WO*' Lx
GAS AND LIQUID MALDISTRIBUTIONS IN PACKED COLUMNS
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus, prof.drs. P.A. Schenck, in het openbaar te verdedigen ten overstaan van een commissie
aangewezen door het College van Dekanen op 27 april 1989 te 14.00 uur
door
Robertus Martinus Stikkelman
geboren te 's-Gravenhage scheikundig ingenieur
TRdiss 1716
Dit proefschrift is goedgekeurd door de promotoren
prof.ir. J.A. Uesselingh prof.dr.ir. J. de Graauw
STELLINGEN
0. Een stelling is slechts te verdedigen binnen een geaccepteerd axiomastelsel.
1. De uitkomst van maldistributiefactormetingen in gepakte scheidingskolommen is afhankelijk van het meetsysteem.
2. Het bepalen van spreidingscoëfficiënten voor de vloeistoffase bij een lage pakkingshoogte is onnauwkeurig.
—3—Voor—de-interpretatie-van-stofoverdrachtgegevens-voor—gepakte kolommen is een nauwkeurige beschrijving van de kolom en de randapparatuur onmisbaar.
4. Het stijgen van de HETP van gestructureerde pakkkingen bij hoge gasbelasting is niet een gevolg van "flooding" maar van
* het optreden van grootschalige maldistributie.
5. De inschatting van Baerends dat de tijd voor het berekenen van atomen/molekulen volgens de "discrete variational Hartree-Fock-Slater" methode in de orde ligt van het aantal electronen In het kwadraat is te optimistisch.
E.J.Baerends, D.E.Ellis and P.Ros, Chem. Phys. 2 (1973) 41
6. Alles is een perpetuum mobile.
7. Filevorming en milieuvervuiling kunnen verminderd worden door het afschaffen van: - de reiskostenvergoedingen en - de overdrachtsbelasting bij de huizenverkoop
8. Het belang van presentatie- en communicatievaardigheden wordt in het huidige studieprogramma voor scheikundig ingenieur onderschat.
9. Fusie leidt tot confusie.
10. De verklaring van het woord stripverhaal heeft tegenwoordig meer met de inhoud dan met de vorm te maken.
R.M.Stikkelman Delft, 27 april 1989
SUMMARY
Packed columns are increasingly used in distillation and absorption/stripping processes. For the application of these. colums a good understanding of the flows in the packing is desirable. This study describes the gas and liquid distributions in random as well as structured packings. The experimental column has a diameter of 0.5 m. It is equipped with a total of 1289 detectors in the top and bottom cross section. These detectors yield a detailed picture of air and water flows through elements of only 25x25 mm2.
Random packings
The observed maldistribution in the gas bulk flow is negligible compared with that of liquid bulk flow. The gas flow rate near the wall equals 1.1-1.3 times the superficial velocity for common 25 mm palEkings. No influence of the liquid flow rate has been observed. The radial spreading coëfficiënt of the gas is in the order of A mm.
The liquid flow in the bulk becomes less uniform when the superficial liquid velocity is diminished. The flow distribution is almost independent of the gas flow. The spreading of liquid shows a srnall increase up to the loading point, above which it increases more rapidly. Values of the radial spreading coëfficiënt of the liquid are approximately 3 mm. In the loading region the liquid wall flow at a fixed packing height tends to lower values at higher gas flow rates.
The measurements of the gas and liquid flow profiles can be simulated with a simple Monte Carlo cell model. It gives a good prediction of the liquid and gas spreading, of the development of wall flow, of the small scale maldistribution and of the effects of the superficial gas velocity.
Two types of tower internals have been simulated: an initial distributor and a wall wiper. Drip point densities of more than 200/m2 hardly improve the liquid distribution in a column with 25 mm packing. In a column with a diameter of 0.5 m the wall flow is only reduced over a small packing height by a wall wiper.
A
The effect of a severe initial liquid maldistribution in a methanol/ethanol distillation column has been investigated. This was done in a 0.45 m diameter column with a packing height of 2.65 m. Sealing half of the distributor resulted in a sharp increase of the height of a transfer unit based on over-all gas-phase resistance.
The Monte Carlo cell model, extended with a simple mass transfer model, gives a realistic simulation of the distillation results.
The separation properties of the first meter of a typical 25 mm packing with different distributors have been simulated. With 30 drip points per m2 a packing height of 0.4 m is effectively lost.
Structured packings
The maldistribution in the gas bulk flow is negligible. Only the observed wall flow can contribute to malperformance. The gas flows parallel to the sheet orientation, thus introducing a radial transport. Together with the change in the orientation of subsequent packing elements, this results in good gas mixing.
It was observed that the liquid wall flow decreases when the gas velocity is higher than 1.7 m/s. Up to the loading point the maldistribution of the liquid is constant. Above this point the quality of the distribution deteriorates rapidly, due to the appearance of large scale liquid segregation.
Appendix A provides a method of characterizing a flow distribution with a relatively small number of parameters. A channel maldistribution factor is defined that indicates which channel sizes contribute most to an overall maldistribution. A newly defined overall maldistribution factor is shown to give a good ranking of different distributions.
B
SAMENVATTING
Gepakte kolommen worden steeds vaker toegepast in de procesindustrie. Het is voor het ontwerpen van zo'n kolom wenselijk dat de stroming in de pakking nauwkeurig beschreven wordt. Dit kan helpen om tegenvallende prestaties te voorkomen. In dit proefschrift is het stromingsgedrag van water en lucht voor zowel losse als gestructureerde pakkingen bestudeerd. In een kolom met 1289 stromingsdetectoren zijn aan de top en bodem van de pakking gedetailleerde profielen gemeten. De kolom heeft een diameter van 0.5 m.
Losse pakkingen
De maldistributie van het gas in de bulk van de pakking is gering ten opzichte van die van de vloeistof. Voor gangbare 25 mm pakkingen is de gemiddelde gassnelheid langs de wand een factor 1.1-1.3 groter dan de_sup_erficiële_snelheid.—De-vloeistofstroming heeft nauwelijks invloed op het-gas. De radiale spreidingscoëfficiënt voor het gas is ongeveer 4mm.
Bij lage vloeistofsnelheden neemt de kwaliteit van de vloeistofverdeling in de bulk van de pakking af. Deze kwaliteit gedraagt zich vrijwel onafhankelijk van de gasbelasting. De spreiding van de vloeistof neemt tot het stuwingspunt enigzins toe; daarboven is er een sterkere toename waargenomen. De waarde van de radiale spreidingscoëfficiënt bedraagt ongeveer 3 mm; In het stuwingsgebied neemt de wandstroming bij een gelijkblijvende pakkingshoogte af.
De meetresultaten van de gas- en vloeistofprofielen zijn gesimuleerd met een Monte Carlo cellenmodel. Dit model beschrijft de vloeistof- en gasspreiding, de ontwikkeling van wandstroming, gemiddelde onregelmatigheden op kleine schaal en het effect van de superficiële gassnelheid.
Voor een beginverdeler is berekend dat voor een aantal sproeipunten van meer dan 200/m2 de kwaliteit van de verdeling nauwelijks verbetert. De simulatie van een wandschraper laat zien dat de wandstroming van de vloeistof slechts over een klein gedeelte van de pakking wordt verminderd.
C
In een destillatiekolom voor een methanol/ethanol mengsel is het effect van een slechte beginverdeler onderzocht. De kolom heeft een diarater van 0.45 ra en een pakkingshoogte van 2.65 m. De hoogte van een stofoverdrachtseenheid gebaseerd op de gasfase weerstand neemt sterk toe als de helft van de beginverdeler af-geblind wordt.
Het Monte Carlo cellenraodel, uitgebreid met een eenvoudig stofoverdrachtsmodel, geeft een reële beschrijving van de destillatieresultaten.
Met het model is het effect van het aantal sproeipunten op de scheidende werking van 25 mm pakkingselementen voor een kolom van 1 m hoogte gesimuleerd. Bij 30 sproeipunten per vierkante meter gaat effectief een pakkingshoogte van ongeveer 0.4 m verloren .
Gestructureerd pakkingen
De maldistributie voor de gasstroming in de bulk van de pakking is verwaarloosbaar. Alleen aan de wand treden er onregelmatigheden op. Het gas stroomt parallel aan de kanalen in de pakking. Doordat de pakkingselementen onderling verdraaid zijn treedt er een goede gasmenging op.
Bij gassnelheden groter dan 1.7 ra/s neemt de wandstroming van de vloeistof af. Onder het stuwingspunt is de maldistributie van de vloeistof constant. Daarboven neemt de kwaliteit van de verdeling sterk af, doordat er grootschalige segregatie ontstaat.
In appendix A is een methode ontwikkeld, waarmee een verdeling gekarakteriseerd kan worden met een gering aantal parameters. Met behulp van een maldistributiefactor voor verschillende kanaalgroottes kan bepaald worden welk kanaal het meest bijdraagt aan maldistributie. Een algemene maldistributiefactor geeft een goede indicatie voor de kwaliteit van verschillende verdelingen.
D
DANKWOORD
I would like to thank the Koninklijke/Shell-Laboratorium for their financial support and Norton Ltd., Raschig GmbH, Julius Montz GmbH and Gebrüder Sulzer AG for supplying us with packing.
Verder wil ik alle collega's bedanken met wie ik prettig heb samengewerkt. Speciale gevoelens gaan uit voor diegene die zich het getal driehonderdtwee-endertig herinnneren: Piet en Peter voor het 0.1 mm werk. Frits en Piet voor de electronica. Arie en Wim voor de ontwerpen. Bram en kornuiten voor de constructie. De uitvoerders van de Centrale Werkplaats.
De beide promotoren Hans Wesselingh en Jan de Graauw hebben me tijdens het onderzoek veel vrijheid gegeven. Dit vind ik belangrijk voor zowel het onderzoek als mezelf. Het zijn vier leerzame jaren geweest.
Gedurende de promotieperiode was het niet altijd even gemakkelijk. Veel ondersteuning heb ik toen gehad van de afstudeerders/stagiaires. Vooral het laatste jaar hebben zij met man (M/V) en macht gewerkt om het project tot een goed einde te brengen. Kees, Aike, Connie, Jos, Krijn, Rens, Maxim, Antonio, Manuela, Aad, Ton, Floris, Ruud, Abdel, Hessel, Jan-Jelle. Zonder jullie was het niet gelukt.
F
CONTENTS
Summary Samenvatting Dankwoord
A C F
CHAPTER I General introduction Scope Earlier investigations Objective of the thesis Structure of the thesis References
1 3 4 5 5
CHAPTER II The experimental setup Introduction Description of the equipment The raeasuring techniques used Data acquisition Characteristics of the equipment References
9 9
12 15 15 17
CHAPTER III Measurements of the gas and liquid maldistri-bution in columns with a random packing
Introduction _._ ~ Literature Velocity profiles
Gas profiles Liquid profiles
Radial spreading Gas spreading Liquid spreading
Interpretation Conclusions Symbols References
-19 19 21 23 25 27 28 28 30 32 33 34
CHAPTER IV Simulation of the gas and liquid distribution Introduction 37 Literature 38 The simulation model 39
Liquid bulk flow 40 Gas bulk flow 43 Wali flow 44
Results 45 Liquid spreading 46 Liquid profiles 48 Gas flow effect on liquid wall flow 48 Drip point density 50 Wall wiper 51
Other cell dimensions 52 Conclusions 53 Symbols 54 References 55
CHAPTER V Measurement and simulation of the influence of maldistribution on distillation in a column with a random packing
Introduction The distillation unit
HTU OG ■value Determination of the Distillation results Simulation of mass transfer Simulation results Influence of the drip point density Discussion Conclusions Symbols References
57 59 60 61 63 67 69 71 71 72 73
CHAPTER VI A study of gas and liquid distributions in structured packings
Introduction 75 Literature survey 76 Gas profiles 78 Gas spreading 80 Liquid profiles 80 Liquid spreading 84 Discussion 85 Conclusions 86 References 87
APPENDIX A Characterlzation of the flow distributions in a cross section of a packed column
Introduction 89 Channel and overall maldistribution 91 Sample distributions and discussion 93
A checkerboard distribution 93 A column with an irrigated outer ring 94 A point source 95 A series of checkerboard distributions 96
Conclusions 96 Symbols 97 References 97
APPENDIX B Description of the computer programs used General information 99 The flow simulation program 99 The mass transfer simulation program 101 The evaluation program 101
CHAPTER I
General introduction
Scope
For a long time gas liquid contact devices have been used in chemical engineering to separate mixtures. Destillation, absorp-tion and stripping are carried out in tray, wetted wall, spray and packed columns. Some typical applications are gas drying, crude oil refinery, monomer purification, alcohol separation and gas cleaning.
One of the problems for a process designer is to choose the economical optimum from the various devices. Although the knowledge on gas liquid contacting has reached a high technologi-cal maturity, innovations in equipment and widening of theoretical backgrounds still happen. Minor improvements can result in large profits becaus_e_of„the-enormous-quant-it-les—involved-
An example to emphasize the importance of innovations is given by the petrochemical industry. During the perlod between 1950 and 1973 the world refinery capacity was rapidly extended from 13xl06
to 6Axl06 barrels a day. After the first oil shock in 1973, oil product demand feil rather sharply, but primary distillation capacity kept rising because of the completion of plant already under construction. Refiners have reacted by closing the least efficiënt and simplest refineries. Due to the second oil shock in 1979 and shift in the oil product demand to light components the utilization rate in 1987 equaled about 75 per cent with low simple refining margins . The most important requirement for refiners became to load fully their complex conversion and upgrading capacity, like visbreaking, flexicoking and the hyconproces, to produce light oil products. Still the margins are small, so refiners have to optimize both supply and refinery operations. Especially the efficiënt use of energy is important, because refinery fuel and electricity costs increased their share of total manufacturing costs from around 20 per cent to over 40 per cent. The worldwide amount of crude oil processed in refineries during
2 the past ten years approximates 60xl06 barrels a day . One of the possibilities to minimize manufacturing costs is to
revamp plate columns with packings. The gain of flexibility and
1
capacity obtained in this marmer is important for the atmospheric distillation of crude oil. A second advantage is the decrease of the pressure drop per mass tranfer stage. Especially for vacuüm distillation this results in lower bottom temperatures and lower energy consumption.
Also in the chemical industry packed columns can improve the column performance. Due to the low pressure drop per mass transfer stage the decomposition of thermolabile products can be suppressed. In general the height of of mass transfer stage is lower than that of a tray tower. Revamping those towers with packing mostly results in better product specifications. For the separation of agressive chemical compounds the ceramic types of structured as well as dumped packings are very suitable.
Many applications of structured as well as dumped packings have -,_ . • -, ■ 3-17 been described in literature
More than fifty varieties of random and structured packings are commercially available on an active market. Roughly they can be divided into three types : conventional dumped packings ( Raschig rings, Pall rings, Berl saddles, Intalox saddles, etc. ), high performance dumped packings ( Intalox Metall Tower Packing, Nutter rings, etc. ) and structured packings ( Mellapak, Gempack, Montz's BI, Ralu-pak, Rombopak, Intalox 2T(C0M), etc. ). Table I outlines typical design data of trays and packings.
Table I Typical design data of trays, dumped packings and struc-18 tured packings according to Chen
F-factor [(kg/m/s2)0•*] HETP [m] AP/HETP [Pa] x 102
Trays
0.3- 2.4 0.6- 1.22 4 -11
Packings
Dumped
0.3 -2.9 0.46-1.52 1.2 -2.4
Structured
0.12 -4.4 0.1 -0.76 0.013-1.0
In the past, the use of packings was limited to columns with a small diameter/packing height ratio, because the performance was considered to be rather unpredlctable. Today this picture has been changed, mainly for two reasons: the availability of carefully deslgned and installed packings and the improved understanding of the flow mechanisms inside the packing.
2
However, there still are a number of disappointing performances of large columns. A nonuniform liquid and gas distribution within the packing is thought to have an negative effect on the separa-tion efficiency. A considerable amount of literature has been produced on the so called maldistribution problem. A general overview is given in the next paragraph. A detailed survey of the studies is presented in the concerning chapters.
Earlier investipations
Many factors that can cause irregular flows in a packed column have been investigated in the literature. The most important factors are summarized below.
A principal cause of maldistribution is the packing itself. The liquid rivulets follow specific paths through the packing. Sometimes they split, sometimes they flow together thus introduc-ing irregularities on a small scale. The equillibrium—f-low-distribution in the bulk of the packing is called natural flow
19 according to Albright . This natural flow has been measured by 20 Hoek . The continuous gas flow is forced through the openings of
the packing. The different orientations and dimensions of these openings result in a natural flow distribution for the gas, which
21 22 has been determined by Ali and Stikkelman
A change in the isotropy of the packing can give a departure from a uniform distribution. Practical examples are void varia-tions due to inproperly installed packing, corrosion, fouling, etc. A serious change in the isotropy is the transition between the packing and the column wall. Liquid moves more easily to the wall than vice versa causing wall flow. Liquid wall flow has been
20 23-30 studied by many authors. ' Gas wall flow received less 31,32 attention.
The quality of the initial distribution of both phases can contribute to a column malperformance. Especially for the liquid an initial maldistribution results in a decrease of the separation
33-35 efficiency . Large scale flow irregularities are diminished by radial spreading. Many spreading data are known for the liquid
3
20 30 36-39 without loading effects ' ' . Rough data of the gas are only 40 available for 250Y Sulzer Mellapak
The interaction between gas and liquid intensifies above the loading point. Many correlations have been proposed for the liquid hold up and the pressure drop, but only few authors studied the
32 41 effect of loading on the flow distribution '
The surface tension and viscosity of the liquid can influence the interfacial area between both phases. Even over the column length the surface tension can vary due to a change in
37 42-48 composition. Although the studies of these phenomenae ' are mostly not integrated with maldistribution, the effects on the separation efficiency can be considerably.
Obiective of the thesis
In literature little attention is given to the gas phase and its effect on the liquid phase. Therefore the two main objec-tives of this thesis are the study of:
Gas flow characteristics
The influence of gas flow on the liquid flow behaviour
Experimental data on spreading and flow profiles of both phases will be measured for structured as well as random packing in an air/water column. This information will contribute to a better understanding of the complex flow mechanism inside the packing.
A flow model will be developed to simulate the experimental results. This model can also be applied to evaluate hypothetical cases. In this way some design failures can be anticipated.
The model, extended with simple mass transfer equations, will be tested to practical distillations with severe maldistribution.
4
Structure of the thesis
In chapter 2 the measuring equipment will be decribed. The next three chapters deal with random packing. They form an
integrated unit starting with basic experimental flow characteris-tics and ending with a complex distillation simulation. The experiraental results will be presented in chapter 3, The flow simulation model will be explained and evaluated in chapter 4. A distillation on a pilot plant scale with severe initial liquid maldistribution is studied in chapter 5.
In chapter 6 the gas and liquid distribution results are given for structured packings.
The first appendix at the end of the thesis concerns a method of characterizing a flow distribution with a relatively small number of parameters. In the second appendix the computer programs used are outlined.
The greater part of the chapters have been submitted for publication. In this thesis their—lay-out—has—been-slrghtly-
modified to give a consistent and readable form.
References
1 De olieprijzen Shell Brochure Series, december 1987 ISBN 90-6644-083-x
2 Energie in kort bestek Shell Brochure Series, august 1987 ISBN 90-6644-079-1.
3 W.Meier, R.Hunkelar, W.D.Stocker I. Chem. E. Symposium Series No.56 (1979) 3.3/1-17
4 R.F.Strigle, K.E.Porter I. Chem. E. Symposium Series No.56 (1979) 3.3/19-33
5 R.F.Strigle, F.Rukovena Chem. Eng. Progr., 75 (1979) 86-91
6 G.K.Chen, L.Kitterman, J.Shieh Chem. Eng. Progr., 79 (1983) 46-49
7 N.P.Lieberman Hydrocarbon Processing, 66 (1984) 143-145
8 R.Billet, J.Mackowiak Chem. Ing. Tech., 57 (1985) 1-3
9 R.F.Strigle Chem. Eng. Progr., 81 (1985) 67-71
10 R.F.Strigle 3rd World Congress of Chemical Engineering, Tokio, 1986 Paper No. 6F-354, 770-773
11 J.R.Sauter, W.E.Younts Oil 6. Gas Journal, 84 (1986) Sept
12 P.Roy, A.C.Mercer I. Chem. E. Symposium Series No.104 (1987) A103-114
5
13 U.Bulhmann I. Chem. E. Symposium Series No.104 (1987) A115-127
14 D.E.Nutter I. Chem. E. Symposium Series No.104 (1987) A129-142
15 H.A.Gangriwala I. Chem. E. Symposiun Series No.104 (1987) B89-99
16 M.Roza, R.Hunkelar, O.J.Berven, S.Ide I. Chem. E. Symposium Series No.104 (1987) B165-178
17 L.Spiege l , P.Bomio Chem. Ing. Tech., 59 (1987) 130-132
18 G.K.Chen Chem. Eng., 91 (1984) 40-51
19 M.A.Albright Hydrocarbon Processing, 9 (1984) 173
20 P.J.Hoek Ph.D. Thesis, Technische Hogeschool Delft, 1983
21 Q.H.Ali Ph.D. Thesis, University of Aston, 1984
22 R.M.Stikkelman and J.A.Wesselingh I. Chem. E. Symposium Series No.104 (1987) B155-164
23 K.E.Porter, J.J.Templeman Chem. Eng. Sci., 20 (1965) 1139-1140
24 K.E.Porter, J.J.Templeman Trans. Instn. Chem. Engrs., 46 (1968) t68
25 E.Dutkai, E.Ruckenstein Chem. Eng. Sci., 23 (1968) 1365-1373
26 V.Stanek, V.Kolar Czech. Chem. Commun., 33 (1968) 1062-1077
27 E.A.Brignole, G.Zacharonek, J.Mangosio Chem. Eng. Sci., 28 (1973) 1225-1229
28 H.C.Groenhof, S.Stemerding Chemie-ing. Techn., 49 (1977) 835
29 M.M.Farid, D.J.Gunn Chem. Eng. Sci., 33 (1978) 1221-1231
30 P.J.Hoek, J.A.Wesselingh and F.J.Zuiderweg Chem. Eng. Res. Des., 64 (1986) 431-449
31 G.Speek Ph.D. Thesis, Technische Hochschule Dresden, 1955
32 R.J.Kouri and J.J.Sohlo I. Chem E. Symposium Series No.104 (1987) B193-211
33 M.Huber, R.Hiltbrunner Chem. Eng. Sci., 21 (1966) 819-832
34 K.J.R.ter Veer, H.W.van der Klooster, A.A.H.Drinkenburg Chem. Engrs. Sci., 35 (1980) 759-761
35 J.G.Kunesh, L.L.Lahm, T.Yanigi I. Chem. E. Symposium Series No.104 (1987) A233-244
36 K.E.Porter, V.D.Barnett and J.J.Templeman Trans. Instn. Chem. Engrs., 46 (1968) t74-85
37 K.Onda, H.Takeuchi, Y.Maeda, N. Takeuchi Chem. Eng. Sci., 28 (1973) 1677-1683
38 V.Stanek, M.Kolev Chem. Eng. Sci., 33 (1978) 1049-1053
39 G.G.Bemer, F.J.Zuiderweg Chem. Eng. Sci., 33 (1978) 1637-1643
40 W.Meier, R.Hunkler and D.Stöcker I. Chem. E. Symposiun Series No.56 (1979) 3.3/1-17
41 E.Dutkai, E.Ruckenstein Chem. Eng. Sci., 25 (1970) 483-488
6
42 F.J.Zuiderweg, A.Harmens Chem. Eng. Sci., Genie Chemiqie, 9 (1958) 89-103
43 R.C.Francis, J.C.Berg Chem. Eng. Sci., 22 (1967) 685-692
44 S.S.Paranik, A.Vogelpohl Chem. Eng. Sci., 29 (1974) 501-507
45 A.B.Ponter, P.Trauffler, S.Vijayan Ind. Eng. Chem. Process. Des. Dev., 15 (1976) 196-199
46 H.W.van der Klooster, A.A.H.Drinkenburg I. Chem. E. Symposium Series no.56 (1979) 2.5/21-37
47 H.Sipma, B.J.Schram, A.A.H.Drinkenburg I2-procestechnologie, 2 (1985) 30-33
48 T.D.Koshy, F.Rukovena Hydrocarbon Processing, (1986) 64-66
7
CHAPTER II
The experimental setup
Introduction
The equipment used in this study has to supply information on the gas flow profile and the liquid flow profile in the top and the bottom cross sections of a packed column. Both the gas and the liquid profiles have to be measured simultaneously. The measuring grid should be able to detect maldistribution on the scale of a reasonable sized column as well as on the scale of the packing element. These requirrements have lead to the design of a column with a diameter of 0.5 m and with a maximum packing height of 3 m. The column is placed on top of an apparatus containing 332 measuring modules. The column is operated at atmospheric pressure, with water flowing downwards and air upwards. ___The_—measjjring_par-t_of_the-equipmenfr;—as—fuHy^described-irT~the following paragraphs, consists of modules with a nominal diameter of 25 mm. This matches with the dimensions of 1 inch dumped packings.
The flows through the modules are collected by an automatic data acquisition system. Data reduction and interpretation are carried out on a personal computer.
Mass transfer experiments are performed on a pilot plant dis-tillation column with a diameter of 0.45 m and packing height of 2.65 m. This column is described in chapter 5.
Description of the equipment
Five different parts can be distinguished in the general flow scheme of the equipment as shown in Figure 1:
- the water circulation unit
A centrifugal water pump feeds a constant head tank. A fraction of the liquid is directed via a flow controller into the liquid distributor. The superficial velocity of the water in the column can be varied between 0 and 15 mm/s. After passing the packed
9
column and the bottom section the water runs into a buffer vessel. The overflow of the constant head tank is lead directly into the buffer vessel.
i constant head tank
air out
Inltlal distrlbutor
coollng water
air cooler
draln' t4 air pump
Figure 1 The general flow scheme of the equipment
the air supply unit
Hot air, supplied by a centrifugal blower, is conditioned to a fixed temperature and humidity by tap water in an air cooler filled with dumped packings. It enters the column via the bottom section, where the gas velocity profile is measured. The superfi-cial air velocity is adjustable between 0 and 3.4 m/s by means of a valve. From the top of the packed column the air is vented into the surroundings.
10
Suppor t g r
L/G sepo.ro.tl
G detec t lon 8< Input
f low d e t e c t o r
yfflEL O ( level detector
level detector
Figure 2 The bottom section with an enlargement of one flow detec-tion module
- the bottom section
The bottom section ( Figure 2 ) combines a number of functions: The upper part ( L/G separation section ) serves as a support
grid. for the packing in the column. This grid divides the cross section of the packed column into 332 modules of 25x25 mm2. Some of the outer modules are partly covered by the column wall. Furthermore, the downcoming water flows around the gas pipes into tubes in the middle part of the bottom section and is in this way separated from the upflowing air.
In the middle part ( G detection & input section ) the gas flow from the cooler is forced through a perforated plate with a rela-tive high pressure drop to obtain a uniform initial gas distribution. The air enters the L/G separation section via 332 small pipes, each of them provided with a flow detector.
11
Water from the L/G separation section falls without inter-ference by air via 332 pipes located in the G detection & input section into square U-tubes in the lower part ( L detection section ). Each of these tubes contains two level sensors and a pneumatic actuated valve. After the filling time between the two levels is registered, the valve can be opened to drain the U-tube and to prepare a new measurement.
- the packed column
The column is built from perspex units with a height of 1 m and a diameter of 0.5 m. Many types of structured as well as random packings, provided by manufacturers, are available.
- the liquid distributor
The distributor, situated above the packing, consists of a hollow plate, perforated with vertical gas tubes and provided with drip points. The initial distribution can be adjusted from 149 drip points ( 760 dp/m2 ) down to one point source with all possibilities in between. A grid of 293 gas tubes, each of them con-taining a gas flow detector, enables the gas to flow out of the packing with a low resistance. A vertical cross section of the distributor is shown in Figure 3.
Figure 3 The liquid distributor
Liquid Gas
t
T T T
A
The measuring techniques used
Gas flows are measured by means of miniature NTC ( Negative Temperature Coëfficiënt ) resistors. The resistance of these
12
devices is dependent on their temperature in such a way, that it decreases when the temperature increases. The resistors are heated by an electrical current so that their temperature is about 180 °C in the absence of gas flow. However, if there is a gas flow, the resistor is cooled by the gas, resulting in a change of temperature and a corresponding change of resistance. The NTC resistor is part of a Wheatstone bridge ( Figure 4 ). The relation between the measured bridge voltage and the gas velocity is given by:
U - A + B x exp(AV) (21)
For each resistor the constants A and B are calculated from calibration measuments with known gas velocities. A typical calibration curve is given in Figure 4.
150 ,15 V
680 n I i ,, . AV . __ !
27 k A
2.7 k / \
TT _i i i i—i i_
4.5 5.5 6.5 Voltage (V)
Figure 4 Response of an NTC resistor in a Wheatstone circuit as a function of the gas flow through one sensor
The relative error with a reliability of 95 % is smaller than 2 X. The measuring system responds within a few seconds, thus measurements can be done almost instantaneously. As the detection of the gas flow depends on the cooling of the resistor, the gas must have exactly the same temperature as during calibration. A
13
convenient temperature is 25.0 °C. The consequences of raeasuring at a different temperature are negligible for differences smaller than 0.2 °C. It turned out to be easy to maintain a gas temperature of 25.0 °C. No deviations were found in the resistor signal under fixed conditions for a period lasting more than six weeks. However, mechanical and electrical shocks should be avoided.
9.6 V
1 k.TL 1 e m l t t e r
l 4. Y -"*+ /A -f- *^^^\
-10 V
i 10 k A
7T g AV ? f •
recelver
Figure 5 The Wheatstone circuit for a liquid level detector
Liquid flows are measured by means of U-formed tubes. Each tube contains a level sensor at the top and a level sensor at the bottom. A sensor consists of a pair of diodes: one LED emitting infrared light and one photodiode whose resistance varies with the amount of infrared light received. This photodiode is included in a Wheatstone bridge ( Figure 6 ) in the same way as the NTC resistor. The absorbance of infrared light in air is different from that in water. When water passes the sensor, a change in the bridge voltage is detected, and the time is registered by means of a computer. At the end of a measurement the liquid flow for each element is computed by dividing the tube volume by the time dif-ference between top and bottom sensor.
For the determination of gas spreading C02 tracer gas is In-jected in the bottom of the packing. The C02 concentrations leaving the packing are measured by an analyzer based on infrared light absorbtion. This analyzer works in a range from 0 to 0.3 vol-%, so the amount of tracer gas required is acceptable.
14
Data acouisition
There are 1289 sensors in the equipment, so an automatic system is indispensable for collecting and processing data. All signals from the sensors are directed to an analogue multiplexer, made out of ordinary CMOS-switches and some address decoding logic. An address, generated by an Olivetti M24 personal computer, selects one of the signals and connects it to an analogue-to-digital converter, which is placed on a LabMaster I/O expansion board. The resulting digitized number is, after some checking and converting, stored into memory and another sensor can be selected. In this way all sensors are measured in sequence.
The entire initial gas distribution, or final gas distribution can be detected ( each NTC is scanned five times ), processed and saved on a floppy disk in about 5 seconds. The time needed to find the distribution of the liquid dependents on the superficial liquid velocity. Typical spans range from 10 to 60 minutes. During the measurement each photodiode is scanned 10 times per second, which—means—that—the—aceuracy—of—f ill—time-determination_is-_quite adequate.
Characteristics of the equipment
For a better understanding of the influence of the gas flow in a packed column an even initial distribution is desirable. From the initial gas distribution measurements without packing it is concluded that for the range of gas flows used the maldistribution is negligible. The Mf-factor as defined by Groenhof equals 0.005.
The variance of the initial liquid distribution, based upon 149 drip points does not depend on the liquid flow rate ( Figure 6 ). The maldistribution of the distributor is so low, that it has no negative influence on Mf-measurements of packings. An inclination of 3° has no effect on the quality of the distribution, except for very low liquid flows. However, the results are still acceptable.
A low pressure drop over the L/G separation section of the bottom section is important, otherwise the gas velocity profile caused by this part dominates the profile caused by the packing. For the same reason a low pressure drop over the liquid distributor is desired. The pressure drops of interest are shown with those of Sulzer 250Y and 25 mm metal Pall rings in Figure 7.
15
O, 0.06 O Ö tö
• f—I
cd 0.04 >
> •i—i
cd 0.02 I—I
0.00
- O
- I 1 1 1 - - i 1 1 1 1 1 r-
* : horizontal O : inclined
* * * * * _i i > i i i i i i i 1 1 1—
10 15 Liquid Velocity ( m m / s )
Figure 6 The variance of the flows through 149 drip points as a function of superficial liquid velocity for a horizontal and a 3° tilted liquid distributor
1000
(Ö Pu
OH O
Q 500
u Vi m <D i-,
CU
' 1
-
.
y s^
-~~~**t^ ^*"~ï"""~^ —
i
Pall /
1 i ~~
' / '
250Y
—" " ,
Init
-
■
^
■
^ ^ . — ■ — :
Bot
0 1 2 3 4 Gas Velocity (m/s )
Figure 7 Pressure drop of the equipment and over 1 m of Sulzer Mellapak 250Y and 25 rara metal Pall rings as a function of superficial gas velocity. ( Init-liquid distributor, Bot=L/G separation section ).
16
These are typical examples of the structured and dumped packings used in this study.For 25 mm dumped packings the differences between measured gas flow profiles and ideal profiles can be atttributed to the packings. For structured packings the measured profiles will be slightly smoothed by the measuring section.
Flooding in the upper part of the bottom section does not occur for the range of gas and liquid velocities used in this study.
References
1 H.G.Groenhof Chem. Eng. J., 14 (1977) 193
17
CHAPTER III
Measurements of the gas and liquid maldistribution in columns with a random packing
Introduction
Research on maldistribution in packed columns has up to now almost only been concerned with the liquid distributlon. The characteristics of the gas as well as its influence on the liquid has received little attention. However, many authors emphasize the
1 2 need of research in this field ' 3 The experience at our laboratory with the liquid has therefore
been extended with gas flow measurements. The equipment has been designed to study the flow of gas and liquid simultaneously on the scale of a packing element. A wide range of loadings can be ap-plied in a column with a diameter of 0.5 m and a packing height up to 3 m. The cross section at the bottom of~the packing is-divïdêd into 332 measuring modules of 25x25 mm2. Each of these modules is provided with liquid and gas flow detectors. The liquid dis-tributor contains 149 drip points and 293 gas flow detectors. The equipment is described in detail in chapter II of this Ph.D. thesis.
The aim of the present work is to determine the maldistribution of the gas and the liquid and their influence upon each other below and above the loading point for the packings as shown in Table I. The shape of the velocity profiles of a cross section of the column and the mixing of both phases will be discussed in terms of wall flow, maldistribution and radial spreading.
Literature
Flow irregularities in the gas distribution can result in a disappointing performance of a packed column.
4 Moore and Rukovena found that the initial gas maldistribution
is a function of the kinetic energy of the inlet gas, the pressure drop in the packed section, and to a lesser extent, the distance between the gas inlet and the bottom of the packed bed. According to Ali , a severe maldistribution at the gas inlet is converted to
19
almost uniform bulk flow within one-half a column diameter. A high pressure drop packing is better with respect to gas redistribution than a low pressure drop packing. Measurements on a small scale showed that flow deviations at the top of a deep bed provided with an elaborate gas distribution system under the bed, are negligible compared with those of the liquid. This observation was also made by Stikkelman and Wesselingh .
An excess of gas flow near the wall was observed by Speek . Q
Krebs modelled the wall flow in columns filled with 15 mm ceramic Raschig rings by taking a bundie of channels with unequal widths, allowing complete mixing between each layer of packing. In a tower
9 of 10.2 cm diameter with 1 cm glass Raschig rings Spedding et al. absorbed ammonia into water. The gas wall flow, combined with the liquid wall flow, resulted in a radial gas concentration profile at the top of the packing. Gas wall flow is also found in beds packed with catalytic particles. Chourhary et al. used two sizes of particles ( 1/16 and 1/8 inch ) to build packed beds with a high resistance core or annulus. Measured distributions were simulated with a vectorial form of the Ergun equation.
Kouri and Sohlo used an 500 mm diameter column fitted with special top and bottom sections for measurement of gas and liquid flow through five or six rings. The results of the gas flow distribution measurements showed that with good initial distributions of the liquid and gas, the gas bulk flow through plastic Pall rings may be expected to be quite uniform and independent of packing height and flow rates. The quality of the liquid distribution for 25 mm Pall rings was said to tend to deteriorate at high gas loads.
12 Baker et al. observed, that the gas flow hardly effects the liquid distribution, even near the loading point. Above the load-ing point, it assists in obtaining a uniform liquid distribution.
13 Dutkai and Ruckenstein came to the same conclusion for Raschig rings and Intalox saddles. Their diffusion model was valid up to 703! of flooding without adjusting the radial spread factor for the liquid and the wall flow parameters. At higher gas loa-dings the radial spread factor increases while the wall flow decreases. Just the opposite occurs when using cocurrent gas flow for 6 mm Raschig rings and Berl saddles .
Stichlmair and Stemmer did not measure flow but temperature profiles. Starting with a good initial distribution, they found
20
that the largest deviations in temperature are located half-way up the packing.
3 From the literature on the liquid phase (discussed by Hoek and recently by Porter ) a nuraber of conclusions can be drawn:
o 1 fi _ 1 ft The maldistribution on the scale of a packing element '
may be considered as an inherent and stabile property of the 19 packing. Albright denoted this by the terra natural distribution
and concluded that an initial distribution that is better than the 20 natural one will degrade to it quickly. Zuiderweg calculated
that the natural distribution appears to have only minor effects on the basic separation efficiency of the packing, due to the influence of radial mixing.
The spreading of liquid in the absence of gas has been studied 21 by many authors, dating back as far as 1893 . Their results can 22-27 be decribed as a random movement of the liquid or as rivulets
3 20 28-31 following specific paths in the packing ' ' . I n either description, a parameter, D , having the unit of length is used in combination with a diffusion like equation assuming axial
~symmëtry:
g £ ^ - D x ( °2f(Z.r) + df(z,r) dz r dr2 rdr
The wall flow, caused by a change in the isotropy of the packing near the wall, is said to find its origin in a difference between the liquid flow towards the wall and the liquid flow from
9fi 97 the wall ' . Solving equation (1), using different boundary conditions for the wall and centre of the column ' ' results in relations between wall flow, initial distribution and packing height.
In summary: there are very few studies in which the mutual influence of gas and liquid have received attention. The studies
11 13 of Kouri and Sohlo and Dutkai and Ruckenstein were based on only a few sampling areas. Furthermore gas spreading data are missing.
Velocitv profiles
The velocity profiles of the gas and liquid, using various packing heights and superficial velocities, have been measured for the packings presented in Table I. Two typical phenomena can be
21
recognized in such a profile as shown in Figure 1: an irregular bulk flow and wall flow.
Table I The types of dumped packings used.
Type
Pall Ring Ralu Ring Ralu Ring Ralu Ring Ralu Ring Torus Saddle Torus Saddle IMTP
Size [mm]
25 25 25 38 38 25 25 25
Material
Stainless Steel Plastic Hydrofilated Plastic Plastic Hydrofilated Plastic Plastic Hydrofilated Plastic Stainless Steel
Code
PR25S R25P R25HP R38P R38HP T25P T25HP IMTP
Figure 1 A three dimensional presentation of a liquid profile showing an irregular bulk flow and pronounced wall flow.
22
The maldistribution factor, Mf, is used to characterize the 32 bulk flow. According to Groenhof this relative factor can be
expressed as:
1 i?n ( u(i)-<u> ) 2
n i=l <u>2 Mf- -^X^fer^- (2) The Mf-value depends on the scale of detection of the local velocities. In this work the scale is based upon the dimensions of the measuring module, which matches the nominal size of most packings used. Furthermore the Mf takes only the variance of the local velocities and not their spatial orientation into account.
33 An alternative method can be applied that overcomes the disad-vantages mentioned above.
The wall flow factor, Wf, is calculated from the average velocity, <u > in a ring adjacent to the wall for both the gas and w the liquid. The ring is chosen to have a thickness of one half of the nominal packing diameter ( •? d ) . This wall flow velocity is compared with the superficial velocity:
<u > W f - - ^ - _ (3)
In the ideal case of plug flow the Mf-value equals 0 and the Wf-value is 1.
Gas profiles
A typical gas velocity profile ( Figure 2 ) shows a smooth bulk flow, with Mf -values smaller than 0.03, and a wall flow, with Wf -values between 1.1 and 1.3. Only for the torus saddles is t, G this value higher. The experimental results for the bottom and the top of the packing are summarized in Table II.
The influence of the gas velocity on the gas wall flow at the bottom of the packing has been investigated with a superficial liquid velocity of 3.4 mm/s. The column contained 1.72 m of IMTP packing. Over a gas velocity range from 1.5 to 3.9 m/s the Wf, -values varied randomly between 1.35 and 1.41. These values approximate the value of the initial gas distribution without packing. It is assumed that the wall profile caused by the measuring equipment determines the Wf, -value.
D , CF
23
r 4
-0.25 0.00 Radius (m)
0.25
re 2 A typical gas velocity profile.
e II Wall flow values at the top, Wf and raaldistribution t, G
factors at the bottom, Mf, „, and at the top, Mf „, for b,G t,u
the gas phase and at the bottora, Mf, , for the liquid phase.
Code
PR25S R25P R25HP R38P R38HP T25P T25HP IMTP
INITIAL GAS
Mf b,G
0.022 0.007 0.030 0.023 0.019 0.018 0.025 0.003
0.011
Mf t,G
0.030 0.026 0.027 0.023 0.014 0.021 0.031 0.004
Wf t,G
1.2 1.3 1.3 1.2 1.2 1.7 1.5 1.1 ...
Mf b,L
0.81 0.55 0.63 0.69 0.81 1.05 1.13 0.57
24
At the top of the packing the Wf -values depend on the type of packing. No significant effects of the liquid load were found up to a superficial gas velocity of 2.5 m/s. Care should be taken to prevent irregularities in packing height, otherwise gas flow channels may occur in which packing particles are lifted.
Liquid profiles
Liquid velocity profiles are less uniform than gas profiles. Small scale maldistribution of the bulk flow is characterized by Mf, -values between 0.5 and 0.8. Again the torus saddles give b,L
higher values of about 1.1. It was found that the gas velocity hardly influences the small scale maldistribution, except for situations close to flooding. At low liquid velocities the quality of the distribution slightly deteriorates ( Figure 3 ). Average Mf, -values for different packings are given in Table II. b, L
Mf
tJ.U
1.5
1.0
0.5
n n
A A
s o
A
?
— 1
A
8 *
A
0
<w
o o
Liquid Velocity ( m m / s ) 10
Figure 2 The maldistribution factor, Mf, of the liquid as afunc-tion of the superficial liquid velocity for 1.72 m of T25P ( ,;, ) , T25HP ( * ) , IMTP (o ), PR25S (<? ) and R25P (o) packings without gas loading.
25
The wall flow tendency at the bottom of the packing is quite remarkable as shown for R25HP packing in Figure 4. For a gas velocity of 0 m/s the wall flow increases rapidly going downwards in the column as a function of the packing height. Starting with an Wf -value of 0.53 at the liquid distributor the Wf developes to 2 for a packing height of 1.72.
Wf
' 1 -
(5
o o <i
i — i 1 —
O
©
o
i , i
■
© ©
Initial
-
-
0 0 0
distributor
0 1 2 3 4 Gas Velocity (m/s )
Figure 4 The development of the liquid wall flow, Wf ,as a function of the gas velocity for a packing height of 0.21 ( e ) , 0.43 ( (?) , 0.86 (of) and 1.72 (©) m of R25HP rings. The superficial liquid velocity is 3.4 mm/s.
The gas velocity stimulates or reduces the wall flow, depending on the height of packing used. At a height of 0.21 m the wall flow increases at higher gas flow rates where as at a height of 1.72 m to wall flow diminishes. Almost all packings show the same behavióur. The Wf, -values of these packings are presented in Figure 5 as a function of u for a packing height of 1.72 m.
G
26
Wf
3^r
2H
1 -
- i 1 1 r
£
# * O O ° < è *
1 2 3 Gas Velocity (m / s )
$_
Figure— 5-The^development— of— the—l-i-qu-id-wall-flow.,_Wf.as __a_func_t.ion. of the gas velocity for a packing height of 1.72 m of PR25S (*), R25P (O), R25HP (•), R38P (□), R38HP (■), T25P (A), T25HP ( O AND IMTP ($) packings. The super-ficial liquid velocity is 3.4 mm/s.
Radial spreading
Radial spreading coefficients have been determined for the gas as well as for the liquid under various loading conditions. Solving equation (1) for a point source with an infinite column radius results in:
f (z,r) 4TTD Z r
exp(- 4D z r
W
where Q is defined as the total flow rate of the point source. The flow rate passing through a circular area, Q , with radius rx
x' is obtained by integration of (4):
r 2?r x
r ,z x -n o o
f(z,r)rdrdfl - Qx(l-exp(-
27 4D z r
-)) (5)
The value of D can be determined from experiments. Measured profiles are fitted to the following expression using Standard statistical methods:
- r2
Dr = 4zxln(l-Q ~^/W (6)
x'
Gas spreading
A point source with a high gas flow cannot be applied to measure gas spreading as with a liquid. A horizontal pressure gradiënt will cause an rapid redistribution effect within a few decimeters of packing. This is not representative for the rest of the packing. This effect can be overcome by using a homogeneous initial distribution of gas, and a point source of tracer gas. The concentration profile which leaves the top of the packing provides information for the calculation of D .-values. The flow rate in
r,G equation (6) should then be substituted by the flow rate of tracer gas.
Carbon dioxide is introduced at the center of the bottom of the packing via a vertical pipe of 10 mm diameter. The concentration profile is measured at 49 points, located at 4 axes on top of the packing, covering the whole cross section. A packing height of 0.9 m was used; with this height the tracer gas does not reach the wall. The relative error for the D „-values is less than 10 %.
r ,G The results of experiments with various gas velocities indicate
that the radial spreading coëfficiënt is almost independent of the superficial gas velocity. Liquid loading has a small positive effect on D „as shown in Tabel III. r ,G
Liquid spreading
A water flow rate of 6.5 l/min is carefully fed via a jet nozzle with a diameter of 10 mm into the center of the top of the packing. A smooth countercurrent gas flow is introduced at the bottom of the packing with various superficial velocities up to 3.2 m/s. The height of packing is chosen to be 0.86 m to avoid wall effects.
28
Table III Gas spreading factors, D , for several dumped pac-r, o
kings with and without liquid loading
Type of p a c k i n g
PR25S R25P R25HP R38P R38HP T25P T25HP IMTP
u . = 0 mm/s
D r , G
[ * ]
0 . 0 0 3 8 0 . 0 0 3 8 0 .0037 0 .0036 0 .0034 0 . 0 0 4 0 0 . 0 0 3 5 0 . 0 0 1 8
u = 3 .4 mm/s
°r,G [ m ]
0.0041 0.0039 0.0040 0.0037 0.0034 0.0045 0.0042 0.0027
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 6 6 5 0 0 0 0 0 0
0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0
0 0 0 0 0 s 10 0 0 0
0 0 0 0 0 0 0 o o o o o o o o 0 0 5 0 0 0 1 1 5 o o o o o o o o 8 5 0 10 0 12 5 7 0 0 22 19 11 0 0 14 0 0 0 0 14 17 6 7 0 0 10 11 34 11 9 9 19 52 14 28 20 1 6 27 6 15 91 0 15 45 41 5 0 13 23 49 11 36 48 6 22
10 15 0 19 14 13 0 7 5 0 0 0 0 0 0 0
0 24 29 15 28 22 0 0 13 0 0 11 11 0 11 12 15 0 0 27 18 0 15 9 0 0 0 0 0 1512 0 0 6 0 7 0 0 15 0 0 0 0 0 6 0 0 0 0 0 0 0 34 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 14 15 23 5 0
0 0 0 0 0 5 0 7 0 11
25 10 0 8 0 6 7 0 0 0
0 0 12 0 0 0 0
0 0 0 0 0 0 0 5 0 6 0 6 0 0 0 0 0 0
0 0 0 0 0 0 0 0 E 0 14 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Figure 6 An example of a liquid velocity distribution (m/sxl04) resulting from a point source after 0.86 m of IMTP packing.
29
An example of a liquid flow distribution ( Figure 6 ) demonstrates the irregularity on a small scale. However, the spreadings are quite reproducible. Even after redumping of the packing the values seldom change by more than ten percent.
The dependency of the liquid spreading factor, D , of all r, L
types of packing upon the superficial gas velocity is presented in Figure 7. All the relations show approximately the same behaviour: The D -value shows a small increase up to a certain gas velocity, above which the spreading increases more rapidly.
0.010
0.005
Q
0.000 1 2 3
Gas Velocity (m/s )
Figure 7 The liquid spreading coëfficiënt, Dr, as a function of the gas velocity for PR25S ( O ) , R25P ( O ) , R25HP ( • ), R38P ( o ) , R38HP ( ■ ) , T25P ( A ) ,. T25HP (*) AND IMTP ( « ) packing.
Interpretation
The results for the gas bulk flow confirm earlier findings ' ' , that small scale maldistribution is not important. The wall flow, however, can be serious especially for the liquid. When it is assumed that a flow equilibrium is established within a few decimeters of packing , the differences in the packing size and form are the most important factors that fix the Wf -value. As an example the IMTP packing and Torus saddles differ in size,
30
although they are denoted with the same nominal diameter. An IMTP element fits in a box with the dimensions of 28x23x14 mm3, whereas a Torus saddle is enclosed by the dimensions 53x28x23 mm3. Therefore an IMTP element is better in filling the non-isotropic zone between bulk and wall and thus yields in a lower Wf -value.
The value for the gas spreading coëfficiënt of the packings tested is around 4 mm. Only the value for the IMTP packing dif-fers, which can be explained as mentioned above in combination with the high porosity of the IMTP packing.
The measured maldistribution of the liquid without gas loading 3 is consistent with the results of Hoek . The increase in mal-
29 distribution for low liquid loadings as found by Bemer has also been observed for the packings used in this work. The bulk distribution quality is almost independant of gas loading.
In the loading zone the flow near the wall is influenced by the gas. The equilibrium value Wf becomes lower. This is illustrated for 25 mm metal Pall rings in Figure 8. At high gas velocities more liquid is transported in a radial direction. Due to the relatively higher gas flow in the wall zone, the returning mechanism from wall to bulk is also stimulated. The higher ex-change rate between wall and bulk leads to different Wf -values.
A striking exception is given by the 25 mm plastic Torus saddle. This saddle has a form which gives only a small contact length with the wall and its material is hydrofobic. Water piek up from the wall is therefore reduced . The hydrofilated version, T25HP, has better wetting capabilities, and thus enables more back flow to the bulk.
The decline of the wall flow in the loading region might be an explanation for the minimum in the HETP-curves for several pack-
34 ings in columns with a relatively small diameter . In literature the minimum is said to be caused by a higher interaction between the gas and the liquid, but no proof is given. The absence of the
35 minimum in columns with a relatively large diameter is another support for the wall effect, because the larger the diameter the less it effects the overall performance.
31
3 |— 1 1 1 1 1 1 r
2 -
Wf . +
1 " $
o I 1 1 1 1 1 1 ■ 0.0 0.5 1.0 1.5 2.0
Packirig Depth (m)
Figure 8 The liquid wall flow, Wf, as a function of the packing depth of PR25S. The superficial gas velocity is 0 (+) or 2.54 (x) m/s and the superficial liquid velocity is 3.4 mm/s.
Conclusions.
A packed bed can be divided in a bulk and a wall zone. In the bulk zone a poor initial distribution of the liquid has
to be considered as the main cause of a possible malperformance of a column. The gas profile plays a secondary role. The gas flow distribution is smooth and much better than that of the liquid. Moreover, maldistribution at the inlet is converted to almost uniform bulk flow within a small packing height.
The maldistribution effect on the scale of a packing element is largely compensated by radial mixing of gas and liquid. However, on a large scale the radial mixing is small for the liquid ( D = 0.3 mm ) as well as for the gas ( D = 0.4 mm ). In r,L r,G practical applications large scale irregularities will persist in the packing for a long distance.
Both phases show a deviation from the average velocity in the wall zone. The liquid wall flow rate develops from the initial
32
wall flow rate to an equilibrium value, which is higher than the average flow rate. This development is accelerated by higher gas velocities, but the equilibrium value is lower. The relative gas wall flow rate is independent of gas or liquid flow and equals 1.1-1.7 times the superficial velocity.
The increase in the spreading coëfficiënt of gas and liquid and decrease of liquid wall flow in the loading region could be an explanation for the minimum in HETP as observed in small columns.
The data as presented in this article can be used to model gas and liquid flow patterns below and above the loading point, in-cluding wall flow , maldistribution and radial spreading.
Acknowledgements
We would like to thank Norton Ltd. and Raschig GmbH for supply-ing us with packing, the Koninklijke/Shell-Laboratorium for the financial support and all students involved for their zest of work.
Symbols
d diameter [m] f(z,r) velocity at position (z,r) [m/s] D radial spreading coëfficiënt [m] Mf maldistribution factor [-] n number of samples [-] Q total flow rate of the point source [m3/s] Q the flow rate passing through a circular
x' area with radius r at a depth z [m3/s]
r radial coordinate [m] Wf wall flow factor [-] u superficial velocity [m/s] z packing height/depth [m]
subscripts: b bottom of packing G gas phase L liquid phase p packing element
33
t top of packing w wall => at infinite height/depth
Greek symbols: 9 polar angle [rad]
References
1 K.E.Porter and M.C.Jones I. Chem. E. Symposium Series No.104 (1987) A245-258
2 F.J.Zuiderweg I. Chem. E. Symposium Series No.104 (1987) A589-596
3 P.J.Hoek Ph.D. Thesis, Technische Hogeschool Delft, 1983
4 F.Moore and F.Rukovena Chemical Plants & Processing, No.8 (1987) 11-15
5 Q.H.Ali Ph.D. Thesis, Univesity of Aston, 1984
6 R.M.Stikkelman and J.A.Wesselingh I. Chem. E. Symposium Series No.104 (1987) B155-164
7 G.Speek Ph.D. Thesis, Technische Hochschule Dresden, 1955
9 C.Krebs Chem. Eng. Process., 19 (1985) 129-142
9 P.L.Spedding, M.T.Jones and G.R.Lightsey Chem. Eng. J., 32 (1986) 151-163
10 M.Choudhary, J.Szekely and S.W.Weller AICHE journal, 22, No.6 (1979) 1021-1032
11 R.J.Kouri and J.J.Sohlo I. Chem. E. Symposium Series No.104 (1987) B193-211
12 T.Baker, T.H.Chilton and H.C.Vernon Trans. Am. Inst. Chem. Engrs., 31 (1935) 296
13 E.Dutkai and E.Ruckenstein Chem. Eng. Sci., 25 (1970) 483-488
14 G.Baldi and V.Specchia Ing. Chim. Ital., 12 (1976) 107-111
15 J.Stichlmair and A.Stemmer I. Chem. E. Symposium Series No.104 (1987) B213-224
16 P.J.Hoek, J.A.Wesselingh and F.J.Zuiderweg Chem. Eng. Res. Des., 64 (1986) 431-449
17 B.Lespinasse and P.le Goff Rev. Inst. Fr. Pét., 17 (1962) 1,21,41
18 H.C.Groenhof Ph.D. Thesis, Rijksuniversiteit Groningen, 1972
19 M.A.Albright Hydrocarbon Processing, 63 (1984) No.9 173-177
20 F.J.Zuiderweg and P.J.Hoek I. Chem. E. Symposium Series No.104 (1987) B247-254
21 F.Hurter J. Soc. Chem. Ind., 12 (1893) 227
22 A.M.Scott Trans. Ind. Che. Engng., 13 (1935) 211
23 R.S.Tour and F.Lehrman Trans. Am. Inst. Chem. Eng., 40 (1944) 79
24 Z.Cihla and O.Schmidt Coll. Czech. Chem. Comm., 22 (1957) 896
25 K.E.Porter and M.C.Jones 34
Trans. Inst. Chem. Eng., 41 (1963) 240 26 E.Dutkai and E.Ruckenstein
Chem. Eng. Sci., 23 (1968) 1365 27 V.Stanek and V.Kolar
Distribution of liquid over a random packing I to X : I Coll. Czech. Chem. Comm., 30 (1965) 1054-1059 X Coll. Czech. Chem. Comm., 42 (1977) 1129-1140
28 K.E.Porter Trans. Inst. Chem. Eng., 46 (1968) T69
29 G.G.Bemer and F.J.Zuiderweg Chem. Eng. Sci., 33 (1978) 1637
30 E.A.Brignole Chem. Eng. Sci., 28 (1973) 1225
31 P.J.Hoftyzer Trans. Instn. Chem. Engrs., 42 (1964) T09-117
32 H.G.Groenhof Chem. Eng. J., 14 (1977) 193
33 R.M.Stikkelman, L.Feenstra, J.de Graauw and J.A.Wesselingh Submitted for publication in Chem. Eng. Res. Des.
34 K.Y.Wu and G.K.Chen I. Chem. E. Symposium Series No.104 (1987) B225-245
35 F.J.Zuiderweg personal communications
35
CHAPTER IV
Simulation of the gas and liquid distribution in a column with a random packlng
Introduction
For the large scale application of random packings a good understanding of the liquid flow in the column is desirable. The effects of gas loading on the liquid distribution have been studied experimentally in a water/air column with a diameter of 0.5 m. The column has been equipped with 332 flow detectors for the liquid as well as for the gas to determine a precise image of the flow distributions. The results were evaluated in terms of a maldistribution factor Mf, a wall flow factor Wf, and a radial spreading coëfficiënt D . These global parameters give an indica-tion of the small and large scale maldistribution and the spreading capabillties of the packing.
Flow patterns can be simulated using the diffusion equation. However, such equations cannot simulate small scale maldistribution. Also it is difficult to stipulate a good set of boundary conditions. These shortcomings can be avoided with cell models i.e. the column is built up with a three dimensional net-work of coupled cells.
In this paper a Monte Carlo cell model is presented which is capable of simulating small scale maldistribution of the liquid, wall flow and spreading of both phases. It includes loading effects. After a survey of the literature on flow modelling the basics of the cell model will be explained. The Mf, Wf and Dr
parameters will be transformed to local parameters for a unit cell. The model will be applied to the simulation of profiles as determined experimentally in the water/air column. Furthermore, the effect of the number of drip points per square meter and the effect of a wall wiper on the quality of the distribution will be discussed.
The model is not llmited to speclfic sizes of the cell or the column. A method to extend the model to any cell dimension will be presented at the end of the paper.
37
Literature
9 -1 ft Several workers have investigated the liquid flow in packed columns and tried to model the distribution of liquid.
In most of the slmulation models the liquid is thought to flow as a continuüm or to flow as rivulets, following speciflc paths through the packing elements. Both mechanisms lead to a diffusion equation analogous to Fick's Law:
df^£) . D x ( *ïg£l + m^ïl ) (1) dz r dr2 rdr
D -values have been calculated from spreading experiments using r. n n. ., . . . . n _,. .. . 6-8,11,12,14-19 a single liquid jet as ïnitial distribution
Equation (1) has been solved for several types of initial dis-20 tribution with a radial symmetry. A survey is given by Prchlik
The boundary condition for the liquid flow near the wall is an 4 important factor of the diffusion based model. Cihla and Schmidt
considered the wall of the column to be a total reflector where liquid is forced back into the packing. This allows the construction of a model of finite dimensions, but it does not predict wall
9 flow. Porter assumed that the wall flow is proportional to the liquid flow rate in the bulk of the packing near the wall. The wall flow was found to be over-estimated for small depths of packing. Dutkai and Ruckenstein introduced an annulus near the wall with a thichness S. They assumed that the penetration of the liquid into the wall region cakes place via an "adsorption-
17 14 desorption" mechanism. Onda et.al and Brignole et.al presented independently a mechanism which is based on the difference between the equilibrium wall flow rate and the actual wall flow rate. Stanek and Kolar introduced a radial transfer coëfficiënt and a distribution coëfficiënt, which denotes the ratio between wall flow and bulk flow at an infinite packing depth.
Another way to simulate the liquid distribution is given by 21 Jameson . He divided the column into layers. Each layer consists of concentric rings. The thickness as well as the height of a ring is equal to the diameter of a packing element. The spread of liquid is controled by a dimensionless factor P and weighted to the circumference of a ring. Although the model predicts the same tendencies as the diffusion models, no complex mathematics is required.
38
Albright and Hoek et.al used a random number numerical model which is not based on diffusion theories. It accounts for the flow onto and from each individual piece of packing. Every packing tends to a "natural" flow distribution i.e. the equi-librium flow in the bulk of the packing without external effects.
The models based on the diffusion equation are capable in solving radial symmetrie distribution problems for the liquid flows. However, they need complex mathematics and do not simulate small scale maldistribution. The numerical method of Jameson is more powerful, but it does not take small scale maldistribution into account and it contains some simplifications. The random number numerical method simulates small scale maldistribution, but misses links to other packing characteristics.
1 2 Gas flow data have only been published recently ' , and model
simulations of the gas have not yet been reported.
The simulation model The aim of the present model is to predict all the observed
characteristics of the liquid and gas flow patterns. This implies simulation of spreading and wall flow of both phases and of small scale maldistribution of the liquid. The simulation of small scale maldistribution of the gas will be omitted because it was found to be negllgible.
The model consists of an orthogonal network of stacked layers of small cells. Each cell has a width, w, of 25 mm and a variable height, h. The height will be chosen in such a way as to produce a proper liquid spreading. The flow model for one cell is outlined in Figure 1.
39
ïsfe i 1 ■ > O s i ^
<1-SG)*G*
<I-SL)*L-
:z_
► 0 .E5*SQ*G gas spl l t t lng
I \ - S L « L liquid spllt-tlng
I O Figure 1 The flow splitting model for liquid (L) and gas (G) .
Liquid bulk flow
Small scale maldistribution is introduced by a random splitting mechanism in a cell. A fraction ST, of the liquid flow L through a cell, leaves the cel randomly to one of the four neighbours of the cell below. The rest of the liquid, (1-S.) L is conducted to the underlying cell. A smoothing effect is caused by the splitting of the liquid: deviations with a relative high flow are reduced with each split. On the other hand maldistribution is promoted by coalescense of the flows: it may happen that a number of side-way flows melt together in one cell, introducing a high flow.
With this model a theoretical prediction of the relative variance, also called the maldistribution factor Mf.for "natural" flow has been calculated by simulations using various S -values and various layer dimensions. From these calculations two conclu-sions can be made:
The calculated Mf values show some scatter. The scatter depends on the number of cells in a layer. The higher the number of cells the less the scatter. For a layer with about 300 cells the calculated Mf-values fluctuate within 10 % of the
40
average value. The average Mf-value is independent of the number of cells in a layer. The average maldistribution can be fitted with formula (2):
S L Mf - 0.42 x — T — (2)
For very small S -values the smoothing character of the splitting mechanism dominates. For an S -value near one the "natural" flow distribution consists of a few cells contaxning all liquid and the rest of the cells without flow. The coalescing effect dominates. The result is a very high Mf-value. For the simulations an explicit function for S is needed. Equation (2) can be rewrit-ten as:
s = » « _ (3) \ 0.42 + Mf (i>
This equation relates a measured global packing characteristic to a—1-oca-l—parameter—of—a—cell I.t-only_ho.lds_if the cell width equals the width of the measuring modules on which the Mf-value is based.
The S -value is a measure for the average fraction of liquid that spreads in a horizontal direction over the surface of a cell. Per layer the liquid is also transported in a vertical direction over the height of a cell. This results in a random-walk type spreading of the liquid.
The spreading coëfficiënt, D can be calculated from:
rf Dr,L 4 x z x ln(l-'x)' (4)
Here x is the fraction of the liquid flowing through a circle with radius r around the axis at a packing depth z. For the simulation model the depth z is equal to the number of layers n multiplied with the height h of one layer
z - n x h (5)
41
The D -value i s a raeasure for the r a t i o between h o r i z o n t a l and r
vertical flow. The distance travelled by the liquid ( vertical flow ) is proportional to the cell height. From equation (4) and (5) it can be seen that for a constant number of layers the D -value is inversely proportional to the cell height. Reniinding that the S -factor spreads over a surface ( in the order of w2 ) in a horizontal direction, a linear relation between D and the ratio S /h is to be expected:
r.L <*ï (6)
The values of D , have been calculated for many ratios of r.L J
S /h. The results, presented in Figure 2, show that equation (6) -4 is satisfactory. The constant C has a value of 1.56x10 [m2]
0.010
0.005
0.000
Sl/h (l/m)
Figure 2 The relation between the splitting per cell height and the radial spreading coëfficiënt for unit cells with a width of 25 mm.
42
The c e l l model r equ i res a value for h. Rearranging equa t ion (6) g ives :
h = C x g - i - (7) r,L
S is already fixed and D -values can be found in literature. Typical h-values for a cell wldth of 25 mm are in the range from 20 to 50 mm for packings with a diameter of 25 mm, i.e. about the height of a packing element, which is satisfactory from a physical point of view.
Gas bulk flow
It was observed that the gas flow maldistribution on the scale of a packing element is very small compared with that of the liquid . For this reason the gas flow model does not contain random aspects; small scale maldistribution is not simulated.
In~Figure—1— the~principle~of—the—gas-spreading—i-s-outs-l-lned—too. A fraction S of the gas flow G through a cell is divided equally over the four neighbours of the cell above. The rest (l-S„)xG flows straigth upwards. Formula (6) can also be applied for a relation between the gas spreading coëfficiënt D and the gas
r, G
splitting factor S . Substituting and rearranging results in:
Dr G X h
The constant C has the same value as in formula (6). A limitation of the gas model is that it only simulates gas
mixing correctly if the Initlal distribution resembles the natural flow profile. The model cannot compensate for radial gas flows caused by horizontal pressure differences. However, it has been shown that a severely maldistributed inlet gas is converted to natural flow wlthin a packing height of one-half a column
25 diameter
43
Wall flow
In this work the wall flow factor, Wf, is arbitrary defined as the ratio between the average velocity in a peripheral ring, adjacent to the wall, with a thickness of 0.0127 m and the super-ficial velocity.
At the column radius the square cells are intersected by the wall. Any liquid flow directed to the wall is deflected downwards. An annulus with a thickness S forms the wall zone ( Figure 3 ). In this zone the splitting factor S is replaced by a splitting factor W This W serves the same function as the S , but can be chosen freely to predict the wall flow. When the W -value is smaller than the S -value the liquid tends to accumulate in the wall zone. The introduction of W causes the development of a constant wall flow at an infinite packing depth. The splitting factors are weighted by the area they occupy in cells that fall partly in the bulk ,in the wall zone or outside the column.
Analogous to the liquid model an W -factor is used for the gas flow in the wall zone instead of an S -factor.
G
w excluded area
wall a r e a
bulk a r e a
Figure 3 The zone near the column wal l . w i s the element width, S i s the wall r i ng th ickness
44
Results
The unit cell as presented above is the building block of a column configuration. A computer program in Turbo Pascal has been written to manage input, layer to layer calculations, output and interpretation. Although the program is able to handle all kinds of configurations, the cells used in this chapter have dimensions equal to those of the measuring equipment .
A layer of 20x20 cells of 25 mm width is used to enclose a circular column cross-section with a diameter of 0.5 m. 332 cells are (partly) covered by the column, the remaining cells are ex-cluded from the calculations.
The wall zone covers 10% of the total cross-sectional area of the column. This results in an annulus thichness of 12.7 mm.
For this configuration several simulations up to the equilibrium profile were performed using various S /W„ ratios. The
G G resultlng wall flow factors are presented in Figure 4. This figure can also be used the other way around. Knowing the wall flow at an ihïihïte packing depth-yi-elds-the-rafei-o-o-f—fehe-spli-tting-fac.tors_ in the bulk and in the wall zone. Figure (4) is also valid to predict the average wall flow of the liquid.
4 I 1 1 1 1 1 1 1 r- 1 1 1 1 r 1 1 1 1 1 1
3 - ^ ^ ^ ^
Wf / ^ 2 - Jf
Q 3 i 1 1 1 1 1 1 1 1 1 1 1——1 1 1 1 1 1 1 1 0 5 10 15 20
S/W
Figure 4 The equilibrium wall flow as a function of the ratio between splitting factors in the bulk and wall zone.
45
The following three examples compare measurments with simulations.
Liquid spreading
Maldistribution factors and liquid spreading coefficients are 2 known for various types of random packings . With this knowledge
the model can be used to simulate the liquid flow distribution resulting from a point source. As an example an S -value of 0.576 and a cell height of 48 mm are obtained by applying equation (3) and (7) on the Mf-value of 0.57 and the D .-value of 1.8 mm as
r, 1 measured for the Intalox Metal Tower Packing (IMTP no.25) in the
2 absence of gas flow . In Figure 5 the results are presented for a liquid spreading
experiment and the simulation for IMTP no.25 in a bed with a height of 1.72 m. Both velocity profiles show small scale irregularities. The overall distributions are similar.
Measured Profile Simulated Profile
o o o o o
o oo o o OO O © Offl©0
oo o© o o o o o o ©0©0 OOOffiOO O O © © ••©O© O
o ©o©o»«©«oo O ©OffiOO©©0 ©O © © • OO © O o © © • • o
©o ©o©© o o oo
o o o o
o oo o
ooooo o ffiffifflO O
o©oo ooo oo ooooeoooo® O ffi©00©ffiffi OO OOO 0©ffi©ffiO©00 ©0«©©«©0©0000
o©©« o»o©ooooo OOOOO •ffiO©ffi© O
OOO « « O OOO OOOffiO©© OO
O O OO o o ooo
O0.3-1.5 ©1.5-3.0 mm/s • } 3.0
Figure 5 Measured and simulated results for the spread of a single liquid jet for 1.72 m of IMTP packing.
46
8 r
Column axis
Liquid Velocity
(mm/s)
Initial liquid distribution
Relative Area
V^/lT^V
j — i i—i i i — i — i i
■ i ■ ' i i i i — 1 _
Simulations
0.21 m
0.43 m
0.86 m
1.72 m
-1 1 1 L_
\J^^\J. J 1 1 1 1 1 1 L.
,.,!, ,.i ■ ■
Measurements Figure 6 Results for the profile developement of a flat in..-_
distribution for 1" pall rings. u_ - 0 m/s. u 3.4 mm/s
47
Liquid profiles
The prediction of large scale effects is illustrated in Figure 6 for a column of 0.5 m diameter filled with 25 mm metal Pall rings (S =0.659, W - 0.146, h = 39.5 mm). The liquid velocity profiles are given for several packing depths. The initial liquid distribution is almost uniform. The measured as well as the calcu-lated profiles show a comparable increase of the wall flow. The fast built up of wall flow causes a shortage of liquid in the annuli close to the wall. The liquid in the centre of the column needs a large column height to equilibrate with the wall flow. This is because of the low spreading capacity of random packings. All profiles show local irregularities.
Gas flow effect on liquid wall flow
The wall flow of the liquid phase decreases at higher gas loads 2 for various types of random packings . For 25 mm metal Pall rings
simulations have been performed to predict the development of the liquid wall flow with and without gas flow as a function of the packing height. In Figure 7 the measuring results are presented for a gas velocity of 0 ( ) and 2.54 ( ) m/s.
The simulation model needs input of global packing characteristics. Spreading and maldistribution data are obtained from the experimental work. At a zero gas velocity D has a value of 2.5 mm and at a gas velocity of 2.54 m/s D equals 3.7 mm. The maldistribution factor for natural flow is 0.81.
The equilibrium wall flow factors of 2.1 and 1.4 are found by extrapolation of the measured data in Figure 7. The equilibrium wall flow factor for the case without gas flow can also be calcu-lated from :
gr= U + O^]- 1 (9) P
For 25 mm Pall rings in a column with a diameter of 0.5 m the right hand side of the equation takes the value of 0.122. Remembering that the wall zone thickness, S, is 12.7 mm leads to a Wf -value of 2.1.
48
Now the global packing characteristics are known the local cell parameters can be determined. Applying equation (3) and (7) results in an S -value of 0.659 and a cell height of 39.5 mm (u - 0 m/s) or 26.7 mm (u = 2.54 m/s). From Figure 4 it follows
G ti
that the splitting factor for the wall zone is: W - 0.146 (u = 0 m/s ) or W = 0.347 (u - 2.54 m/s).
Wf
Packing Depth (m)
Figure 7 The developement of the liquid wall flow as a function of the packing height for 1" metal pall rings with and without gas loading. Measurements ( J> , u -2.5 m/s ) and
G u =0 m/s )). Result of one simulation (• ( O average result (
■) and
-).
The initial liquid distribution for the simulation was chosen to be equal to that used for the measurements. The superficial liquid velocity is 3.4 mm/s.
In Figure 7 two types of simulation results are presented. The dotted curves show the results for a single simulation. The random aspect of the model introduces a scatter in the Wf-factor. The solid curves give the average value of the Wf-factor. The curves are in agreement with measurements.
49
In the following two examples the model will be applied to predict the effect of tower internals:
Drip point density
A practical problem with the design of packed columns is the choice of the number of drip points per square meter of the initial distributor. A high number of drip points is desired but there are constructional limits: it hard to obtain a flat profile on a large scale. Also fouling can destroy the quality of the distribution. A low number of drip points however could wörsen the separation efficiency of the packing.
The simulation model has been used to study the development of the quality of the distribution as a function of the packing depth for eight different initial distributions. The number of drip points per square meter has been varied between 40/m2 and 1600/m2. The spreading coëfficiënt of the liquid is 3 mm and the Mf-value for natural flow was chosen to be 0.6. These are realistic values for 25 mm packings. The wall effect has been neglected.
10
Mf
■IV w \ c \
T '
\
\ ) • \
\
ZÜ
\
\
g \
\
\
1 1 1 ] 1 1 1 -
\ h
\
\
l
0.0 0.5 1.0 Packing Depth (m)
Figure 8 The effect of the drip point density on the maldistribu-tion as a function of the packing height. a=1600, b=800, c=400, d=200, e=100, f=80, g=60 and h=40 drip points/m2
50
Figure 8 shows the behaviour of the maldistribution factor as a function of the packing depth. The lower the drip point density the more packing depth is needed to reach natural flow. A very high density hardly improves the quality of the distribution. On the other hand a poor density penetrates deeply into a packed bed.
Wall wiper
Wall wipers are said to be required when the tower efficiency is reduced due to a large percentage of liquid flowing down the column wall. Their principal function is to remove the liquid from the column wall and transport it to the interior of the bed. This problem occurs most often in small diameter towers.
In this example the effect of a wall wiper on the wall flow is investigated. The column has a packing height of 6 m and a diameter of 0.5 m. The initial liquid distribution has 100 drip points per square meter. The wall wiper is located at three meters front the top of the bed\ The dimensions~cTf~the modeT~8"5"8 me tal-
26 "Rosette" wall wipper as manufactured by Norton Ltd . were used as an indication for the redistribution of the liquid. The wall flow is spread over 16 regulary orientated points which are about 6 cm from the wall. The same values were chosen for the liquid spreading coëfficiënt, the "natural" flow maldistribution and the equilibrium wall flow as in the previous example.
Figure 9 shows the results of the simulation. The solid line gives the wall flow factor as a function of the packing height. Starting at the top of the packing the Wf-factor develops within 1 m of packing from zero to about 1.5. After the wall wiper has reduced the wall flow to zero at a packing depth of 3 m, the same kind of development is calculated again. This means that the effect of this wall wiper lasts for about 1 m of packing.
It should be emphasized that a Wf-factor of one is ideal. Therefore the absolute difference between ideal wall flow and simulated wall flow should be regarded as a measure for a wall effect. This difference is represented by a dashed line in the figure. From this line it can be concluded that the wall wiper only gives a small improvement in the total wall effect of the column. Moreover, an increased maldistribution factor for some layers just below the wall wiper, caused by the 16 points, also diminishes the improvement.
51
3
2
Wf
1
O 0 1 2 3 4 5 6
Packing Depth (m)
Figure 9 The wall flow ( ) as a function of packing depth with a wall wiper for a typical 25 mm packing. Column height = 6 m. Diameter - 0.5 m. The deviation from the ideal value is denoted by ( ).
Other cell dimensions
The simulations presented above were all based on a unit cell size of 25 mm. This size corresponds to the dimensions of the experimental collectors. For large columns the number of simula-tion cells per layer increases rapidly. A larger unit cell can be applied to avoid'time and memory consuming simulations. However, the measured maldistrlbution factor on a scale of 25 mm cannot be directly used; the flow maldistributions are smaller with larger unit cells.
The maldistribution factor for a unit cell with the size of w, Mf , can be calculated from: w
Mfw " Mf0.025* < ̂ >* <10)
where Mf. n_s is the maldistribution factor as measured on an 25 mm scale. and w is an integer multiple of 0.025 m. w must always be
52
1 1 i r
smaller than the column diameter. The Mf -value can be substituted w
in equation (3) to find the S -value of the larger unit cell. A second problem to be solved is the dependency of the spread-
ing coëfficiënt upon the cell width. As stated before a fraction of the flow in each cell is spreaded in a horizontal direction over a surface in the order of the cell area ( w2 ). The total amount of sideward flow is proportional to the splitting factors, S and S . During this spreading the flow is also transported in a vertical direction over a distance of the cell height, h. Radial spreading is proportional to the ratio between horizontal and vertical transportation. If the constant C is replaced by C'xw2, then C'= 0.25. Equation (6) can now be rewritten as:
Dr,L " ° - 2 5 * Ï T x S L (11)
The constant C can be substituted in equation (7) and (8) to give more general formulae for the determination of the cell height and S
Attemps were made to derive the new constant from theoretical considerations. A value of 0.25 was thought to be equal to the probability of flow in one of the four horizontal directions. An acceptable explanation has not yet been found.
The use of larger cells implicates a loss in the accuracy of the flow description on a small scale. However, the spreading behaviour and maldistribution on the scale of the unit cell still agree with practice.
Conclusions
A Monte Carlo cell model based on simple principles has been developed to simulate flow distributions in columns with a random packing. Measured quantities such as the maldistribution factor, spreading coefficients and wall flow at equilibrium are trans-formed to parameters for a unit cell. The column to be simulated is formed by a three dimensional network of cells.
The agreement between simulations and measurements of radial spreading of liquid, of velocity profiles and of the development of wall flow as a function of packing height and gas velocity demonstrates the capabilities of the model.
53
The model can also be used to evaluate the drip point density of a initial distributor, the redistribution caused by a wall wiper, etc.
There are some limitations. The model cannot compensate for radial gas flow caused by horizontal pressure differences as they may occur in the lower part of the packing. Also axial mixing is not taken into account.
Note
The simulation program suitable for IBM compatible personal computers is available for people who are interested in this subject.
Svmbols
C constant used in equation (6) [m2] d column diameter [ml c d nominal packing diameter [m] D radial spreading coëfficiënt [m] h height of a unit cell [m] Mf maldistribution factor [-] Qtot total liquid flow [m3/s] Qwall wall flow [m3/s] r radius [m] S splitting factor for the bulk zone [-] u velocity [m/s] w cell width [m] W splitting factor for the wall zone [-] Wf Wall flow factor [-] x fraction of the flow through
a circle with radius r [-] z packing depth [m]
subscripts:
G L w 00
gas phase l iquid phase for cel ls with a width of w at equilibrium
54
References
R.M.Stikkelman -this thesis chapter 3 -Chem. Eng. Res. Des. submitted for publication R.S.Tour, F.Lerman Trans. Am. Inst. Chem. Engrs. R.S.Tour, F.Lerman Trans. Am. I n s t . Chem. Engrs. Z.Cihla, O.Schmidt
22
23
, 35 (1939) 719-42
, 40 (1944) 79-103
(1957) 896-907
(1958) 569-578
42 (1977) 1129-1140
Coll. Czech. Chem. Comm. 5 Z.Cihla, O.Schmidt Coll. Czech. Chem. Comm.
6 V.Stanek and V.Kolar Distribution of liquid over a random packing I - X I Coll. Czech. Chem. Comm., 30 (1965) 1054-1059 X Coll. Czech. Chem. Comm..
7 E.Dutkai, E.Ruckenstein Chem. Eng. Sci., 23 (1968) 1365-1373
8 E.Dutkai, E.Ruckenstein Chem. Eng. Sci., 25 (1970) 483-488
9 K.E.Porter, M.C.Jones Trans. Instn. Chem. Engrs.,
10 J.J.Templeman, K.E.Porter Chem. Eng. Sci., 20 (1965) 1139-40
11 K.E.Porter, J.J.Templeman Trans. Instn. Chem. Engrs.
12-P-. J-.—Hofty z er — Trans. Instn. CHem. Engrs.
13 D.J.Gunn Chem. Eng. Sci., 33 (1978) 1211-19
14 E.A.Brignole, G.Zacharonek, Chem. Eng. Sci. , 28 (1973)
15 P.J.Hoek Ph.D. Thesis, Technische Hogeschool Delft, 1983
16 P.J.Hoek, J.A.Wesselingh and F.J.Zuiderweg Chem. Eng. Res. Des., 64 (1986) 431-49
17 K.Onda, H.Takeuchi, Y.Maeda, N.Takeuchi Chem. Eng. Sci., 28 (1973) 1677-83
F.J.Zuiderweg 33 (1978) 1637-43
41 (1963) 240-47
46 (1968) t86
42 (1964) tl09-17
J.Mangosio 1225-29
18 G.G.Bemer Chem. Eng. Sci.
19 D.E.Nutter I. Chem. E
20 J.Prchlik, Symposium Series No.104 (1987) A129-42
J.Soukop, V.Zapletal, V.Ruzicka Coll. Czech. Chem. Comm., 40 (1975) 845-55
21 G.J.Jameson Trans. Instn. Chem. Engrs., 45 (1967), T44
22 M.A.Albright Hydrocarbon Processing, 9 (1984) 173
23 Q.H.Ali Ph.D. Thesis, University of Aston, 1984
24 Packed Tower internals Norton, Bulletin TA-80R, revised december 1976
55
CHAPTER V
Measurement and simulation of the influence of maldistribution on distillation in a column
with a random packing
Introduction
Vfhen a column with a random packing is designed some important choices have to be made concerning the type of packing, the dimen-sions of the column, the liquid distributor, redistributors, etc. Insight in the development of flow in a packed bed can contribute to an optimal design. Malperformance can be predicted or possibly diminished.
The effect of liquid flow patterns on the performance of packed columns has been studied in the literature mostly using an artifi-j;ially defined maldistribution . Controlled maldistribution studies by using an adjustable distributor were performed by Kunesh . Discontinuities or step changes in the flow from various zones of the distributor have the most severe impact on efficiency, while modest amounts of tilting and sagging are said to be tolerable. The separating efficiency of beds of a random packing as influenced by large scale liquid maldistribtuion has been described by Zuiderweg in a zone-stage mass transfer model in-cluding radial spreading. The model can calculate the effects of relative volatility, concentrations, reflux ratio, packing size, column diameter, bed height and redistribution. It has been ap-plied to simulate results of the controlled maldistribution
Q
studies by Fractionation Research Ine . The study on maldistribution effects in packed columns at the
Delft Unlversity of Technology is divided into two projects. One project, concerning the liquid and gas flow patterns in a
9 water/air column, has been presented in a previous paper . The second project deals with the effects of a severe initial liquid maldistribution on the separation efficiency of a pilot distillation plant.
57
After a short description of the pilot plant the experimental results of distillations with and without liquid maldistribution will be presented. These results will be simulated by a Monte Carlo cell model. The flow simulation part of this model has already been described . By including simple mass transfer equa-tions the model is capable of simulating concentration profiles.
A practical example of the model is given in a study of the effect of the drip point density of a liquid distributor on the overall separation efficiency.
The distillation unit
In a pilot plant distillation column the influence of large scale initial liquid maldistribution has been investigated for 25 mm plastic hydrofilized Ralu rings (RALU) and the no.25 stainless steel Intalox Metal Tower Packing (IMTP). The plant is situated at the Laboratory of Process Equipment at the Delft University of Technology. It consists of the following major parts ( Figure 1 ):
an isolated column, made of stainless steel, with an internal diameter of 0.45 m. The height of the packing is 2.65 m. An adjustable liquid distributor is situated at a distance of 2 cm above of the packing. a reboiler, where heating is supplied by steam with a maximum pressure of 0.5 MPa. a condensor, where the vapour phase is condensed with water from a cooling tower.
The distillations were performed with a methanol/ethanol mixture at atmospheric pressure and at total reflux. During each run the pressure in the column was measured and temperatures ( top, bot-tom, cooling water in, cooling water out ) were recorded. Also the cooling water flow was measured. When the whole system was stable liquid samples were taken at the bottom of the column and at the reflux pipe.
The liquid samples have been analyzed by a refractive index meter ABBE 60 at 20 "C. The maximuin error in the determination of a mole fraction is 0.01. Some samples have also been checked by GLC. The results fall within the error of the refractive index analysis.
59
The loading of the column ( the F-factor ) can be obtained from a heat balance over the condensor. Experiments were perforraed twice for several F-factors between 0.5 and 1.6 Pa ' .
The liquid distributor used ( Figure 2 ) consists of eight troughs. The liquid is evenly distributed over 52 drip points (330/m2 on the column crossectional area). This will be called the ideal distribution. The maldistributed feed is introduced by covering the left four of the eight troughs. Then only one half of the top of the packing is irrigated.
® ® ® ® ®
® ® ® ® ® ® ® ® ® ® ® ® ® ® ® ®
® ® ® ® ®
drip points in the O maldistributed
feed
• ideal distributor
. 1 o cm
Figure 2 The configuration of the drip points of the liquid distributor. Ideal (•). Maldistributed feed (O).
Determination of the HTU -value OG
The number of over-all gas-phase transfer units, Nn„, is given by:
OG -i t dy ( y e q - y > ( i )
where:
yeq 1 + (a-1) x (2)
60
The relative volatility, Q, can be obtained be applying the Van Laar and Antoine equations. At a pressure of 10s Pa the value of a is almost constant over the whole composition range and equals 1.72.
The experiments were all under total reflux so the overall operating line is given by:
y - x (3)
Therefore liquid compositions as measured at the bottom and the top of the packing can be used in equation (1) instead of the vapour compositions.
The height of a transfer unit based on the over-all gas-phase resistance, HTU , is calculated according to:
The relative error in the determination of the HTU-„-value_is — — — — — ~"— Uu less than 10 %.
Distillation results
The results for the two types of packing are summarized in Table I for a packing height of 2.65 m. For several F-factors the concentrations at the top ( x ) and at the bottom ( x, ) of the packed bed have been determined. Applying equation (4) results in the height of an over-all gas-phase transfer unit.
The effect of the maldistributed feed is considerable: the average HTU.„-value is 2.7 (IMTP) or 1.6 (RALU) times higher than
UG the average HTU - value in the case of the ideal distributor. The IMTP packing is more sensitive to initial liquid maldistribution than the RALU packing.
Increasing the F-factor leads in general to lower values of the height of a transfer unit. Due to practical limitations, it was not possible to apply higher F-fact< experiments are without loading effects. not possible to apply higher F-factors than 1.6 Pa ' so all
61
, 01 m on
r-l r-t 01 I
u o > o
o n P
C O
T l (11 0) Ifl
43
P •H C 1
u <i>
M-l w C ra ^ p
ra M-l 0
P 43 on
•H 0)
43
Cl * H
X> dl N
•i-t i - l •i-I 4-1 O IJ •o ■>!
43
O - H 4J 01 ra r-l O.
Ë £
m Ol
4-1 o
-u c
e 33
~ <11 O c ra •p oi
•H 01 <U I-I
s> 01 ra
X. o.
C o
•H -P O e 3
M-l
ra Ol ra
PH H T\ t-i
i - l o o •u UI
01 01 o
r-l a •rH ra 4J 0]
i n CM
o c
T> e ra 01 on c •r-l U
3 I - I ra os
Ui
•o ra o r-l
01 ra 00
ai e ■p
T3 C ra
(3 O
■r-l 4-> ra i-i 3 00
■w 1H c o o
I-I 0 u 3
43 ■r-l U P 01
•H T l
01 43 p
4-1 O
i 01 o R ■p
ra ■P
ra I - I o e ra 43 4-1 w 1 i-H O C ra 43 ■P O a
E 01 ■p 01 ^ co
e i n v£>
CSI
4-1 43 00
• f l 01
43
on e ■ H
4<! O ra 04
T) d ra o. o 4J
ai 43 4J
(3 •r-l
01 c o
•r-l P O ra u 4-1
r-l O c ra 43 P 01 E
o 43 H
01 c o •H P •i-l •0 c 0 o
ü •r-i u o
43 O.
c 3 l-i
43 O ra 01
u o
i p
c 01 > •H 00
01 u ra
-/ "
TJ c ra
4-> X
» H 0 p p o
43
O O
X — '
43
P
i n
o
ra 04
O v O V D i n o O C f i r H O O v o m o r ~ o o i - i r - - C T > r o m r O C M C M r o t M c M o o o o o o o o
C A v o r H t n r ^ t - i c N i n ■ ? ( n i o m r o \ o o » o CNCV4CMC*JCMCMr- ICN
o o o o o o o o
v o a \ c o c s i c o o m o
C T i C ^ C T * C J \ C A O ï C A C T *
o o o o o o o o
C T i O v O O O r - l t n C N C T l < r m o o o o c M < N i n m
O O O O r - 4 r - l r - l r - l
H Q 01
•d ■I-I
n p i o i N c i v o i / i o » O O O M S O H O O i n C A o o o o % c ^ o o r ^ r ^
O O i - I O O O O O
i n i - t c M t n c M r ^ o ^ ? l O i n M T l l O f i l M N
O O O O O O O O
o \ o i e o o \ o i t N m M i — i r - i r ^ c O r - i m m < r
O O O O O O O O
N t n f n i o - J i o i s p i 4 > ï » 0 0 H H ^ i n
O O O O I H I - I I - I I - I
mal
distri
buted
feed
M S H 04
o o c r v c N i n i - i < f o a \ O l O O i n N C l O M N
o o o o o o o o
i n m o o - ^ - ^ c o c A N U N N H r i O O
o o o o o o o o
O C M ^ t i - H C M C T i C M r - -f O < J i o i o i n n H M 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0
o o o o o o o o
O ^ O l l / l l O H H C M O
O O O O r H l H r - l r H
ra m
•H
y ï w H r i m c g i n n
o o o o o o o o
t n p ^ O N i n m i n c M O O O V O V O t M C M O N O O n c - l C M C M C M C s l i H i - H
O O O O O O O O
r - - i n o o o o i n r o r H o o m t n o o O i - i r H O r - * r - r ^ r ^ r - * r ^ r * i ^
O O O O O O O O
M « ^ l O O \ H O H ■ l ^ o t o H o j m i n
O O O O r H i - l i - l i H
mal
distri
buted
feed
PÜ < J D
43 ra H
Simulation of mass transfer
A Monte Carlo cell model has been used to simulate mass transfer. The column is thought to be built of cells with dimen-sions in the order of a packing element. The flows of the gas and the liquid through these boxes are calculated by a flow splitting model capable of predicting wall flow, spreading and mal-
9 distribution as determined experimentally . Small scale maldistribution in the liquid phase is introduced
by a random splitting mechanism. A fraction S , of the liquid flow through a cell, leaves the cell randomly to one of the four neigh-bours of the cell below. Maldistribution is caused by coalescense of flows. The height of the box is chosen to produce a proper liquid spreading.
Gas spreading is descrlbed by a gas splitting factor S . A fraction S p of the gas flow G through a cell is divided equally over the four neighbours of the cell above. The rest (1-S )xG
Cr
flows straigth upwards. An annulus near the wall with a thickness of half a packing
diameter forms the wall zone in which the splitting factors are adjusted to correctly predict wall flows.
Measurement Simulation
Figure 3 Measurement and simulation of the liquid velocity profile leaving the bottom of 2.16 m IMTP with the maldistributed f eed.
63
The flow simulation model is illustrated for 2.16 m IMTP with the maldistributed feed. Figure 3 shows a measured as well as a simulated velocity profile of the liquid leaving the bottom of the packing. The velocity measurements were performed in an air/water column with a diameter of 0.5 m.
Both profiles show .that the liquid is not equally distributed over the column cross section. About 78 % of the total amount of liquid still remains in the half of the cross section which was originally irrigated. This would be 50 % in the ideal case. From this it can be concluded that a low performance of a distillation column with such a configuration is to be expected.
The Monte Carlo model has been extended to include mass transfer. For each cell the incoming flows are known from the flow simulation. When the mole fractions of these flows are also known, the mole fractions entering the cell, x. and y. , can be calcu-
° i n y i n lated applying mixing rules . The fractions which leave the ce l l , x ,_ and y , are found by using two re la t ions: out ■'out J b
The material balance over a cell
y - - y . = TT ( x. - x ) (5) •'out -'m V in out'
and
An estimation of mass transfer in a cel l
W y i n " HTÏÏ- ^eq - ^in ' ( 6 )
UG
in which h is the height of a cell. For a mixture of methanol ethanol at atmospheric con
proximated by formula (2) using x and ethanol at atmospheric conditions the y -value can be ap
out Substracting equation (6) from (5) and substituting equation (2), with a — 1.72, results in a quadratic expression in x which can be solved analytically. Y can be calculated using the material balance.
Many correlations are known for the HTU . In general the HTU is fairly insensitive to the gas and liquid velocities. For this reason it was decided to introducé a flxed HTU on a local scale
OG for all cells.
No correlations such local transfer units are known. Therefore the local HTU has been fitted to experiments with the ideal
OG 64
distributor i.e. the height of the transfer unit is chosen to produce top and bottom concentrations corresponding to experimen-tal results. The resulting HTU is then applied to the case of the maldistributed feed.
The simulation procedure is as follows:
- The initial distributions for the liquid as well as the gas are chosen according to the experimental set up.
- The correct parameters of the splitting model are entered and the flow profiles are calculated
- The composition of the liquid in the distributor is used to make a first guess of the concentration profiles of both phases. The composition of the liquid at the top is known from the experiments.
- Better estimates for the concentrations are calculated using the simulated flow profiles and the mass transfer according to eqüation (6). The column is simulated from the top to the
—bottonn - In the case of total reflux the amount of each component leaving the packing as liquid should equal the amount of that component entering the packing as gas. The average initial gas concentration can be calculated.
- Even better estimates are calculated. The column is simulated from the bottom to the top.
The last three steps of this procedure are repeated until the mass balance at the top of column is correct. During the simulation the bottom concentration slowly converges to a constant value. This value can be compared to the experimental bottom concentration.
65
H
o • • f t ü
0.5 -
O Ö
0.0 1 2
Packing Depth (m)
Figure 4 Simulated gas concentration profiles for IMTP ( ) and RALU (—) with and without an initial liquid maldistribution. The measured top and bottom fractions are denoted by ($) for IMTP and (•) for RALU.
N
K
2
1 1 r-
1 \ N 1 \ ^"V
'. \ maldis \
V ~ *~ ~ ^ v ** \ ideal
r
" - - ^ * 5 r s - ^ .
-
-
-
- ^ ~
0 1 2 Packing Depth (m)
Figure 5 Simulated HETP(z) values for IMTP ( ) and RALU (---) with and without an initial liquid maldistribution as a function of the packing depth, z.
66
Simulation results
Mass tranfer simulations were performed for both types of packing with and without covering the distributor for an F-factor of about 1.5 Pa ' . Figure 4 shows the radially averaged simulated gas concentration profiles as a function of the packing depth. The measured top and bottom concentrations are presented on the Y-axis' . The top concentrations are forced to correspond to the experimental values by the simulation procedure. The bottom concentrations for the case of the ideal distributor are fitted by the local height of an over-all gas-phase transfer unit. This equals 0.19 m for the IMTP and 0.31 m for the RALU packing. With these values the separation with the maldistributed feed has now been calculated.
It is found that the calculated bottom concentration for the IMTP is exactly the same as the measured value. For the RALU packing the calculated value is 5 % higher.
Figure 4 shows that in the top of the packing only one half is effectively irrigated. This part operates in a pinch condition, and very little separation is achieved. Further downwards the. liquid distribution improves and the methanol concentration falls more rapidly.
The performance of the packing as a function of the packing depth can be better expressed by an HETP-value than by a concentration decrease. The HETP-value can be derived from the Fenske Equation:
H * ln(a) H E T P . _ _ ^ ( 7 )
ln( [riH* !ir^ > 1 x t *b
This relation can also be applied on a small height of packing resulting in a local height equivalent to a theoretical plate HETP(z):
h * ln(a) HETP(z) = j (8)
z z+h
These l o c a l HETP-values are given in Figure 5. The s imula t ions
67
B D
Figure 6 Longitudinal cross-sections of simulated gas concentra-tion profiles. RALU (A) and IMTP (B) vith the ideal distributor and RALU (C) and IMTP (D) with the mal-distributed feed. Column height - 2.65 m. Diameter «0.45 m.
68
with the ideal distributor rapidly reach a stable HETP-value. Only in the first 0.2 m of the packing the initial liquid distrlbution causes a malperformance of the column. The large scale effect caused by the maldistributed feed penetrates deeply into the packing. The HETP-value slowly improves.
The difference of the HETP development between the RALU and the IMTP packing with the maldistributed feed is remarkably. Due to a
9 higher liquid spreading coëfficiënt ( Dr - 3.4 mm ) the RALU packing gives a better smoothing of the maldistribution than the IMTP packing. The liquid spreading coëfficiënt for IMTP packing equals 1.9 mm. Also the difference in gas spreading coëfficiënt contributes to this difference. IMTP: Dr = 2.4 mm, RALU: Dr„ = 4 . 0 mm.
The concentration profiles of the gas over a longitudinal cross-section of the simulated column are presented in Figure 6 for both types of packing.
For the case of the ideal distributor the profiles contain some irregularities caused by small scale maldistribution and wall flow. The profiles for the maldistributed feed show that the concentration in a part of the gas hardly changes. This by-pass of gas is more profound in the top of the packing: iso-concentration lines are almost vertical. The higher spreading capacity of the RALU packing is illustrated by the more regular transition from vertical to horizontal iso-concentration lines.
Influence of the drip point densitv
The number of drip points per square meter is an important parameter for the design of a packed column. In the ideal case the initial liquid profile is flat, but then an distributor is needed with a high drip point denslty. Such a distributor is difficult to manufacture. A low number of drip points could worsen the separa-tion efficiency of the packing. A large packing depth is required before the jets properly overlap.
The effect of the drip point density has been studied with the mass tranfer simulation model for random packings. For a column with a height of l m and a diameter of 0.5 m the concentration profiles have been calculated for different distributor configurations. The number of drip points per square meter has been varied between 220/m2 and 5/m2. The drip points were evenly
69
distributed over the column cross section. The spreading coëfficiënt equals 3 mm for the liquid and 4 mm for the gas. The maldistribution factor, Mf, is 0.6. These parameters are realistic for 25 mm packings. Wall flow has not been taken into account. For all simulations a constant local HTU -value of 0.3 m has been used. The relative volatility a is also constant and equals 1.725.
The calculated concentration profile as a function of the packing depth has been transformed into the local HETP-values using equation (8). These values are given in Figure 7.
N
E-W
Packing Depth (m)
Figure 7 The development of the HETP as a function of packing depth and drip point density: a = 200 dp/ra2; b = 60 dp/m2; c - 30 dp/m2; d = 15 dp/m2; e = 5 dp/m2
For the simulations with a drip point density equal to or higher than 15/m2 the local HETP reaches a stable value within 1 m of packing. The simulation with 5 drip points per square meter is still developing after 1 m. Here large scale maldistribution caused by the liquid distributor results in an inferior column performance. The height of packing lost to stabilize the liquid distribution varies from 0.2 m for 200 dp/m2 to 0.4 m for 30 dp/m2. This last value is low compared to the common number of 100/m2 in practical applications.
70
Figure 7 holds only for the parameters mentioned above. Packings with lower spreading capacities, like no.25 IMTP, will need more packing depth to stabilize.
Discussion
The experiments with 50 % of the distributor covered show that when maldistribution is large the column performance is primarily controlled by flow parameters and not by mass transfer parameters. Although the potential HETP value of IMTP is lower than that of RALU, the distillation result is worse for the maldistributed feed. Kunesh did comparable experiments for a column with 3.7 m of 1 inch Pall rings and a diameter of 1.2 m. In the case in which liquid flow to one-half of the bed was shut off no recovery of the potential separation efficiency was found.
The cross sections as presented in Figure 6 show that con-centration profiles are not flat even for the ideal distributor. These profiles are very simular to the temperature profiles measured by Stichlmair . Due to the irregularities one should not evaluate the performance of a column by taking samples at a single point in the packing.
The proposed model is capable of simulating all types of random packing and column configurations. This paper only gives some examples. The mass transfer model can be easily replaced by more elaborate models to simulate less ideal mixtures.
The model has however one serious disadvantage. The calcula-tions are time consuming. The simulation of one distillation experiment took about 16 hours on an 8086 Personal Computer. No extra effort was given to speed up the convergence so improvements are possible.
Conclusions
Experiments have shown that the effect of the initial liquid distribution on distillation can be considerable. A flow simulation model combined with simple mass transfer equations has been verified for two distillations with an extreme initial maldistribution. Calculated and experimental results agree. As an example the model has been used to predict the packing efficiency
71
of common 25 mm packings. For a drip point density of 30 dp/m2
only 0.4 m of the top of the packing is lost compared with the ideal initial distribution.
The model can be used for all kinds of packed column configurations. A disadvantage is the long time needed for the simulation to converge.
Acknowledgements
This work was made possible by the support of the Koninklijke/Shell-laboratorium, Amsterdam and the Delft Laboratory for Process Equipment. We would also like to thank Norton Ltd. and Raschig GmbH for providing us with packings.
Svmbols
d packing diameter [m] Dr gas spreading factor [m] Dr liquid spreading factor [m] h height of a simulation cell [m] H height of packing [m] HETP height equivalent to one
theoretical plate [m] HTUn„ height of a transfer unit based on
over-all gas-phase resistance [m] L molar liquid flow [mol/s] Mf maldistribution factor [-] N number of over-all gas-phase
transfer units [ - ] ST liquid split factor [-] S„ gas split factor [-] V molar gas flow [mol/s] x liquid mole fraction [-] y gas mole fraction [-] z packing depth [m]
relative volatility [-]
72
subscripts
b bottom eq at equilibrium t top
References
1 J.W.Mullin Ind. Chemist., 33 (1957) 408
2 R.E.Manning, M.R.Cannon Ind. Engng. Chem. 49 (1957) 347
3 G.A.Morris Proc. Int. Symp. Dist. London (1960) 146
4 M.Huber, R.Hiltbrunner Chem. Engng. Sci. 21 (1966) 819
5 H.C.Yuan, L.Spiegel, Proc. 2nd. World Congr. Chem. Engng. Montreal 4 (1981) 274
6 J.G.Kunesh, L.L.Lahm, T.Yanigi Ind. Eng. Chem. Res. 26 (1987) 1845-1850
7 F.J.Zuiderweg, P.J.Hoek and L.L.Lahm I. Chem. E. Symposium Series No.104 (1987) A217-231
8 J.G.Kunesh, L.L.Lahm, T.Yanigi I—Chem7-ET-Symposium-Series-No-.-104-(1987-)-A233^24_4 I
9 R.M.Stikkelman, J.de Graauw, J.A.Wesselingh - this thesis chapter 3 - chem. eng. res. des. submitted for publication
10 R.M.Stikkelman, J.de Graauw, R.F.de Ruiter, J.A.Wesselingh - this thesis chapter 4 - chem. end .res. des. submitted for publication
11 J.Stichlmair and A.Stemmer I. Chem. E. Symposium Series No.104 (1987) B213-224
73
CHAPTER VI
A study of gas and liquid distributions in structured packings
Tntroduction
Structured packings are being used in separation columns on a large scale. Even so their behaviour is not fully understood. The flow in such columns appears to be fairly ideal, both in the liquid and in the gas phase up to the loading point. However little is known about the loading region. Therefore, it was decided to determine the flow profile in both the gas and the liquid using a water/air column with a diameter of 0.5 m. The measuring equipment enables the detection of the gas and liquid flow rates simultaneously in about 300 measuring sections of 25x25 mm2 in both the top and the bottom of the column. A detailed description of the equipment is given in the Ph.D. thesis of Stikkelman .
The aim of the present work was to study in detail the gas and liquid flows and their interaction for the varous types of structured packings listed in Table I. The results of the measurements will be discussed in terms of two parameters:
the maldistribution factor Mf the wall flow factor Wf The maldistribution factor, which is a measure for small scale
maldistribution, is equal to the relative Standard variation of the flows in the bulk of the packing. For a uniform distribution it has a value of zero.
The wall flow factor is defined as the ratio between the flow rate in an annulus near the wall and the average flow rate. For an ideal distribution this factor equals one.
Also attention will be given to the occurrence of large scale segregation of the liquid flows and spreading of both phases in the column.
75
Table I The structured packings used.
Type of packing
Sulzer Mellapak 250Y
Sulzer Mellapak 500Y
Sulzer plastic BX
Julius Montz BI-250
Julius Montz BS-450
Raschig Ralupak 250YC
Material
perforated metal sheet
perforated metal sheet
plastic gauze
metal sheet
gauze-like metal sheet
slitted metal sheet
Code
250Y
500Y
BX
Bl-250
BS-450
Rpak
Literature survey
The major part of the literature on maldistribution in columns equipped with structured packing has been produced by the coworkers of Sulzer Brothers Ltd.
Dealing with the fractionation of heavy water they designed a 2 3 4 5
gauze packing ' . Huber and Hiltbrunner and Flatt developed a theory to obtain quantitative results for columns with an ar-bitrary maldistribution. These results have been confirmed experimentally by Meier and Huber in a column with BX packing and artificially generated maldistribution at total reflux. Huber and
4 Hiltbrunner showed that in a column with a small diameter the influence of maldistribution is compensated by lateral mixing in the vapour phase. The total reflux experiments with BX packing are summarized by Meier and Huber . An initial maldistribution is maintained in the top of the packing, causing a decrease in the separation efficiency. If the initial liquid distribution is good enough the packing will keep the liquid evenly distributed, except when the liquid does not wet the packing. No substantial influence
o of the column diameter was found. Experiments by Zogg showéd that the gas-side mass transfer coëfficiënt is independent of the packing height and column diameter. Low pressure distillation under partial reflux using an artificial liquid maldistribution
9 resulted in a complex system of relationships. Yuan and Spiegel
76
indicated that the influence of maldistribution on column performance is less at partial reflux than at total reflux.
The corrugated metal sheet packing "Mellapak" was introduced in 1977 . The HETP seems to be almost independent of the column diameter, packing height and reflux ratio. For the Mellapak 250Y the HETP is fairly constant up to a F-factor of 2.8, above which it increases rapidly . The lateral mixing of the gas was studied by adding one percent of carbon dioxide in the column axis upstream of the packing. Lateral mixing of the gas is said to be
12 much better than in dumped packing In general the Sulzer coworkers concluded that, assuming cor
rect working conditions and no initial maldistribution, design data obtained from calculations or pilot tests can be transferred to industrial scale with a large degree of certainty.
Measurements of the flow distribution on a small scale for the 13 liquid, in the absence of gas loading, have been made by Hoek
The quality of the distribution was found to be better for BX, 250Y and 500Y packings than for random packings. The natural flow profile, assuming a reasonable initial distribution, was establi-shed within three elements. Liquid was spread rapidly in the direction parallel to the sheet orientation.
14 Gas-only studies were performed by Stikkelman and Wesselingh The maldistribution of the gas was found to be negligible compared to that of the liquid. However, initial maldistribution of the gas phase was retained longer in structured packings than in random packings.
Interaction between gas and liquid has been the subject of studies by Kouri and Sohlo . For their investigations they used 5 concentric rings as collecting sections in a column with a diameter of 0.5 m. Gas distribution was found to be close to ideal for the BX packing, unless a poor initial liquid distribution was employed. Liquid distribution was uniform at moderate gas loads, but it deteriorated at gas velocities above 1.7 m/s.
From the literature it can be concluded that the influence of the gas flow near or in the loading region on maldistribution has sofar received little attention.
77
Gas profiles t
Measurements for the gas phase showed that the local flow distribution over a cross section is almost uniform at the top as well as in the bottom of the packing. The maldistribution factor , Mf, based on about 300 measuring areas of 25x25 mm2, is an order of magnitude smaller than the average value of 0.25 for the liquid phase. Values for the packings tested are given in Table II.
The relative gas flow rate near the wall, Wf, in the bottom of the packing is about 1.25 times the average bulk flow rate. In the equipment wall flow at the gas inlet is partially caused by the reduced area of the measuring elements at the periphery. However, due to the sleeves around the packing element, one can discuss whether the wall flow is constant over the height of one packing element. The two sleeves per element form obstacles for the gas flowing upwards through the annulus between the wall and the packing.
Table II Maldistribution and wall flow factors for the gas and the liquid phase at the top ( Mf , Wf ) and in the bottom ( Mf, , Wf, ) of a 0.5 m diameter column with 4 elements b b of structured packing (BX 5 elements).
Code
250Y
500Y
BX Bl-250
BS-450
Rpak
Gas
Mf b
0.016
0.020
0.014
0.013
0.019
0.023
Wf b
1.25
1.33
1.35
1.18
1.24
1.30
Mf t
0.011
0.011
0.015
0.006
0.011
0.018
Wf t
-
-
-
1.3
-
-
Liquid
Mf b
0.24
0.21
0.22
0.23
0.44
0.71
u Gmax
[m/s]
2.6
2.1
2.4
2.6
2.2
>3.2
u Lmin
[mm/s]
3
1.5
1.5
3
<0.8
5
78
Figure 1 The gas velocity near the wall as a function of~the location on the circumference, denoted by a polar coor-dinate a, for Montz BI-250 packing.
Furthermore, the orientation of the sheets plays a role in the variation of wall flow over the cross section of the column. For example, for the BI-250 packing the gas flow through the annulus is higher at the open ends of the sheets perpendicular to the column wall. This is illustrated in Figure 1.
79
Gas snreading
Gas spreading has to be determined by means of a tracer gas due to the fact of gas being the continuous phase. Quantitative inves-tigations have been performed for one element of each of the packings in Table I. A relative small amount of carbon dioxide was injected at a point in the center of the bottom of the packing. The concentration profile leaving the sheet upright of the point source was analyzed channel by channel, thus providing a detailed picture of the mixing pattern of the gas.
Two adjacent sheets, having opposite orientations of their channels, show mirror image concentration profiles. The average result for each type of packing is presented in Figure 2.
Two mechanisms of gas transport, both independent of the gas flow rate, can be distinguished: lateral transport of the gas due to the flow through one channel, and mixing- of the gas between channels of two adjacent sheets. For 500Y the mixing effect dominates, while for the other packings lateral transport is the most important mechanisra.
Lateral transport alone does not flatten a gas concentration deviation in one element. However, it results in large scale mixing because successive elements are rotated 90 degrees.
Liquid profiles
The "natural" flow of the liquid, with a gas flow rate beneath the loading point, is almost established after the third element from the top, as illustrated in Figure 3. The liquid distributor used had 760 drip points/m2 and had a Mf .value of 1.01. This is because the number of drip points - although large - is smaller than the number of measuring elements.
The "natural" flow maldistribution is constant within certain limits. No influence of the superficial liquid veiocity was found down to a critical veiocity, u. . . Beneath this value the dis-
J Lmin tribution quality deteriorates. The "natural" flow maldistribution factors, together with the critical liquid veiocity, are presented in Table II.
80
X cd
u o o cd
u CU o cd u
0 25 r (cm)
Figure 2 The relative tracer concentration profile, resulting from a central point source injection up stream of one packing element, measured channel by channel over the length of one sheet. 1.0
Mf 0.5
0.0
■ ■
ir
1 2 3 Number of Elements
Figure 3 The maldistribution factor for the liquid phase as a function of the number of packing elements between the liquid distributor and the bottom of the packing for Sulzer 250Y (■) and Montz Bl-250 (*) (superficial gas velocity 1.8 m/s, superficial liquid velocity 3.4 mm/s).
81
1 2 3 Gas Velocity (m/s )
Figure 4 The maldistribution factors of the liquid phase for 4 elements of Sulzer 250Y (■), 500Y (O), Montz Bl-250 (•), BS-450 (o) and Raschig RaluPak 250YC (« ) and 5 elements of Sulzer BX ( • ) as a function of the gas velocity at a superficial liquid velocity of 3.4 mm/s.
In Figure 4 the Mf values of the liquid distributions are plotted as a function of the superficial gas velocity. The packing height is about 0.8 m. For all packings except Rpak the quality of the distributions deteriorates rapidly above a critical gas velocity, uGmax.
It was observed that partial flooding started at the bottom sleeve of the elements at a gas velocity of about 1.7 m/s for all packings except Rpak and BX. The liquid is prevented to flow through the sleeves. This results in a decrease of the wall flow factor, from 1.1 down to about 0.5. The effect of partial flooding deteriorates the overall distribution and is intensified at higher gas rates. The sleeves of Rpak are highly permeable to the liquid and are not folded round the bottom of the packing element. This may explain why we have measured that the wall flow for four elements varies from 3 in the absence of flowing gas, down to 1 at a gas velocity of 3 m/s.
82
Above the critical gas velocity liquid segregation occurs, as shown for 250Y In Figure 5. Near the wall two areas with a high liquid load can be distinguished. About 70 percent of the total flow passes through these areas.
Perhaps unexpectedly, the orientation of the areas is parallel to (and not perpendicular to) the sheet orientation of the bottom element. Moreover, the areas with a high liquid flow rotate by 90 degrees as the orientation of successive packing elements is changed. However, a flow rotation of 90 degrees between each element is impossible, because liquid cannot flow through a sheet. Only spreading parallel to the sheets is allowed. This "paradox" can be explained by assuming liquid migration over the sleeves around an element.
Figure 5 Three dimensional presentation of a liquid velocity profile leaving the bottom of the packing at a gas velocity of 2.0 m/s (upper part of the figure, Mf — 0.26) and of 3.2 m/s (Mf = 1.74) for 4 elements of Sulzer 250Y.
83
This assumption has been checked by removing the Standard sleeves and by wrapping the packing with tape to prevent any horizontal migration along the wall. This resulted in an improved liquid distribution, as illustrated in Figure 6.
Figure 6 Three dimensional presentation of a liquid velocity profile leaving the bottom of the packing at a gas velocity of 3.1 m/s (Mf = 0.43) for 4 elements of wrapped Montz BI-250.
Liquid spreadine
Liquid spreading experiments were performed by injecting a single jet of liquid in the centre of one element, so that the spreading between the sheets could be studied.
Two mechanisms of spreading, simular to those for gas spreading, are proposed to explain the observed phenomena.
The first mechanism only translates liquid: a jet remains in the channel through which it is flowing, without liquid crossing to the adjacent sheet. This mechanism is thought to be the main cause of spreading in Rpak, 250Y and 500Y (Figure 7).
The second mechanism redistributes the liquid to some extent at each contact point of adjacent sheets. This mechanism may explain the profiles measured for BX, BS-450 and Bl-250 (Figure 7).
Because successive elements are rotated by 90 degrees both mechanisms lead to large scale mixing.
- 84
Figure 7 Three dimensional presentation of a liquid velocity profile after one element of packing resulting from a liquid point source for Sulzer Mellapak 500Y (left) and Montz BI-250 (right).
Discussion
The observed low maldistribution of the gas agrees well with 14 15 previous studies ' . Gas spreading profiles comparable to the 12 results of Meier are found for 500Y, but for the other types of
packing a different spreading mechanism can be involved. Because of the combination of mixing and translation inside the packing element and large scale mixing between the rotated elements, one should be careful to use the diffusion model to express the mixing performance of the packing.
13 The assumption of Hoek that below the loading point the "natural" flow is not influenced by the gas was confirmed by our experiments. The "natural" flow is rapidly established, provided a properly designed liquid distributor is used. Also an increase of maldistribution was found when applying low liquid loadings. A
85
reduction of wall flow by gas flow was observed below the loading point.
The results for the liquid spreading show the same behaviour as 13 the measurements of Hoek . The gas has a negligible effect on the
spreading. For the various types of packing two spreading mechanisms, comparable with the gas phase, are suggested. Again the definition of the radial spreading coëfficiënt, as used in the diffusion model, is not appropriate for low height of structured packings.
The origin of the large scale liquid segregation is related to the nature of the sleeves between the packing and the column wall. The precise flow paths of both phases are as yet not known. However, local flooding on the sleeves, horizontal liquid migra-tion through the annulus and an excess of gas wall flow at the open sides of the sheets are thought to be the main contributions to the development of this large scale maldistribution.
A similar break point F-factor is observed when comparing Figure 4 with HETP-curves from literature . It seems probable that the increase in HETP-value is caused by the large scale liquid segregation.
Conclusions
Gas maldistribution can be ignored at all gas velocities as compared to the liquid maldistribution. The flow through the annulus between packing and column wall is about 1.25 times higher than the superficial gas flow. Gas spreading is caused by mixing inside an element (between adjacent sheets) and by a combination of translation and rotation between two elements.
Up to the loading point no severe disturbances were found in the liquid. Only a reduction of the wall flow was observed. Above the loading point the maldistribution increased rapidly, accom-panied by large scale liquid segregation for some packings. The spreading of the liquid is a result of translation and mixing between adjacent channels.
Although not quit well understöod, the sleeves around the packing . elements play an important role in the formation of large scale maldistributions above the loading point. Optimization of the sleeve performance can result in a higher loading limit for structured packings.
86
Acknowledgements
We would like to thank Julius Montz GmbH, Raschig GmbH and Gebrüder Sulzer AG for supplying us with packing. Furthermore this research project would not have been possible without the finan-cial support of the Koninklijke/Shell-Laboratorium, Amsterdam and the zest for work of all students involved.
References
1 R.M.Stikkelman, This thesis chapter II
2 M.Huber, A.Sperandio Chem. Ing. Tech., 36 (1964) 221-227
3 A.Sperandio, M.Richard, M.Huber Chem. Ing. Tech., 37 (1965) 322-328
4 M.Huber, R.Hiltbrunner Chem. Eng. Sci., 21 (1966) 819-831
5 P.Flatt Chem. Ing. Tech., 38 (1966) 254-259
6 W.Meier, M.Huber Cnei7~IngTTëcïT7~3 9~( 19 6 7 )^7 9 7 ^ 8 0 2 — —
7 W.Meier, M.Huber I . Chem. E. Symposium Ser ies No.32 (1969) 4.31-37
8 M.Zogg Chem. Ing. Tech., 45 (1973) 67-74
9 H.C.Yuan, L.Spiegel Chem. Ing. Tech., 54 (1982) 774-775
10 W.Meier, W.D.Stöcker, B.Weinstein Chem. Eng. Progr., 73 (1977) Vol.11 71-77
11 W.Meier, R.Hunkelar, W.D.Stöcker Chem. Ing. Tech., 51 (1979) 119
12 W.Meier, R.Hunkelar, W.D.Stöcker I. Chem. E. Symposium Series No.56 (1979) 3.3/1
13 P.J.Hoek Ph.D. Thesis, Technische Hogeschool Delft, 1983
14 R.M.Stikkelman, J.A.Wesselingh I. Chem. E. Symposium Series No.104 (1987) B155-164
15 R.J.Kouri, J.J.Sohlo I. Chem. E. Symposium Series No.104 (1987) B193-211
16 H.G.Groenhof, S.Stemerding Chem. Eng. J., 14 (1977) 193
17 L.Spiegel, W.Meier I. Chem. E. Symposium Series No.104 (1987) A203-215
APPENDIX A
Characterization of the flow distributions in a cross section of a packed column
Introduction
Designers often assume plug flow for both the gas and liquid in packed columns1"4. However non-ideal flow of both phases^has often been observed and described as wall flow and channeling . Non-Ideal flow is also thought to contribute to the malfunctioning of industrial packed columns.
Better design procedures will have to take the non-ideality in account. This requires a description of the flow, preferably in terms of only a few parameters. In the past measurements have been made in which the flow rates , *[j], in a cross section have been determined in a large number of raeasuring elements . The mal-
"^Jj^^^^TwIs^delcHbêd^by" a raTldi¥tFiblïtioïr"faceor^(Mf) ~:
Mf - i *s ( ~i$;— )
This Mf is related to the Standard deviation of the flows in the measuring elements. It has a value of zero if all elements have the same flow; its value increases rapidly as the flow be-comes less regular.
The Mf has one serious disadvantage; it does not convay any information on the pattern of the distribution. As an example a number of cross sections are shown in Figure 1. In these illustra-tions the black elements all have a flow twice the average, while the white elements have no flow. It should be obvious that the flow distribution deteriorates going from pattern E to pattern A. Yet the Mf's of all these pattems are equal!
A better description of the maldistribution is developed in this paper. The proposed method discriminates between patterns such as those in Figure 1.
89
A B
D
Figure 1 Five "checkerboard" distributions having an equal value of the maldistribution factor (Kf=l).
W
Figure 2 A typical measuring grid. The enlargeraent of the wall section shows superimposed point sources (.) and excluded points (o)
90
Channel and overall distribution
Consider a cylindrical column with a diameter D. The cross section is divided into a large number of square measuring elements with a side of length b (Figure 2). The flow $[j] in each of these elements is assumed to be known. For calculational purposes each element is considered to contain a number (here sixteen) of equal point sources. Some elements fall partly outside the column. The flow in the elements that fall partly outside the column is distributed over those point sources in the element which lie inside the column boundary.
We define a channel as the circular part of the cross section with a diameter d ( see Figure 3 ). This diameter may have any value between zero and the column diameter. Each channel is characterized by the coordinates (x,y) of its centre and by its diameter. The flow through a channel *(x,y,d) is the sum of the point sources inside the channel.
Figure 3 Possible (x,y)-positions for a channel of size d in a column of diameter D.
91
For each channel size d, a channel maldistribution ra(d) is determined. The centre of the channel is moved in discrete steps from point source to point source. All positions are covered in which the channel does not intersect the column wall. At each position (there are n(d) of them) the channel flow tf(x,y,d) is calculated. A channel maldistribution is then defined as:
m(d) = [ S( *(x,v d) - <*(d» ) 2 ]0.5 * _ A _ ( 2 )
<*(d)> * (n(d)-l) A(d)
The left-most contribution in formula (2) is the relative Standard deviation of the channel flows. The right hand factor is chosen such that the m(d)'s of different channel sizes are all based on the same area.
In the calculations presented here the channel diameter has been increased in steps equal to the side b of the measuring elements. The smallest channel has a diameter b, and the largest D-b. A channel with a diameter 0 has no physical meaning, and a channel with diameter D always yields m(D)=0.
An overall channel factor is defined as:
b M - - * S m(d) ( 4)
D
The diameter ratio in this factor makes the value relatively insensitive to the choice of the size b of the measuring element.
Two main concepts have been introduced in this paper:
the channel maldistribution of different channel diameters m(d) and the overall maldistribution M
As will be shown m(d) gives an impression of which channel sizes are the major contributors to the maldistribution. The overall maldistribution can be used to rank the quality of différents distributions: this ranking generally agrees well with what one would decide on the basis of visual observations.
92
Sample distributions and discussion
A number of more or less regular distributions will be analysed to see what this method of characterization yields. A relative diameter, d , expressed in units of one measuring element length, is introduced.
A checkerboard distribution
This distribution is shown in Figure 4. The flow carrying areas have dimensions of 5*5 measuring elements. Some areas contain fewer measuring elements because they are cut off by the column wall.
0 10 20 Relative Channel Diameter
Figure 4 Channel maldistributions as a function of a relative channel diameter for a "checker board" distribution
The channel maldistributions are shown for different channel diameters. These are given in terms of the diameter of the measuring elements. As is to be expected, there is a major contribution from channels of five or less element diameters. Channels with a diameter of about nine elements contain approximately equal amounts of "black" and "white". They hardly contribute to the
93
maldistribution. Relatively large contributions are found for channels of around 12 and 17 element diameters.
In the previous case all elements have either twice the average flow or no flow. Figure 4 also shows the same distribution but with all elements having either one and a half times or one half of the average flow. As to be expected the distribution of the channel maldistributions is the same, but their values are lower.
A column with an irrigated outer ring
The distributions, shown in Figure 5, simulate some of the aspects of wall flow. The outer ring has a width from one to ten measuring elements. In the last case there is a uniform flow over the whole of the column. To obtain an understanding of the be-haviour of this example first consider the case with an outer ring one element wide. There are then two channel sizes contributing largely to the maldistribution:
channels with a diameter of 18 elements (the inner circle) channels with a diameter of 1 element (portions of the outer ring).
0 5 10 15 20 Relative Channel Diameter
Figure 5 Channel maldistribution for rings with a different inner radius
94
As the width of the outer ring is increased the two channel sizes move towards each other and the overall maldistribution decreases. Note that channels with a diameter larger than the inner circle do not contribute to the maldistribution.
A point source
Figure 6 gives the overall maldistribution of a single point source as a function of its position. First of all notice the very high values of the overall maldistribution. A single point is of course a very poor distribution. Note also that this distribution is termed best if the point source is in the centre of the column.
5 10 Relative Displacement
Figure 6 Overall maldistributions for a point source as a function of its position
95
A series of checkerboard dlstributions
The overall maldistributions of the checkerboard distributions as presented in Figure 1 are:
Distribution Overall Maldistribution M
A 3.027 B 0.847 C 0.490 D 0.268 E 0.127
What one percieves to be a good distribution does indeed have a low value of M.
Conclusions
Two entities have been constructed to describe the ideality of a distribution:
the channel maldistribution m(d), which depends on the chan-nel size chosen and the overall maldistribution, M, which is a summation of contributions from the different channel sizes. The channel maldistribution function m(d) gives an impression
of which channel sizes give the major contribution to the overall maldistribution M. The latter parameter permits different flow distributions to be ranked.
Thls method is not restricted to flow distributions but can also be applied to other distributions such as those of concentra-tion or temperature.
We are currently evaluatlng maldistributions for different types of packing, different packing heights, initial distributions and operating conditions on a pilot plant scale in a column with a diameter of 0.5 m. It is hoped that such figures will contribute to a more rational design of packed columns.
96
Svmbols
2 A Area of column cross section [m ] 2 A(d) Summation of all the areas of [m ]
used channels b size of measurlng element [m] D column diameter [m] d relative channel size [-i relative distance [? M overall maldistribution factor [-m(d) maldistribution factor for a channel [•
with size d n number of areas [-]
3 -1 <$> average flow through the measuring areas [m s ] 3 -1 $[i] flow through the i-th measuring area [m s ] 3 -1 <*(d) Average flow through all channels [m s ]
with size d
References 1 W.H.Walker, W.K.Lewis and W.H.McAdams Principles of Chemical Engineering (1927) McGraw-Hill Book Co., New York
2 T.H.Chilton and A.P.Colburn Ind. Eng. Chem., 27 (1935) 255
3 R.H.Perry and D.Green Perry's Chemical Engineers' Handbook (1985) 13-96/97 and references therein McGraw-Hill Book Co., Singapore
4 T.Baker, T.H.Chilton and H.C.Vernon Trans. AIChE., 31 (1935) 296
5 J.C.Charpentier. Ph.D. Thesis, University of Nancy, 1968
6 K.E.Porter, V.D.Barnett and J.J.Templeman Trans. I. Chem. E., 46 (1968) 69
7 H.G.Groenhof Chem. Eng. J., 14 (1977) 193
8 G.G.Bemer and F.J.Zuiderweg Chem. Eng. Sci., 33 (1978) 1637
9 P.J.Hoek Ph.D. Thesis, Technische Hogeschool Delft, 1983
10 J.Szekely and J.J.Poveromo AIChE. Journal, 21 (1975) 769
11 Q.H.Ali Ph.D. Thesis, University of Aston, 1984
12 C.Krebs Chem. Eng. Process., 19 (1985) 129-142
13 P.L.Speddlng and G.R.Llghtsey Chem. Eng. J., 32 (1986) 151-163
97
14 F.J.Zuiderweg, P.J.Hoek and L.Lahm Jr. I. Chem. E. Symposium. Series No.104 (1987) A217
15 J.G.Kunesh, L.L.Lahu and T.Yonagi I. Chem. E. Symposium. Series No.104 (1987) A233
16 R.M.Stikkelman and J.A.Wesselingh I. Chem. E. Symposium. Series No.104 (1987) B155
17 R.J.Kuri and J.J.Sohlo I. Chem. E. Symposium. Series No.104 (1987) B193
18 J.Stichlmair and A.Stemmer I. Chem. E. Symposium. Series No.104 (1987) B213
19 F.J.Zuiderweg and P.J.Hoek I. Chem. E. Symposium. Series No.104 (1987) B247
20 P.J.Hoek, F.J.Zuiderweg, J.A.Wesselingh Chem. Eng. Res. Des., 64 (1986) 431-449
21 H.C.Groenhof Chem. Eng. J., 14 (1977) 193
98
APPENDIX B
Descriptlon of the computer programs used
General information
This appendix provides a global reference to the options ap-pearing within the programs that were used for the simulations as described in chapter IV and V. Three programs, written in Turbo Pascal code, are available: FLOWSIM.PAS, MASSSIM.PAS and ANALYZE.PAS. The most important parts of these programs are docu-mented in the source code.
The compiled versions, FL0WSIM.COM, MASSSIM.COM and ANALYZE.COM, can be used within the MSdos environment. However, there are some limitations: - A IBM compatible computer with 8087 processor is needed -—A-c:driye^is_assumed (preferably a virtual disk) The number of cells on the column diameter should be lower than 21. This limitation can be changed by adjusting the first lines of the source code of all programs.
The flow simulation program
After starting the program information is gathered to fix the column configuration and packing characteristics. First you have to choose the type of simulation by highlighting a option. The following message appears:
iquid flow simulation
as flow simulation
ass transfer preparation
C ontinue
Make your choice and press fc] to continue.
99
Enter all data needed for the simulation
Column diameter Number of element on radius Thickness of wall zone Mf on 0.025 m scale Liquid D -value W.-value Gas D -value
r W„-value Column height
[m]
[m]
[m]
[m]
[m]
The constants can be found in chapter III. Automatically the corresponding splitting factors S , S and cell heigth h are calculated.
Next the initial distributions have to be entered. Earlier saved distributions can be loaded. New distributions can be created with the Edit option and eventually saved on disk. Just select your option.
oad data
dit data c:all_zero.liq
ave data
C ontinue
In the editing mode you can move the active cell by the cursor keys. The active cell points the place where you can enter an integer value of the flow or velocity (press [x]). The distribu-tion can be altered quickly by using the [Tjrace option; the active value shadows the active cell, or by using the [XJdjust all option.
After the editing sessions the program asks for the filenames to store simulations results. Finally all values can be printed. The simulation starts, showing all intermediate results on the screen. Typical run times are in the order of 10 minutes. Top or bottom distributions can be saved on disk. The calculated profiles can be used in MASSSIM.COM to simulated mass transfer or can be analyzed with ANALYZE.COM.
100
The mass transfer simulation program
This program uses the data produced by the flow simulation. It starts with a questionnaire which is partially filled in. With the average liquid and gas densities and the average molecular weigth the flow profiles are transformed to moles. Eventually the L/V ratio can be forced to unity by adjusting the molar gas flows to guarantee total reflux. Up to now the program only handles total reflux distillation. After entering the volatility, the local HETP, the top concentration of the liquid and some filenames the concentration profiles are calculated. The simulation time is in the order of 10 hours. The results can be visualized with ANALYZE.COM.
The evaluation program
The ANALYZE.COM program is a powerful tooi to study the simu-lated profiles. The options are:
[cjrnf-factor The channel maldistribution of different channel diameters and the overall maldistribution M conform the definition in Appendix A.
[F]requencies Histogram of the distribution. [S]f-factor The relative variancy ( maldistribution factor ) of
the distribution. [R]ings The average velocity through concentric rings with
a constant thickness or area and the Wall flow factor Wf.
|s|egments The average velocity through segments. [Ï]hree-D A three dimensional presentation. [vjalues The simulated values of a cross section scaled by a
factor [Y] .
The programs are available for people who are interested in this subject.
101