SEPARATIONS
Experimental and Computational Fluid Dynamics Investigation of the Flow in and around Once-Through Swirl Tubes
Siri Jacobsson,Trond Austrheim,and Alex C.Hoffmann*
Department of Physics and Technology,Uni V ersity of Bergen,Allegaten55,5007Bergen,Norway
The gas flow in and around once-through swirl tubes s also called axial flow cyclones s for gas demisting has
been investigated by flow visualization using a neutrally boyant tracer and by computational fluid dynamics
(CFD).The neutrally boyant tracer was helium-filled soap bubbles,which were made visible by illumination音标的英文
with sheets of white light and recorded by a variety of photographic techniques.The flowpatterns in the inlet
swirl vanes,the paration space,and the core of the vortex are shown as streak patterns.Additionally,it was
possible to study the breakdown of the vortex beyond the swirl tube outlet.Using high-speed photography
and dedicated software,velocity information was gleaned from the experiments.The experimental results
were compared with CFD simulations wherein the renormalization group k- turbulence model was ud to
simulate the flowpattern in the same configuration.The simulations matched the experimental results well in
the paration space itlf but could not properly reproduce the complicated flow around the tube outlet.
1.Introduction
This rearch project was carried out as part of a larger rearch program,HiPGaS,aimed at the improved modeling and design of natural gas scrubbers.
The development of oil and gas fields moves toward smaller and more remote fields.The technology push is to carry out more process,such as cleaning and paration,remotely and with minimal intervention.The long-term vision is to carry out most offshore gas/liquid and liquid/liquid paration suba. Incorrect design of paration equipment would be particu-larly damaging for remote and suba installations,so that a better basis for design is crucial.
Separation of droplets from gas/vapor streams is a ubiquitous operation in the processing of natural gas.Liquid must be parated from natural gas at veral stages,for instance to protect downstream equipment from corrosive liquids,lower the hydrocarbon or water dew point of the gas,or recover valuable liquefied petroleum gas(LPG)products from the gas stream.
There are many different types of demisting equipment on the market.Often different types of equipment are combined to give optimal liquid removal efficiency.One typical arrange-ment is a gas-liquid scrubber that consists of three stages:a preparation stage at the scrubber inlet ction is followed by an agglomeration stage,often a mist mat,and finally by a demisting stage.The demisting
stage often consists of a bank of once-through swirl tubes,also called axial flow cyclones, working in parallel.
There is a large body of rearch literature dedicated to modeling,characterizing,and optimizing traditional rever-flow tangential-inlet cyclones,but despite their popularity,few publications exist in the open literature dealing with efficiency predictions for once-through swirl tubes.Their paration efficiency has been discusd by Bu¨rkholz,1Ramachandran et al.,2and Brunazzi et al.,3while different geometries have been more thoroughly discusd ,Swanborn,4Bu¨rkholz,1 Verlaan,5Nieuwstadt and Dirkzwager,6and Hoffmann and Stein.7Of the publications,only Verlaan has given a descrip-tion of the geometry of a tested swirl tube with sufficient details to reproduce it.
Most of the published work has been performed under near-atmospheric air-water conditions,and the design of parators for high-pressure natural gas scrubbing has,therefore,often been bad on either extrapolated low-pressure data or confidential data provided by the suppliers,which also in many cas may have been produced under low-pressure conditions.
In this paper,we are concerned with some outstanding questions around the flowpattern in and around once-through cyclones,which must be answered to predict their performance. One issue is th
e radial profile of the tangential velocity.In rever-flow cyclones,the connsus in the literature is that the tangential velocity distribution is near solid-body , with a constant angular velocity)in a narrow core,surrounded by a near loss-free ,a decreasing tangential velocity with increasing radius);thus,the rotation rembles a Rankine-type vortex.
In once-through cyclones,the picture is not so clear. Ramachandran et al.,2while investigating a rotary-flow once-through cyclone,state that in contrast to a rever-flow cyclone there are no outer and inner vortices in a rotary-flow cyclone and that the tangential velocity profile is also different:instead of the velocity decreasing from the central core to the wall in the outer part,they assumed it to increa from zero at the axis to a maximum at the wall.Also,Stenhou and Trow8in their model for an axial flow cyclone assume,and verify using lar Doppler anemometry,that the tangential velocity is zero at the wall and increas toward the wall in a more-or-less solid-body rotation,apart from a“boundary layer”at the wall,where the tangential velocity again decreas.Maynard,9in his model for once-through cyclones,calculated the tangential velocity profile
*To whom correspondence should be addresd.Tel.:004755582876. Fax:004755589440.E-mail:alex.hoffmann@6525
Ind.Eng.Chem.Res.2006,45,6525-6530
10.1021/ie051200s CCC:$33.50©2006American Chemical Society
Published on Web08/22/2006
assuming the flow to be plug flow in a helicoidal channel.This also results in a tangential velocity that increas from zero at the axis to a maximum at the wall.
On the other hand,Verlaan 5found that the tangential velocity profile follows a modified Rankine-type vortex profile.
An added complexity is that many once-through cyclones are equipped with vertical slits to aid in the liquid paration.The slits may create extra turbulence in the vortex and cau the tangential velocity distribution to be different from that in conventional cyclones.
Another outstanding issue is the flow at the exit from the axial flow cyclones.Does the vortex break down rapidly,or does it persist in the space above the axial flow cyclones,attaching to some solid wall in the scrubber space above the cyclone bank?Little is known about this issue at prent.2.Objective
The objective of this work was to answer the two questions outlined above:Is the tangential velocity
pattern esntially different from that of rever-flow cyclones?Does the vortex break down immediately as it exits the tube?The object was also to gain insight into the detailed mechanisms of liquid paration and possible reentrainment and to asss the potential and shortcomings of computational fluid dynamics (CFD)simulation of the flow.3.Experimental Section
3.1.Swirl Tubes.Figure 1shows a diagram and a photograph of the swirl tube tested.The swirl tube is configured in accordance with the tube of Verlaan.5Two different sizes of the swirl tube were made,with lengths of 250and 400mm,respectively,and diameters of 50and 80mm.The material of construction is perspex,except for the vane arrangement,which is of stainless steel (SS 316).
The swirl vanes have a blade exit angle of 45°and contain six vane blades.The central body diameters are 30and 48mm,respectively,and the total lengths,125and 200mm.The axial length of the blades themlves are 50and 80mm.
Four vertical drainage slits in the cyclone tube wall,111and 176mm long,respectively,and spanning the paration ction between the swirl device and the exit are ud,as shown in Figure 1.The slits were sharp-edged and expanding through the tube wall,having inner diameters of 5and 8mm in the two models,respectively,and outer diameters of 7.5and 12mm,respectively.
One possible modification to make the liquid drainage more efficient is to draw off a condary gas flow from the surrounding collection chamber to help the liquid through the
drainage slits.In general,there are two ways of doing this:by recycling some gas back to the low-pressure zone upstream of the swirl element;by adding small condary outlets directly at the top of the drain chamber.
In this study,only the last one was ud by opening one circular condary outlet of 15mm,giving a cross-ctional area of 176mm 2,as arrowed in the figure.
3.2.Rig.The rig was produced in-hou,and a diagram is shown in Figure 2.The swirl tubes and piping were manufac-tured in transparent material (perspex)to make visualization possible.
The bubble generator that makes neutrally buoyant soap bubbles (e below)is connected to the inlet tube,which has a length of 1000mm and a diameter of 80mm.Thus,when the smallest swirl tube was installed,the inlet tube went from 80to 50mm before the swirl device.The gas with tracer bubbles enters the swirl device and moves,while spinning,through the paration ction.
At the outlet of the tube,the tracer-laden gas exits into a rectangular chamber with a cross ction of 150mm ×150mm.The outlet from the box was higher up,at one side,as indicated in the diagram.
The flow is regulated using a hand valve and then pasd through a venturimeter and a pump.
3.3.Flow Visualization Equipment.The bubble generator,supplied by Sage Action Inc.,Ithaca,NY,produces helium-filled neutrally buoyant bubbles (soap bubbles)of a controlled size of 1-2mm.At the core of the generator,there are two concentric stainless steel hypodermic tubes.Helium pass through the inner tube and bubble film solution through the outer one.An air flow blows the bubbles off the tip of the tube.A mini vortex filter removes bubbles that deviate from neutral
Figure 1.Diagram (adapted from ref 5)and photograph of one of the swirl tubes ud.The condary outlet is arrowed.
Figure 2.Diagram of the experimental rig.
Figure 3.Diagram showing the positions of horizontal light sheets ud.Measurements on the left are for the 80mm diameter tube,and tho on the right are for the 50mm diameter tube.
6526Ind.Eng.Chem.Res.,Vol.45,No.19,
2006
density.The bubbles follow the flow streamlines without bursting or impacting on objects within the flow.
To visualize the flow pattern,continuous white light from a 300W halogen lamp,equipped with cooling fan,in combination with either digital film/image cameras or a high-speed camera was ud.When required,puld light was created using a fan (a plate with slits)in front of the lamp.4.Experimental Method and Analysis
4.1.Flow Visualization and Analysis.The flow was visualized using continuous light combined with either image or video photography.The streak length and shape gave a good impression of the flowpattern and the local velocity distribution.The light source and camera were always at 90°to each other for optimal clarity.The light source was placed both at the top and bottom of the device,and horizontal light sheets were also ud at a range of axial positions,as shown in Figure 3.
For velocity measurements,two different methods were tried:(1)the u of continuous light together with a high-speed camera and (2)pul illumination in combination with a digital image camera,using a long exposure time compared to the illumination puls.
For the cond of the two methods,the above-mentioned fan was ud to create puld light.The slits in the fan were of different angular sizes to “code”the streak pattern to obtain not only the velocity from the length of the streaks but also the direction of flow from the pattern of streaks.Although we managed to create software to make this method work,the first of the two methods gave the best results with the software as it is developed at the prent time,and we will therefore only describe the first method below.
DiaTrack,a commercial software package,was ud for extracting velocity information.The principle is to recognize individual tracer particles on successive frames and calculate velocities from the distance between images on successive frames and the known interval between the frames.To do this successfully,the high-speed camera needed to record as much as 2000frames per s when high-velocity regions,such as the region within the tube itlf,were analyzed.
Of the 64films taken,33films were analyzed:24films from the 50mm cyclone and 7films from the 80mm cyclone.Most films were from the 50mm cyclone,since this was the tube configuration ud for the CFD simulations.
Films were obtained in the 50mm diameter tube at three inlet velocities: 3.8,4.6,and 5.8m/s.
In principle,threshold and filters could be t in the software to recognize the tracer particles automatically.Despite this,recognition sometimes had to be done manually due to variations in the strength of the light source,making the frames unevenly expod.Reflections also caud some problems.
Once the trackings had been analyzed,the software extracted velocity information that could be displayed either as a vector plot or in color coding.Figure 4shows an example of a t of tracks and the resulting vector plot of the average velocity 36cm above the inlet of the 50mm ,in the vesl of square cross ction above the tube (e Figure 3).
4.2.CFD.3-D CFD simulations of the flowpattern in the 50mm swirl tube,at the three inlet velocities ud experimentally,were carried out using the commercial finite-volume software package FLUENT 6.1,having built the computational grid using the associated package GAMBIT.The turbulence model ud was the RNG k - turbulence model,10with the swirl-dominated flow option.
The computational grid consisted of 502000unstructured tetrahedral cells.Figure 5gives an impression of the grid resolution relative to the equipment.
Figure 4.Horizontal light sheet,placed 36cm above the inlet of the 50mm swirl ,in the vesl
of square cross ction above the tube,photographed from above.The inlet velocity is 3.8m/s.(left)Vector plot of the averaged flow pattern.The scale en in the top left-hand corner corresponds to 0.89m/s.(right)Trackings that are included in the calculations.
Figure 5.Swirl vanes with the computational grid indicated.Table 1.Model Settings for the CFD Simulations
Cmu C1-epsilon C2-epsilon swirl factor 0.0845
1.42
1.68
0.07
Ind.Eng.Chem.Res.,Vol.45,No.19,2006
6527
The model ttings are given in Table 1.At the inflow boundary,the turbulence intensity was taken as 5%;this is slightly more than the 4.3-4.5%estimated using the well-known empirical expression: 1.6Re -1/8.However,we considered that the flow had been through a sharp turn just before arriving at the inflow boundary.The turbulence length scale was taken as 0.,of the order of magnitude of the container,as is often done.
5.Results and Discussion
We first show the most interesting results from the flow-visualization studies and then the calculated velocity fields,comparing them with the results of the CFD simulations.5.1.Flow-Visualization.Figure 6shows the flowpattern around the swirl V anes .Due to reflection from the vanes,it was difficult to obtain streak-lines of the flow in the swirl vanes themlves,but the images clearly show the swirling flowpattern generated by the vanes and how the shaping of the central body helps in avoiding turbulence creation in the flow exiting the vanes. our knowledge,the first pictures in the published literature to show the flowpattern in a swirl-vane asmbly with neutral-density tracer.The method is clearly suitable for studying recirculation or turbulence in,and around,swirl-vane designs.
In the paration space ,the tracer bubbles tended to concentrate in the center of the vortex,although some remained in the outer regions of the vortex.This issue,and the wider issue of the faithfulness with which the tracer bubbles follow the flowpattern,including the turbulent eddies,is discusd in ref 11.The conclusion is that evaporation from the bubble surface may cau them to become slightly less than neutrally boyant as they move through the flow,resulting in them tending to centralize in the strongly swirling flow in the paration space.The advantage of this is that the core is clearly visible.
The strongly swirling flow will exacerbate any deviation from neutral boyancy of the tracer.Tho tracer bubbles that are too den will be centrifuged to the wall and vanish as they are crushed on the wall,while tho that are too light will asmble in the core,where they will form a highly visible band.Thus under all circumstances,one is likely to e the core very clearly using this type of tracing.
We would like to stress,however,that the photos we are showing here are meant to show the core.By fine-tuning the
Figure 6.Photos showing streak-lines of He-bubble tracer particles around the inlet swirl vanes.Each line is double due to the two reflection points on a bubble.
犹太人的故事
Figure 7.Photo showing the flowpattern in the paration space of the 80mm tube.The inlet velocity is 2.9m/s.The light is from the top.The shutter speed of the camera is 1/100s.This photo is meant to show the core of the vortex clearly;by fine-tuning the bubble generator,a much higher bubble concentration in the outer part of the vortex could be obtained.
Figure 8.Photos of the 80mm cyclone 41cm above the inlet,just above the outlet.The inlet velocity is 2.9m/s.The camera exposure times are (left)1/80and (right)1/50s.
6528Ind.Eng.Chem.Res.,Vol.45,No.19,
鱼缸理论
2006
教师资格证笔试
bubble generator,it was possible to obtain a much higher bubble concentration in the outer part of the vortex.
The flowpattern is en in Figure 7.The core of the vortex (there are paration slits just in front and behind the core in the photo)is en to be almost straight,right from the swirl vanes to the exit from the tube.Some streaks from tracer particles further out in the vortex can be en.
Above the outlet ,the vortex dissipates.Some of the prent authors recently published a study of the “end of the vortex”phenomenon in rever-flow cyclones and swirl tubes.12They
found that this phenomenon bore no relation to vortex break-down internally in the fluid but was due to the vortex core bending to the wall,attaching to it,and rotating around it.In the prent ca,however,in a once-through configuration,we think that we are eing vortex breakdown above the tube.The flow configuration and the tube expanding to the containing vesl is very similar to the configuration in which studies of vortex breakdown are carried out,where the breakdown is normally induced by inrtion of a diverging ction in the tube.Faler and Leibovich,in a classic paper,13discuss a number of different vortex breakdown modes,and Lucca-Negro and O’Doherty 14
review the issue of vortex breakdown in a recent paper.
Figure 8shows the flow pattern just above the outlet of the 80mm swirl tube.It can clearly be en that the individual tracer bubbles spread out from their centered position,probably around a recirculatory flow (a “bubble”),such as in the axisymmetric type “0”breakdown discusd by Faler and Leibovich.
The flow at the stations further above the outlet showed a high degree of turbulence,and a pattern was not immediately evident from this type of photo,although,as Figure 4shows,on the average the flowpattern did exhibit swirl.
We can compare the streakline patterns shown in the previous figures with the pattern obtained from the CFD simulations.Figure 9shows pathlines from the simulations of the 50mm swirl tube.
Obviously,the flow around the swirl element and the swirling flow with a straight vortex core in the paration ction are matched by the CFD results,but the vortex breakdown just after the outlet does not em to be well reproduced in the simula-tions.One pathline is en to escape through the drainage slits.The reason that the vortex breakdown was not reflected in the simulation was probably that the turbulence model was too simple and/or the grid too rough.As many recent papers show,
Figure 9.Pathlines of the simulated flow in the 50mm swirl tube.The inlet velocity is 3.8m/s.
Figure 10.Distribution of tangential velocity in a horizontal light sheet placed 13cm above the inlet in the 50mm swirl tube with an inlet velocity of 3.8m/s.(Left)Experimental.(right)CFD simulation.
Figure 11.Distribution of tangential velocity in a horizontal light sheet placed 19cm above the inlet in the 50mm swirl tube with an inlet velocity of 3.8m/s.(left)Experimental.(right)CFD simulation.
Ind.Eng.Chem.Res.,Vol.45,No.19,2006
十神详解
6529
for instance,refs 14and 15,CFD simulations can reproduce even the fine details of type 0vortex breakdown.The paper of Derkn 15is interesting in showing such a breakdown just after a sudden c
onstriction ,as in the vortex finder in a rever-flow cyclone or swirl tube.
5.2.Velocity Distribution.The left plate of Figure 10shows the measured tangential velocity distribution in the paration ction of the swirl tube,just over the swirl element.The velocity is clearly low in the core,as one would expect,and it is fairly uniform over the rest of the cross ction.The right plate shows the CFD simulations.The ranges of the color scales in the two plates are the same.The two flows are somewhat similar,although the CFD simulations em to indicate an increasing velocity with an increasing radius,while this is not so in the experiments.
会展组织Figure 11shows the measured velocity distribution higher up in the tube.Clearly,the flow is changing into the familiar pattern of a near solid-body rotation surrounded by a near loss-free vortex motion,with the velocity going through a maximum clo to the core of the vortex.This is similar to the tangential velocity distribution in Burgers’vortex (sketched in in Figure 12together with the data from Figure 11).As en in the figure,the CFD simulations at this level reflect this transition somewhat less well,and we must conclude that the experiment tangential velocity distribution was not very well matched by the CFD simulations.6.Conclusions
This project answered the questions asked at the outt.Some itemized conclusions are the following:
•Studying the flowpattern in once-through swirl tubes using neutral boyancy tracer has revealed features of the flow not yet en.
•The swirl tubes have been shown to posss an almost axisymmetric flow with the core of the vortex almost straight,right from the swirl vanes to the outlet of the tube.
•The experiments show that the cross-ctional distribution of tangential velocity is somewhat flat just after the swirl element,developing into the familiar tangential velocity distri-bution,rembling Burgers’vortex and the Rankine vortex further from the swirl element.
粉身碎骨•The vortex appears to suffer axisymmetric,bubble-type breakdown at the exit of the tube and does not persist above the tube.
•The CFD simulations reflected the velocity distribution in the tube just after the swirl vanes reasonably well but less well further on in the tube,and they failed to match the breakdown of the vortex at the tube exit.A more sophisticated simulation is clearly required,a complicating factor being the prence of the vertical slits.
Acknowledgment
The authors wish to thank Dr.Carl Birger Jensn of Statoil for tting up the configuration for the CFD,Dr Weiming Peng for helpful advice for experimental work and the calculations,and members of the rearch group of Prof.Hallvard Svendn at NTNU for assistance.We also thank the HiPGaS industrial partners:NFR,Statoil,FMC Kongsberg Suba,Conoco Philips,Norsk Hydro AS,ABB,and Kværner Process Systems for sponsorship and assistance.Literature Cited
(1)Bu ¨rkholz,A.Droplet paration ;VCH Publishers:New York,1989;ISBN 0-89573-879-1.
(2)Ramachandran,G.;Raynor,P.C.;Leith,D.Collection efficiency and pressure drop for a rotary-flow cyclone.
Filtr.Sep.1994,31,631-636.
(3)Brunazzi,E.;Paglianti,A.;Talamelli,A.Simplified design of axial-flow cyclone mist eliminators.AIChE J.2003,49,41-51.
(4)Swanborn,R. A.A new approach to the design of gas-liquid parators for the oil industry.Ph.D.Thesis,Delft University of Technology,Delft,The Netherlands,1988.
(5)Verlaan,C.C.J.Performance of novel mist eliminators.Ph.D.Thesis,Delft University of Technology,Delft,The Netherlands,1991;ISBN 90-370-0054-1.
(6)Nieuwstadt,F.T.M.;Dirkzwager,M.A.Fluid mechanics model for an axial cyclone parator.Ind.Eng.Chem.Res.1995,34,3399-3404.(7)Hoffmann,A.C.;Stein,L.E.Gas Cyclones and Swirl Tubes -Principles,Design and Operation ;Springer-Verlag:Heidelberg,2002;ISBN 3-540-43326-0.
(8)Stenhou,J.I.T.;Trow,M.The behaviour of uniflow cyclones.In Proceedings of the Second World Filtration Congress ;The Filtration Society:London,1969;pp 151-155.
(9)Maynard,A.D.A simple model of axial flow cyclone performance under laminar flow conditions.J.Aerosol Sci.2000,31,151-167.
(10)Ma,L.;Ingham,H.D.B.;Wen,X.Numerical modelling of the fluid and particle penetration through small sampling cyclones.J.Aerosol Sci.2000,31,1097-1119.
(11)Peng,W.;Hoffmann,A.C.;Dries,H.W.A.;Regelink,M.;Foo,K.-K.Neutrally boyant tracer in gas cleaning equipment.A ca study.Meas.Sci.Technol.2005,16,2405-2414.
(12)Peng,W.;Hoffmann,A.C.;Dries,H.W.A.;Regelink,M.;Stein,L.E.Experimental study of the vortex end in centrifugal parators:The nature of the vortex end.Chem.Eng.Sci.2005,60,6919-6928.
(13)Faler,J.H.;Leibovich,S.Disrupted states of vortex flow and vortex breakdown.Phys.Fluids 1977,20,1385-1400.梅干菜怎么泡发
(14)Lucca-Negro,O.;O’Doherty,T.Vortex breakdown:A review.Progr.Energy Combust.Sci.2001,27,431-481.
(15)Derkn,J.J.Simulations of confined turbulent vortex flow.Comput.Fluids 2005,34,301-318.
Recei V ed for re V iew October 28,2005Re V id manuscript recei V ed June 10,2006
Accepted July 20,2006
IE051200S
Figure 12.Sketch of the tangential velocity distribution in Burgers’vortex together with the data in Figure 11.
6530Ind.Eng.Chem.Res.,Vol.45,No.19,2006
