Chemical Engineering Journal 286 (2016) 436446

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Coupled simulation of convection section with dual stage steam feed

mixing of an industrial ethylene cracking furnace
Guihua Hu, Benfeng Yuan, Liang Zhang, Jinlong Li, Wenlin Du, Feng Qian
Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University of Science and Technology, Shanghai 200237, China

h i g h l i g h t s

A complete coupled simulation method of convection section is proposed.

Velocity fields are not uniform along width direction due to asymmetrical structure.

Recirculation zones cause a longer residence time of flue gas and local overheating.

Process gas and tube skin temperature and heat flux have axial and radial profiles.

Changes of flow pattern are effected by gravity and centrifugal force.

a r t i c l e

i n f o

Article history:
Received 30 July 2015
Received in revised form 26 October 2015
Accepted 28 October 2015
Available online 2 November 2015
Keywords:
Coupled simulation
Convection section
Dual stage steam feed mixing
Cracking furnace
Computational fluid dynamics
Evaporation

a b s t r a c t
A complete coupled simulation of the convection chamber and tubes with dual stage steam feed mixing
of an industrial ethylene cracking furnace has been carried out with the computational fluid dynamics
(CFD) method for the first time. In the convection chamber, the standard ke model and discrete ordinates (DO) radiation model were respectively used in the descriptions of turbulence characteristics
and radiative heat transfer. In the tubes, renormalization group (RNG) ke model and volume of fluid
(VOF) model were respectively applied to the turbulence flow and the liquidvapor two phases flow.
Simulation results agree well with the industrial data. Based on the coupled result, a dynamic simulation
was calculated in the feedstock preheater (FPH). Simulation results show that the velocity and temperature fields are inhomogeneous distributions along the width direction due to the asymmetrical structure
of convection chamber. Two recirculation zones occur at the corner both near and away from the
entrance to the convection chamber, which will cause a longer residence time of flue gas and local overheating in furnace wall of convection chamber. The process gas temperature, tube skin temperature and
heat flux profiles are respectively different along the axial and radial direction of the high temperature
coil (HTC-I). The changes of flow pattern from bubble flow to spray flow are effected by gravity and centrifugal force during evaporation. The results will be helpful for the design and operation in cracking
furnace.
2015 Elsevier B.V. All rights reserved.

1. Introduction
Ethylene cracking furnace is the key equipment in the production of ethylene, which influences the yield efficiency of feedstock,
selectivity of important products, equipment energy consumption,
etc. In the world 99% of ethylene production has been adopted by
tubular cracking furnace [1].
Tubular cracking furnace mainly includes two parts: radiation
section and convection section. For the study of radiation section,
many achievements have been gained at home and abroad.
Corresponding author at: No. 130 of Meilong Road, Shanghai 200237, China.
E-mail address: fqian@ecust.edu.cn (F. Qian).
http://dx.doi.org/10.1016/j.cej.2015.10.093
1385-8947/ 2015 Elsevier B.V. All rights reserved.

Heynderick et al. [2], Oprins et al. [3,4], Stefanidis et al. [5,6],

Coelho [7] and Habibi et al. [8] early used the numerical method
to study the information of velocity and temperature fields of
radiation section in the cracking furnace. The influences of the grid
formation, combustion model and radiation model on the numerical simulation of cracking furnace were suggested. Lan et al. [9]
and Han et al. [10] simulated different types of cracking furnace
using computational fluid dynamics (CFD) software, and got a
detailed velocity, temperature and concentration fields. Liu et al.
[11] improved the CFD models of the firebox, and thus improving
the computational efficiency. Hu et al. [12,13] improved coupled
method in which CFD software was used to simulate the firebox,
and the software Coilsim1D was used to simulate the tube.

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

437

Nomenclature
I
radiation intensity (J/m2/s)
_ pq , m
_ qp mass transfer from phase p to phase q and from phase q
m
to phase p respectively (kg/m3/s)
_ l!v , m
_ v !l rates of mass transfer due to evaporation and conm
densation respectively (kg/m3/s)
n
refractive index ()
~
r
position vector ()
~
s
direction vector ()
~
s0
scattering direction vector ()
s
path length (m)
Saq
mass source term of the phase q (kg/m3/s)
Su
source term ()
t
time (s)
T
local temperature (K)
Tl
temperature of the liquid phase (K)
Tv
temperature of the vapor phase (K)
u
fluid velocity (m/s)

The results show that the method can greatly improve the computational efficiency of the process gas side.
Convection section study early focused on the calculation of the
macroscopic phenomenon. For example, He et al. [14] developed a
simulation software of convection section based on Pro-II. Liu et al.
[15] and Zhou and Yang [16] respectively established their convection section programs based on Aspen Plus [17] platform. However,
the previous studies focused on the process modeles of convection
section, and greatly simplified the fluid flow and heat transfer processes inside the tube, thus, the improved understanding of the
occurring processes could not be taken into account. With
the development of CFD technique and computer technology, the
research in microscopic phenomena of the convection section has
also been paid more and more attention. The difficulty of numerical simulation study in the convection section lies in vaporization
of hydrocarbon feedstock in the tube. The existence of two-phase
flow makes the fluid flow, heat transfer and mass transfer become
more complicated. De Schepper et al. [18] added source terms of
the energy and mass source to the control equations by using the
volume fraction method, thereby preliminarily studying the fluid
flow boiling process of convection section. Mahulkar et al.
[19,20] and De Schepper et al. [21] studied numerical simulation
of coking phenomena of heavy feedstock pyrolysis, which provided
suggestions for reducing thickness of coke layer. Mertinger et al.
[22] studied the cause of heat tube corrosion in the convection section with ANSYS FLUENT 14.0 software, providing theoretical reference for the dangerous case judgement.
In the convection section, vaporization process of feedstock
hydrocarbon is an endothermic process, and the vaporization heat
required is provided by the waste heat of the flue gas. Therefore,
flue gas flow field has a close relationship with heat transfer and
vaporization inside tubes. Only efficient coupling heat transfer process between the convection chamber and tubes does accurately
get the vaporization and heat transfer and other such characteristics inside tubes, thus accurately predicting outlet temperature of
flue gas and process gas in convection section. De Schepper et al.
[23] used the CFD method to carry out coupled simulation of flue
gas side and the internals of the heat exchanger tubes in the convection section. However, they simulated convection section with
only one steam flux, and did not consider the heat changes of feedstock in the evaporation section, which may impact the coupled
simulation results.

Greek letters
absorption coefficient (1/m)
volume fraction of the phase q ()
gas density (kg/m3)
density of the liquid phase (kg/m3)
density of the vapor phase (kg/m3)
density of the phase q (kg/m3)
Cu
generalized diffusion coefficient ()
r
StefanBoltzmann constant (r = 5.672  108 W/m2K4)
rS
scattering coefficient (1/m)
u
dependent variable ()
U
phase function ()
X0
solid angle ()

a
aq
q
ql
qv
qq

In this work coupled convection section/tubes with dual stage

steam feed mixing simulations have been performed for an industrial naphtha cracking furnace. The highly complex coupling process such as fluid flow, heat transfer and mass transfer in the
convection section is considered in the process of calculation. In
the convection chamber, the compressible formulation of the
Reynolds-averaged NavierStokes (RANS) equations is adopted to
simulate the fluid flow. The standard ke model is used for closure.
The discrete ordinates (DO) radiation model is used for modeling
the radiative heat transfer. For the tubes simulations, the Renormalization Group (RNG) ke model is applied to turbulence flow
inside tubes. VOF model is used for two-phase flow. Temperature
and velocity distributions of the flue gas in the convection chamber
and process gas temperature, heat flux, tube skin temperature profiles along the tubes are obtained. The simulation results are in
agreement with the industrial data.
2. Mathematical models
2.1. Flow model
The fluid flow in the convection section complies with the law
of nature conservation, including the conservation of mass, energy
and momentum, and control equations are described by mathematics method. In this paper, NavierStokes Reynolds average is
used to model the fluid flow of convection chamber and tubes.
For convection chamber, the standard ke model is used for closure. For tubes, because there is amount of swirling flow, RNG k
e model is more suitable for turbulence flow inside tubes than
the standard ke model [24]. Control equations are mass, energy
and momentum conservation equations, and convection chamber
also includes species transport equation. Because each equation
has the similarity in form, so the following general form of representation can be described as:

@qu
divquu divCu gradu Su
@t

2.2. Multiphase flow model

In the FPH, hydrocarbon feedstock is heated and vaporization
occurs from liquid into vapor. Therefore, the volume of fluid

438

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

(VOF) model based on EulerianEulerian approach is applied

to simulation of the multiphase flow in the feedstock preheater
(FPH).
In the VOF model, the volume fraction of the phase is added in
the computational cell. In each control volume, the volume fractions of all phases sum to unity. Thus, if the qth fluids volume fraction in the cell is denoted as aq, then the following three conditions
are possible:
d aq = 0: the cell of the qth fluid is empty;
d aq = 1: the cell of the qth fluid is full;
d 0 < aq < 1: the cell contains the interface between the qth fluid
and one or more other fluids.
For the qth phase, the continuity equation of volume fraction
has the following form:

"

qq

n




X


!
@

_ pq  m
_ qp
m
aq qq r  aq qq v q Saq
@t
p1

#
2

The volume fraction equation will not be solved for the primary
phase; the primary-phase volume fraction will be computed based
on the following constraint:

aq 1

q1

This paper adopted the evaporationcondensation model proposed by Lee [25]. The model added the mass and energy exchange
source terms of between vapor and liquid to the control equations,
and correlated fluid saturation temperature. Based on the following temperature regimes, the mass transfer can be described as
follows:
If Tl > Tsat

_ l!v bl al ql
m

T l  T sat
T sat

If Tv < Tsat

_ l!v bv av qv
m

T sat  T v
T sat

In this work, the saturation temperature Tsat is assumed

to the average temperature of bubble and dew points of
hydrocarbon feedstock, 366.65 K. The heat transfer corresponding
with the evaporation or condensation process can be
calculated by multiplying the calculated mass transfer with the
latent heat.
In equations above, bl and bv is the mass transfer time relaxation parameters, 100 s1 [26], which make the difference between
the interfacial temperature and the saturation temperature the
minimum under meeting the calculation convergence.
2.3. Radiation model
The DO model is probably the best suited for computing radiation for the problems with localized sources of heat. The DO model
has a broader application range, and is to be preferred [27]. The DO
radiation model solves the radiative transfer equation (RTE) for a
finite number of discrete solid angles, each associated with a vector
direction ~
s fixed in the global Cartesian system (x, y, z). The DO
model solves for as many transport equations as there are directions ~
s. The solution method is identical to that used for the fluid
flow and energy equations. The DO model uses a conservative variant called the finite-volume scheme, and its extension to unstructured meshes. The mathematical expression of the DO model is as
follows:

r  I~
r;~
s~
s a rs I~
r;~
s an2

rT 4 rs

p 4p

Z 4p

I~
r;~
s0 U~
s ~
s0 dX0

6
Domain-based weighted-sum of gray gas model (WSGGM) is
used for the calculation of the emissivity of the flue gas mixture
[13].
2.4. Physical properties
In the industry, there are many kinds of hydrocarbon
feedstock which usually include naphtha, diesel oil of ordinary
pressure, light hydrocarbon, heavy-vacuum gas oil, etc. And
online detection instrument is expensive, which is seldom
adopted in actual production, therefore it is difficult to get the
detailed physical property data of hydrocarbon feedstock in industry, thus leading to difficulty for the numerical simulation of
convection section.
In this work, the combinations of COILSIM1D [28] and Aspen
plus give rise to the detailed information of material using industrial data. The detailed steps are described as follows: first of all,
the collected industrial data such as the density, distillation range,
PIONA value of naphtha are input to COILSIM1D. The industrial
indices of the naphtha feedstock are described in Table 1. Secondly,
COILSIM1D is adopted to build feedstock model and to obtain the
detailed feedstock component data. Thirdly, these components
are input to Aspen Plus, which is used to calculate the physical
properties of mixture based on SoaveRedlichKwong equation
[29]. Finally, based on the data of calculation above, polynomial
functions expressing the variation of the physical properties with
temperature are obtained by least square method, as shown in
the Supporting information. Among these physical properties, the
polynomial functions of the flue gas, the dilution steam and the
mixture of naphtha and dilution steam are calculated by Aspen
Plus, similar to the last two steps of the physical properties calculation of naphtha. These polynomial functions of physical properties are input into Ansys Fluent 14.0 [30] for the numerical
simulation work in the next step. The detailed calculation process
is shown in Fig. 1.
3. Simulation procedure
3.1. Geometry and operating conditions
The structure diagram of dual stage steam feed mixing of convection section of ethylene cracking furnace is shown in Fig. 2. This
section consists of eight tube sections, which are the FPH, economizer (ECO), high temperature coil I (HTC-I), high temperature coil

Table 1
Industrial indices of the naphtha feedstock.
Specific density (20 C) (g/cm3)
Specific density (15 C) (g/cm3)

0.715
0.7196

ASTM D86 boiling points (K)

IBP
10%
50%
90%
95%
FBP

318.15
342.15
365.15
434.15
442.15
464.15

PIONA (wt%)
n-Paraffins
i-Paraffins
Olefin
Naphthenes
Aromatics

30.16
41.56
1.37
16.13
10.78

439

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Computing feedstock
information with
COILSIM1D

Input
information

1. Density
2. Distillation range
3. PIONA value

NAP

FPH
BFW

ECO

Drum

DS

Computing feedstock
properties with Aspen Plus

Output
information

HTC-I

1. Density
2. Thermal conductivity
3. Heat capacity
4. Dynamic viscosity

HTC-II
Drum

HPSSH-I

Fig. 1. Model structure of hydrocarbon feed.

Desuperheater

HPSSH-II

DS

High pressure steam

DSSH
HTC-III
Radiation section

Inlet of convection
section
Fig. 3. Diagram of convection section process with dual stage steam feed mixing.

The convection section dimensions and operating conditions

are given in Table 2. As there are a large number of straight tubes
and heat flux change of each straight tube is taken into consideration, the journal and executable exe files are compiled, which are
used to treat with the larger repetition works such as automatic
processing of the conversion of heat flux. The detailed structure
of each tube sections is described as follows.
3.1.1. Convection chamber
The convection chamber is the part of convective heat transfer
which depends on the flue gas from the outlet of radiation section.
In the convection chamber, the multiple rows of tubes are densely
arranged, and the flue gas scours these tubes by the great velocity,
and carries out effective convection heat transfer. Hexahedral cells
are used to discretize the convection chamber. Grid refinement is
used in the tube skin zones with small diameter such as FPH and
ECO. The number of grid cells of the convection chamber is
1,563,462.

Fig. 2. Schematic diagram of convection section with dual stage steam feed mixing.

II (HTC-II), high pressure steam super-heater I (HPSSH-I), high

pressure steam super-heater II (HPSSH-II), dilution steam superheater (DSSH) and high temperature coil III (HTC-III), respectively.
Feedstock is heated in the FPH and gradually from the liquid phase
to gas phase. Boiler feed water is heated in the ECO. The first steam
flux is carried out before the feedstock enters the HTC-I. The second steam flux enters the DSSH and mixed with the mixture of
HTC-II outlet, then entering the HTC-III to be heated. The detailed
process diagram of convection section is shown in Fig. 3.

3.1.2. Tube sections

FPH tube section adopts staggered arrangement, as shown in
Fig. 2. It consists of 6 horizontal tubes, and each tube has 24 tube
passes, so this tube section contains a total of 144 straight tubes.
There are two kinds of tube structure whose layer numbers are
all 16. In the first structure, the number of tube passes s of each
layer from the top to the bottom is 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1,
2, 1, 2, 1, respectively. In the second structure, the number of tube
passes of each layer from the top to the bottom is 1, 2, 1, 2, 1, 2, 1, 2,
1, 2, 1, 2, 1, 2, 1, 2, respectively. The grid cell number of each tube is
431,483.
ECO tube section also adopts staggered arrangement, as shown
in Fig. 2. The straight tube number of each layer from the top to the
down is two. The grid cell number of each tube is 305,501.
Cracking feedstock heated in FPH, together with dilution steam,
will enter HTC-I, therefore, the inner diameter of straight tube
HTC-I (high temperature section) is bigger than that of FPH and
ECO. HTC-I tube section also consists of 6 tubes where each tube
consists of 6 straight tubes and 5 bend tubes. The grid cell number
of each tube is 212,848. HTC-II tube section consists of 6 tubes

440

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Table 2
Convection section dimensions and operating conditions.
Convection chamber specifications
Length (z-direction) (m)
Width (x-direction) (m)
Height (y-direction) (m)
Number of tube sections
Number of total tubes
Straight length of each tube pass

10.3689
2.26
13.385
8
432
10.368

FPH tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of naphtha (kg/s)
Tube inlet temperature (K)
Tube outlet pressure (g) (MPa)

24
6
0.0603
0.00554
8.4861
333.15
0.48

ECO tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of boiler feed water (kg/s)
Tube inlet temperature (K)
Tube outlet pressure (g) (MPa)

18
6
0.0603
0.00874
10.4475
418.15
11.92

HTC-I and HTC-II tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of naphtha (kg/s)
Mass flow rate of dilution steam (kg/s)
Tube inlet temperature in HTC-I
HTC-2 tube outlet pressure (g) (MPa)

6
6
0.1143
0.00602
8.4861
0.8486
406.15
0.40

HPSSH-I and HPSSH-II tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of high pressure steam in HPSSH-I and HPSSH-II
(kg/s)
Mass flow rate of boiler feed water in HPSSH-II (kg/s)
HPSSH-1 tube inlet temperature (K)
HPSSH-2 tube inlet temperature (K)
HPSSH-1 tube outlet pressure (g) (MPa)
HPSSH-2 tube outlet pressure (g) (MPa)

3
10
0.0603
0.00874
10.2428
0.5850
597.15
656.15
11.69
11.55

DSSH tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of dilution steam (kg/s)
Tube inlet temperature (K)
Tube outlet pressure (g) (MPa)

4
6
0.1143
0.00602
4.2431
458.15
0.40

HTC-III tube section

Number of tube passes
Number of tubes
Tube outer diameter (m)
Thickness of tube (m)
Mass flow rate of naphtha (kg/s)
Mass flow rate of dilution steam (kg/s)
Tube inlet temperature (K)
Tube outlet pressure (g) (MPa)

4
6
0.1413
0.00655
8.4861
5.0917
719.15
0.33

where each tube consists of 6 straight tubes and 6 bend tubes. The
grid cell number of each tube is 101,036.
HPSSH-I tube section consists of 10 tubes where each tube consists of 3 straight tubes and 2 bend tubes. The grid cell number of
each tube is 48,650.
HPSSH-II tube section structure is the same as the HPSSH-I. The
grid cell number of each tube is 48,804.

DSSH tube section consists of 6 tubes where each tube includes

4 straight tubes and 3 bend tubes. The grid cell number of each
tube is 13,417.
The mixture of hydrocarbon feedstock and dilution steam from
outlet of HTC-II and dilution steam from outlet of DSSH are mixed
to enter HTC-III. HTC-III tube section consists of 6 horizontal
arrangement tubes where each tube includes 4 straight tubes
and 3 bend tubes. There are two structures which respectively
have 241,970 and 323,273 grid cells.
3.2. Boundary condition and computing method
The real industry data are applied to boundary conditions of different modules of convection section. For the convection chamber
and tube sections, mass flow inlets are respectively adopted as
inlet boundary, and pressure outlet is adopted as outlet boundary
condition.
The non-linear governing equations are discretized implicitly
through a second-order upwind scheme and linearized to produce
a series of equations for the dependent variables in every computational cell. Semi-Implicit Method for Pressure-Linked Equations
(SIMPLE) algorithm is used to solve the coupled momentum,
energy and species transport equations. The convergence was
judged by monitoring residuals and the overall mass and energy
balances. The convergence criterion is that all scaled residuals
decrease to 103. The exceptions are the energy and the radiation
equations, where the criterion is 106.
3.3. Coupled convection chamber and tubes simulation
In order to make the computation results more accurate and
effective, a coupled simulation method of convection chamber
and tube sections is adopted. The coupled calculation algorithm
is shown in Fig. 4. In the first iteration, an external tube skin temperature profile from industrial measurement is adopted in the
convection chamber simulation. The simulated heat flux profile
to the tubes is then applied in the tubes simulation. Because of
the complex structure for tubes, it is necessary to respectively
divide grids and compute for each tube section. For example, the
inlet boundary condition of HTC-II is the outlet boundary condition
of HTC-I. After finished all the tube sections calculation, updated
external tube skin temperature profiles for the furnace simulation
are output. This procedure is repeated until the maximum difference of the external tube skin temperature between two successive
iterations is below a specified predefined threshold value of 1 K.
4. Results and discussion
A naphtha is adopted hydrocarbon feedstock of convection section of the dual stage steam feed mixing. Table 3 is the comparison
of simulation results and industrial data. The simulated average
outlet temperature of each tube section is in good agreement with
the industrial data. The exhaust gas temperature of convection
chamber is 385.15 and 388.65 K for industry data and simulation
result, respectively. The relative error is 0.9%. The cross section
temperature of tube section is 884.15 and 887.38 K for industry
data and simulation result, respectively. The relative error is 0.36%.
4.1. The simulation results of convection chamber
Fig. 5 shows flue gas velocity contours at section z = 0.5, 5 and
10 m of convection chamber respectively for dual stage steam feed
mixing. The same velocity distributions are discovered in the three
sections, thus showing the flue gas velocity field is basically

441

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Construct geometric model

and grid of tubes

Start

Establish initial conditions of

tubes

Construct geometric model

and grid of convection
chamber

Calculate
equations of
DSSH

Calculate
equations of FPH

Establish initial conditions of

convection chamber

Convergence?

Input the measured tube skin

temperature profiles

N
Convergence?

Y
Calculate equations
of HTC-I

Calculate
equations of
HPSSH-I
N

Convergence?

Calculate
equations of
ECO
N

Convergence?

Calculate equations of
HPSSH-II

Calculate equations of
mass, momentum, energy
and species

Convergence?

Convergence?

Calculate equations of
HTC-II

Y
N

Convergence?
Y

Convergence?

Output heat flux of tube skin

Calculate equations
of HTC-III

Convergence?

N
N

Convergence?

Output new temperature

of tube skin

End
Fig. 4. Coupled calculation procedure of tubes and furnace with dual stage steam feed mixing.

Table 3
Comparison of simulation results and industry data.
Items

Industry data

Simulation results

Exhaust gas temperature (K)

Outlet temperature of FPH (K)
Outlet temperature of ECO (K)
Outlet temperature of HTC-I (K)
Outlet temperature of HTC-II (K)
Outlet temperature of HPSSH-I (K)
Outlet temperature of HPSSH-II (K)
Outlet temperature of DSSH (K)
Outlet temperature of HTC-III (K)

385.15
430.15
523.15
609.15
677.15
695.15
793.15
844.15
884.15

388.65
437.41
527.78
615.75
681.61
698.66
792.44
843.18
887.38

consistent along z axial direction. At the corner both near and away
from the entrance to the convection chamber, flue gas velocity is
less than that in the other places. Fig. 6 shows flue gas velocity vector field at cross section z = 5 m. At the corner both near and away
from the entrance to the convection chamber, there are two recirculation zones obviously. The recirculation zone will make residence time of high temperature flue gas lengthen in the
convection chamber. At the same time, it will cause local overheating in furnace wall of convection chamber, thus being not conducive to the stable operation of the cracking furnace. When
high temperature flue gas entering the convection chamber, its
velocity is larger and will scour the furnace wall of opposite side.
In the convection chamber bottom, flue gas velocity distribution
along the x axial direction is not uniform. This is because the structure of the convection chamber is asymmetrical. With the increase
of height the flue gas passes through the staggered tube bundles

and its velocity field along the x axial direction is gradually uniform. While arriving at the outlet of convection chamber, the flue
gas velocity is basically consistent along x axial direction.
Fig. 7 shows flue gas temperature contours at section z = 0.5, 5
and 10 m of convection chamber respectively for dual stage steam
feed mixing. One can see that flue gas temperature distribution
along the x axial direction is not uniform in the bottom of convection chamber. With the increase of height, the flue gas passes
through the staggered tube bundles and its temperature field along
the x axial direction is gradually uniform. Fig. 8 shows flue gas temperature profiles along x and z axial direction. In Fig. 8a, when the
height of 3.10 m, flue gas temperature at x = 0.1 m is smaller than
that of x = 1.0 m. This is because the flue gas velocity of x = 0.1 m
which is near the entrance to convection chamber is smaller than
that of x = 1.0 m (this is can be seen in Fig. 5), leading to a smaller
heat transfer coefficient, and thus a lower flue gas temperature.
The flue gas of upward flow carries out heat exchange with the
tubes, and the flue gas temperature field also gradually becomes
uniform. From Fig. 8a, the temperature basically arrives at
consistence at y = 8.88 m, showing that the flue gas temperature
field gradually becomes uniform through the heat exchange with
staggered tubes. While arriving at the outlet of convection
chamber, the flue gas temperature is basically consistent along x
axial direction, showing staggered tube facilitates the exchange
of heat. Both figures show with the increase of height, flue gas
temperature gradually decreases. As can be seen from Fig. 8b, the
flue gas temperature field is basically consistent along z axial
direction, which is caused by the uniform flow field along z axial
direction in Fig. 5.

442

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Fig. 7. Flue gas temperature contours with dual stage steam feed mixing. (a)
z = 0.5 m; (b) z = 5 m; (c) z = 10 m.
Fig. 5. Flue gas velocity contours of furnace with dual stage steam feed mixing. (a)
z = 0.5 m; (b) z = 5 m; (c) z = 10 m.

Flue gas temperature (K)

1400
1200
1000
800
600
y=3.10 m
y=5.76 m
y=7.31 m
y=10.34 m

400
200

(a)

y=4.84 m
y=6.53 m
y=8.88 m

0
0

0.2

0.4

0.6

0.8

1.2

1.4

Width (m)

Flue gas temperature (K)

1400
1200
1000
800
600

200

(b)

0
0
Fig. 6. Flue gas velocity vector filed at cross section z = 5 m. (a) Velocity vectors
enlarged diagram at corner away from the entrance to the convection chamber; (b)
velocity vectors enlarged diagram at corner near the entrance to the convection
chamber.

y=3.10 m
y=5.76 m
y=7.31 m
y=10.34 m

400

y=4.84 m
y=6.53 m
y=8.88 m

10

Length (m)
Fig. 8. (a) Flue gas temperature distribution along x axial direction in convection
chamber; (b) flue gas temperature distribution along z axial direction in convection
chamber.

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Fig. 9. External tube heat flux profiles of different passes in HTC-I.

Fig. 10. Process gas temperature contours of different cross sections along z axial
direction in HTC-I. (a) 3 m; (b) 6 m; (c) 9 m.

4.2. The simulation results of tubes section

In view of the similarity of the simulation results of different
tube sections, this paper will take the simulation results of HTC-I
section as example to describe and analyze.
Fig. 9 shows external tube heat flux profiles of different passes
along tube axial position in HTC-I. The heat flux at the two ends of
tube is obviously lower than that of the other place. This is because

Fig. 11. Process gas temperature profiles of different passes in HTC-I.

443

Fig. 12. External tube skin temperature profiles of different passes in HTC-I.

the two ends of tube lie in near the furnace wall, where the flue gas
temperature is low, leading to a low difference between the flue
gas temperature and tube skin temperature, thus resulting in a
low heat flux. One can also see from this figure that in the Pass
1, Pass 3 and Pass 5, the heat flux increases along tube axial position, while in the Pass 2, Pass 4 and Pass 6, the heat flux decreases
along tube axial position.
Fig. 10 shows the process gas temperature contours of
different cross sections along z direction. Along the hydrocarbon
feedstock flow direction, the temperatures of the hydrocarbon
feedstock near the tube wall firstly rise, then the heat is gradually
transferred to the center of tube. It shows that temperatures near
the tube wall are higher than that of tube center. When reaching
the tube outlet, the temperature profile has been average in z axial
direction.
Fig. 11 shows the process gas temperature profiles of different
passes along tube axial position in HTC-I. The process gas temperature increases from Pass 1 to Pass 6. This is because the process
gas absorbs the heat transferred by flue gas, resulting in the
increase of temperature.
Fig. 12 shows the external tube skin temperature profiles of different passes along tube axial position in HTC-I. The external tube
skin temperature increases from Pass 1 to Pass 6 except near
the two ends of tubes, that is, in the crossover position of two tube
passes, the staggered external tube skin temperature exists. This is
because the two ends of tubes connects the bend tubes that is
outside of the convection section, which are not heated by flue
gas, leading to the lower tube skin temperature than inside the
convection section.

Fig. 13. Mass fraction profiles of liquid along FPH tube.

444

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

Fig. 14. Mass fraction profile of FPH straight tube: (a) bubble flow; (b) elongated
bubble flow; (c) stratified flow; (d) spray flow.

4.3. Dynamic simulation results

In the previous research of FPH [21], a dynamic model was used
to calculate. The advantage of the dynamic model is accurate calculation, however, the disadvantage of this model is long computational time. In the course of the coupling, the dynamic
calculation will greatly increase the time of the coupled calculation. In order to speed up the calculation efficiency, a piecewise linear interface calculation (PLIC) method [31] is used in the interface
interpolation of the coupled calculation in this paper. After the
coupled calculation, the tube skin heat flux profile is defined as
the boundary condition, and the PLIC method is used as the interface reconstruction one in the dynamic calculation. They can result
in a more accurate gasliquid two-phase change result.
Fig. 13 shows mass fraction profile along FPH tube. The feedstock vaporization begins in the fifteenth pass of tube. At the outlet
of tube, the liquid phase has been basically changed into the gas
phase.
The mass fraction profile of FPH straight tube is shown in
Fig. 14. Because of the long tube length, in order to clearly show
the calculation results, the x, y, z axes are adopted as the proportion
of 5:5:1. The flow pattern change of hydrocarbon feedstock has a
close relationship with the tube wall temperature. In the tube,
the liquid near tube wall first reaches the saturation temperature,
and forming the bubble that distributes in a continuous liquid
phase, which is called the bubble flow, as shown in Fig. 14a. With
the influence of gravity, the bubble will move to the top of tube,
and thus the liquid is gradually divided by long and thin bubble.
At this time the fluid will become chaotic, forming the elongated

Fig. 16. Velocity distribution in horizontal bend tube of FPH.

bubble flow, as shown in Fig. 14b. When a lot of bubbles gather

in the top of tube and form a continuous gas phase, leading gas
liquid two-phase to form a significant interface. Then, the liquid
phase lies below the gas phase, forming the stratified flow, as
shown in Fig. 14c. When most of liquid phase finish vaporization,
a small amount of liquid phase disperses in the continuous gas
phase, forming the spray flow, as shown in Fig. 14d. In the vaporized part of the feedstock, the tube wall temperature will sharply
increase, and thus tube wall temperature in this place will be
higher than that near the tube wall.
Fig. 15 shows temperature profile of vaporliquid mixture along
FPH tube. From the entrance into the tube, the hydrocarbon feedstock gradually absorbs the heat of the flue gas, and the temperature gradually increases. The outlet temperature of tube is in
agreement with the value of the industrial plant.
Fig. 16 shows velocity distribution in horizontal bend tube of
FPH. In Fig. 16, x, z coordinate directions are the horizontal ones,
y coordinate direction is the vertical one. 100 points are chosen
at the equal intervals in the horizontal bend tube of FPH, and the
horizontal and vertical velocities of these points are calculated.
The horizontal velocity is much higher than that of the vertical
velocity. Fig. 17 shows flow pattern in horizontal bend tube of
FPH. From the figure, we can see gas phase velocity is different
from that of liquid phase, causing liquid phase the waves in the
flow direction. The mass fraction of the liquid phase at the position
A is much higher than that of the vicinity. Combined with the analysis of Fig. 16, it shows that after entering bend tube, the liquid
phase is influenced by the action of the centrifugal force, and leading to a large horizontal velocity, thus resulting in the erosion of
the tube wall.
5. Conclusions
A complete coupled convection chamber/tube simulation of an
industrial cracking furnace with dual stage steam feed mixing is
carried out for the first time. The detailed simulation results
including the flue gas velocity and temperature fields in the convection chamber and the process gas temperature, tube skin temperature and heat flux profiles in the tubes are obtained. The
simulation results are in good agreement with the industrial data,
which explains the accuracy of model and could be used for the
further research:

Fig. 15. Temperature profile of vaporliquid mixture in FPH tube.

(1) There is recirculation zone at the corner both near and away
from the entrance to the convection chamber. The recirculation zone will make residence time of high temperature flue

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

445

Fig. 17. Mass fraction of liquid in horizontal bend tube of FPH.

gas lengthen in the convection chamber. It will cause local

overheating in furnace wall of convection chamber, thus
being not conducive to the stable operation of the cracking
furnace.
(2) There is a high degree of coupling between the velocity field
and temperature field of the convection chamber. The higher
the flue gas velocity, the higher the heat transfer coefficient,
the higher flue gas temperature. With the increase of height,
the flue gas passes through the staggered tube bundles and
its temperature field along the x axial direction is gradually
uniform. It shows the staggered tube facilitates the exchange
of heat.
(3) In the HTC-I, the heat flux at the two ends of tube is obviously lower than that of the other place. The heat flux
increases along tube axial position in the Pass 1, Pass 3
and Pass 5, whereas the heat flux decreases along tube axial
position in the Pass 2, Pass 4 and Pass 6. The process gas
temperature has different profiles along radial direction,
that is, at the beginning of several passes the temperature
near the tube wall is higher than that of the center of tube,
at the outlet of tube the temperature becomes average.
The process gas temperature increases along axial direction
from Pass 1 to Pass 6. The external tube skin temperature
increases from Pass 1 to Pass 6 except near the two ends
of tubes.
(4) A dynamic model is established for the FPH with dual stage
steam feed mixing on the basis of the coupled simulation.
The PLIC method is applied to the interface reconstruction
algorithms. The simulation results show the changes of flow
pattern in the FPH. Meanwhile, the influences of gravity on
two-phase flow are studied. The influence of centrifugal
force on flow pattern changes in bend tube section is also
analyzed.

Acknowledgments
This work is supported by Major State Basic Research Development Program of China (2012CB720500) and National Natural
Science Foundation of China China (Key Program: U1162202).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.cej.2015.10.093.
References
[1] J.L. Qian, Tubular Heating Furnace, China Petrochemical Press, Beijing, 2003.
[2] G.J. Heynderickx, A.J.M. Oprins, G.B. Marin, E. Dick, Three-dimensional flow
patterns in cracking furnaces with long-flame burners, AIChE J. 47 (2001) 388
400.

[3] A.J.M. Oprins, G.J. Heynderickx, G.B. Marin, Three-dimensional asymmetric

flow and temperature fields in cracking furnaces, Ind. Eng. Chem. Res. 40
(2001) 50875094.
[4] A.J.M. Oprins, G.J. Heynderickx, Calculation of three-dimensional flow
and pressure fields in cracking furnaces, Chem. Eng. Sci. 58 (2003) 4883
4893.
[5] G.D. Stefanidis, G.J. Heynderickx, G.B. Marin, Development of reduced
combustion mechanisms for premixed flame modeling in steam
cracking furnaces with emphasis on NO emission, Energy Fuels 20 (2006)
103113.
[6] G.D. Stefanidis, B. Merci, G.J. Heynderickx, G.B. Marin, CFD simulations of
steam cracking furnaces using detailed combustion mechanisms, Comput.
Chem. Eng. 30 (2006) 635649.
[7] P.J. Coelho, Numerical simulation of radiative heat transfer from non-gray
gases in three-dimensional enclosures, J. Quant. Spectrosc. Radiat. Transfer 74
(2002) 307328.
[8] A. Habibi, B. Merci, G.J. Heynderickx, Impact of radiation models in CFD
simulations of steam cracking furnaces, Comput. Chem. Eng. 31 (2007) 1389
1406.
[9] X.Y. Lan, J.S. Gao, C.M. Xu, H.M. Zhang, Numerical simulation of transfer and
reaction processes in ethylene furnaces, Trans IChemE A Chem. Eng. Res. Des.
85 (2007) 15651579.
[10] Y.L. Han, R. Xiao, M.Y. Zhang, Combustion and pyrolysis reactions in a naphtha
cracking furnace, Chem. Eng. Technol. 29 (2006) 112120.
[11] S.T. Liu, H.G. Wang, F. Qian, G.H. Hu, Coupled simulation of combustion with
heat transfer and cracking reaction in SL-II industrial ethylene pyrolyzer, CIESC
J. 62 (2011) 13081318.
[12] G.H. Hu, H.G. Wang, F. Qian, K.M. Van Geem, C.M. Schietekat, G.B. Marin,
Coupled simulation of an industrial naphtha cracking furnace equipped
with long-flame and radiation burners, Comput. Chem. Eng. 38 (2012) 24
34.
[13] G.H. Hu, C.M. Schietekat, Y. Zhang, F. Qian, G. Heynderickx, K.M. Van Geem, G.
B. Marin, Impact of radiation models in coupled simulations of steam cracking
furnaces and reactors, Ind. Eng. Chem. Res. 54 (2015) 24532465.
[14] X.O. He, B.W. Fang, C. Shao, The development and application of the simulation
software of the convection section in cracking furnace, China Petrochem. Des.
18 (2011) 16.
[15] J.J. Liu, Y. Guo, L.J. Zhang, Process simulation of the convection section of
cracking furnace based on Asepen Plus user model, Tianjin Chem. Ind. 23
(2009) 2529.
[16] Y. Zhou, D.Z. Yang, Simulation and optimum design for convection section of
ethylene cracking furnace, Chem. Eng. (China) 38 (2010) 99102.
[17] Aspen Plus, Ten Canal Park, Cambridge (MA) 02141, Aspen Technology Inc.,
2009.
[18] S.C.K. De Schepper, G.J. Heynderickx, G.B. Marin, Modeling the evaporation of a
hydrocarbon feedstock in the convection section of a steam cracker, Comput.
Chem. Eng. 33 (2009) 122132.
[19] A.V. Mahulkar, G.J. Heynderickx, G.B. Marin, Simulation of coking in
convection section of steam cracker, Chem. Eng. Trans. 29 (2012) 1375
1380.
[20] A.V. Mahulkar, G.J. Heynderickx, G.B. Marin, Simulation of the coking
phenomenon in the superheater of a steam cracker, Chem. Eng. Sci. 110
(2014) 3143.
[21] S.C.K. De Schepper, G.J. Heynderickx, G.B. Marin, Modeling the coke formation
in the convection section tubes of a steam cracker, Ind. Eng. Chem. Res. 49
(2010) 57525764.
[22] V. Mertinger, M. Benke, S. Szab, O. Bnhidi, B. Bollo, . Kovcs, Examination of
a failure detected in the convection zone of a cracking furnace, Eng. Fail. Anal.
18 (2011) 16751682.
[23] S.C.K. De Schepper, G.J. Heynderickx, G.B. Marin, Coupled simulation of the flue
gas and process gas side of a steam cracker convection section, AIChE J. 55
(2009) 27732787.
[24] C.M. Schietekat, D.J. Van Cauwenberge, K.M. Van Geem, G.B. Marin,
Computational fluid dynamics-based design of finned steam cracking
reactors, AIChE J. 60 (2014) 794808.
[25] W.H. Lee, A pressure iteration scheme for two-phase flow modeling, in: T.N.
Veziroglu (Ed.), Multiphase Transport Fundamentals, Reactor Safety,
Applications, Hemisphere Publishing, Washington, DC, 1980.

446

G. Hu et al. / Chemical Engineering Journal 286 (2016) 436446

[26] Z. Yang, X.F. Peng, P. Ye, Numerical and experimental investigation of two
phase flow during boiling in a coiled tube, Int. J. Heat Mass Transf. 51 (2008)
10031016.
[27] A. Habibi, B. Merci, G.J. Heynderickx, Multiscale modeling of turbulent
combustion and NOx emission in steam crackers, AIChE J. 53 (2007) 2384
2398.
[28] K.M. Van Geem, R. Zajdlk, M.F. Reyniers, G.B. Marin, Dimensional analysis for
scaling up and down steam cracking coils, Chem. Eng. J. 134 (2007) 310.

[29] R.R. Akberov, Calculating the vaporliquid phase equilibrium for

multicomponent systems using the SoaveRedlichKwong equation, Theor.
Found. Chem. Eng. 45 (2011) 312318.
[30] ANSYS FLUENT Users Guide, Release 14.0, ANSYS Inc., 2011.
[31] S.C.K. De Schepper, G.J. Heynderickx, G.B. Marin, CFD modeling of all gas
liquid and vaporliquid flow regimes predicted by the Baker chart, Chem. Eng.
J. 138 (2008) 349357.