Thermodynamic modelling of hydrocarbon-chains and light-weight supercritical solvents

Lombard, James Edward (2015-03)

Thesis (MEng)--Stellenbosch University, 2015.

Thesis

ENGLISH ABSTRACT: Long-chain hydrocarbons are of value to numerous lucrative industries. Due to the low volatility and close melting and boiling points of these solutes, traditional fractionation methods lack the required selectivity for separation and cause thermal degradation of the product. This project investigates the feasibility of Supercritical Fluid Extraction (SFE) for processing these systems, with the primary objective of modelling the high-pressure vapour-liquid equilibrium (VLE) properties of hydrocarbon solutes with a light-weight solvent using a semiempirical equation of state (EOS). Pure component vapour pressures and saturated liquid volumes are also investigated. A thorough investigation into the phase behaviour of the n-alkanes, 1-alcohols, carboxylic acids and esters in light weight supercritical solvents CO2, ethane and propane revealed that the solute structure and temperature largely influence the solute solubility and process feasibility. Good selectivity amongst the various solutes was observed for all three solvents, but very high pressures were required for complete miscibility using CO2 (exceeding 30 MPa). The quadrapole moment of CO2 further leads to complexities in phase behaviour such as temperature and density inversions (CO2/alkanes and CO2/alcohols) and 3-phase regions within the operating range. Simple linear trends in pressure vs. carbon number and temperature were observed for all the considered series using ethane and propane and these solvents were thus selected for conducting the modelling for this study. A thorough review of semi-emperical EOS models from literature revealed that the simple cubic equations of state (CEOSs) provide a promising modelling approach for SFE applications due to their simplicity, flexibility and reliability. The simple Peng-Robinson (PR) and Soave-Redlich-Kwong (SRK) EOSs provide good correlation of vapour pressure (%AAD below 5 %) for all the series over a large carbon number range (up to nC20), provided a two parameter alpha function is used. A 3rd parameter in the volume dependence for Patel-Teja (PT) EOS provides considerable improvement over the PR and SRK EOSs for satureate liquid volume correlations of the non-polar solutes (alkanes and esters), but offers virtually no advantage for the more polar alcohols and acids. The CEOSs therefore suffer clear limitations in simultaneous representation of these saturation properties (vapour pressure and liquid molar volume) for the systems of interest. Good correlations of high pressure binary VLE data were obtained using CEOSs available in the Aspen Plus ® simulator (% AAD in P, T and X2 generally below 1 % and ranging from 4 to 12 % for Y2 for all series) provided that two binary interaction parameters (BIPs) are used in the model mixing rules, irrespective of the model used. Aspen Plus ® was further validated as a reliable thermodynamic tool by comparing model fits using the RK-ASPEN model with parameters obtained from the Aspen Plus ® data regression routine and computational methods used in self-developed MATLAB software. Very similar results were obtained for both computational methods, which encourages the use of Aspen Plus ® for process modelling in SFE applications. A statistical sensitivity analysis into the relative effect and interactions between 6 modelling factors in applying the CEOSs revealed that the mixing rules, temperature and solute structure had the largest effect on the correlation of the high pressure VLE, with the pure component limit having negligible effect once BIPs are fitted to data. A significant interaction was, however, observed between the pure component model and the solute structure and temperature, which suggest that accurate correlation of mixture VLE does not solely rely on appropriate mixing rule selection, but also the pure model. Binary interaction parameters (BIPs) in model mixing rules were found to become intercorrelated when more than one are used, greatly impeding the development of generalized correlations. BIPs were also found to be sensitive to the pure component limit (alpha function and pure constants used), the temperature, the combining rules used and possibly the fluid density. These factors should all be taken into account systematically for developing generalized correlations which therefore fell outside the scope of this study. Recommendations were, however, made on how the MATLAB software developed in this study can be used to both expand the size of the statistical analysis already conducted into relevant modelling factors and to develop new generalized correlations for BIPs and new mixing rules.

AFRIKAANSE OPSOMMING: Lang-ketting koolwaterstowwe is van waarde in talle winsgewende industriële toepassings. Vanweë die lae vlugbaarheiden ooreenstemmende kook- en smeltpunte van hierdie molekules, toon tradisionele fraktioneringsmetodes nie die nodige selektiwiteit vir ekstraksie nie en veroorsaak bonop termiese degradering van die produk. Hierdie projek ondersoek dus die lewensvatbaarheid van superkritiese ekstraksie vir die prosesering van hierdie sisteme, met primêre fokus op die modellering van die hoë-druk damp-vloeistof ewewig eienskappe van koolwaterstowwe opgelos in ‘n lae-massa oplosmiddel met gebruik van ‘n semi-empiriese toestandsvergelyking. Suiwer-komponent dampdrukke en versadigde vloeistof volumes word ook ondersoek. ‘n Deeglike ondersoek na die fasegedrag van die n-alkane, 1-alkohole, korboksiel-sure asook esters in lae-massa superkritiese oplosmidds CO2, etaan en propaan toon dat die struktuur van die opgeloste stof en die temperatuur ‘n groot invloed het op die oplosbaarheid en proses lewensvatbaarheid. Goeie selektiwiteit tussen die verskillende koolwaterstowwe was waargeneem vir al drie oplosmiddels, alhoewel baie hoë drukke nodig was vir totale vermenging van die fases in CO2 (hoër as 30 MPa). Die quadrupool moment van CO2 veroorsaak verder ongewenste kompleksiteite in fase gedrag soos temperatuuren digtheid inversies (CO2/alkane en CO2/alkohole) en 3-fase-gebiede in die bedryfs-kondisies. Eenvoudige lineêre tendense in druk tenoor die koolstofnommer van die opgeloste stof asook die temperatuur was waargeneem vir al die ondersoekte koolwaterstof reekse in etaan en propaan en hierdie oplosmiddels was dus gekies vir die modellering vir hierdie studie. 'n Deeglike oorsig van semi-empiriese toestandsvergelykings uit die literatuur het getoon dat die eenvoudige kubiese toestandsvergelykings ‘n belowende modelleringsbenadering bied vir superkritiese ekstraksie toepassings vanweë hul eenvoudigeid, buigsaamheid enbetroubaarheid. Die eenvoudige Peng-Robinson (PR) en Soave-Redlich-Kwong (SRK) toestandsvergelykings bied goeie korrelasie van suiwer dampdruk (foute laer as 5 %) vir alle koolwaterstowwe oor ‘n groot koolstofnommer gebied (tot by nC20), met die voorwaarde dat ‘n 2 parameter alpha funksie gebruik word. ‘n 3rde parameter in die volume afhanklikheid van die Patel-Teja (PT) toestandsvergelyking bied ‘n beduidende verbetering in die passing van die versadigde vloeistof volume vir die nie-polêre koolwaterstowwe (n-alkane en die esters), maar bied geen voordeel vir die meer polêre alkohole en karkoksiel sure nie. Die kubiese modelle toon dus duidelike beperkings vir die gelyktydige voorstelling van hierdie versadigingde eienskappe (dampdruk en vloeistof volume) vir die sisteme van belang. Goeie korrelasie van hoë druk binêre damp-vloeistof ewewig data was verkry deur gebruik van die kubiese toestandsvergelykings beskikbaar inAspen Plus ® (fout in P, T en X2 tipies laer as 1 % en van 4 tot 12 % vir Y2 vir alle sisteme), met die voorwaarde dat 2 binêre interaksie parameters gebuik word in die model mengreëls, onafhanklik van die model. Aspen Plus ® was verder bekraktig as ‘n betroubare termodinamiese hulpmiddel deur model passings te vergelyk met die RK-ASPEN model tussen gevalle waar parameters verkry is deur die beskikbare regressie metode in Aspen Plus ® en metodes gebruik in self-ontwikkelde MATLAB sagteware. Eenderse resultate was verkry vir beide berekeningsmetodes, wat die gebruik van Aspen Plus ® vir prosesmodellering in superkritiese ekstrasie toepassings aanmoedig. ‘n Satistiese sensitiwiteits analise op die relatiewe effek en interaksies tussen 6 modelleringsfaktore in die toepassing van die kubiese toestandsvergelykings het gevind dat die mengreëls, temperatuur en die stuktuur van die opgeloste stof die grootste effek op die korrelasie van hoë druk binêre damp-vloeistof ewewig het, met ‘n weglaatbare effek vandie suiwerkomponent limiet waargeneem sodra binêre interaksie parameters gepas is aan data. ‘n Beduidende interaksie was wel waargeneem tussen die suiwerkomponent model en die struktuur van die opgeloste stof asook die temperatuur, wat daarop dui dat akurate korrelasie van mengsel damp-vloeistof ewewig nie slegs afhanklink is van ‘n gepaste keuse van mengreëls nie, maar ook die suiwer-komponent model. Binêre interaksie parameters in die model mengreëls ondergaan inter-korrelasie wanneer meer as een interaksie parameter gebruik word, wat die ontwikkeling van algemeen toepaslike korrelasies grotendeels belemmer. Binêre interaksie parameters was ook bevind om sensitief te wees tot die suiwer component limiet (alpha funksie en suiwer konstantes wat gebruik is), die temperatuur, die kombineringsreëls en moontlik die vloeistof digtheid. Hierdie faktore moet dus almal sistematies in ag geneem word wanneer algemeen toepaslike korrelasies ontwikkel word, wat dus buite die omvang van die huidge studie val. Aanbevelings was wel gemaak vir hoe die MATLAB sagteware ontwikkel vir hierdie studie gebruik kan word om beide die betaande statistiese analise uit te brei, asook nuwe korrelasies vir binêre interaksies parameters en nuwe mengreëls te ontwikkel.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/96844
This item appears in the following collections: