You are on page 1of 11

Pergamon

PII: bUUU~-Z~U~(97)UU~gl-I

('hcmilul litlHineerittq Scieme, V o l 52, Nt~ 24, pp. 4461 4471, 1997 1997 Flsevier S~:icncc Ltd. All rights lcscrvcd Printed m ( h e a t Britain

01)119_'5o997 S17t~o ~tl~lll

Gas-liquid flow in slightly inclined pipes


Eric Grolman* and Jan M. H. Fortuin*
Department of Chemical Engineering, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands

Abstract The modified apparent rough surface (MARS) model is presented, which facilitates prediction of liquid holdup and pressure gradient in gas liquid flow through horizontal and slightly sloping pipes. The model is based on two steady-state, one-dimensional momentum balance equations, which include the superficial velocities and transport properties of gas and liquid, the diameter of the pipe and its angle from the horizontal. Separate correlations are proposed for the interfacial friction factorJl, the liquid-to-wall friction factor j}, the interfacial perimeter S~ and the wetted perimeter SL. Predicted values of liquid holdup and pressure gradient are verified against results of 2400 carefully performed, laboratory experiments in horizontal and sloping glass pipes ( - 3 ~< fl ~< + 6') of 15, 26 and 51 mm diameter. The liquid holdup ct, ranges from 0 to 0.42, and the superficial liquid velocity ULSfrom 0 to 0.06 m s 1 and the superficial gas velocity UGs from 1.8 to 34 m s- ~. The gas liquid systems used are air~water and air/tetradecane (n-C~4H3o) at atmospheric pressure and room temperature. Significant effects of small inclination angles were found at low gas flow rates, leading to an eightjbld increase in liquid holdup and more than afimrlohl increase in pressure gradient compared to horizontal flow. Nevertheless, the average relative error in the prediction of liquid holdup and pressure gradient is less than 10%. ,C 1997 Elsevier Science Ltd

Keywords: Gas-liquid pipe flow; inclined; holdup; pressure gradient.

INTRODUCTION

Liquid holdup (~:L) and pressure gradient ( - dP/dL) are thc key parameters in the design, scale-up and control of pipelines carrying gas and liquid simultaneously. Whereas for steady-state horizontal flow, these variables can be estimated with relatively high confidence, this is not the general case for inclined flow. This study shows that angles of inclination as 0.1 can significantly affect liquid holdup and pressure gradient in the low gas velocity range. Therefore, many commercial pipelines are bound to include inclined sections. The present study is restricted to low liquid loading (ULS < 0.(/6 m s ~) and slightly inclined ( - 3 ~< fl + 6 ) gas liquid pipe flow, such as typically encountered in gas transportation pipelines, where small amounts of liquid hydrocarbons are formed by retrograde condensation (Stewart, 19881. Conditions of low liquid loading may also be found in condensing (steam) flows, flow of lubricant oil in gas and occasionally in two-phase oil/gas transportation pipelines. The effect of small flows of liquid on the pressure gradient, is tinderpredicted by many models found in

* Current address: DSM Research, P.O. Box 18, 6160 MD Geleen. The Netherlands. ('orrcsponding author.

literature (Hamersma and Hart, 19871. When condensate turns up in natural gas, suppliers are frequently charged by pipeline-operating companies for the estimated increased pressure drop and decreased gas throughput. Correct estimates require accurate models for these low liquid-loading flows. The gas flow rate at which liquid will still be carried out of lower areas in pipeline systems is an important operating limit, which defines the minimum gas flow rate for steady-state operation. Lower gas flow rates result in periodic pipe blockage and excessive liquid production when throughput is increased. The daily and annual cycles in the demand of natural gas are known to pose these problems in pipeline networks, requiring models for the higher as well as the lower range of gas flow rates. An important and well-recognized concept in dealing with gas liquid pipe flow is that of flow patterns (Barnea and Taitel. 1986). The present paper is restricted to separated flow of gas and liquid, i.e. to the stratified-smooth, wavy and annular flow regimes, for which average interracial (gas-to-liquidK liquid-towall and gas-to-wall perimeters are easily identified, and corresponding friction factors can be defined (see Fig. 1). Although transitions to slug flow were occasionally encountered, no attempt was made to do measurements in oi- to model the intermittent flow regimes. Prediction of the flow pattern transition

4461

4462

E. Grolman and J. M. H. Fortuin

sG

~ G w

~_.._.._.~

- - M o d e l s for the interracial shear stress based on wave properties, have also been found insufficiently accurate. This is due to the fact that in horizontal pipes no waves are formed at low gas velocities, while in (upward) inclined pipes they always are. The fact that present-day theoretical models do not predict trends that adhere to our experimental results from inclined pipes has posed some extraordinary difficulties in describing our data. The problem was solved with a correlation for the interracial friction factor that treats the gas-liquid interface as a mobile wall of apparent, continuously variable (sand) roughness. The correlation is based on governing physical parameters, does not contain the angle of inclination explicitly, is dimensionally consistent and does the job well for 2400 sets of measurements under a wide range of conditions. Nevertheless, it presently lacks an appealing theoretical foundation. During the experiments, 13 different angles of inclination between - 3 and + 6 , three pipe diameters, i.e. 15, 26 and 51 mm, and the two systems air/water and air/tetradecane were used. For all positive angles of inclination, the lower bound of the gas flow rate was limited only by the onset of slug flow. MODEL
Steady-state momentum balance

~LW

Fig. 1. Schematic view of stratified-wavy flow in an inclined pipe, showing the conventions used.

boundaries was presented elsewhere (Grolman et al., 1996), while the transients leading to flow pattern transition were also modeled in a previous paper (Grolman and Fortuin, 1996). The scope of the present paper is limited to the prediction of liquid holdup and pressure gradient in steady-state, slightly inclined ( - 3 - < - <+ 6 ) gas-liquid pipe flow in the separ% fl % ated flow regimes, based on phenomenological concepts. In lowering the gas flow rate in upward sloping pipes, there exists an increasingly fine balance between the frictional and gravitational contributions to the loss of momentum. The gas-liquid interfacial shear stress zl is the main frictional parameter to determine whether or not liquid will flow back and accumulate. Conventional models for the interfacial shear stress are usually based on results obtained in horizontal pipes. These models do not lead to acceptable predictions of liquid holdup and pressure gradient in inclined pipes at decreased gas flow rates, for the following reasons: - - S m o o t h - p i p e friction factors for calculating the interracial shear stress are insufficient, because waves on the gas-liquid interface generate a roughened surface, with increased interracial friction. Whereas in horizontal flow, waves may not appear at low gas velocities, in inclined pipes they do. - - Models for the interfacial friction factor that are based on the liquid velocity or on the liquid-phase Reynolds number, fail on the current set of data, because ReL can become arbitrarily small in inclined gas-liquid flow. The latter can be deduced from the fact that, given sufficient time, any (high) liquid holdup value can be achieved with a minute liquid feed, if the pipe is inclined and the gas velocity is sufficiently small. Obviously, interfacial friction is then responsible for counterbalancing the tendency of the liquid to flow back on account of gravity. At these small liquid feeds and high holdups, the liquid Reynolds number bears little or no relation to the eventual liquid holdup obtained, hence ReL cannot be used to predict interfacial friction. Furthermore, under such circumstances much larger volumes of liquid are involved in oscillatory wave motion than in the actual transport of liquid, which makes ReL a poor parameter to describe the governing physical processes of interfacial friction.

The modified apparent rough surface (MARS) model is based on the steady-state gas and liquid phase momentum balance equations (1) and (2) for the stratified smooth, wavy and annular flow regimes. Wavy flow is shown schematically in Fig. 1 and with additional measuring equipment in Fig. 2. The two balances are

A~ - ~

AL - d- L = ZLwSL -- ziSi + ALPLg sin ft.

[dq Edq

G:

"~GwSG "4- z i S i -~- A o p ~ g s i n f l

(1)

(2)

Equations (1) and (2) may be combined to give the equation commonly solved for the steady-state liquid holdup, provided the steady-state pressure gradients in gas and liquid are equal, i.e.
[-- dP/dL]G = [ - dP/dL]L = [ - dP/dL]

and for steady-state flow,

+G
+ (PL -- PO)g sin fl = 0.

(3)

The assumption of equal average pressure gradients in gas and liquid is generally valid in the absence of interfacial liquid level gradients, on both experimental and theoretical grounds, for horizontal and inclined flow alike.

Gas-liquid flow in slightly inclined pipes

4463

MEASURING EQUIPMENT : Angle gauge (pipe) Angle gauge (liquid) C(ontrol) F(low) P(ressure) T(emperature) W(eight measuring equipment) O Connected to computer
Turbine meter Cyclone Heat Exchanger Waterring Compressor Cyclone Pump I Precisi~lance 1 Ambient temperature . . . . . . "-'~ L~ Precisi~"~lance 2

4+
Fig. 2. Schematic diagram of the experimental setup. Equations (1) (3) contain:
-three cross-sectional areas,
At,,, =

rcD2/4,

At. = ct.Atot,

A~ = (1 - eL)Atot;

(1989) model. Judicious examination of the data and the original model, allowed a rewrite of the governing parameters, leading to the following correlation with liquid-phase Weber and gas-phase Froude numbers:
0MARS = (JO(Gwalcr/O.)O. 15

(4)
- t h r e e parameters, where s~ and 0 are the dimensionless perimeters to be modeled,

o.2s - o.s ,~ + gels. Fr<; p<;., ~(PL -- p(;)cosfl} with

(7)

Si - sdzD,

Sc = O~zD,

S<; = (1 - O))zD

(5)

WeLs =
-three shear stresses, where the friction factors f are to be modeled:
~ /,, ~ Z(iW =.IG2[ GU&S. {1 -- EL}-, b/GS ~. 1 )

and

Fro; =

(1 - cLt~gDJ

TLW =.[L~pLUt.s/EL,

2 / 2

(6) r~=.I1,; l - c 0
ui

Equation (7) is successful in describing both the original data of Hart et al. (1989), as well as those from inclined pipes in the present paper. In a pipe of circular cross section, the m i n i m u m attainable wetted wall fraction 00 is, on geometrical grounds, related to the liquid holdup ~:L by q. = 0o - {sin(2rC0o)}/127r). (8)

The circumferential and inter[acial perimeters S L and S(; and Si it is an experimental fact that in wavy gas liquid pipe flow, liquid creeps up against the inner sides of the pipe to varying extent. Although this effect is rarely accounted for in literature, it significantly affects the interracial perimeters &, Sc and Sa, as well as the interfacial friction factor 1). Hart et al. (1989) used the wetted wall fraction 0( = SL/{~zD}) to successfully model the average perimeters in horizontal gas-liquid pipe flow. Their model was based on the partial conversion of liquid-phase kinetic energy into potential energy, leading to a liquid-phase Froude n u m b e r as the governing parameter for increments in 0. For non-horizontal flow, the wetted wall fraction deviates considerably and systematically from the Hart et al.

When the liquid holdup eL is known, the m i n i m u m wetted wall fraction 0o is calculated iteratively from eq. (8). However, 0o can be approximated without significant loss of accuracy by the following explicit equation: 0o ~ 0.624c~,"3v~. (9) In Fig. 3, measured and modeled values of the wetted wall fraction 0 are compared, with 0MARSfrom eq. (7), using 0o from eq. (9). The average interfacial perimeter Si may vary between a m i n i m u m value of {D sin0z0o)} for stratifiedsmooth flow (0 = 0ot and {(1 - t:L)'STrD} for cylindrically symmetric, closed annular flow ( 0 = 1), depending on the value of 0. A linear interpolation

4464
i i iii I I i i ~ i I ~ I I I I /

E. Grolman and J. M. H. Fortuin


r J I

+ +
-

+ ++
+ +

/O

0.

I"RSl

l+0%, ,+6oi

fl = 0

'

/00/4

+~ +++~+}+
7+ +

A -3~; < 0 [ fl 0 fl=O


+

0< fl ' + 6

D = 26 mm

0.6

0.4

0.2

.1
i i I III 100
I i I

I ~1 1000

O~

0.2

0.4

0.6

0.8

Fig. 3. The modeled wetted wall fraction 0MARS from eq. (7) vs the measured wetted wall fraction 0~xpfor air~water pipe flow.

Fig. 4. The measured liquid-phase friction factor fz,xp, as a function of the liquid Reynolds number ReL for air~water pipe flow.

between these two extremes renders the best estimate of Si:


Si =

0.04 0.03 ~ 0
+
--0.05 ;

k/D

= 0.2- - - - - - - -~+~ + - --

Si ~tD

1 - 0 sin n0o 0 - Oo - ( l - - e L ) '5q 1 -- Oo 1 -- 0 0 (10)

+
%

++ +
~ ~

+
B
o,, ,

0.02

The dimensionless interfacial perimeter s~ has also proven to be important for modeling the interfacial friction factor f~.

The measured friction factors fL.oxp and fi,~xp


Having measured the pressure gradient - d P / d L , liquid holdup eL and wetted wall fraction 0, four parameters in the m o m e n t u m balance equations (1) and (2) remain unknown: Si, raw, Zzw and q. A good estimate for S~ is obtained with eq. (10). For want of a better method, it is customary to treat the gas-towall shear stress Z~w as that of a dry pipe wall and to calculate the gas-phase friction factor f~ in eq. (6) as a function of Re~, using an existing, empirical, singlephase friction factor equation for smooth pipes (Eck, 1978): 0.07725

0.01 "-1+ no.a..-~ol

0.01~

10~

5.10 4

10 5

Fig. 5. The measured interfacial friction factor f~.,xp, as a function of the gas-phase Reynolds number Rea, showing several lines of constant relative roughness k/D for air~water pipe flow.

p~u~sD
with Re~ = qo( 1 _ 0 + si)" (11) lines of constant relative roughness k/D, calculated with the empirical, rough pipe friction-factor equation (16) (Eck, 1978). It is clear that neither of the two friction factorsfL a n d f can be modeled as a function of merely a Reynolds number.

f6 = {logxoERea/7]}2

With plausible estimates of S~ and Vow, the remaining parameters %w and ~ are calculated with eqs (1) and (2), using measured values of - dP/dL and eL. The measured values of the friction factors,fL,xp andf~. ~p, are then obtained by rearranging eq. (6). Figures 4 and 5 showfz.oxp a n d ~ . o ~ as a function of ReL and Re~, respectively. In Fig. 4, the well-known laminar and turbulent friction factor lines for single-phase pipe flow are also drawn. In addition, Fig. 5 shows several

The liquid friction factor fL,Chan, in downward openehannel flow


F r o m Fig. 4 it is clear that the liquid-phase friction factor fz in inclined gas-liquid pipe flow generally is much larger than its single-phase counterpart. The

Gas-liquid flow in slightly inclined pipes v',r i~11 r i i , ~ i Fi I ~ i J tll

4465

/x f l = - I "~ A X fl =-2
i

01_ U - 2

to= 6m

t
i

I
1 .

I )1.

[. 0.01

I ,I

~[
100

I I

IIII
1000

! ~li~l
0.1

I I iJl!l
1

Fig. 6. The liquid-phase friction factorfL,Cha, in open-channel down flow of water, as a function of the liquid Reynolds number ReL.

Fig. 7. The measured liquid friction factorJk.~xp, as a function of the modeled liquid friction factorfL. MARS air~water for pipe flow.

latter is represented by the solid lines marked '16/ReL" (laminar flow) and 'eq. (11)' (turbulent flow). Strangely, extremes in fL are obtained for inclined pipes, at relatively low Reynolds numbers. For such conditions, recirculation currents in the liquid-phase would be expected, possibly with countercurrent liquid flow near the bottom pipe wall. This would locally reverse the liquid-to-wall shear stress rLW and, on average, lead to a decrease in the liquid friction factor ft. However, the results reveal completely the opposite effect. The matter was pursued further with measurements of a wavy liquid layer flowing down the sloping glass pipe on account of gravity only. Under such conditions, the interfacial shear stress q and applied
/

cocurrent gas-liquid pipe flow there exists a gas-phase

contribution to the liquidfriction factor.[k. The MARS model fi~r fL Hart et al. (1989) introduced the correlation fL/f = 108Rez~)726, which leads to very good results
in the prediction of liquid holdup (Franca and Lahey, 1992; Spedding and Hand, 1995). Among other desirable properties, this correlation contains a gas-side contribution to the liquid friction factor through the interfacial friction factor.~. The correlation was modified by G r o l m a n (1994) for the low liquid Reynolds range, but otherwise could not be faulted, giving the following equation fOrJL.MARS:
\0.274

f/202
IL, MARS

Cwar)
~-L

OoReLsRe~ ReLs -- pLULsD


ilL

w h e n ReL < 2100 (13) w h e n ReL >1 2100.

[ filO8 Re[s'T26;

pressure gradient - dP/dL are zero, reducing eq. (2) In Fig. 7, the measured liquid friction factorJ~.cxv is to just two terms. The friction factor in open-channel plotted as a function of fL. MARS. The friction factor liquid flow fL.Chan, is obtained as a function of eL, O, fL. MARS is calculated with eqs (9)-(11) and (13), in ULS and fl after substituting eqs (4)-(6) into eq. (2): which f is obtained from eqs (14) (16). Comparing Figs. 4 and 7 shows that Jk, MARSfrom eq. (13) yields 4 D better values for fL.oxp than can be obtained with Z L w S L ~- --ALPLgSinfl, f L , C h a n - 0 2U2s gsin(--fl). a friction factor based on purely the liquid Reynolds (12) number. Figure 6 shoWSfL, Chanas a function of ReL. The range of Reynolds numbers in Fig. 6 is comparable to that of Fig. 4. Clearly, wavy open-channel liquid flow without a flowing gas phase, does not exhibit friction factor values systematically above those generally encountered in single-phase pipe flow. Consequently, in

7"he MARS model for fi


The interracial friction factorf, is the most difficult to model frictional parameter in gas-liquid pipe flow. Extensive study of all the available data has resulted in a set of three equations, from which f~ has to be solved iteratively. The iterative, implicit nature makes

4466

E. Grolman and J. M. H. Fortuin Equation (15) represents the hyperbolic tangent curve that was fitted through the data when [k/O]s]5/cs. (instead of the original k/D) was plotted as a function of Fn. The interfacial friction factor .// follows from the regular rough-pipe friction factor (16) and is substituted back into eq. (14) until convergence is reached. The iterative nature of the algorithm allows f to become very sensitive to small changes in ~:L and U~s under conditions close to instability, but much less so under more common conditions of higher U~s. Moreover, this sensibility of//to C, and U~s allows the L effects of all 14 small angles of inclination to be incorporated. The correlation properly discriminates between angles of inclination differing as little as 0.25, even in areas where the sensitivity to small changes in UGs is quite extreme (Figs 10-12). Before the interfacial shear stress z~ can be calculated from Jl and eq. (6), the average interracial velocity ui must be estimated. Based on wave velocity considerations, the following empirical equation was obtained (Grolman, 1994):

this the most complicated correlation presented here, and probably the most complicated correlation for f~ available today. However, after having generated over 40,000 plots in a relentless search for good correlations, the chances of still finding a general, explicit correlation forJ~ have become rather small. The MARS-model interfacial friction factor.//is obtained by applying the following iterative procedure: calculate the friction number Fn, initially using an estimate f o r f (e.g. f = f a )

Fn

f,.

(u~s~

(0.05 +f)(1 - ~,,)1., \ . , / ~ /

x\~,~

\-----~ j

(14)

--calculate the frictional relative roughness k/D: k -- = 0.5145eLSf 1.5 D x {tanh[0.05762(Fn - 33.74)] + 0.9450} (15) using eqs (7), (9), (10) and (14) for si and Fn, calculate the interfacial friction factor using the following rough pipe friction-factor equation (Eck, 1978): 0.0625

I I.8UI~s/eL, Ui = (ULs/et.,

ReL < 2100 ReL >1 2100.

(17)

Generally, the effect of ui on liquid holdup and pressure drop is relatively small.

- - u s e this new estimate off~ in eq. (14) and repeat the process until convergence is reached, e.g. until the change in f~ between consecutive iterations is smaller than 0.1%. This correlation for f~ results in values for the gas liquid interfacial friction factor, which range from those of a smooth pipe (kiD = 0), to those of a pipe with highly roughened walls (k/D .~ eL~O).

Liquid holdup and pressure gradient The liquid holdup is solved iteratively from eq. (3) by substitution of the equations presented. A useful starting value for the holdup in the iteration is obtained from the correlation of Hart et al. (1989), which was derived for horizontal flow: initial estimate:
eL ~ - - " 1 + Uc,s \Pc . (18)

How did we arrive at this correlation? Originally, the relative (sand) roughness k/O calculated from measured pressure drop values, was found to show some correlation with the interfacial shear stress ri. Since interfacial waves are amplified by interfacial shear and vice versa, the existence of some correlation between kiD and ri was not surprising. More remarkable was the fact that after fitting a curve through the correlation observed, an iterative algorithm could be built that actually had predicting power, i.e. it would converge to a single, non-trivial value. Later incorporation of effects of pipe diameter, surface tension, liquid viscosity, requirements of dimensional consistency and the process of fine tuning to the data available, has led to the use of Fn [eq. (14)] instead of the original ri. However, Fn still contains the remnants of the interracial shear stress to the power one-half ("co.5 = [0.5fP6]'5U6s/(1 -- ~L)).

The pressure gradient is obtained by solving eq. (l) for ( - d P / d L ) when the liquid holdup and the shear stresses therein are known.
EXPERIMENTS

The experiments were performed in carefully aligned, slightly inclined glass pipes of 15ram (L=llm),26mm(L=llm)and51mm(L= 15m) diameter, using water-saturated air/water and air/tetradecane systems at atmospheric pressure. The transport properties of gas and liquid were calculated using measured temperatures and pressures where relevant, and approximately amount to: p a ~ 1.18kgm-3; q ~ 1 . 8 x l 0 - S p a s ; demineralized water: PL 998kgm-3; q L ~ 1 0 3pas; a ~ 0 . 0 7 2 P a m and tetradecane; PL ~ 762kgm-3; qL ~ 2.1 X 10 3 Pas; a ~ 0.028 Pa m. A schematic diagram of the experimental setup is shown in Fig. 2. The volumetric gas flow rate, relevant temperatures, the weight of the liquid to be injected into the pipe and the weight of the

Gas-liquid flow in slightly inclined pipes liquid collected at the end of the pipe, were sampled at frequencies of 0.2, 0.1 and 5 Hz, respectively, and recorded. The gas flow rate was measured with an Instromet turbine meter (+0.5%), the temperatures with Pt- 100 thermometers ( + 0.05 K) and the weights with digital Mettler PM-16 and PM-30 precision balances t _+0.5 g). The pipe was positioned at 14 different angles of inclination (_+0.005% using two specially designed angle gauges, one at each end. The experiments were performed isothermally ( + 1 K) at room temperature. Pressure drop was measured across the gas phase with inclined precision alcohol manometers, connected to pressure taps on the pipe which were either 3 or 6 m apart. The wetted wall fraction was determined by eye, using angle gauges fixed to the circumference of the glass pipes. RESULTS AND DISCUSSION Figures 8 and 9 show the modeled liquid holdup and modeled pressure gradient as a function of their measured equivalents, for two systems (air/water and air/tetradecane) and two pipe diameters (26 and 51 mm). The liquid holdup ~t~ ranges from 0 to 0.42, the pressure gradient - d P / d L from 0 to 400 Pa m - 1 , the superficial liquid velocity ULS from 0 to 0.06 m s - 1, the superficial gas velocity UGs from 1.8 to 34 m s and the inclination angle/3 from - 3 to + 6 . In Figs 10, 11 and 12, holdup, pressure gradient and wetted wall fraction from inclined gas-liquid pipe flow are plotted relative to the values obtained for horizontal flow, respectively. These figures clearly demonstrate the relative effect that small angles of inclination have, especially in the low range of superficial gas velocities. The modeled curves were calculated for seven distinct angles of inclination ( - 3, - I, 0, 0.25, 0.5. 1.0 and 3.0) and a fixed liquid flow rate of ii II
i , , i ii I ,

4467 ~,,:~

. ( -. ~./.d L ) L A R S M lOO Pa m "1 J

/ -2

:/

Pa m -1

lO

lOO

Fig. 9. The modeled pressure gradient --dP./dLMARS as a function of the measured pressure gradient - d P / d L ~ x p. Larger markers: D = 51 mm, others: D = 26 mm; bold markers: air/tetradecane, others: air~water.

s!

I /~

I '

I ~ I F I ~ I

/--

'

I q-i
L _ _ ,

2li<0- o,=o
-- ~J>O

7
6 II/ ~ } ,,

L,hor. [

slug.flow

4: 3

o.1 2 4 6 8 10 12 14 16 18 Fig. 10. The liquid holdup et divided by the holdup in horizontal air~water pipe flow ec, hor, as a function of the superficial gas velocity Uas; lines: calculated for ULS 2.5 x 1 0 . 3 m s - 1; points: measured.
=

o.{}1 ULS=2.5xlO-3ms-1. The markers represent measurements performed at the selected angles of inclination and at fixed liquid flow rates that lie within _+20% of ULS = 2.5 X 10 -3 m s -1. F o r the measured points, the matching horizontal-flow values were calculated with the M A R S model. Since predictions are relatively easy for horizontal gas-liquid flow, these graphs represent a fair comparison of model and measurements.

0.01

0.1

Fig. 8. The modeled liquid holdup EL.MARSas a function of the measured liquid holdup eL,exp. Larger markers: D = 5 1 m m , others: D = 2 6 m m ; bold markers: air/tetradecane, others: air/water.

4468
..v,....,.-

E. Grolman and J. M. H. Fortuin


I i I r I i I i

(dP/dL).
u

Shor. [
I

o #:o
+ fl>O

slugflow [] semi-slugflow
2K2~K 3

- +~ ~ 9

\i

E semi-slugflow

-/+2 5t t x V

mm,

~ ~

~.+

' m s "1

I I ~ I I I I I
2 4 6 8 10 12 14 16 18 2 4 6 8

I i I I I
10 12 14

I I
16 18

Fig. 11. The pressure gradient -(dP/dL) divided by the pressure gradient in horizontal air/water pipe flow -(dP/dL)ho, as a function of the superficial gas velocity U6s; lines: calculated for ULS=2.5xl0-3ms-~; points: measured.

Fig. 12. The wetted wall fraction 0 divided by the wetted wall fraction in horizontal air~water pipe flow 0ho~,as a function of the superficial gas velocity U~s; lines: calculated for Ucs = 2.5 x 10-3 m s-L; points: measured.

Figure

10 shows the ratio of liquid holdups

eL/eL,ho r. It can be seen that the relative effect of

inclination angle can be quite remarkable, and that the effect is correctly modeled, right up to the transition from wavy to slug flow. The ratio does depend on the superficial liquid velocity ULS. Ratios eL/g'L, hor of up to eight were measured for ULS < 2.5 X 10 3 m S 1 while smaller ratios are generally obtained for U1.s > 2.5 x 10 3 m s- 1. These effects of ULs are not shown explicitly in this paper, but they also follow from the M A R S model and were also verified experimentally (Fig. 8). The curves in Fig. 10 show a maximum in the ratio eL/eL, ho,. The maximum does not represent a maximum in the value of the calculated holdup eL. Rather, while decreasing Uas beyond the maximum in a curve, eL.hor increases more rapidly than EL. This is due mainly to differences in the rates of change of interfacial friction. In any case, the experiments show that a transition from wavy to slug flow occurs before the predicted maxima are reached. The transition from wavy to slug flow and a model to predict its occurrence, were presented previously (Grolman et al., 1996). Figure 11 shows the ratio of pressure gradients (dP/dL)/(dP/dL)ho~. As before, there is a remarkable effect of inclination angle on this ratio in the low range of gas velocities, which is correctly incorporated. In the range 12 < U~s/(ms- 1) < 18, the markers lie somewhat below the horizontal line. Since the circles, which represent horizontal flow, show the same trend, the pressure drop of horizontal flow is somewhat overestimated by the M A R S model in this range. It can be said that this figure represents just one pipe diameter, one liquid system and one liquid flow

rate o f u L s ~ 2 . 5 x l 0 - 3 m s 1 and that the M A R S model was built to a much wider range of conditions. Even so, there remains some room for improvement of the M A R S model in some areas. Figure 12 contains the ratio of wetted wall fractions 0/Got. Measurements of the wetted wall fraction were performed by aligning the eye simultaneously both with the angle gauge attached to the glass pipe and with the gas-liquid glass dividing line inside the pipe. Of all the methods used, estimating the average value of 0 in this way introduces the largest relative measurement error. This is reflected by the scatter of the points in Fig. 12 and also in Fig. 3. F r o m eq. (7), the effects of small angles of inclination on 0 appear to be negligible [1/cos fl at most], but this is not true. We have seen that the liquid holdup eL is severely affected by [t g=O. Whence, the main effects of fl on the wetted wall fraction 0 are caused by the liquid holdup eL, which acts on both terms of eq. (7), i.e. through 0o ( ~ e '37'~) and through Fr~ ( ~ [ 1 - - E L ] 1"6t. By this mechanism, both terms of eq. (7) contribute significantly to the effects of fl on the wetted wall fraction 0. This also explains the maxima observed in the modeled curves of Fig. 12. In Fig. 13 the average relative and absolute errors in the prediction of liquid holdup and pressure gradient are displayed for the M A R S model, the ARS model of Hart et al. (1989) and the ' S M O O T H ' model. The latter assumes smooth-pipe friction factors of the form f = a R e b, with a = 16, b = 1 for laminar (Re < 2100) flow and a = 0.046, b = - 0.2 for turbulent flow, for all three friction factors (Taitel and Dukler, 1976). The average and absolute error

Gas-liquid flow in slightly inclined pipes

4469

err (%)
_ _ .~3o2

16

~a

6a.

34

148

89

107

316

80

1285

75

17 =

MARS

I !
0.2 0.25 o 0..35 L_F 6 =,8

I.,t [] m(-<~/~)

302

341

610

34

148

89

1107 316

80

285

75

17 = N

~i
-5o

o.2~. 1

11;5~~.~(_am/~l
[] ~(-dP/dL)

/ ' / / / / / / / / / /
34 148 89 10~ 316
80

artetal.(1

i -loo
1~ooo A 3 2 I~ i~1 161o t28 Fs t~ =N
3 5 6 =']~

0.35

'Smooth'

model

Fig. 13. Comparison of average relative errors (%) in the predicted liquid holdup eL and pressure gradient -(dP/dL), sorted on inclination angle/7; D = 26 and 51 mm; systems = air~waterand air/tetradecane

4470 parameters have been defined as: - average

E. Grolman and J. M. H. Fortuin the average liquid-layer depth, was found to produce good results at higher gas velocities. However, at lower gas velocities and inclined or declined pipes, this model does not succeed either [Fig. 13(B)]. The wide use of smooth-pipe friction factors for non-horizontal gas liquid pipe flow, as in the model by Taitel and Dukler (1976), is unacceptable. This can be judged from Fig. 13 (C). The interactions between the turbulent flow field of the gas phase and waves on a laminar or turbulent liquid phase in gas liquid pipe flow, are extremely complicated and are likely to be the subject of studies for decades to come. Interfacial waves are the main cause of increased friction between gas and liquid. The process of wave formation and growth is primarily determined by the rate of transfer of mechanical energy from the gas-phase to the liquid phase, and the rate of viscous dissipation within the liquid. These processes are affected by the angle of inclination in a way that present-day models for wave formation and interfacial shear cannot predict. The methods introduced in this paper, enable the calculation of liquid holdup and pressure gradient for 2400 well-controlled and accurate experiments, including many sets of marginally stable conditions, to within accuracies hitherto unavailable. We hope this will aid the development of more theoretically based models for the frictional parameters in inclined gas-liquid pipe flow.
Acknowledgements

"e'~(eL)~ 200~m(~'L'jmd-~;L'jexPlN ; ~=1 \ EL'] mod eL,j e x o /


E-if(- dP/dL) = - N
200 (19)

N
//
- -

x j2 1 \ =

( ( - de/dC)j~oa -- ( - dP/dL)j ~xp~ ~ + ( dV/dL)j exp/]

absolute

200~]~(eL'jmd--eL'jexpl2
lerr(eDI = ~ le r r ( - dP/dL)l 200
= , \ e L , j . mod

7eL, j exp/
(20)

~ f j ((( dP/dL)j~"a - ( - dP/dL)j exp'~2 =1 - dP/dL)j~od + ( - dP/dL)f~p] "

In these error parameters, the differences between the measured and modeled values are scaled with the average of the measured and modeled values. In this way a fair representation of the errors across a number of orders of magnitude is obtained. Figure 13 shows that for gas liquid pipe flow in sloping pipes and a large range of variables (D, ULS, UOS,fl, tlL,a, pL), the MARS model provides a considerable improvement over earlier models in the prediction of eL and - d P / d L . Furthermore, the preference of the other models for particular angles of inclination is not noticeable in the MARS model.

DISCUSSIONAND CONCLUSIONS The experiments reported were performed in glass pipes at small angles of inclination, with a wide range of superficial gas velocities, pressure gradients, liquid holdups and low liquid flow rates. At relatively low superficial velocities in inclined pipes, the friction fac- A tor equations used are crucial for the quality in the dh predictions of liquid holdup and pressure gradient, dP/dL Furthermore, the wetted wall fraction 0 plays a more D important role than is generally recognized. We have f not been able to obtain an explicit equation for the Fn interracial friction factorf. In addition, f~ appears to Fr~ result from a rather empirical set of equations. However, all present-day friction factor equations for tur- g bulent, single-phase flow with which we are familiar, k are also empirical in nature. N Literature models for the interfacial friction factor Re used by Cohen and Hanratty (1968), Miya et al. Rea (1971), Agrawal et al. (1973), Andritsos and Hanratty (1987), Andreussi and Persen (1987), Kowalski (1987) ReL and Andreussi et al. (1985, 1993) were tried on the data from inclined pipes, but failed. The model of Hart si et al. (1989), which essentially estimates the relative S roughness of the gas liquid interface k/D as 2.3 times

The authors are grateful for the financial support by The Netherlands Foundation for Chemical Research (SON), The Netherlands Technology Foundation (STW), Shell Research B.V. (Amsterdam, The Netherlands), Energiebeheer Nederland/DSM (Heerlen, The Netherlands) and Gasunie N.V. (Groningen, The Netherlands). This work would not have succeeded without the active support from our Mechanical Workshop, headed by Mr J. Zoutberg, and our Glass Workshop, headed by Mr Clewitz. NOTATION cross-sectional area, m 2 hydraulic diameter (= 4A/S), m average axial pressure gradient, Pa m inner diameter of pipe, m Fanning friction factor, dimensionless friction number, eq. (14), dimensionless gas-phase Froude number, eq. (7), dimensionless gravitational acceleration, m s- 2 equivalent sand roughness, m number of experiments, dimensionless Reynolds number (= puD/rl), dimensionless gas Reynolds number [ = ReGs/(1 -- 0 + si)], dimensionless liquid Reynolds number (= ReLs/O), dimensionless interfacial perimeter, eq. (10), dimensionless perimeter of interface (gas liq., gas wall or liq. wall), m

Gas-liquid flow in slightly inclined pipes


U

4471

Wel.s

average velocity, m ssuperficial liquid Weber number, eq. (7), dimensionless

Greek letters // angle of inclination;/~ > 0 for upward flow, des c holdup, i.e. average volume fraction, dimensionless q dynamic viscosity, Pa s 0 wetted-wall fraction [-SL/(nD)], dimensionless p density, k g m - 3 interfacial tension between gas and liquid, Pa m shear stress, Pa Subscripts G gas exp based on measured values of eL, 0 and - d P/dL hor the value that is obtained when fl = 0 is assumed gas-liquid interface i experiment identification number J liquid L MARS according to the modified apparent rough smface model presented here mod modeled S superficial, i.e. (volume flow rate)/(7~D2/4)
REFERENCES

Agrawal, S. S., Gregory, G. A. and Gorier, G. W. (1973) An analysis of horizontal stratified twophase flow in pipes. Can. J. Chem. Engng 51, 280 286. Andreussi, P., Asali, J. C. and Hanratty, T. J. (1985) Initiation of roll waves in gas-liquid flows. A.I.Ch.E.J. 31, 119-126. Andreussi, P., Minervi, A. and Paglanti, A. (1993) Mechanistic model of slug flow in near horizontal pipelines. Int. J. Multiphase Flow 39, 1281-1291. Andreussi, P. and Persen, L. N. (1987) Stratified gas liquid flow in downwardly inclined pipes, lnt. J. Multiphase Jtow 13, 565-575. Andritsos, N. and Hanratty, T. J. (1987) Interfacial instabilities for horizontal gas-liquid flows in pipelines. Int. J. Multiphase Flow 13, 565-575.

Barnea, D. and Taitel, Y. (1986) Flow pattern transition in two-phase gas liquid flows. In Encyclopedia of Fluid Mechanics, ed. N. P. Cheremisinofl; Part 3. Gulf publishing Co., Houston. Cohen, L. S. and Hanratty, T. J. (1968) Effects of waves at a gas-liquid interface on a turbulent air flow. J. Fluid Mech. 31, 467 469. Eck, B. (1978) Technische Str6mungslehre, Band 1, p. 107. Springer, New York, U.S.A. Franca, F. and Lahey, R. T. Jr. (19921 The use of drift-flux techniques for the analysis of horizontal two-phase flow. Int. J. Multiphase Flow 18, 787-801. Grolman, E. (1994) Gas-liquid flow with low liquid loading in slightly inclined pipes. Ph.D. thesis, University of Amsterdam, The Netherlands, ISBN 90-9007470-8. Grolman, E., Commandeur, C. J., de Baat, E. C. and Fortuin, J. M. H. (1996) Wavy-to-slug flow transition in inclined gas-liquid pipe flow. A.I.Ch.E.J. 42, 901-909. Grolman, E. and Fortuin, J. M. H. (1996) Transient gas-liquid flow in upward sloping pipes, approaching the wavy-to-slug flow transition. ASME J. Fluids Engng 118, 729-735. Hamersma P. J. and Hart, J. (1987J A pressure drop correlation for gas-liquid pipe flow with a small liquid holdup. Chem. Enqnq Sei. 42, 1187 1196. Hart, J., Hamersma, P. J. and Fortuin, J. M. H. (19891 Correlations predicting frictional pressure drop and liquid-holdup during horizontal gas-liquid pipe flow with a small liquid holdup. Int. J. Multiphase Flow 15, 947- 964. Kowalski, J. E. (1987) Wall and interfacial shear stress in stratified flow in a horizontal pipe. A.I.Ch.E.J. 33, 274 281. Miya, M., Woodmansee, D. E. and Hanratty, T. J. (1971) A model for roll waves in gas-liquid flow. Chem. Engng Sci. 26, 1915-1931. Spedding, P. L. and Hand, N. P. (1995) Prediction of holdup and frictional pressure loss in two-phase co-current flow. In Two-Phase Flow Modelling and Experimentation, eds G. P. Celata and R. K. Shah, Vol. 1, pp. 573--581. Edizioni ETS, Pisa, Italy, ISBN 88-7741-852-4. Stewart, P. (1988) Subsea separation and transport. Chem. Em,t 22-25. Taitel, Y. and Dukler, A. E. (1976) A model for predicting flow regime transitions in horizontal and near horizontal gas-liquid flow. A.I.Ch.E.J. 22, 47-55.

You might also like