Professional Documents
Culture Documents
Fuel
journal homepage: www.elsevier.com/locate/fuel
h i g h l i g h t s
" Modeling and simulation of crude oil owing inside an horizontal tube of petrochemical pre-heaters.
" An eulerianeulerian two-phase ow model is used in this formulation.
" A group of seven pseudo-components is used to predict the thermal cracking of the petroleum.
" Vaporization is considered in this model.
a r t i c l e
i n f o
Article history:
Received 30 March 2012
Received in revised form 4 September 2012
Accepted 17 September 2012
Available online 13 October 2012
Keywords:
CFD modeling
Two-phase ow
Phase change
Coke formation
Fired heater
a b s t r a c t
The transport of uids in industrial units takes place usually inside cylindrical ducts. Multiphase systems
owing inside pipes are very common and many times there are also mass, energy and momentum transfer. Two-phase ow is found in many petrochemical processes, as is the case of preheating furnaces
where normally the liquid vaporizes as it ows inside the heater. In the great majority of these heaters,
there is a phase change from liquid to gas. The objective of this work is to simulate the two-phase gas
liquid ow of crude oil inside pipes of petrochemical red heaters with the use of a Computational Fluid
Dynamics (CFD) model to be later used in the prediction of coke formation through a thermal cracking
model and a ternary solubility diagram for the petrol feed, according to Souza et al. [28]. The CFD free
software OpenFOAM was used. There is a growing interest in the use of the OpenFoam project and many
successful models have been implemented using this software. New routines were implemented to estimate temperature and concentration inside the tube, taking into consideration the interaction between
the phases. To estimate the momentum it was used a phase intensive formulation for dispersed twophase ow already implemented in OpenFOAM. The temperature proles were predicted. The k-epsilon
model was used to describe the turbulence and a vaporization model was implemented to estimate the
phase change. A kinetic reaction net for crude oil with seven lumps was used in order to predict the thermal cracking of the crude oil.
2012 Elsevier Ltd. All rights reserved.
1. Introduction
The uid owing inside circular tubes, or ducts, is the most
common type of ow in industrial units. The transport of uids
among operation units of an industry usually takes place inside
cylindrical ducts. In the petrochemical industry a problem that occurs during such transport is the formation of coke inside the pipes
of red heaters due to thermal cracking of the oil, which happens
in most operations under high temperatures. This cracking reaction leads to the deposition of coke on the walls of these ducts.
The deposition, in turn, increases pressure drop and causes the
Corresponding author. Tel.: +55 19 3521 3967; fax: +55 19 3521 3910.
E-mail address: nunhez@feq.unicamp.br (J.R. Nunhez).
0016-2361/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.fuel.2012.09.065
197
Nomenclature
ai
Yi
Cl
C1
C2
Cp
D
d
f
g
H
h
k
M
Nu
p
Pk
Pr
Q
R
Re
Sc
T
T
t
U
concluded that the kinetic model based on these pseudocomponents reproduced the data found in the literature and can
be applied to different oils and in different conditions of operation.
The asphaltenes are present in crude oils and is an important
class of compounds to be considered in this work. They represent
a complex mixture of hydrocarbons that are dened by their solubility characteristics and not by an specic chemical classication.
The precipitation of asphaltenes depends on the stability of complex colloidal systems. The composition acts as the main cause
for the deposition of asphaltenes, but the pressure and temperature are also important factors in the occurrence of this phenomena, especially temperature. Thus, any change in pressure,
temperature or composition can cause polymerization of asphaltenic molecules, generating clusters which increase in size and can
lead to coke deposition on the tube walls. Asphaltenes tend to settle in areas close to regions of low temperature or low uid
velocity.
Mansoori and Leontaritis [17] proposed a model to predict the
occulation of asphaltenes, known as colloidal model, which is
based on the assumption that the asphaltenes exist in oil as suspended solids, in colloidal form and stabilized by resins adsorbed
on its surface. In a given system, in which the type and amount
of oil and of asphaltenes are xed, a change in concentration of resins will cause a variation in the amount of resins that are adsorbed
on particles of asphaltenes. If this change is such that the amount
of resins is insufcient to cover the entire surface of the particles,
such resins will aggregate and end up precipitating. There are currently some works where kinetic nets are proposed for the thermal
cracking reaction of the oil. In the study by Souza et al. [28] the kinetic net present in Mendona Filho [19] was used.
Souza et al. [28] developed a numerical model for single-phase
ow inside cylindrical ducts applied to petrochemical red heaters.
The ow equations that govern this phenomena were numerically
integrated using a nite volume scheme. An interesting feature of
Greek letters
volume fraction ()
turbulence dissipation rate (m2 s3)
j
thermal conductivity (W m1 K1)
k
thermal diffusivity (m2 s1)
l
molecular viscosity (Pa s)
m
kinematic viscosity (m2 s1)
q
density (kg m3)
rk
constant (1.00)
r
constant (1.30)
198
@ ea
M v ap Mcond
r Uea r Ur ea 1 ea
@t
qa
where U = elUl + egUg is the mean velocity; Ur = Ug Ul is the relative velocity; Mvap is the mass rate of vaporization; Mcond is the mass
rate of condensation; q is the density; g represents the gas phase; l
indicates liquid phase and e is the volume phase fraction.
2.1.2. Momentum equations
The gas and liquid phase momentum equations may be expressed as
@Ua
rea eff
rp
Fa
T
g
Ua rUa r Teff
a
@t
ea a
qa
ea qa
2
2
eff
Teff
rUa rUTa Ir Ua Ika
a la
3
3
@kl
Ul rkl r
@t
meff
l
rk Pk l Sk
rk l
@ l
Ul rl r
@t
meff
l
r l C 1 Pk C 2 l S
r l
kl
lturb
C l ql
l
where rk; r; Cl; C1 and C2 are constants of the k model and Pk
is the turbulence production, which is given in
Pk lturb
rUl rUl rUl T
l
@T a
rea
keff
Ua rT a r keff
a rT a
a rT a
@t
ea
Q w;a
Ca
ai conv
Hv ap
ha T b T a
ea qa
ea qa
ea qa
ai
6el eg
el dl eg dg
10
keff
a
jeff
a
qa Cpa
11
jeff
a is given
12
The vaporization occurs only near the wall and starts when the liquid temperature Tl is greater then the average boiling temperature
Tb. The evaporated mass is calculated according to
M v ap
Q w;l
Hv ap
13
Q w;l
14
and the convective heat transfer coefcient inside pipes for the liquid phase is described in Bergman et al. [1] and expressed by
conv
hl
eff
kl Nul
15
Dt
Nul
f =8Re 1000Pr
16
f 0:7 ln Re 1:642
17
Q g;l
Hv ap
M cond
18
Q g;l
conv
ai hg T g
Tb
19
conv
hg
hg
ql Cpl
re
@Y i;a
a
reac
Deff
Ua rY i;a r Deff
a rY i;a
a rY i;a Ri;a
@t
ea
21
For turbulent ows, it is consider that the Schmidt number approaches unity.
Sc
4
conv
T w T l
ea hl
Dt
s
p kUr k jl
1
4 dg ql Cpl 1 jturb
=jl
l
20
199
meff
l
Dl
1
22
Dl meff
l
23
Rdistillate;l k1 Y saturated;l
Rsaturated;l k2 Y aromatics;l k1 Y saturated;l
24
25
26
27
28
29
Rcoke;l k6 Y asphmeso;l
30
ki
1
B
exp A
3600
RT l
31
In this equation A and B are parameters that differ for each reaction
and R is the universal gas constant. The subscripts asphsol and
asphmeso indicate soluble asphaltenes and asphaltenes in
mesophase.
3. Simulation
The free opensource software for CFD OpenFOAM was used generate the model. The momentum and turbulence equations were
already implemented in OpenFOAM. New subroutines where
added to estimate the phase change, heat and mass transfer, so a
new eulerianeulerian two-phase model was developed. An horizontal cylindrical geometry of 25 m of length and a diameter of
0.152 m was constructed to represent a segment of the industrial
red heater tube as shown in Fig. 2.
A grid independence test was carried out with three hexaedrical
meshes containing 442,000, 552,000 and 663,000 elements using
the O-grid technique to avoid mesh distortion. The pressure drop
relative error between the three meshes was analyzed showing a
relative error of 3.6% for the pressure drop when comparing the
Fig. 1. Kinetic net of the thermal cracking of petroleum based on the modied model of Kseoglu and Phillips [12] describe in Mendona Filho [19].
200
1
2
3
4
5
6
29.9
23.43
13.57
17.07
19.34
27.54
175000
136000
85000
96000
103000
168000
Table 2
Arabian light mass fraction for the pseudocomponents at the inlet.
rst and second meshes and 1.7% between the second and third. It
should be highlighted that pressure is much more mesh sensitive
than the velocities. It was considered that the 552,500 mesh elements was adequate. The mesh domain is a cylindrical tube geometry of 25 m in length and diameter of 0.152 m. The mesh contains
500 elements in the axial direction, 13 elements in each direction
of the internal o-grid block and 18 elements in radial region for the
other four external blocks composing the o-grid. The mesh quality
indicates a average non-orthogonality of 4.74, maximum aspect ratio of 56.55 and the maximum skewness was around 0.48. The
mesh can be seen in Fig. 3.
Due to the phase separation, the gas tends to accumulate at the
top of the tube and due to this reason the phase intensive formulation was adopted.
For simulation purposes, a wall temperature of 723 K is considered. The reaction rate depends heavily on the temperature. Table 1
shows the kinetic constant of reaction parameters.
The mass fraction of the pseudo-components are presented in
Table 2 for the Arabian light oil and in Table 3 for the Athabasca
bitumen.
The same strategy of Souza et al. [28] to predict coke formation
using a ternary diagram (Takatsuka et al. [29]), shown in Fig. 4, was
used instead of using a thermodynamic model. The model considers that the pseudo-component asphaltene is the only one that
leads to coke formation and its formation is associated with the
amount of asphaltene formed. The logic is that the more the
asphaltenes are formed the more coke will tend to precipitate at
the tube walls.
In order to evaluate the phase change and coke formation two
situations were carried out and compared. Table 4 presents the
operational conditions using a average boiling temperature of
603K for the Arabian light oil and Table 5 uses an average boiling
temperature of 615 K for the Athabasca bitumen.
Pseudo-component
Yi
Distillate
Resins
Saturated
Aromatics
Asphaltenes
Coke
0.00
0.37
0.15
0.37
0.11
0.00
Table 3
Athabasca bitumen mass fraction for the
pseudo-components at the inlet.
Pseudo-component
Yi
Distillate
Resins
Saturated
Aromatics
Asphaltenes
Coke
0.00
0.17
0.05
0.23
0.55
0.00
Table 4
Initial and boundary conditions used in the simulations.
Variable
Inlet
Outlet
Walls
Ul
Ug
(0 0 1.5)
(0 0 1.5)
1.0
0.0
rp = 0
583
603
rUl = 0
rUg = 0
rel = 0
reg = 0
0.0
0.0
el
eg
p
Tl
Tg
r el = 0
r eg = 0
r p=0
rTl = 0
rTl = 0
At the beginning of the tube there is only liquid present. However the coupled equations for the liquid and gas phase have to be
Inlet
Outlet
Walls
Ul
Ug
(0 0 1.5)
(0 0 1.5)
1.0
0.0
rp = 0
595
615
rUl = 0
rUg = 0
rel = 0
reg = 0
0.0
0.0
el
eg
p
Tl
Tg
r el = 0
r eg = 0
rp = 0
rTl = 0
rTl = 0
201
Fig. 5. Comparison of the estimated pressure gradient by the CFD model and the
correlation using friction factor for turbulent ow.
Fig. 6. Comparison of the estimated velocity by the CFD model and the experimental data in the radial position for fully turbulent ow by Laufer [16].
202
Fig. 9. Streamline proles of liquid Ul and gas Ug for Arabian light crude oil at Tb = 603.15 K.
Fig. 10. Streamline proles of liquid Ul and gas Ug for Athabasca bitumen at Tb = 615 K.
This rst model considers that the boiling of the liquid phase is
limited to a single boiling temperature, whereas the petrol vaporizes in a large temperature range. This aspect is being implemented
in a future model and it is important to be considered since the
boiling temperature of the petrol increases as it ows inside the
tube.
Fig. 9 shows the gas and liquid streamline proles for the Arabian crude oil. It can be observed in Fig. 9(a) the turbulent movement of the liquid showing the interaction with the gas phase.
Fig. 9(b) shows the gas being formed and rising quickly to the
upper part of the tube.
Fig. 10 shows the gas and liquid streamline proles for the Athabasca bitumen. The same behavior observed for the Arabian light
was noticed for the Athabasca bitumen. However, a smaller gas
volume fraction was observed.
This rst model simulates only a tube length of 25 m at the entrance of the red heater. The predicted uid temperatures did not
reached values high enough where high conversion rates and higher coke formation would be expected. Due to this a negligible formation of coke was predicted.
5. Conclusions
The proposed eulerianeulerian model showed robustness for
the simulated cases. Further renement of the model will require
a more comprehensive vaporization model.
The phase intensive formulation used in this work applies the
phase fraction to only one term of the transport equation reeaa ,
making it possible to attain to a numerical solution giving stability
to the model even with high phase fractions. The energy and multicomponent mass transport equations were solved numerically
according to this formulation and it was observed a good stability
in the estimation of temperature and mass fractions.
The use of hydrostatic pressure boundary conditions for this
two-phase ow was implemented for a stable pressurevelocity
coupling.
Coke concentration is higher in regions where liquid temperature is higher. This behavior is expected since coke is generated
in regions of high temperature.
The non symmetric behavior of this two-phase three-dimensional ow inside red heater tube was readily observed as shown
by the streamlines.
Acknowledgement
The authors acknowledge the nancial support of the Brazilian
sponsoring agencies CAPES and FAPESP for this research.
References
[1] Bergman TL, Incropera FPE, Lavine AS, Dewitt DP. Fundamentals of heat and
mass transfer. USA: John Willey and Sons; 2011.
[2] Bhramara P, Rao VD, Sharma KV, Reddy TKK. CFD analysis of two phase ow in
a horizontal pipe-prediction of pressure drop. Int J Aerospace Mech Eng
2009;3(2):728.
[3] Carneiro JE, Kaufmann V, Polifke W. Numerical simulation of droplet
dispersion and evaporation with a moments-based CFD model; COBEM 1520, 2009. In: 20th International Congress of Mechanical Engineering. Gramado,
RS, Brazil; 2009.
[4] Akshay Gandhir A, Wardle KE. CFD analysis of uid ow above the upper weir
of and annular centrifugal contactor. Sep Sci Technol 2012;47:110.
[5] Gosman AD, Lekakou C, Politis S, Issa RI, Looney MK. Multidimensional
modelling of turbulent two-phase ows in stirred vessels. AIChE
1992;38(12):194656.
[6] Gopalakrishnan S, Schmidt DPA. Computational study of ashing ow in fuel
injector nozzles. SAE Paper-2008-01-0141 2008.
[7] Ishii M, Hibiki T. Thermo-uid dynamics of two-phase ow. New
York: Springer; 2011.
[8] Issa RI. Solution of the implicitly discretised uid ow equations by operatorsplitting. J Comp Phys 1986;62:4065.
[9] Komatsu S, Hori Y, Shimizu S. VISABC process yields high conversion.
Hydrocarbon Process 1986;65:424.
[10] Kseoglu R, Phillips CR. Hydrocracking of Athabasca bitumen. Kinetics
formation gases. Fuel 1988;67(4):5526.
203
[11] Kseoglu R, Phillips CR. Kinetic models for the non-catalytic hydrocracking of
Athabasca bitumen. Fuel 1988;67:90615. 1988b.
[12] Kseoglu R, Phillips CR. Kinetcs and product yield distributions in the CoO
MoO2/Al2O3 catalyzed hydrocracking of Athabasca bitumen. Fuel
1988;67:14116.
[13] Kseoglu R, Phillips CR. Effect of reaction variables on the catalytic
hydrocracking of Athabasca bitumen. Fuel 1988;67(9):12014.
[14] Kseoglu R, Phillips CR. Kinetics of non-catalytic hydrocracking of Athabasca
bitumen. Fuel 1988;66(6):7418.
[15] Krishna R, Saxena AK. Use of an axial dispersion model for kinetic description
of hydrocraking. Chem Eng Sci 1989;44:70312.
[16] Laufer J. The structure of turbulence in fully developed pipe ow. National
Advisory Committee of Aeronautics 1174. 1952. National Aeronautics and
Space Administration of The USA, Washington DC
[17] Mansoori GA, Leontaritis KJA. Colloidal model for asphaltene occulation from
petroleum uids. Iranian J Sci Technol 1992:16.
[18] Matos EM, Guirardello R. Modeling and simulation of a pseudo-two-phase
gasliquid column reactor for thermal hydrocracking of petroleum heavy
fractions Brazilian. J Chem Eng 2002;19:31934.
[19] Mendona Filho. Modelagem e Simulao de Reator Trifsico Aplicado leos
Pesados com Reaes de Hidrocraqueamento Trmico e Cataltico,
Desmetalizao de Coque na Mistura e no Catalisador, Msc thesis. Unicamp
Brazil; 1998.
[20] Mercado ER, Souza VC, Guirardello R, Nunhez JR. Modeling ow and heat
transfer in tubes using a fast CFD formulation. Comput Chem Eng
2001;25:71322.
[21] Michita E. Modeling of subcooled nucleate boiling with OpenFOAM. Master
degree thesis, division of nuclear reactor technology. Royal Inst of technology,
Stockholm; 2011.
[22] Migliaccio M, Beatrice C, Briani M, Fraioli V. Numerical simulations of diesel
engine combustion by means of OpenFOAM CFD ToolBox. SAE paper 2009-240002; 2009.
[23] Nunhez JR. The CFD in chemical engineering and industrial process
symposium the 8th world congress of chemical engineering. Can J Chem Eng
2011;89:6257.
[24] OpenCFD ltd. The open source CFD toolbox: programmers guide; 2011.
[25] OpenCFD ltd. The open source CFD toolbox: users guide; 2011.
[26] Rusche H. Computational uid dynamics of dispersed two-phase ows at high
phase fractions. Ph.D. Thesis, Imperial College of Science, Technology and
Medicine, London, UK; 2002.
[27] Silva LFLR, Lage PLC. Development and implementation of a polydispersed
multiphase ow model in OpenFOAM. Comput Chem Eng 2011;35:265366.
[28] Souza BA, Mato EM, Guirardello R, Nunhez JR. Predicting coke formation due to
thermal cracking inside tubes of petrochemical red heaters using a fast CFD
formulation. J Pet Sci Eng 2005;51:13848.
[29] Takatsuka T, Wada Y, Hirohama S, Fukui Y. A prediction model for dry sludge
formation in residue hydroconversion. J Chem Eng Jpn 1989;22:289303.
[30] Wardle KE. Open-source CFD simulations of liquid-liquid ow in the annular
centrifugal contactor. Sep Sci Technol 2011;46:240917.
[31] Wolfert K. Non-equilibrium mass transfer between liquid and vapor phases
during depressurization processes in transient two-phase ow. In: Proc 2nd
CSNI specialists meeting, vol. 2. Paris; 1978. p. 13771387.