Oxygen transfer in a model hydrocarbon bioprocess in a bubble column reactor

Cloete, Jannean Christelle (2015-03)

Thesis (MEng)--Stellenbosch University, 2015.

Thesis

ENGLISH ABSTRACT: The expansion of the global fuels industry has caused an increase in the quantity of hydrocarbons produced as a by-product of refinery gas-to-liquid processes. Conversion of hydrocarbons to higher value products is possible using bioprocesses, which are sustainable and environmentally benign. Due to the deficiency of oxygen in the alkane molecule, the supply of sufficient oxygen through aeration is a major obstacle for the optimization of hydrocarbon bioprocesses. While the oxygen solubility is increased in the presence of hydrocarbons, under certain process conditions, the enhanced solubility is outweighed by an increase in viscosity, causing a depression in overall volumetric oxygen transfer coefficient (KLa). The rate at which oxygen is transferred is defined in terms of a concentration driving force (oxygen solubility) and the overall volumetric oxygen transfer coefficient (KLa). The KLa term comprises an oxygen transfer coefficient (KL) and the gas-liquid interfacial area (a), which are dependent on the uid properties and system hydrodynamics. This behaviour is not well understood for hydrocarbon bioprocesses and in a bubble column reactor (BCR). To provide an understanding of oxygen transfer behaviour, a model hydrocarbon bioprocess was developed using a BCR with a porous sparger. To evaluate the interfacial area, the Sauter mean bubble diameter (D32) was measured using an image analysis algorithm and gas holdup (ϵG) was measured by the change in liquid height in the column. Together the D32 and ϵG were used in the calculation of interfacial area in the column. The KLa was evaluated with incorporation of the probe response lag, allowing more accurate representation of the KLa behaviour. The probe response lag was measured at all experimental conditions to ensure accuracy and reliability of data. The model hydrocarbon bioprocess employed C14-20 alkane-aqueous dispersions (2.5 - 20 vol% hydrocarbon) with suspended solids (0.5 - 6 g/l) at discrete super ficial gas velocity (uG) (1 - 3 cm/s). For systems with inert solids (corn our, dp = 13.36 m), the interfacial area and KLa were measured and the behaviour of KLa was described by separation of the in uences of interfacial area and oxygen transfer coefficient (KL). To further the understanding of oxygen transfer behaviour, non-viable yeast cells (dp = 5.059 m) were used as the dispersed solid phase and interfacial area behaviour was determined. This interfacial area behaviour was compared with the behaviour of systems with inert solids to understand the differences with change in solids type. In systems using inert solids, a linear relationship was found between G and uG. An empirical correlation fo rthe prediction of this behaviour showed an accuracy of 83.34% across the experimental range. The interfacial area showed a similar relationship with uG and the empirical correlation provided an accuracy of 78.8% for prediction across the experimental range. In inert solids dispersions, the KLa increased with uG as the result of an increase in interfacial area as well as increases in KL. An increase in solids loading indicated an initial increase in KLa, due to the in uence of liquid-film penetration on KL, followed by a decrease in KL at solids loading greater than 2.5 g/l, due to diffusion blocking effects. In systems with yeast dispersions, the presence of surfactant molecules in the media inhibited coalescence up to a yeast loading of about 3.5 g/l, and resulted in a decrease in D32. Above this yeast loading, the fine yeast particles increased the apparent viscosity of the dispersion sufficiently to overcome the in uence of surfactant and increase the D32. The behaviour of G in yeast dispersions was similar to that found with inert solids and demonstrated a linear increase with uG. However, in yeast dispersions, the interaction between alkane concentration and yeast loading caused a slight increase in dispersion viscosity and therefore G. An empirical correlation to predict G behaviour with increased uG was developed with an accuracy of 72.55% for the experimental range considered. Comparison of yeast and inert solids dispersions indicated a 37.5% lower G in yeast dispersions compared to inert solids as a result of the apparent viscosity introduced by finer solid particles. This G and D32 data resulted in a linear increase in interfacial area with uG with no significant in uence of alkane concentration and yeast loading. This interfacial area was on average 6.7% lower than interfacial area found in inert solid dispersions as a likely consequence of the apparent viscosity with finer particles. This study provides a fundamental understanding of the parameters which underpin oxygen transfer in a model hydrocarbon bioprocess BCR under discrete hydrodynamic conditions. This fundamental understanding provides a basis for further investigation of hydrocarbon bioprocesses and the prediction of KLa behaviour in these systems.

AFRIKAANSE OPSOMMING: Die uitbreiding van die internasionale brandstofbedryf het 'n toename veroorsaak in die hoeveelheid koolwaterstowwe geproduseer as 'n deur-produk van raffinadery gas-tot-vloeistof prosesse. Omskakeling van koolwaterstowwe na hoër waarde produkte is moontlik met behulp van bioprosesse, wat volhoubaar en omgewingsvriendelik is. As gevolg van die tekort aan suurstof in die alkaan molekule, is die verskaffing van voldoende suurstof deur deurlugting 'n groot uitdaging vir die optimalisering van koolwaterstof bioprosesse. Terwyl die suurstof oplosbaarheid verhoog in die teenwoordigheid van koolwaterstowwe, onder sekere proses voorwaardes is die verhoogde oplosbaarheid oortref deur 'n toename in viskositeit, wat 'n depressive veroorsaak in die algehele volumetriese suurstofoordragkoëffisiënt (KLa). Die suurstof oordrag tempo word gedefinieer in terme van 'n konsentrasie dryfkrag (suurstof oplosbaarheid) en KLa. Die KLa term behels 'n suurstofoordragkoëffisiënt (KL) en die gas-vloeistof oppervlakarea (a), wat afhanklik is van die vloeistof eienskappe en stelsel hidrodinamika. Hierdie gedrag is nie goed verstaan vir koolwaterstof bioprosesse nie, asook in kolom reaktors (BCR). Om 'n begrip van suurstof oordrag gedrag te voorsien, is 'n model koolwaterstof bioproses ontwikkel met 'n BCR met 'n poreuse besproeier. Om die oppervlakarea te evalueer, is die gemiddelde Sauter deursnit (D32) gemeet deur 'n foto-analise algoritme en gas vasvanging ( G) is gemeet deur die verandering in vloeibare hoogte in die kolom. Saam is die D32 en G gebruik in die berekening van die oppervlakarea in die kolom. Die KLa is geëvalueer met insluiting van die meter se reaksie sloering, om n meer akkurate voorstelling van die KLa gedrag te bereken. Die meter reaksie sloering was gemeet op alle eksperimentele toestande om die akkuraatheid en betroubaarheid van data te verseker. Die model koolwaterstof bioproses gebruik n-C14-20 alkaan-water dispersies (2.5 - 20 vol% koolwaterstof) solide partikels (0.5 - 6 g/l) op diskrete oppervlakkige gas snelhede (1 - 3 cm/s). Vir stelsels met inerte solides (koring meel, dp = 13.36 m), is die oppervlakarea en KLa gemeet en die gedrag van KLa beskryf deur skeiding van die invloede van oppervlakarea en KL. Om die begrip van suurstof oordrag se gedrag te bevorder, is nie-lewensvatbare gisselle (dp = 5.059 m) gebruik as die verspreide solide fase en oppervlakarea is bepaal. Hierdie oppervlakarea gedrag is vergelyk met die van stelsels met inerte solides om die verskille met verandering in solide tipes te verstaan. In stelsels met inerte solides, is 'n line^ere verwantskap gevind tussen G en uG. 'n Empiriese korrelasie vir die voorspelling van hierdie gedrag is opgestel met 'n akkuraatheid van 83.34% in die eksperimentele reeks. Die oppervlakarea het 'n soortgelyke verhouding met uG en die empiriese korrelasie verskaf 'n akkuraatheid van 78,8% vir die voorspelling van oppervlakarea oor die eksperimentele reeks. In inerte solide dispersies, het die KLa toegeneem met uG as die gevolg van 'n toename in grens oppervlak asook stygings in KL. 'n Toename in solides belading het n aanvanklike styging in KLa aangedui, as gevolg van die invloed van die vloeistof-film penetrasie op KL, gevolg deur 'n afname in KL op vastestowwe ladings groter as 2.5 g/l, te danke aan diffusie blokkeer effekte. In stelsels met gis dispersies, het die teenwoordigheid van benattings molekules in die media samesmelting geïnhibeer tot 'n gis lading van ongeveer 3.5 g/l, en het gelei tot 'n afname in D32. Bo hierdie gis lading, het die fyn gis partikels die skynbare viskositeit van die verspreiding verhoog genoegsaam om die invloed van benattings molekules te oorkom en die D32 te verhoog. Die gedrag van G in gis dispersies was soortgelyk aan die van inerte solides en dui op 'n lineêre toename met uG. Maar in gis dispersies, het die interaksie tussen alkaan konsentrasie en gis lading 'n effense toename veroorsaak in die verstrooiing viskositeit en dus in G. 'n Empiriese korrelasie is ontwikkel om G gedrag te voorspel en het 'n akkuraatheid van 72,55% vir die eksperimentele verskeidenheid beskou. Vergelyking van gis en inerte patrikel dispersies wys 'n 37.5% laer G in gis dispersies in vergelyking met inerte vaste stowwe as 'n gevolg van die skynbare viskositeit bekendgestel deur fyner vastestowwe partikels. Hierdie G en D32 data het gelei tot 'n linere toename in grens oppervlak met uG met geen beduidende invloed van alkaan konsentrasie en gis lading nie. Die oppervlakarea was gemiddeld 6.7% laer as oppervlakarea gevind in inerte partikel dispersies as 'n waarskynlike gevolg van die skynbare viskositeit met fyner partikels. Hierdie studie bied 'n fundamentele begrip van die veranderlikes wat die suurstof oordrag definieer in 'n model koolwaterstof bioproses BCR onder diskrete hidrodinamiese voorwaardes. Hierdie fundamentele begrip bied n basis vir verdere ondersoek van koolwaterstof bioprosesse en en die voorspelling van KLa gedrag in hierdie stelsels.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/96981
This item appears in the following collections: