HYDRODYNAMIC MODELING OF A DIRECT CONTACT HEAT EXCHANGER USED FOR BIO OIL CONDENSATION

Transcription

1 Ninth International Conference on CFD in the Minerals and Process Industries CSIRO, Melbourne, Australia 10-1 December 01 HYDRODYNAMIC MODELING OF A DIRECT CONTACT HEAT EXCHANGER USED FOR BIO OIL CONDENSATION Kiran Kumar PVS 1,, K PAPADIKIS 1 and Sai GU * 1 1 Civil Engineering Deartment, Xi an Jiao tong-liverool University, Suzhou, Jiangsu Province, CHINA School of Engineering, University of Liverool, Liverool, UK *Corresonding author, address: SaiGu@xjtlueducn ABSTRACT Quenching columns are used for condensing yrolysis vaours to roduce liuid bio-oil using the direct contact heat exchange method The fluid flow inside this tye of euiment is uite comlex involving the interaction between multile hases In the current aer, the hydrodynamic behaviour of a uenching column utilised in ablative fast yrolysis exeriment is modelled Liuid octane is circulated inside the condenser as a cooling medium Pyrolysis gaseous comonents were modeled by non-condensable gas nitrogen This gas is introduced at the bottom of the column where it comes in direct contact with the cooling liuid The Eulerian - Eulerian model with immiscible otion is used to analyse the gas liuid interactions The CFD redictions validated the flooding henomena occurred in the uenching column during the exeriments Design modifications are also discussed together with the imlications of those on the final erformance NOMENCLATURE A c Curtain area m A w Window area m d Drolet/ bubble diameter m g Gravitational acceleration vector m/s h Secific enthaly of the hase J/kg Pressure Pa Re Reynolds number T Temerature K t time s V c Curtain velocity m/s V w Window velocity m/s Greek letters Q Volumetric flow rate m 3 /s a Volume fraction Density kg/m 3 Dynamic viscosity Pa s ν Velocity vector m/s σ Surface tension N/m Subscrits, hase index of gas and liuid resectively ν volume averaged INTRODUCTION Generating Bio Oil from biomass by using yrolysis method is gaining oularity worldwide as this is from renewable resources and in contrast to other renewables, that give heat and ower, Bio mass reresents only source for liuid solid and gaseous fuels Bio mass yrolysis rocess rovides liuid fuel that can be substitute for fuel oil in any static heating or electric generation alication (AV Bridgewater and GVC Peacock, 000) However, its main success deends on the yield of the bio oil er unit biomass This intern will give an idea on how economical the articular rocess is while roducing the bio oil Bio oil is roduced by raidly removing the heat from biomass yrolysis vaours in heat exchangers There are several tyes of heat exchangers available for the urose Simle indirect heat exchange can cause referential deosition of lignin-based comonents leading to liuid fractionation and eventual choking in ielines and heat exchanges (AV Bridgwater, 01) Quenching in roduct bio-oil or in an immiscible hydrocarbon solvent is widely ractised More often uench columns are used for the urose of condensing the yrolysis vaours to get bio oil Quench columns are direct contact heat exchangers that will condense bio oil from yrolysis vaour by removing heat from it raidly These direct contact heat exchangers rovide cost effective heat transfer and hence it is suitable for making the rocess more economical and viable on commercial scale The main idea behind this rocess is to get advantage of interfacial area between two working fluids in the absence of solid wall between them gives raid heat transfer Here it is worth noting that the raid cooling of yrolysis vaour is at most imortant as it dictates the ercentage yield er biomass Slower cooling rates encourage secondary reactions where the yrolytic vaour comounds can crack further into smaller molecular weight fragments and/or olymerize into larger fragments, both at the exense of fragments which makeu the desired liuid roduct resulting to lower yield of bio oil (Diebold, JP, 1980 and Bradbury et al, 1979) Designing uench columns using the standard emirical relations is difficult as it involves comlex flow aths of the fluid inside as well as comlicated heat transfer and hase change henomena The gas liuid interaction may generate different flow regimes, inturns determine how the fluids will behave Hence the gas liuid interface and the effectiveness of the uench column are strongly deendent on the fluid dynamics inside the euiment Comutational methods are widely used to get extensive Coyright 01 CSIRO Australia 1

2 understanding on the fluid behaviour inside this kind of euiment and it is necessary to design economically The basic underlying rincile on the flow regimes inside the uench column lates are very much similar to trays used in distillation columns Various CFD modelling studies have been undertaken to model the hydrodynamics and flow atterns in the sieve tray columns However, according to author s knowledge no modelling attemt has been resented in the literature in modelling hydrodynamics of a uench column This is robably due to very limited exerimental data available and the very comlex hysics involved as mentioned before In this aer an attemt has been done to get information on the hydrodynamics of the fluids inside the uench column using CFD methods Eulerian - Eulerian multihase model with immiscible fluid otion was used to model the gas liuid interactions The CFD redictions demonstrated the flooding henomena occurred in existing uench column design This was observed in the exeriment and was reorted earlier The exeriment conducted at Aston University was to roduce bio oil based on ablative yrolysis method (Peacocke GVC et al, 1995) The detailed flow ath of the rocess can be seen in figure 1 and exerimental setu shown in figure This euiment is designed to oerate at 5kg/hr throughut, but due to feeder limitation it was limited to 3kg/hr In these exeriments (Robinson, 000), a uench column and an electrostatic reciitator were emloyed to condense the yrolysis vaours to get the bio oil The uench column is sized based on the gas flows and flooding factors 1 l g Q Vw, max 0 58 g Aw Liuid octane at -5 o C was sulied as coolant It was reorted that, at the design flow rates the uench column started flooding This roblem was rectified by lowering the flow rates of the liuid and reducing the diameter of the discs Figure : Exerimental setu (adated from Robinson 000) In this work, the effect of multile chemical comounds resent in the yrolysis vaour while condensing has not been considered As 70% of the yrolysis roduct on weight basis has nitrogen which is sulied as urging as well as carrier gas, majority of the hydrodynamics effects inside the euiment were accounted The interhase heat transfer is based on the Ranz-Marshall correlation Some design iterations briefly discussed here and the imlication of these changes on final erformance also noted () Figure 1: Ablative yrolysis rocess [US atent 76553] The design of the liuid collection system is as shown in figure 3 This euiment is designed for the vaour volumetric flow of 00011m 3 /s at a temerature of C The Disc and Donut within the column has the secifications mentioned in table 1 The dimensions were chosen based on curtain and window velocities calculated near the bottom of the column where maximum flow rate was exected Correlations for these velocities are shown in euations 1 and The values 115 and 058 are the recommended Souders-Brown coefficients 1 l g Vc, max 1 15 g Q A c (1) Figure 3: Sketch of the exerimental uench column Coyright 01 CSIRO Australia

3 MODEL DESCRIPTION CFD modelling The commercial CFD ackage, ANSYS Fluent 130 was used to model the gas liuid motions in the baseline and modified uench columns As the coolant used in the uench column was immiscible fluid, the interface area between gas and liuid is imortant in evaluating the fluid dynamics within the uench column This makes the modelling more challenging There are several methods available to trace the interface between two fluids Among those the oular methods are volume of fluid (VOF) method and level set method Both aroaches are based on imlicit free surface reconstruction methods More recently, the immiscible fluid model for Eulerian multihase was introduced in Fluent and is based on Geo- Reconstruct with exlicit VOF otion The immiscible Eulerian model overcomes the limitation of shared velocities at the interface, resent at the VOF model In this way heat transfer modelling can be much more accurate, since a sli velocity between the different hases exists at the interface The heat transfer coefficient between hases was calculated based on the Nusselt number using the Ranz-Marshall correlation The euations are as follows: The continuity euation for the hase is a t a v 0 Note that in this roblem, two hases are resent, namely nitrogen (gas) and octane (liuid) Hence the conservation euations are solved for each hase individually Conservation of momentum yields the euation 4 mentioned below for hase a v a vv a t a g R F In euation 4, is the stress-strain tensor, R is the interaction force between two hases given by euation 5 R K( v v ) (5) Here K is the interhase momentum exchange coefficient, defined as a ( a a ) f K, (6) (3) (4) where f is the drag function, defined asc D Re 4, and C is based on Schiller Naumann drag model C D D (1 015Re ) / Re Re 1000, 044 Re 1000 (7) v v d Re (8) The term used in euation 6 is the articulate relaxation time and is defined as ( a a ) d 18( a a ) (9) Euations 5 to 9 are based on the symmetric model This model is recommended when the disersed hase in one region become continuous hase in another region of the domain Here the subscrits and are interchangeable for gas and liuid based on the volume fractions F in euation (4) is the source term, which accounts the surface tension forces at the interface The formulation for the surface tension is based on works of Brackbill et al (199) F vka 05 (10) Here k is defined as the curvature and is comuted from the unit normal which are defined as the gradient of the volume fraction of the liuid hase k l n (11) n n a (1) Conservation of energy for hase gives euation 13 ah a vh t Q a : v t (13) Here Q is volumetric rate of energy transfer between two hases, defined by euation 14 Q h ( T T ) (14) Where h is the volumetric heat transfer coefficient between gas and liuid hases and is calculated by using the Ranz - Marshall correlation Geometry and mesh generation The baseline model was constructed based on existing exerimental uench column dimensions which was shown in table 1 The sketch of the exerimental uench column is shown in figure 3 The original design has 18 sets of disc and donuts Flooding is caused by choking flow near the flow area either at disc lates or donut late s ieuflow of the vaour restricts the down flow of the liuid coolant (FJZuiderweg, 198) In order to assess this henomenon in the comutational model, it is sufficient to consider the bottom three stages of donut arrangement as the maximum exected flow rates of the vaour is exected in this area For this reason, three donuts and two disc late arrangement considered in the baseline model as shown in figure 4 Based on the observations from the results of the baseline model, a new design was roosed and is modelled The modified design is identical to the baseline model excet for the Coyright 01 CSIRO Australia 3

4 donut late configuration In this model, eight holes with liners (herein after are called as cas) were laced circumferentially This is to facilitate byassing of some gas to next stage The lining was about 6mm in height Parameters Exeriment Baseline Volumetric flow rate, m 3 /s Temerature, 0 C Column diameter, cm Donut inner annular diameter, cm Disc diameter, cm 9 77 Sacing between disc and donut Number of discs Number of donuts Table 1: Quench column design arameters Figure 4: Disc and Donut late arrangement 3 Figure 6: Grid sacing section taken at mid height Initial and Boundary conditions Initially D models were considered as these will give uick insight into the hysics But due to flow which is highly unsymmetrical and random in nature, 3D transient analysis was adoted for this modelling task The initial volume fraction of coolant was set to zero and entire euiment filled with nitrogen at room temerature ie at 5 0 C Initial velocity and ressure set to zero (gauge) in the whole domain A mass flow rate of 005 kg/s of octane is sulied for the coolant inlet at -5 0 C The initial run was continued until the coolant liuid reaches to the bottom surface of the euiment Once the liuid flow was develoed, Nitrogen was injected from the vaour inlet at C The initial flow rate sulied is about 10% of the designed flow rate Here the designed flow rate for this euiment is 00044m 3 /s, since the detailed chemical comonents of the yrolysis vaours not included and the flow is assumed to be nitrogen, the calculated mass flow rate is about 0003 kg/s based on its density at 400 o C For the baseline model two studies conducted with different flow rates one is with 10% flow rate and another is with 50% flow rate Whereas the modified model is run with 50, 75 and 100% flow rates resectively The uench column wall was considered as adiabatic and heat transfer from it is neglected Atmosheric ressure outlet was maintained at the gas outlet The coolant outlet is modelled as a mass flow inlet with negative flow rate Based on the liuid levels at the bottom, the flow rate changes to either 0 or -005 kg/s in order to avoid rising of the liuid coolant beyond secified levels RESULTS Figure 5: CFD Domain - Quench column Duo to the size of uench column (fig 5) and its comlexity of the flow behaviour, the grid size was chosen as 15 mm with uniform sacing Hexahedral cells were laced inside the domain, as hexahedral mesh gives better accuracy than tetrahedral meshes while comuting surface tension effects The section of the meshed model is shown in the figure 6 Total number of cells for the baseline model is 066 million whereas for modified design model is 073 million All the models started with 0001 s time ste After monitoring the gas temeratures at the exit and the liuid flow rate at the intermediate section which was located in between disc late and donut, the steady state was achieved after 8s max in all cases excet baseline with 50% flow case where the steady state solution is not ossible due to flooding of the coolant The resented data in this art corresonds to data obtained after this time Figure 7 shows the coolant liuid inside the domain for the baseline case According to the figure, the coolant liuid is clearly flowing downwards for the 10% case (50% to designed flow) It also clearly shows that in 50% flow case flooding is haening at the to donut late as Coyright 01 CSIRO Australia 4

5 reorted in the exeriment mentioned earlier The coolant liuid level is constantly rising on donut and very small uantity is flowing towards the bottom as this is evident from liuid level at the bottom for these two cases The liuid is forced to flow uwards through the donut due to the higher ressure at the lower stages (due to rising the liuid level above the gases), however the ath is blocked by choked flow of gases causing the accumulating liuid at the to of the donut further increasing the resistance of the coolant on smooth assage of the gas This revents the smooth flow of the coolant to the lower stages of the column and eventually leads to flooding at high gas and liuid flow rates through these holes so that the flooding will be avoided as observed in earlier case Figure 9 shows the coolant liuid flow attern in the modified model for different flow conditions From this icture it is evident that the flooding is avoided comletely by this new configuration and is able to erform u to its designed limits It was also observed that the disc lates are not holding the coolant comletely and coolant is forming atterns on it As this will reduce the heat transfer considerably, it would be better if the liners are rovided for the disc lates as well so that constant level of coolant will be always maintained on the disc lates Figure 9: Iso-surface showing coolant in modified models Figure 7: Iso-surface showing coolant in Baseline models Velocity vectors for the gas combined with temerature contours varying from maximum 400 o C to minimum -5 o C resented in the figure 8 This section created by slicing the euiment vertically at the centre which asses through the vaour inlet From this, it can be observed that in 50% case of the base line model, the maximum velocity reached by the vaour is about 5m/s which is well below the ermissible velocity mentioned in the literature based on which the euiment was designed Since the flooding is observed at 50% flow of the designed flow itself, the 100% flow case on the baseline model was not conducted Figure 8: velocity vectors in Baseline models As observed in the figure 8, the flow is highly unsymmetrical and fluctuating from one side to the other over time Based on the results of the baseline model, it was observed that the main issue lies with the donut late design The flow area for the donut late is less than the flow area available for disc in the baseline model In order to get same oen area for disc and donut lates, eight holes with liners were introduced on donut late The liners on the hole ensure that only vaour will ass Figure 10: Velocity vectors and temerature contours in modified models Figure 10 shows the combined lot of velocity vectors coloured by velocity magnitude and the temerature contours for the all three cases of the modified model From this figure it is evident that a art of the vaour is byassing through the holes rovided on the donut lates The holes are acting as nozzles which are directing the vaour towards the disc late After hitting the disc lates the vaour starts circulating and flowing back towards the donut late and contacts with coolant there by maintaining high heat transfer rate and higher effectiveness It is also observed that the flow directions/vectors are unchanged irresective of the flow conditions and this gives the euiment a scoe for scaling u The average temeratures of the gas for the different cases are resented in figure 11 For the modified model at 100% flow condition, the minimum temerature attained was about 68 O C which indicates that a significant amount of cooling can be achieved only by just three stages Further cooling can be achieved if more stages are introduced inside the column This deends on the nature of the rocess and the degree of cooling that is reuired From this grah it is noted that major cooling is haening Coyright 01 CSIRO Australia 5

6 at disc late interface as it resenting two tyes of cooling, one is imingement cooling which comes under indirect cooling and another is direct contact cooling After assing through the holes of the donut late the vaour is directly iminging on the disc late which is highly desired for faster heat transfer The residence times for the gas flow for the different cases were tabulated in table Here the to row shows the er cent of the total inlet flow which directly assing within the given time 85% of the flow in modified model is going to outlet within 1 sec and thus raid cooling of yrolysis vaours can be achieved CONCLUSION Figure 11: Average temerature of the gas along the uench column height Figure 1: Average ressure of the gas along the uench column height Pressures shown are gauge ressures Figure 1 shows the ressure lot along the length of the uench column for different runs From this, it is clearly evident that in the baseline model 50% flow case, large ressure dro occurs at the interface of to most donut location This is due to the increasing levels of the coolant over the vaour due to flooding Whereas in case of 10% flow case, the maximum volume flow is occurring at the bottom donut late level as the temerature of the vaour is high and hence the ressure dro is high at that location In modified model cases the maximum ressure dro was limited to aroximately 11 Pa as against 90Pa in the baseline case \CFD Run 5% 50% 75% 80% 85% %of inlet flow -> Baseline model with % Modified model with % Modified model with % Modified model with 100% Table : Gas residence times (sec) inside the uench column An Eulerian-Eulerian method with immiscible fluid model for modelling uench column hydrodynamics has been resented The baseline model simulated for two different flow conditions, one with 5% gas flow rate and another with 50% gas flow rate In 50% condition, flooding of the liuid coolant started on the to most donut late as observed in the exeriments It was identified that the flow area near the donut late is not sufficient for this magnitude of gas flow rates even though it was designed by considering the design correlations A modified model was develoed by giving extra flow area to the gas The results were comared against the baseline model The modified model successfully demonstrated in handling 100% designed flow rate as well as giving better heat transfer with less ressure dro It is also shown that in less than 1 s, the vaours can be cooled down to 77 O C (temerature dro of 33 O C) Understanding of the condensation rocess and modeling of it can rovide the detailed information on the bio oil extraction while validating with existing as well as future exeriments For this, a comrehensive condensation model is under develoment and results will be ublished by the authors in the near future REFERENCES AVBRIDGWATER (000), Fast yrolysis rocess for biomass, Renewable Sustainable Energy Rev, 4, 1-74 AVBRIDGWATER (01), Review of fast yrolysis of biomass and roduct ugrading, Biomass Bioenerg, 38, PEACOCKE GVC, AVBRIDGWATER, (1995), Ablative late yrolysis of biomass for liuids Biomass Bioenerg, 7, AVBRIDGWATER, PEACOCKE GVC, ROBINSON NM US atent number US 7,65,53 DIEBOLD, JP, (1980), Preliminary Results in the Fast Pyrolysis of Biomass to Lower Olefins, Petroleum Chemistry division of ACS Symosium on Alternate Feed stocks for Petrochemicals, Las Vegas, NV USA BRADBURY, AGW, SAKAI, Y, SHAFIZADEH, F (1979) A Kinetic Model for Pyrolysis of Cellulose, J Al Polym Sci, 3, 371 NMROBINSON, (000), Design, Modelling and Construction of a Novel Ablative Fast Pyrolysis Reactor and Product Collection System, PhD thesis, Aston University, UK JUBRACKBILL, DBKOTHE, and CZEMACH, (199), A continuum method for modelling surface tension, J Comut Phys, 100, F I ZUIDERWEG, 198, Sieve trays a view on the state of the art, Chem Eng Sci, 37, Coyright 01 CSIRO Australia 6