Chemical Engineering and Processing 42 (2003) 867 /878

www.elsevier.com/locate/cep

Estimation of tray efficiency in dehydration absorbers

Lars Erik Øi *
Telemark University College, TF, Box 203, N-3901 Porsgrunn, Norway

Received 3 April 2002; received in revised form 21 November 2002; accepted 21 November 2002

Abstract

In triethylene glycol (TEG) dehydration, a bubble cap tray efficiency between 25 and 40% has normally been recommended in
literature. Measured Murphree efficiencies, based on performance data from literature and improved phase equilibrium data, have
been calculated to values between 51 and 76%. The AIChE method has been used to estimate tray efficiencies under the same
conditions. The estimated Murphree efficiencies varied between 57 and 70%, the estimated overall efficiencies between 31 and 62%.
The AIChE method seems to give reasonable efficiency estimates for TEG absorbers. Both measured and estimated efficiencies
indicate that the actual efficiencies are higher than recommended in literature. The Murphree efficiencies are less sensitive to changes
in parameters than the overall efficiencies. A general design recommendation is suggested: using 50% Murphree efficiency based on
accurate phase equilibrium data should give a conservative design at normal high pressure dehydration conditions.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Tray efficiency; Dehydration absorbers; AIChE method

1. Introduction reliable performance data from natural gas dehydration,

especially from newer plants.
Dehydration by absorption in triethylene glycol The aim of this paper is to calculate estimated tray
(TEG) is the traditional way to remove water from efficiencies based on general methods and compare them
natural gas to avoid ice and hydrate and to avoid with calculated tray efficiencies based on available
corrosion (Fig. 1). performance data from literature.
In designing a dehydration absorber, a key parameter
is the estimated tray efficiency (or HETP for structured
packing). In literature, the recommended overall tray
efficiency for design has until now varied between 25 2. Definitions of tray efficiency
and 40%. However, there are reasons to believe that
these estimated values are too low. One reason is that The overall tray efficiency is defined as the number of
equilibrium data used for design have been very ideal equilibrium trays divided by the actual(real)
inaccurate. This results in inaccurately calculated tray number of trays:
efficiency values. This problem has been known for
many years, but has not been regarded as a major Nideal
Eo
(1)
problem because uncertainties in other factors like Nreal
column efficiency has been larger than the uncertainties The Murphree tray efficiency related to the gas side
in equilibrium data. (for tray number n) is defined by the equation:
Another explanation for the low estimated tray  
efficiencies, is the large uncertainty in measurement of y  yn1
EM
(2)
water content in natural gas. There is a serious lack of y  yn1
where y is the mol fraction in the gas from the tray, yn1
* Tel.: /47-3557-5000; fax: /47-3557-5001. is the mol fraction from the tray below and y * is in
E-mail address: lars.oi@hit.no (L.E. Øi). equilibrium with the liquid at tray n (see Fig. 2).
0255-2701/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 2 5 5 - 2 7 0 1 ( 0 2 ) 0 0 1 8 7 - 3
868 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878

general by Prausnitz and Wichterle [2,3], and for TEG/

water/natural gas by Parrish and Øi [4,5].

3.1. TEG/water equilibrium data

The data describing the equilibrium between TEG

and water are most important. Much data on the TEG/
water (and natural gas) system have been reported.
However, there are severe discrepancies in reported
equilibrium values.
Parrish [4] gives an extensive review of the data up to
1986. Many of the sources are obviously inaccurate.
Among these are the much used data from Worley [6],
which are shown by Parrish to be inaccurate and
inconsistent. The most consistent data seems to be
Fig. 1. Dehydration column.
from Herskowitz [7].
Parrish measured activity coefficients at infinite dilu-
tion. These data were combined with the Herskowitz
data in a correlation. This is now the industry standard,
recommended when using the simulation programs
ASPEN PLUS and PRO/II.
Bestani [8] has also measured activity coefficients at
infinite dilution, and the difference is about 15%
compared with Parrish. The Bestani data are close to
the extrapolated values from the Herskowitz data
(shown in Fig. 3). Also, the temperature dependence
of the activity coefficients measured by Bestani is more
consistent with heat of mixing data than the Parrish
activity coefficients. For these reasons, the Herskowitz
and Bestani data are regarded as the most reliable [5].

Fig. 2. Definition of Murphree efficiency.

3.2. Total vapour/liquid equilibrium model

Overall and Murphree efficiency are connected by the A combination of a Van Laar liquid activity model,
general Eq. (1): the Virial equation for the gas phase and a Poynting
correction for the pressure dependence has been im-
[1  EM (mV =L  1)] plemented in the process simulation program HYSYS.
Eo
log (3)
log(mV =L) The Van Laar parameters are based on the Herskowitz
and Bestani data. HYSYS recommends an extended Peng
where V and L are the molar vapour and liquid flow Robinson equation of state based on data from Union
rates. Carbide [9], but when calculating measured tray effi-
Because the definitions of tray efficiencies are con- ciencies in HYSYS, a combined model is regarded as
nected to gas/liquid equilibrium, uncertainties in gas/ more accurate [5].
liquid equilibrium will influence on the uncertainties in The basic equation used to describe the vapour/liquid
tray efficiency. equilibrium is:
gH2 O xH2 O f H2 O Poy
yH2 O 8 H2 O P (4)

3. Gas/liquid equilibrium The activity coefficient of water (gH2O) as a function

of mole fraction water and the Van Laar parameters, is
The most accurate way to correlate data in the given by the equation:
nonideal system TEG/methane/water at high pressure A12
is probably by the help of an equation of state model ln gH2 O
(5)
(1  (A12 xH2 O )=(A21 (1  xH2 O )))2
describing the gas phase, an activity model to describe
the liquid phase and a Poynting correction to describe The Van Laar parameters from Herskowitz at
the total pressure dependence. This is discussed in 24.4 8C have been confirmed, and the resulting infinite
 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878 869

Fig. 3. Herskowitz, Bestani and Parrish data at 24.4 8C.

activity coefficient is shown in Fig. 3 to be close to the binary parameter was sufficient to fit the experimental
extrapolated Bestani value (0.61). The infinite activity data with an accuracy (with respect to 8H2O) about 1/
coefficient from Parrish is also shown in Fig. 3. The 2%. A traditional equation of state like PR and SRK is
fitted Van Laar parameters at 24.4 and 59.4 8C are also probably sufficient to describe the gas phase
given in Table 1. For absorber calculations, the tem- nonidealities for water/hydrocarbon systems with only
perature dependent Van Laar parameters in HYSYS are one parameter for each binary pair.
fitted to give the parameters from Øi [10] in Table 1 at The Poynting correction factor (Poy in Eq. (4)) is
24.4 and 59.4 8C. This covers the usual range of defined as
dehydration absorption conditions.
The temperature dependent Van Laar parameters in
HYSYS are given by P  
v H2 O
A12
a12 b12 T (6)
Poy
exp
g P RT
dP (8)

A21
a21 b21 T (7)

where T is the absolute temperature in Kelvin. where vH2O is the molar volume of water in the mixture.
The gas phase fugacity coefficient (8H2O) is calculated P * is the pressure at the reference state (pure water at
by HYSYS using the Virial equation. Rigby [11] has the given temperature). The effect of the total pressure
measured the gas phase system water/methane between on fugacity is negligible at low pressures, but is about
25 and 100 8C and between 20 and 100 atm. They found 5% at 60 bar [5]. fH2O  is the fugacity of water at the
that the Virial equation with one temperature dependent reference state and is very close to the vapour pressure
of water.
Table 1 Solubilities of hydrocarbons, especially methane, is
Van Laar parameters, water/TEG also part of the liquid phase calculation. The solubilities
A12 A21 can be given as Henrys constants which may be
temperature dependent. The solubility of methane in
24.4 8C /0.4855 /1.822
TEG is given as a Henrys constant of 2000 bar from
59.4 8C /0.3710 /0.850
HYSYS (a12 and a21) /1.455 /24.46 GPSA [12]. In the HYSYS efficiency calculation, all
HYSYS (b12 and b21) 0.00326 0.0761 hydrocarbons are calculated as methane. This results
in negligible differences in the efficiency calculations.
870 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878
4. Design and performance data in literature
4.1. Design recommendations
Traditional dehydrators contain bubble cap trays.
When designing a TEG absorber, an estimated value of
the tray efficiency is normally used. GPSA [12] suggests
an overall efficiency of 25 /35%, Kohl [13] suggests 25/
40%. It is natural to hide a safety margin in the tray
efficiency. Then a statement of a tray efficiency ‘recom-
mended for design’ is normally given.
An interesting point is that the normally used
equilibrium data from Worley [6] are very optimistic.
This should result in a great number of performance
failures. However, the use of optimistic Worley equili-
brium data and a conservative tray efficiency value may
give a reasonable design.
Sulzer gives a very approximate value of 0.5 theore-
tical stages for each meter of structured packing in
dehydration [14], very dependent on the conditions. This
is equivalent to an approximate height equivalent to
theoretical plate (HETP) of 2 m. However, Sulzer
recommends the use of transfer units rather than
theoretical stages in designing dehydrators with struc-
tured packing. Information from Koch-Glitsch [15]
states that 0.3 theoretical stages per feet of packing are
provided in a typical TEG contactor with their struc-
tured packing. This is equivalent to an HETP of 1 m.
4.2. Review of performance data
There are not much open available data about the
actual performance of dehydrators. Evaluated sources
are Worley, Kirsten, Kean, Holm, Polderman, Swerdl-
off, Peahl and Øi [5,6,16 /21].
Only a few of the sources give enough information to
calculate a measured tray efficiency. Some of the data
also have a serious lack of reliability. A main problem is
that accurate measurement of water content is extremely
difficult. The problem is complicated by the ambiguity
in the term dew point. A dew point measured by the
detection of the first drop, may be measured reprodu-
cible, but this will probably be far from the thermo-
dynamical dew point.
Many of the dehydration design methods have been
based on performance data from Worley [6]. These data
are, however, inconsistent with newer TEG/water equi-
librium data. The claimed water dew point achieved at
100 8F contact temperature is much better than possible
with the actual TEG concentrations. This is also the case
with the data from Polderman [19].
Data from Kirsten [16] was obtained in a laboratory
jet tray column under stripping conditions at 1 atm. The
water concentration was 53% by weight and thus much
less concentrated in TEG compared with natural gas
dehydration. The measured Murphree efficiencies were
in the range between 47 and 70%.
Performance data from the Frigg field in the North
Sea is given by Holm [18]. The dehydration was
performed at very high pressure (140 /160 bar) and the
measured overall efficiencies were in the area of 15/
25%. These efficiency calculations were, however, based
on Worley data which normally results in too low values
on calculated efficiencies. It was not enough details to
calculate efficiencies based on the given information.
ARCO [17] performed a test with how packing works
in dehydrators. The calculated HETP ranged from 3.7 to
8.5 feet (1.1 /2.6 m). It was not given what equilibrium
data were used in these efficiency calculations, and it
was not enough details to calculate efficiencies based on
the given information.
Measured tray efficiencies have been calculated based
on the data from the most complete and reliable data
from Peahl, Swerdloff and Øi (Gullfaks) [20,21,5].
Performance data are given in Table 2 (Swerdloff) and
Table 3 (Gullfaks).
4.3. Peahl
Data from Peahl [20] measured in 1949/1950 is the
only reference given in Kohl [13] on performance data in
sweet natural gas dehydration with TEG.
The absorption column had four bubble cap trays and
a diameter 0.915 m. The operating pressure was 27.5
bar(a), the vapour load was 556 kmol/h and the water in
TEG weight fraction was 1.72%. The gas inlet tempera-
ture was 12.9 8C, but out of the column the temperature
was 18.9 8C due to high temperature on incoming
glycol.
4.4. Swerdloff
Data from Swerdloff [21] are given from six different
plants in operation before 1957 with values for con-
tactor pressure, wet gas temperature, glycol rate (in gal/
lb), reboiler temperature and dew point depression. The
columns had four actual trays except plant 3 with five
trays and plant 4 with six actual trays. The given data
have been converted to SI units and the converted
performance data are given in Table 2. TEG concentra-
tion in regenerated glycol was calculated from the
reboiler temperature based on equilibrium data from
Dow [29] at 1 atm.
4.5. Gullfaks
Data sets from the Gullfaks C platform have been
measured in 1995 [5]. The absorption column had six
bubble cap trays and a diameter 2.35 m. The operating
pressure was 63 bar(a), the vapour load was 8500 kmol/
h and the water in TEG weight fraction was 0.4%. In
 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878 871

Table 2
Swerdloff performance data (converted)

Plant TG (8C) P (bar(a)) Lin (m3TEG /kgH2O) TEG (wt.%) Tdew (8C)

1 30.6 46.9 0.0292 97.8 /1.1

2 26.1 46.9 0.0728 97.8 /8.3
3 25.6 35.5 0.0253 97.9 /11.1
4 30.6 40.7 0.134 97.5 /10
5 35.0 66.1 0.0511 98.8 /0.6
6 35.6 71.3 0.0921 96.0 /2.8

Table 3 Table 4
Gullfaks performance data Measured and AIChE estimated tray efficiencies

Data set TG (8C) Lin (kg/h) TL (8C) Tdew (8C) Data set TG (8C) EM,Meas (%) EM,Est (%) EO,Est (%)

1 22.5 1103 48 /23 Peahl

2 27.8 1179 49 /19 18.9 65 70 41
3 29.9 1103 49 /17 Swerdloff
4 32.1 1145 51 /16 1 30.6 57 57 41
5 23.0 1030 48 /21 2 26.1 63 58 31
6 23.0 1500 53 /22
3 25.6 72 57 40
7 22.8 1700 57 /23 4 30.6 /100 (66) (31)
8 22.9 1045 43 /21 5 35.0 51 60 42
9 23.0 1150 50 /19
6 35.6 /100 (62) (30)
Gullfaks
1 22.5 60 62 51
run 9, no stripping gas was used, so the weight fraction 2 27.8 66 64 56
water was 1.1%. 3 29.9 71 66 60
4 32.1 76 66 62
5 23.0 58 63 52
6 23.0 55 62 47
5. Calculation of measured tray efficiencies 7 22.8 55 61 45
8 22.9 57 63 53
9 23.0 68 62 51
All the measured Murphree efficiencies have been
calculated using the process simulation program HYSYS.
The method is similar to the calculation of Gullfaks
performance data claim that the dehydrated gas con-
efficiencies in Øi [5]. The calculations are based on the
tains less water than in equilibrium with incoming
combined vapour/liquid equilibrium model in the va-
glycol. This is not regarded as probable, so these data
pour/liquid equilibrium section. The HYSYS calculations
sets are discarded.
were performed by simulating the absorption column
If the efficiencies had been calculated with a model
with the wet gas inlet stream and glycol inlet stream
based on Parrish data, the calculated efficiencies had
from given data. The Murphree efficiency for all the
been slightly less because the Parrish activity coefficients
actual trays was then adjusted to achieve the ppm water
are lower at infinite dilution. The efficiency based on the
in the dried gas stream. The dew points in Tables 2 and 3
Gullfaks data has also been calculated by Øi [10] with
were converted to ppm water by a McKetta diagram
the HYSYS recommended (but not accurate) Peng
[22].
Robinson model, and the resulting Murphree efficien-
The glycol temperatures in the Swerdloff calculations
cies varied from 59 to 95%.
were assumed to be 10 8F above the wet gas tempera-
ture (this is traditional). The Swerdloff performance
data did not give any information about gas flow or
column diameter. A diameter of 0.915 m as in the Peahl 6. Calculation of estimated tray efficiencies
calculations was assumed with 50% flooding (without
any foaming factor). 6.1. General methods
The calculated efficiencies are summarised in Table 4.
The range in calculated Murphree efficiency is from 51 There are several general methods for estimating tray
to 76%. The Swerdloff plants 4 and 6 give Murphree efficiency without using performance data for the actual
efficiencies higher than 100%. This is because the system. Tray efficiency methods have been reviewed in
872 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878
Coulson, Vital and Perry [1,23,24]. The most recom-
mended methods are:
/ O’Connell [25].
/ MacFarland/Sigmund/Van Winkle [26].
/ AIChE [27].
General rules of thumb are also available. In Perry
[24], it is claimed that the efficiency in gas controlled
absorption (like in dehydration) is normally 50 /100%.
6.2. O’Connell
The O’Connell method [25] is based on absorption
conditions with the viscosity and gas solubility as the
main parameters. In Coulson [1] the parameter correla-
tion is converted to SI units:
rS
X
0:062103
(mS mMS )
In Eq. (9), the equilibrium constant for the solute in
Coulson [1] has been replaced by m (the slope of the
equilibrium curve). This is equivalent in the Henrys law
area. The factor 10 3 has also been included to make
the equation consistent with viscosities in Ns/m2. Sub-
script S in Eq. (9) is for the solvent.
The physical properties are calculated based on the
methods given in the Section 6.5. Solvent density is 1120
kg/m3 and molecular weight is 150.2. In the Gullfaks
case at 296 K, liquid viscosity is 0.036 Ns/m2 and m is
0.00034. At Peahl conditions, the viscosity is 0.041 Ns/
m2 and m is 0.00023. Eq. (9) gives an X parameter 38 for
the Gullfaks conditions and 49 for the Peahl conditions.
Overall estimated efficiency is found by a curve as a
function of the X parameter.
These X parameters in the dehydration cases are
outside the normal range in the O’Connell curve due to
the very high gas solubility in the system. The curve
predicts an estimated overall efficiency higher than 60%,
with an order of magnitude 70%.
6.3. MacFarland/Sigmund/Van Winkle
The two suggested equations in this method [26] are
correlations based on dimensionless groups:
EM
7:0(NDg )0:14 (NSc )0:25 (NRe )0:08
EM
6:8(NRe NSc )0:1 (NDg NSc )0:115
The dimensionless groups are defined in the symbol
list.
Estimated tray efficiencies were calculated based on
the conditions in the Gullfaks normal case. The
estimated values varied from 200 to 280%. Then the
test example in the original paper [26] was calculated
and verified. After inspection of the equations in the
original paper, it was found that the efficiency increased
with increased viscosity and increased with decreasing
liquid diffusivity. Both effects were opposite to what was
expected. This results in unrealistically high estimated
efficiencies because of the high liquid viscosity and low
liquid diffusivity. Because of this, the MacFarland/
Sigmund/Van Winkle estimation method is regarded as
not suitable for estimating tray efficiencies in dehydra-
tion absorbers.
6.4. The AIChE method
The AIChE method from the AIChE Bubble Cap
Design Manual [27] is regarded as the most reliable
method for estimating plate efficiency in bubble cap
columns [24]. The notation and equations used here are
from Coulson and Richardson [1] in a slightly revised
form. The equations in Coulson and Richardson are
(9)
equivalent to the equations in the original AIChE
manual, converted to SI units. (There is, however, a
minor misprint in equation 11.74 in Coulson and
Richardson (here Eq. (16)) where a minus sign is
changed to a plus sign.) In this work, the main change
compared with the original AIChE method is that the
use of figures for calculation is avoided and replaced by
fitted equations valid for the actual conditions. This
makes the method suitable for EXCEL spreadsheet
calculations.
The method is based on calculating the point effi-
ciency (Emv) from the physical conditions:
 
1
Emv
1exp (12)
(1=NG  mV =(LNL ))
The number of transfer units are calculated for the gas
and liquid films:
 
0:776  4:57  103 hW  0:24Fv  105Lp
NG

(mv =(rv Dv ))0:5
(13)
NL
(4:13108 DL )0:5 (0:21Fv 0:15)tL (14)
Liquid contact time is calculated:
ZC ZL
tL
(15)
Lp
Liquid hold-up on the plate is correlated to an
(10)
equation for bubble cap trays:
(11)
ZC
0:0420:19103 hW 0:014Fv 2:5Lp (16)
The Peclet number is based on a correlation for the
eddy diffusivity (De):
ZL2
Pe
(17)
(De tL )
De
(0:00380:017ua 3:86Lp 0:18103 hW )2 (18)
 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878 873

In Coulson [1], the Murphree efficiency is found by a
figure giving the ratio EmV/Emv as a function of the
product Emv mV /L and the Peclet number. For low
values of Emv mV /L , the curves are straight functions of
Emv mV /L for each Peclet number. The slope (SL ) of
the curves was curve fitted by the least square method as
a function of the Peclet number (valid for Emv mV /L
values below 1.0):
1:9
SL
0:62
(Pe  2:8)
Then the Murphree efficiency (not corrected for
entrainment) can be calculated:
 
SLEmv mV
EmV
Emv 1 (20)
L Gullfaks conditions.
To correct the efficiency for entrainment, the actual
(Murphree) efficiency is calculated by the equation:
EmV
Ea

[1  EmV (c=(1  c))]
Overall efficiency (Eo) can be calculated by replacing
EM in Eq. (3) with Ea calculated from Eq. (21).
log[1  Ea (mV =L  1)]
Eo

log(mV =L)
The Eqs. (12) /(22) are set up in an EXCEL spreadsheet
program to calculate estimated efficiencies at different
conditions.
6.5. Estimation of physical properties
The tray efficiency estimation methods are based on a
number of physical properties. It is recommended to use
estimating methods from the book: The properties of
Gases and Liquids by Reid, Prausnitz and Poling [28].
Densities and liquid flows have been taken from the
HYSYS calculations.
Liquid viscosities are calculated with the following
linear equation in absolute temperature based on data
from Dow [29]. The equation can be used between 288
The vapour viscosities have been calculated based on
tabular values for methane from Reid [26] at low
pressures and corrected by a pressure correction corre-
lation also from Reid. The resulting gas viscosities was
1.2 /105 Ns/m2 at Gullfaks conditions, 9.6 /106 at
Peahl conditions and between 9.7 /106 and 1.3 /
105 at Swerdloff conditions.
The vapour diffusivities were calculated by the Fuller
method and corrected for pressure by a correlation
(19) recommended in Reid [28]:
1:26  109 T 1:75
DV
(25)
P (bar)
The resulting gas diffusivity was multiplied by a
pressure correction factor of 0.88 (from Reid [28]) at
The m factor (slope of the equilibrium curve) is equal
to yH2O/xH2O as long as Henry’s law is valid. The m
values are calculated by HYSYS based on the specified
vapour/liquid equilibrium model.
(21)
6.6. Estimated efficiencies
The plate efficiencies have been calculated in an
EXCEL spreadsheet program with the AIChE method
(22)
(Eqs. (12)/(22)) for the Peahl, Swerdloff and Gullfaks
conditions.
The input to the spreadsheet calculation is sum-
marised for Peahl, Swerdloff (plant 1) and Gullfaks
(run 1) conditions in Table 5. The liquid path length
(ZL) is estimated to 0.4 m for Swerdloff, 0.6 m for Peahl
and 1.0 m for the Gullfaks calculations. The weir heights
in the Peahl and Swerdloff calculations are assumed to
be 110 mm as in the Gullfaks case. Entrainment (c ) is
set to 30% at Gullfaks conditions based on a figure
originally from Fair [31] in Perry [24], assuming
approximately 60% flooding. The Peahl and Swerdloff
calculations are based on 50% flooding giving approxi-
Table 5
Input to AIChE estimation calculations
Peahl Swerdloff Gullfaks
(plant 1) (run 1)

and 313 K (15 /40 8C), and covers the most usual range T (K) 292 303.8 295.65
of dehydration absorption conditions. UV (m/s) 0.17 0.38 0.16
rL (kg/m3) 1120 1120 1120
mL
0:0250:0015(303T) (23) rG (kg/m3) 27 32.3 69
hW (mm) 110 110 110
Liquid diffusion coefficients have been calculated LP (m3/(s m)) 0.00023 0.00066 0.00030
using the Wilke /Chang equation from Geankoplis [30]: ZL (m3/m2) 0.60 0.4 1.0
m (dimensionless) 0.00053 0.0011 0.00034
1:173  1016 (8 MW )0:5 T V (kmol/h) 556 1820 8500
DL
(24) L (kmol/h) 3.71 7.7 7.3
(mL v0:6
H2 O ) c (dimensionless) 0.2 0.2 0.3
mL (Ns/m2) 0.042 0.024 0.036
The 8 parameter is estimated to 1.5 which is the same DL (m2/s) 1.34 /10 10 2.44/10 10 1.57 /10 10
mG (Ns/m2) 0.0096/10 3 0.011 /10 3 0.012 /10 3
as for ethanol. vH2O (the molar volume of water) is
DV (m2/s) 9.45 /10 7 5.39/10 7 3.76 /10 7
0.0188 m3/kmol.
874 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878
mately 20% entrainment. There is high uncertainty in
these values. Parrish [4] states an entrainment value
about 22 /25% for ordinary dehydration conditions.
The resulting estimated efficiencies are given in Table
4. Estimated Murphree efficiency is 57/70% compared
with measured 51 /76%. Estimated overall efficiency is
31 /62%. In Fig. 4, estimated Murphree efficiency is
plotted as a function of measured Murphree efficiency.
7. Calculation of parameters influence on tray efficiency
According to Fig. 4, the only source giving a
significant parameter influence is the data from Gull-
faks. Fig. 5 shows measured and estimated Murphree
efficiency as a function of temperature at Gullfaks. Both
curves show an increase in efficiency with temperature,
but the measured dependency is stronger than the
estimated.
Fig. 6 shows the liquid flow dependence. Both the
measured and estimated curves show a slight decrease in
Murphree efficiency with increased liquid flow.
There is one Gullfaks data set at higher water
concentration. The measured efficiency is slightly higher
than for lower concentration at the same conditions.
Concentration did not influence on the estimated
efficiency in the estimation calculations.
Fig. 4. Estimated vs. measured Murphree efficiencies.
Pressure was varied from 1 to 130 bar in the
calculation of estimated efficiency. Increased pressure
resulted in a slightly reduced estimated efficiency.
Increased vapour flow leads to slightly reduced
Murphree efficiency, but increased overall efficiency
calculated by the AIChE method. If increased entrain-
ment due to increased vapour flow had been included,
the overall efficiency would also decrease with increased
vapour flow.
8. Discussion
8.1. Absolute value of measured and estimated efficiencies
All the measured and estimated Murphree efficiencies
in this work are higher than 50%. The estimated overall
efficiencies vary from 31 to 62%. Both measured and
estimated efficiencies indicate that actual efficiencies are
higher than recommended in literature (25 /40%).
Fig. 4 shows more scatter in the measured efficiencies
(51 /76) than in the estimated efficiencies (57 /70). This
is as expected because of significant uncertainties in
measurements and because all factors can not be
included in the estimation method.
There are uncertainties connected to the calculation of
measured and estimated efficiencies. The accuracy of the
 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878 875

Fig. 5. Efficiency as a function of temperature at Gullfaks.

measured efficiencies is mainly dependent on the mea- accuracy of measured efficiencies is also dependent on
surement accuracy. Especially important is the measure- the vapour/liquid equilibrium model used in the calcula-
ment of water content in the dehydrated gas. The tion. The uncertainty due to this is, however, not

Fig. 6. Efficiency as a function of liquid flow at Gullfaks.

876 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878
expected to account for more than a few percent in
measured efficiency. A greater difference can be ex-
pected when the dehydrated gas is very close to
equilibrium with incoming glycol.
The uncertainties in estimated efficiency are mostly
due to the applicability of the estimation methods. The
significance of efficiencies calculated with the O’Connel
method is limited, even though the values show good
agreement with the measured efficiencies and the AIChE
method. The MacFarland/Sigmund/Van Winkle method
predicts efficiencies far from realistic values.
Even though the AIChE method is the most reliable
method available, it is in this work used outside the
normal range. The data basis for the method is for
pressures below 6 bar and liquid viscosities below 1 cP
(10 3 Ns/m2). In the AIChE calculations, the factor
influencing most on the efficiency, is the fraction of
entrainment. This is difficult to estimate because it is
influenced by unknown details on the bubble cap trays.
Effects on efficiency due to different tray types can be
quite important, and this can not be taken into account
in the AIChE method.
The Murphree efficiencies are less sensitive to changes
in the parameters than the overall efficiencies. This
indicates a recommendation to use Murphree efficiency
and not the overall efficiency in design calculations.
Calculation with Murphree efficiencies is easy to imple-
ment in modern process simulation tools. The recom-
mendation of using Murphree efficiencies is analogous
to Sulzers recommendation [14] to use transfer units and
not theoretical stages in the design of structured
dehydration absorbers.
8.2. Temperature dependence
The estimated dependence on temperature is less than
the measured dependency. The estimated efficiency
increase calculated by the AIChE method is probably
the most realistic.
8.3. Liquid flow dependence
Both the calculated and the estimated Murphree
efficiency show a slight negative dependence on liquid
flow. The overall efficiency is significantly reduced by
increased liquid flow. Collins [14] predicts a slight
positive effect on efficiency with increased liquid flow.
There is not shown any significant effect on Murphree
efficiency on liquid flow, but the overall efficiency will
be reduced by increased liquid flow.
8.4. Concentration dependence
Concentration does not influence directly on the
estimated efficiencies in traditional correlations. How-
ever, the concentration will influence indirectly because
the liquid viscosity is concentration dependent. The
slightly higher measured efficiency at higher water
concentration (and lower viscosity) is consistent with
this. At normal dehydration conditions, the water
concentration is expected to have a minor effect on the
efficiency.
8.5. Pressure dependence
It is expected that the calculated slight decrease in
efficiency with increased pressure is realistic. This effect
is also claimed by Collins [14].
8.6. Column diameter dependence
The AIChE method shows that increased tray dia-
meter gives increased efficiency. This is probably
realistic under normal conditions.
8.7. Vapour flow dependence
Traditionally, efficiency will increase with increasing
gas flow until an optimum point where the efficiency will
decrease with increasing gas flow. Kean [17] claims that
HETP decreases (overall efficiency increases) with
increasing vapour flow in their performance test of
structured packing. Even though an increased efficiency
with increased gas flow under some conditions is
predicted, it is expected that under normal operating
conditions, increased vapour flow will lead to reduced
efficiency. This is also predicted for structured packing
by Collins [14].
9. Conclusions
Tray efficiencies in dehydration bubble cap absorbers
have been calculated based on performance data from
literature and improved phase equilibrium data. Mea-
sured Murphree efficiencies, based on improved phase
equilibrium data, have been calculated to values be-
tween 51 and 76%.
Measured tray efficiencies have been compared with
estimated tray efficiencies based on available methods.
The AIChE method seems to be the most consistent
estimation method available. The estimated Murphree
efficiencies varied between 57 and 70%, the estimated
overall efficiencies between 31 and 62%. The AIChE
method seems to give reasonable efficiency estimates for
TEG absorbers. The estimated parameter dependencies
calculated by the AIChE method were consistent with
measured dependencies.
Both measured and estimated efficiencies indicate that
the actual efficiencies are higher than recommended in
literature. The Murphree efficiencies are less sensitive to
changes in parameters than the overall efficiencies. A
 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878 877

general design recommendation is suggested: using 50% x mol fraction (liquid) (dimensionless)
Murphree efficiency based on accurate phase equili- y mol fraction (gas) (dimensionless)
brium data should give a conservative design at normal yn1 mol fraction in gas from tray below (dimension-
high pressure dehydration conditions. less)
y* mol fraction in gas in equilibrium with liquid on
tray (dimensionless)
Appendix A: Symbol list ZC liquid hold-up pr. active area on plate (m3/m2)
ZL length of liquid path (m)
Aij Van Laar binary parameter (dimensionless)
aij Van Laar binary parameter(temperature depen- Greek letters
dent) (dimensionless) gi activity coefficient (of component i) (dimension-
bij Van Laar binary parameter(temperature depen- less)
dent) (dimensionless) 8 parameter (in Wilke Chang equation) (dimen-
De eddy diffusion (m2/s) sionless)
DL liquid diffusivity (m2/s) 8i fugacity coefficient (dimensionless)
DV vapour diffusivity (m2/s) c entrainment fraction (dimensionless)
Ea Murphree efficiency (entrainment corrected) (di- r density (kg/m3)
mensionless, %) m viscosity (N s/m2)
Eo Overall tray efficiency (dimensionless, %)
EM Murphree tray efficiency (dimensionless, %) Subscripts
EmV Murphree efficiency (not entrainment corrected) i component index
(dimensionless, %) G gas
Emv point efficiency (dimensionless, %) L liquid
FA fractional free area (dimensionless) n tray number
fi fugacity (of component i) (N/m2) S solvent
FV vapour F -factor
/uar0.5
v (/) V vapour
hW Weir height (mm)
L liquid molar flow (kmol/h)
LP volumetric liquid flow pr. plate width (m3/(s m))
m slope of equilibrium curve (dimensionless)
Ms molecular weight of solvent (g/mol) References
MW molecular weight (g/mol)
NDg surface tension number
/sL/(mLUv) (dimension- [1] J.M. Coulson, J.F. Richardson (Eds.), Chemical Engineering, vol.
6, Pergamon Press, 1991, p. 445.
less) [2] J.M. Prausnitz, R.N. Lichtenthaler, E.G. de Azevedo, Molecular
NG number of transfer units (gas side) (dimension- Thermodynamics of Fluid-Phase Equilibria, Prentice-Hall, 1986.
less) [3] I. Wichterle, High pressure vapour /liquid equilibrium, V, quan-
Nideal number of equilibrium trays (dimensionless) titative description, part 3, Fluid Phase Equilib. 2 (1978) 143.
NL number of transfer units (liquid side) (dimen- [4] W.R. Parrish, K.W. Won, M.E. Baltatu, Phase behavior of the
triethylene glycol /water system and dehydration/regeneration
sionless) design for extremely low dew point requirements, Proceedings
NRe Reynolds number
/hWUVrV/(mLFA ) (dimen- of 65th Annual Convention of Gas Processors Association, 1986,
sionless) p. 202.
Nreal number of actual trays (dimensionless) [5] L.E. Øi, Calculation of dehydration absorbers based on improved
phase equilibrium data, Proceedings of the 78th Annual Conven-
NSc Schmidt number
/mL/(rLDL) (dimensionless)
tion of Gas Processors Association, 1999, p. 32.
P pressure (N/m2) [6] S. Worley, Super-dehydration with glycols, Proceedings of the
P* vapour pressure (of water) (N/m2) Gas Conditioning Conference, Norman, OK, 1967.
Pe Peclet number
/Z2L/(DetL) (dimensionless) [7] M. Herskowitz, M. Gottlieb, Vapor /liquid equilibrium in aqu-
Poy Poynting factor (dimensionless) eous solutions of various glycols and poly(ethylene glycols). 1,
R gas constant (J/(kmol K)) Tri-ethylene glycol, J. Chem. Eng. Data 29 (1984) 173.
[8] B. Bestani, K.S. Shing, Infinite-dilution activity coefficients of
SL curve slope (in AIChE figure) (dimensionless) water in TEG, PEG, glycerol and their mixtures in the tempera-
T temperature (K) ture range of 50 /140 8C, Fluid Phase Equilib. 50 (1989) 209.
ua vapour velocity based on bubbling area (m/s) [9] Union Carbide, Glycols-for Anti-freezes, Coupling Agents, Hu-
UV vapour velocity (/) mectants, Liquid Coolants, Solvents Resin Intermediates, Union
Carbide, 1978.
tL liquid contact time (s)
[10] L.E. Øi, T. Bråthen, Calculation of TEG dehydration in HYSYS,
V vapour molar flow (kmol/h) Conference Hyprotech 2000, Amsterdam, 2000.
vH2O molar volume of water (m3/kmol) [11] M. Rigby, J.M. Prausnitz, Solubility of water in compressed
X parameter in O’Connell method (/) nitrogen, argon and methane, J. Phys. Chem. 72 (1) (1968) 330.
878 L.E. Øi / Chemical Engineering and Processing 42 (2003) 867 /878
[12] GPSA, GPSA Engineering Data Book, Section 20, Dehydration
and Treating, 10th ed., Gas Processors Suppliers Association,
Tulsa, OK, 1987.
[13] A. Kohl, R. Nielsen (Eds.), Gas Purification, fifth ed, Gulf
Publications, 1997.
[14] P. Collins, K. Breu (Sulzer Brothers Ltd.), Structured packings for
TEG contactors-design methods and operating characteristics,
Ninth Continental Meeting, European Chapter of Gas Processors
Association, 1992.
[15] Koch-Glitsch, Mass Transfer Technology. Improving TEG Con-
tactor Performance with Structured Packing, Koch-Glitsch In-
corporation, 1999.
[16] R.A. Kirsten, M. Van Winkle, Efficiency of jet trays, air /water /
triethylene glycol system, Ind. Eng. Chem. Proc. Des. Dev. 9 (1)
(1970) 100.
[17] J.A. Kean, H.M. Turner, B.C. Price (ARCO), How packing works
in dehydrators, Hydrocarbon Processing, April (4) (1991) 47.
[18] P. Holm, Offshore TEG dehydration unit performance exceeds
design, Oil Gas J., January (25) (1993) 96.
[19] L.D. Polderman, Dehydrating natural gas with glycol, Oil Gas J.,
September (23) (1957) 107.
[20] L.H. Peahl, Installation, operation and performance of a skid-
mounted gas dehydration plant, Oil Gas J., July (13) (1950) 92.
[21] W. Swerdloff, What we have learned in 20 years about gas
dehydrators, Oil Gas J., April (29) (1957) 122.
[22] J.J. McKetta, A.H. Wehe, Use this chart for water content in
natural gases, Petroleum Refiner, August (8) (1958) 153.
[23] T.J. Vital, S.S. Grossel, P.I. Olsen, Estimating separation
efficiency, Part 2 */plate columns, Hydrocarbon Processing,
November (11) (1984) 147.
[24] H.R. Perry, C.H. Chilton (Eds.), Chemical Engineering Hand-
book, seventh ed, McGraw-Hill, 1997.
[25] H.E. O’Connell, Plate efficiency of fractionating columns and
absorbers, Trans. Am. Inst. Chem. Eng. 42 (1946) 741.
[26] S.A. MacFarland, P.M. Sigmund, M. Van Winkle, Predict
distillation efficiency, Hydrocarbon Processing, July (7) (1972)
111.
[27] AIChE, Bubble-Tray Design Manual, AIChE Distillation Sub-
committee of the Research Committee, New York, 1958.
[28] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases
and Liquids, fourth ed, McGraw-Hill, 1988.
[29] Dow, A Guide to Glycols, Dow Chemical Company, 1992.
[30] C.J. Geankoplis, Tranport Processes and Unit Operations, third
ed, Prentice Hall, 1993.
[31] J. Fair, How to predict sieve tray entrainment and flooding, Pet./
Chem. Eng. 33 (10) (1961) 45.