 % 3,7-dimethyl-1-octanol+ 25 % 2,6-dimethyl-2-octanol).Polar parameters were regressed from pure component vapour pressure data predicted withcorrelations available in Aspen Plus®. Binary interaction parameters were regressed fromexperimental bubble and dew point data. Binary bubble and dew point data were measured for anumber of systems containing ethane or CO2 and a C10-alkane or C10-alcohol isomer at temperaturesbetween 308 K and 353 K, and compositions ranging between 0.01 and 0.7 mass fraction solute. Acomparison between the phase equilibrium data measured for these systems revealed that the structureof the molecule, and not only the molecular weight, influences its solubility in the supercriticalsolvent. The phase transition pressures of n-decane, 2-methylnonane, 3-methylnonane and4-methylnonane did not differ significantly in CO2 or ethane, and these compounds will in alllikelihood not be separated in a supercritical fluid fractionation process. The phase transitionpressures measured for the C10-alcohol isomers decreased in both CO2 and ethane in the followingorder: 1-decanol, 3,7-dimethyl-1-octanol, 2-decanol, 2,6-dimethyl-2-octanol and3,7-dimethyl-3-octanol. The position of the hydroxyl group and the number, length and position ofthe side branches, all influence the solubility behaviour and phase transition pressures of the isomericalcohols in the supercritical solvent. Since the use of ethane did not show any significant benefitswith regard to selectivity, the use of the less harmful and less expensive solvent, CO2, in furtherinvestigations was justified.The RK-ASPEN thermodynamic model, with the inclusion of the regressed polar and binarysolute-solvent interaction parameters, was implemented in the process model and the separationperformance of the process was simulated at different operating conditions for the CO2 +(25 % n-decane + 25 % 1-decanol + 25 % 3,7-dimethyl-1-octanol + 25 % 2,6-dimethyl-2-octanol)mixture. A comparison to experimental pilot plant data revealed that the model cannot be used topredict the separation performance at low fractionation temperatures (316 K) due to shortcomings inthe thermodynamic model. However, the performance of the process at high fractionationtemperatures (344 K) could be predicted well, with an error of 10 – 36 %. Simulations for the CO2 +(25 % n-decane + 25 % 1-decanol + 25 % 3,7-dimethyl-1-octanol + 25 % 2,6-dimethyl-2-octanol) andCO2 + (20 % n-dodecane + 70 % 1-decanol + 10 % 3,7-dimethyl-1-octanol) mixtures showed that thecomposition of the feed mixture have a significant effect on the location and size of the operatingwindow and optimum operating conditions. The optimum operating conditions were defined as theconditions where an acceptable selectivity ratio and alcohol recovery occurred simultaneously. Since the selectivity ratio and alcohol recovery have opposing optimization approaches, a number ofpossible optimum operating conditions exist, based on the product specifications. When an alcoholand an alkane with similar phase behaviour exist in a mixture, a distinct minimum selectivity ratiowill occur at a point within the extract-to-feed ratio limits of the process. When the alkanes andalcohols present in a mixture do not have similar or overlapping phase transition pressures, theminimum selectivity ratio will typically cover a small range of extract-to-feed ratios at the high endlimit of the extract-to-feed ratio range.To summarize: A process model was established in Aspen Plus® that can be used to determinethe feasibility and separation performance of a supercritical fractionation process for a feed mixture ofdetergent range alkane and alcohol isomers. The model was used to prove that an SFF process is afeasible alternative process to consider for the removal of alkanes from mixtures of detergent rangealcohol isomers, even where overlapping boiling points or low relative volatilities occur. During thedevelopment of the process model, the following significant novel contributions were made:· New phase equilibrium data were measured for C10-alkane and C10-alcohol isomers insupercritical ethane, as published in The Journal of Supercritical Fluids 58 (2011) 330 –342.· New phase equilibrium data were measured for C10-alkane and C10-alcohol isomers insupercritical CO2, as published in The Journal of Supercritical Fluids 59 (2011) 14 – 26.· A thermodynamic model was developed in Aspen Plus® that can accurately predict thephase transition pressures of binary, ternary and multi-component mixtures of detergentrange alkanes and alcohols in supercritical CO2, as published in The Journal ofSupercritical Fluids 84 (2013) 132 – 145.· A process model was developed in Aspen Plus® that can be used to predict the separationperformance of a supercritical fluid fractionation process for the separation of mixtures ofdetergent range alkanes and alcohols.· Experimental and simulated results indicated that a supercritical fluid fractionation processcan be implemented successfully to separate an alkane from a mixture of alcohol isomers,as was shown for two mixtures: CO2 + (25 % n-decane + 25 % 1-decanol + 25 %3,7-dimethyl-1-octanol + 25 % 2,6-dimethyl-2-octanol) and CO2 + (20 % n-dodecane +70 % 1-decanol + 10 % 3,7-dimethyl-1-octanol).

[发布机构] Stellenbosch University