Professional Documents
Culture Documents
com
Abstract
In the petrochemical industry, separation of oil from water is a very important process. Wells produce mixtures of gas, oil, and water
which undergo a primary stage of separation inside horizontal gravity separators. The performance of these vessels is evaluated by measuring
mean residence time (MRT) and residence time distribution (RTD). Although many researchers studied flow characteristics in horizontal
separators, limited number of articles exist that discuss separator MRT and RTD with varying water-cuts. In this article, the authors study an
experiment using a horizontal gravity separator by previous researchers and perform computational fluid dynamics (CFD) simulations on the
same geometry under similar conditions. The simulation results show qualitative agreement with the experiments by previous researchers.
As shown by experiments before, CFD results showed that MRT of the organic phase increased with increase in water-cut. In addition, the
RTD characteristics show very similar trends between CFD and experimental results.
© 2020 Shanghai Jiaotong University. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Keywords: Multiphase separators; Computational fluid dynamics; Mean residence time; Residence time distribution.
Flow inlet
Vapor outlet
Continuous progress in the field of numerical methods has
led to computational fluid dynamics (CFD) being an alterna-
tive tool gaining more confidence among the industrial com-
diverter Mesh pad munity [23–26]. CFD is more flexible and less expensive
than fluid flow experiments that require costly instrumenta-
Liquid level Weir tion. It is also more universal than semi-empirical models
Oil-water interface used in designing of separators. Previous research in this
area generally involved model development accounting for
droplet size distribution in addition to coalescence and break-
up [27–30]. Fewer articles focused on assessing effects of sep-
Water outlet Oil outlet arator internals on separator performance [31–33]. However,
there is a lack of published CFD studies that directly eval-
Fig. 2. Horizontal separator components. uate phase MRT based on RTD. Such studies are extremely
critical towards understanding flow characteristics in multi-
is shown by Eq. (1) [16]: phase separators and more accurate assessment of separator
∞ performance.
t c (t )dt
MRT = 0∞ , (1) While very few articles exist that evaluate MRT based on
0 c (t )dt RTD experimentally [16,22], seemingly none of the previ-
where MRT is the mean residence time, c(t) is the concen- ously published CFD work report RTD based MRT evalua-
tration at the outlet, and t is time in seconds. With increase tions in oil-water mixtures with varying volume-fractions of
in MRT, a given phase has more time to disengage from the water. In the present work, the authors perform CFD simu-
other phase. lations to study liquid-liquid flow characteristics in a simple
Residence time distribution (RTD) is defined as the fraction horizontal multiphase separator. The authors use CFD results
of elements leaving the separator between times t and t + t. to obtain RTD and MRT as functions of varying water-cuts.
Following is the equation for RTD [16]: These results are further compared with experiments using a
similar geometry by previous researchers.
c (t )
RT D (t ) = ∞ . (2)
0 c (t )dt 2. Simulation geometry and methods
Hence
∞ The geometry contains an inlet, two outlets, a weir, and
RT D (t ) = 1. (3) a porous plate but doesn’t accommodate components such
0
as the demister, mesh-pad, and the diverter. The geometry
Eq. (3) suggests that over a long period of time, cumulative
is shown in Fig. 3. A weir separates the water outlet from
residence time distribution will be equal to unity. Therefore,
the oil outlet. Two different heights of weir are used for the
while evaluating MRT, the upper limit of the integral is chosen
purpose of simulations. From CFD results, the authors report
such that cumulative RTD tends to unity.
MRT and RTD with varying water-cuts.
Through the past several decades, researchers have had
Two phase simulations are performed with mixtures of wa-
varying opinions about optimal MRT for multiphase separa-
ter and kerosene in a horizontal separator. The separator is
tors [3,17–21]. On field operators usually measure MRT using
0.6 m in diameter and 2.5 m in length. The simulations are
an alternate formula shown by Eq. (4):
performed using two different weir heights: 0.22 m and 0.3 m.
VA The separator inlet and outlet diameters are 0.05 m each. The
MRTA = , (4)
∀˙ separator is classified into three zones – an inlet zone with
where MRTA is mean residence time evaluated using active the flow inlet, a porous zone, and the separation/outlet zone
vessel volume, VA is the active vessel volume involving each with the two outlets. The length of the outlet zone of the sep-
phase, and ∀˙ is the volumetric flowrate. The volume occu- arator is 2.18 m, and the weir is placed halfway between the
pied by each phase is calculated assuming that the height of water and oil outlets. The height of the inlet and the two out-
oil-water interface can be extrapolated from the weir back to lets is 0.3 m. The porous zone is modeled as a 0.20 m thick
the separator inlet. This method may involve substantial er- perforated plate with a viscous resistance of 5 × 108 m−2 .
rors associated with oil-water interface height obtained using Simulations are performed with four different volume frac-
level controllers [16,22]. Therefore, evaluation of MRT using tions of water (water-cuts): 10%, 21%, 32%, and 57%. In the
RTD information, as shown by Eq. (1) is desirable for better mixture of kerosene and water, for each water-cut, the mass
accuracy. flowrate of water is kept constant at 4 kg/s. For each water-
On field RTD and corresponding MRT measurements are cut, transient simulations are performed through a flowtime
difficult because they may require complicated instrumenta- of 40 s.
tion and flow visualization. This may sometimes be virtually 2-dimensional transient, multiphase CFD simulations are
impossible because field vessels operate under pressure and performed on the horizontal separator using ANSYS Flu-
are constructed of steel [22]. ent 18.2. The code uses finite volume method to solve
264 T. Acharya and L. Casimiro / Journal of Ocean Engineering and Science 5 (2020) 261–268
plate. The perforated plate is closer to the inlet and provides 0.4
10 % water-cut
0.35
resistance to flow. It is useful towards increasing MRT. 21 % water-cut
0.3
The RTD and MRT are evaluated as functions of water-cut. 0.25
32 % water-cut
RTD(t)
Eq. (2) is used to calculate RTD at each water-cut. Figs. 9 and 0.2
10 show RTD versus time with the low weir and high weir 0.15 57 % water-cut
value very close to the start. With both weir configurations, 0.05
0
the RTD peak is the highest with 10% water-cut. The RTD 0 5 10 15 20 25 30 35 40 45
-0.05
peak reduces as water-cut is increased to 21%. This suggests time (seconds)
that with larger volume fraction of water in the mixture, lesser
amount of kerosene leaves the separator from start through Fig. 9. RTD versus time with low weir.
a given time corresponding to the peak. However, with the
low-weir configuration, the RTD peak increases marginally
from 21% water-cut to 32% water-cut. With both configura- calculations. At 57% water-cut smaller secondary RTD peaks
tions, the RTD peaks marginally shift towards the right with are observed with the high weir. The presence of secondary
increasing water-cut. This indicates that as amount of wa- peaks generally indicates secondary flows within the separator
ter in the mixture increases, kerosene takes longer to leave due to considerable differences between the individual phase
the separator. These results could further be verified by MRT velocities [16].
266 T. Acharya and L. Casimiro / Journal of Ocean Engineering and Science 5 (2020) 261–268
0.25
57 % watercut
0.2
0.15
0.1
0.05
0
-0.05 0 5 10 15 20 25 30 35 40 45
time (seconds)
MRT versus watercut Using CFD results, the oil droplet rising velocity is calcu-
8 lated using Eq. (13):
7 lw
vr = , (13)
Mean Residence Time (MRT)
6 MRT
5
where lw is the water layer thickness, and MRT is mean res-
idence time [29]. Kerosene rising velocity in water obtained
4 using CFD results is compared against the rising velocity
3 obtained using Stokes Law assuming a droplet diameter of
High Weir 500 μm. Fig. 12 shows a comparison of kerosene rising ve-
2
Low Weir locity in water obtained using both methods.
1 Both methods show that kerosene rising velocity in wa-
0
ter increases with increase in water-cut. However, velocities
0 10 20 30 40 50 60 obtained from CFD results are slightly larger than the veloc-
percentage water-cut ities obtained from Stokes Law. Therefore, the classic design
methods maybe somewhat more conservative when estimating
Fig. 11. MRT versus water-cut. rising velocity of kerosene droplets. This observation agrees
with results published by previous researchers [29].
Eq. (1) is used to calculate MRT at each water-cut, for 4. Experimental validation
both weir configurations. Fig. 11 shows that MRT of the
organic phase (kerosene) increases with increase in water-cut The authors compared their numerical results against very
for both weir configurations. With increase in weir height, similar experiments performed by previous researchers [16].
MRT increases for each water-cut. Therefore, as water-cut in- However, since some pieces of information about the experi-
creases, the organic phase is retained within the separator for mental geometry were unavailable, only a qualitative compari-
longer duration. With increase in retention time, the organic son between the results was possible. The details are provided
phase has more time to disengage from the aqueous phase. in the following sections.
When the weir height is increased, the organic phase is also Simmons et al. performed pilot-scale experiments using a
retained for durations longer than the lower weir height. horizontal separator and measured MRT and RTD for vary-
During phase separation oil rising velocity in water can be ing water-cuts with two different weir-heights [16]. Like the
obtained using Stokes law, as shown by Eq. (12): simulation geometry, the separator used by them also had a
diameter of 0.6 m and a length of 2.5 m and used two differ-
2 ρ p− ρ f ent weir heights: 0.22 m and 0.3 m. A two-phase mixture of
vr = gR2 , (12)
9 μ kerosene and water was used. To measure RTD, they injected
tracer dyes that were soluble in only one phase at a time.
where vr is the oil droplet rising velocity, ρ p is the density of MRT and RTD were measured using Eqs. (1) and (2).
the discrete phase, ρ f is the density of the continuous phase, They showed that with both weir configurations, MRT of
μ is the dynamic viscosity of oil water emulsion, g is the the organic phase (kerosene) increased with increase in water-
acceleration due to gravity, and R is the radius of oil droplet cut. Also, with the high weir, MRT increased with all three
in water. Previous studies show that diameters of freely ris- water-cuts than the low weir. Simmons et al. measured RTD
ing stable untreated oil droplets range from 50 μm to 1 mm as a function of water-cut and showed that RTD peaks were
[35,36] shifted towards the right with increase in water-cut. The RTD
T. Acharya and L. Casimiro / Journal of Ocean Engineering and Science 5 (2020) 261–268 267
5. Conclusion [1] Gas Processors Suppliers Association, GPSA Engineering Data Book,
Volume 1, eleventh ed., Gas Processors Association, Tulsa, 1998.
Computational fluid dynamics (CFD) simulations are per- [2] G. Skelton, P.A. Stockil (Ed.), British Petroleum Company Limited,
1977.
formed to assess kerosene-water flow characteristics inside [3] F. Evans, Equipment Design Handbook for Refineries and Chemical
a horizontal multiphase separator. Kerosene mean residence Plants, Volume 2, 1974 Gulf Houston, ISBN: 978-0-87-201255-4.
time (MRT) is evaluated based of residence time distribution [4] K. Arnold, M. Stewart, Surface Production Operations, third ed., Else-
as this approach leads to more accurate results. Simulations vier, New York, 2008 ISBN: 978-0-7506-7853-7.
[5] H.V. Smith, Petroleum Engineering Handbook, Society of Petroleum
show that in a two-phase mixture of immiscible liquids such
Engineering, Richardson, TX, 1987.
as water and kerosene, MRT of the organic phase (kerosene) [6] S.M. Walas, Chemical Process Equipment Selection and Design, But-
increases with increase in water-cut. Therefore, separator per- terworth-Heinemann, Houston, 1990, pp. 713–715.
formance improves with increase in water-cut. Also, between [7] E.O. Grødal, M.J. Realff, Optimal Design of Two- and Three-Phase
the two weir heights used, higher MRT values are obtained Separators: A Mathematical Programming Formulation, Society of
Petroleum Engineering, 1999, pp. 1–16. 56645.
with the higher weir for each water-cut. This is intuitive
[8] R.N. Watkins, Hydrocarbon Process. 46 (11) (1967) 253–256.
as with the higher weir, the liquids are retained within the [9] A. Hallanger, F. Soenstaboe, T. Knutsen, in: SPE Annual Technical
separator vessel for longer durations. On the RTD versus time Conference and Exhibition, Denver, CO, 1996, pp. 695–706, doi:10.
plot, RTD peaks shift towards the right with increase in water- 2118/36644-MS.
cut. Since RTD is defined as the fraction of a given phase that [10] T.T. Le, S.N. Ngo, Y. Lim II., C.K. Park, B.-.D. Lee, B.-G. Kim, D.-
H. Lim, J. Petrol. Sci. Eng. 171 (2018) 731–747, doi:10.1016/j.petrol.
stays within the separator between times t and t + t, this
2018.08.001.
suggests that with less quantity of water, larger volume frac- [11] P. Yu, S. Liu, Y. Wang, W. Lin, Z. Xiao, C. Wang, Procedia Eng. 31
tions of kerosene leave the separator closer to the start. Also, (2012) 145–149, doi:10.1016/j.proeng.2012.01.1004.
at larger water-cuts such as 57%, secondary smaller RTD [12] A. Ghaffarkhah, M. Shahrabi, M. Moraveji, H. Eslami, Egypt. J. Petrol.
peaks are observed. This suggests secondary flow behavior 26 (2) (2017) 413–420.
within the separator which may rise due to large differences [13] N. Kharoua, L. Khezzar, H. Saadawi, in: ASME 2013 Fluids Engineer-
ing Division Summer Meeting, 2013 V01CT17A013-V01CT17A013.
in velocities between the phases. The simulation results by [14] E. Hansen, ASME-PUBLICATIONS-PVP 431 (2001) 23–30.
the authors agree with previous experimental observations by [15] A.P. Laleh, W.Y. Svrcek, W.D. Monnery, Can. J. Chem. Eng. 90 (2012)
Simmons et al.. While almost all previous CFD articles in this 1547–1560.
268 T. Acharya and L. Casimiro / Journal of Ocean Engineering and Science 5 (2020) 261–268
[16] M.J.H. Simmons, J.A. Wilson, B.J. Azzopardi, Chem. Eng. Res. Des. [27] M. Abdulkadir, V.H. Perez, World Acad. Sci. Eng. Technol. 61 (2010)
80 (5) (2002) 471–481. 35–43.
[17] J.R. Couper, W.R. Penney, J.R. Fair, S.M. Walas, Chemical Process [28] N. Kharoua, L. Khezzar, H. Saadawi, in: Proc. ASME FEDSM, 2013,
Equipment- Selection and Design, Gulf Professional Publishing, 2012 p. 16321.
ISBN: 978-0-12-396959-0. [29] B.A. Grimes, J. Dispers. Sci. Technol. 33 (2012) 578–590.
[18] A. Gerunda, Chem. Eng. 88 (9) (1981) 81–84. [30] B.A. Grimsea, C.A. Doraob, N.V.D.T. Opedala, I. Kralovaa,
[19] W.Y. Svrcek, W.D. Monnery, Chem. Eng. Prog. 89 (10) (1993) 53–60. G.H. Sørlande, J. Sjöbloma, J. Dispers. Sci. Technol. 33 (2012)
[20] R.K. Sinnott, Chemical Engineering Design, Elsevier, 2014 ISBN: 591–598.
978-1-48-329470-4. [31] D. Wilkinson, B. Waldie, M.I. Mohamed Nor, H. Yen Lee, Chem. Eng.
[21] K. Arnold, M. Stewart, Surface Production Operations, third ed., Else- J. 77 (2000) 221–226.
vier, New York, 2008. [32] A.P. Laleh, W.Y. Svrcek, W.D. Monnery, Oil Gas Facil. 1 (6) (2012)
[22] M.J.H. Simmons, E. Komonibo, B.J. Azzopardi, D.R. Dick, Chem. Eng. 57–68.
Res. Des. 82 (10) (2004) 1383–1390. [33] A.P. Laleh, W.Y. Svrcek, W.D. Monnery, Oil Gas Facil. 2 (1) (2013)
[23] O. Abu Arqub, M. Al-Smadi, Numer. Methods Partial Differ. Equ. 34 52–59.
(5) (2018) 1577–1597. [34] Fluent, A.N.S.Y.S., 2013, ANSYS Fluent Theory Guide 15.0. ANSYS,
[24] O. Abu Arqub, Numer. Methods Partial Differ. Equ. 34 (5) (2018) Canonsburg, PA
1759–1780. [35] S. Hu, R.C. Kinter, AIChE J. 1 (1) (1955) 42–48.
[25] O. Abu Arqub, Fundam. Inform. 166 (2) (2019) 87–110. [36] P.J. Brandvik, Ø. Johansen, F. Leirvik, U. Farooq, P.S. Daling, Mar.
[26] O. Abu Arqub, Fundam. Inform. 166 (2) (2019) 111–137. Pollut. Bull. 73 (1) (2013) 319–326.