You are on page 1of 11

Computers and Chemical Engineering 38 (2012) 24–34

Contents lists available at SciVerse ScienceDirect

Computers and Chemical Engineering


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

Coupled simulation of an industrial naphtha cracking furnace equipped with


long-flame and radiation burners
Guihua Hu a , Honggang Wang a , Feng Qian a,∗ , Kevin M. Van Geem b , Carl M. Schietekat b , Guy B. Marin b
a
Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University of Science and Technology,130 Meilong Road, Shanghai
200237, China
b
Laboratory for Chemical Technology, Ghent University, Krijgslaan 281 (S5), 9000 Gent, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: A coupled reactor/furnace simulation has been conducted for a 100 kt/a SL-II naphtha cracking furnace
Received 24 January 2011 containing both long-flame and radiation burners. The computational fluid dynamics approach was used
Received in revised form 28 October 2011 to simulate the flow, combustion and radiative heat transfer in the furnace. The software packages COIL-
Accepted 4 November 2011
SIM1D and SimCO were used to account for the cracking process in the reactor coils. The simulation
Available online 15 November 2011
provides for the first time detailed information about concentration, velocity, and temperature fields for
these types of furnaces. Comparison of the calculated product yields against measured industrial data
Keywords:
validates the simulation and shows that the difference with using a predefined normalized heat flux
Reactor/furnace simulation
Cracking furnace
profile is limited. The results show that the design of radiation section outlet leads to an asymmetric
Computational fluid dynamics flue gas-temperature, concentration and velocity profile. Large recirculation zones exist near the reactor
Product yields tubes, making the temperature in the middle of furnace more uniform.
Recirculation © 2011 Elsevier Ltd. All rights reserved.

1. Introduction 2009; Bockelie, Adams, Cremer, 1998) more recently. Due to the
continuous improvement of CFD software and the continuously
In a naphtha cracking furnace many complicated physical increasing calculation power CFD has steadily grown to become the
and chemical processes occur simultaneously such as momentum favorite and most accurate methodology for simulating the crack-
transfer, heat transfer, mass transfer and combustion. All these pro- ing furnaces. Detemmerman and Froment (1998) and Heynderickx,
cesses are directly influenced by the endothermic reactions taking Oprins, Marin, and Dick (2001) were the first to combine the zone
place in the reactor tubes, i.e. the steam cracking coils. Combustion method of Hottel (Hottel & Sarofim, 1967) and CFD calculations to
of the fuel gas generates a large quantity of flue gas that flows in the assess the coupled simulation of the radiative heat transfer in the
furnace and transfers heat to the process gas inside the reactor coil, furnace and the cracking reactions occurring in the reactor coils
giving rise to the complicated cracking reactions of the hydrocar- of the furnace. Oprins, Heynderickx, and Marin (2001) and Oprins
bon/steam mixture. Obviously, what is happening inside the coil is and Heynderickx (2003) simulated the non-symmetric flow in a
strongly affected by how the heat is transferred from the furnace steam cracking furnace. More recently Stefanidis, Heynderickx, and
to the coil. Because all the processes in the furnace and coil are Marin (2006) and Stefanidis, Merci, Heynderickx, and Marin (2006)
strongly coupled an accurate description of the heat transfer pro- studied detailed and reduced combustion mechanisms for com-
cesses occurring in the furnace radiation section and reasonable putational fluid dynamic simulation of steam cracking furnaces.
allocation of the energy within the furnace are crucial for design Stefanidis, Merci, Heynderickx, and Marin (2007) and Stefanidis,
and optimization of cracking furnaces. Van Geem, Heynderickx, and Marin (2008) also looked at the impor-
Over the past decades significant progress has been made in tance of gray/non-gray gas modeling for steam cracking furnaces.
modeling the fire gas side and convection side of steam cracking Habibi, Merci, and Heynderickx (2007) studied the choice of the
furnaces, from zero-dimensional models (Lobo & Evans, 1939), over radiation model and concluded that the discrete ordinate model
multi-dimensional models (Hottel & Sarofim, 1967; Howell, 1968; (DOM) is the preferred radiation model for modeling radiative heat
Roesler, 1967) to a computational fluid dynamics (CFD) descrip- transfer in steam cracking furnaces. Han, Xiao, and Zhang (2006),
tion (De Schepper Sandra, Heynderickx Geraldine, & Marin Guy, Lan, Gao, Xu, and Zhang (2007) and Zhang, Huang, and Wu (2008)
used Fluent for simulating the fire-side of a naphtha cracking fur-
nace. In all these studies a non-premixed PDF model was adopted
∗ Corresponding author at: No. 130 of Meilong Road, Shanghai 200237, China. to simulate the combustion of the fuel gas that may lead to unex-
E-mail address: fqian@ecust.edu.cn (F. Qian). pected results for the premixed combustion process when radiation

0098-1354/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2011.11.001
G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34 25

burners are used. Wang and Zhang (2005) used the simple P1 radia- the furnace are simulated with the commercial CFD package Fluent
tion model to calculate radiation heat transfer in the furnace. Wang (Fluent, 2006), while COILSIM1D (Van Geem, Marin, Hedebouin, &
and Zhang (2005) showed that the calculated precision of the model Grootjans, 2008; Van Geem, Reyniers, & Marin, 2008) is used for
was poor for the large-scale industrial furnace with complex geom- the reactor side. The necessity of using CFD for the furnace side has
etry boundary. Zhou and Jia (2007) used empirical formula in the been addressed by comparing our CFD results with those obtained
calculation of medium radiation characteristic. This method may using a simpler and faster approach based on a simple mathe-
result in large errors for the absorption coefficient of the flue gas matical model for predicting heat flux profiles in steam cracking
mixture due to the empirical formula for the radiation. It is also furnaces (Colannino, 2007). The coupled simulation of furnace and
surprising that in all of the currently published articles either fur- reactor allows to obtain on the one hand temperature, velocity and
naces solely fired with radiation burners or furnaces solely fired flue gas concentration fields in the furnace. On the other hand the
with long-flame burners in the floor have been studied. The com- calculated heat flux to the reactor allows to determine the product
bined firing with both long flame burners and radiation burners yields of over 150 molecules.
has remained unstudied although this is the furnace design that is
currently being installed in most of the modern olefin production 2. Furnace/reactor/feedstock calculation
plants. Moreover, in all of the cited articles relatively simple noz-
zles arrangements are used for the burners in the cracking furnace. 2.1. Furnace model
Even more surprisingly is that according to the authors’ knowl-
edge in none of the cited papers there is a direct comparison to the 2.1.1. Flow model
industrially measured product yields. The calculation of the three-dimensional statistically stationary
Crucial for obtaining accurate simulation results when crack- turbulent flow field in the furnace is based on the solution of the
ing a complex mixture such as naphtha is the level of detail Reynolds-averaged Navier–Stokes (RANS) equations in compress-
considered to describe the chemical reactions of the feed in the ible formulation, closed with the standard k–ε model. The transport
reactor. Although there is a general consensus about the free radi- equations for mass, momentum, energy, turbulent kinetic energy
cal mechanism (Rice, 1931, 1933; Rice & Herzefeld, 1934), several and dissipation rate of turbulent kinetic energy can be expressed
different types of kinetic models have been developed over the by the following general form (Tao, 2001):
years to simulate the steam cracking process. A distinction can be  
made between three different types of models: empirical, global, ∂() ∂(Uj ) ∂ ∂
+ =  + S (1)
and detailed kinetic models. The simplest models correlate the ∂t ∂xj ∂xj ∂xj
product yields with some parameter such as the cracking sever-
ity index (Shu & Ross, 1982), or fitted empirical correlations for 2.1.2. Combustion model
reaction rate coefficients (Davis & Farrell, 1973). Global kinetic The combustion of the long-flame burners is non-premixed,
models postulate a series of molecular reactions based on the whereas the combustion of the radiation burners is premixed. In
free-radical mechanism, starting from an overall primary reac- this work the turbulence–chemistry interaction model based on
tion followed by a set of secondary reactions between the primary the work of Magnussen and Hjertager (1976), called the Finite
products (Belohlav, Zamostny, & Herink, 2003; Kumar & Kunzru, Rate/Eddy-Dissipation model, is used. The finite-rate model com-
1985; Pant & Kunzru, 1996). However, neither the empirical nor putes the chemical source terms using Arrhenius expressions, and
the global kinetic models consider sufficient detail to deal with ignores the effects of turbulent fluctuations. The net source of
the continuously varying naphtha composition in industrial steam chemical species i due to reaction is computed as the sum of the
crackers. That is why in industrial practice only the most advanced Arrhenius reaction sources over the NR reactions that the species
kinetic models such as SPYRO (Dente, Ranzi, & Goossens, 1979; participate in:
van Goethem, Kleinendorst, van Leeuwen, & van Velzen, 2001) and
CRACKSIM/COILSIM1D (Van Geem et al., 2007; Van Geem, Zajdlik, 
NR

Reyniers, & Marin, 2007) are used for simulation and optimiza- Ri = Mw,i R̂i,r (2)
tion of naphtha furnaces. Traditionally a 1-dimensional reactor r=1

model is employed, however, in some cases the implementation


where R̂i,r is the Arrhenius molar rate of creation/destruction of
of 2- (Van Geem, Heynderickx, & Marin, 2004) or 3-dimensional
species i in reaction r and is given by:
(DeSaegher, Detemmerman, & Froment, 1996) reactor models ⎛ ⎞
becomes inevitable, e.g. for cracking coils with internally finned 
N
R̂i,r =  ( i,r −  i,r ) ⎝kf,r ⎠
 
tubes. Because of the high Reynolds numbers (>105 ) in the tubular [Cj,r ]( j,r + j,r ) (3)
reactors used for naphtha steam cracking radial gradients can be j=1
neglected (Van Geem et al., 2007; Van Geem, Zajdlik, et al., 2007)
and therefore the assumption of plug flow can be safely accepted. The net rate of production of species i due to reaction k, Ri,k ,
In the present work, an industrial SL-II naphtha cracking fur- by the Eddy-Dissipation model is given by the smaller (i.e. limiting
nace with 100 kt/a capacity is simulated. This type of furnace is value) of the two expressions below:
not only equipped with long-flame burners located in the furnace  
floor, but also with radiation burners located in the side-walls of the  ε YR
Ri,k =  i,k Mw,i A min (4)
furnace. The feedstock composition is calculated from the on-line k  R,k Mw,R
measured global characteristics of the cracked naphtha. The fur-

ε YP
nace calculations are based on the computational fluid dynamics Ri,k =  i,k Mw,i AB
N (5)
(CFD) approach. To account for the bottom firing a non-premixed k  j,k Mw,j
j
combustion mechanism is adopted, while a premixed combus-
tion mechanism for the radiation burners in the front and the The following reaction scheme is considered for the combustion
rear wall. The discrete ordinate model is used to calculate radia- of CH4 /H2 /air mixture:
tive heat transfer in the furnace. A weighted-sum-of-gray-gases CH4 + 1.5O2 → CO + 2H2 O (6)
model (WSGGM) is used to calculate the absorption coefficient of
the flue gas mixture. The flow, combustion and heat transfer in CO + 0.5O2 → CO2 (7)
26 G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34

H2 + 0.5O2 → H2 O (8) The energy equation is given by

In this model C2 H4 is not explicitly accounted for. In our sim- dT


dt
dt2 
=
q(z) +
(− Hfj0 )Rj (12)
ulations we have assumed that all hydrocarbons in the fuel gas is dz Fc 4 Fc
j j pj j j pj j
methane because the C2 H4 content in the fuel gas is small and using
C2 H4 instead of methane has a small effect on fuel gas combustion.
The mechanical energy equation, accounting for pressure changes
The intrinsic reaction rate of the first reaction comes from
due to friction and changes in mechanical energy, is given by
the two-step mechanism describing methane combustion derived
by Westbrook and Dryer (1981). The set of kinetic parameters dpt
2 f
du
= + g u2 + g u (13)
of H2 combustion is found to give reaction rates close to those dz dt
rb dz
determined based on the parameters in the ASPEN, KDB and
NIST databases (Aspen, 2009; Kang et al., 2003; Mallard, Westley, The process gas temperature profile, conversion, and concentra-
Herron, & Hampson, 2005). tion profiles can be calculated based on an imposed heat flux profile
or an imposed external tube skin-temperature profile. In this work,
calculations are performed using a heat flux profile obtained from
2.1.3. Radiation model the CFD furnace calculation or using a predefined normalized inci-
Due to the high temperature of the combustion gases and the dent heat flux profile as explained by Colannino (2007). Colannino
furnace walls, radiation is the predominant mode of heat transfer (2007) has shown that it is possible to characterize normalized heat
in the furnace. Since radiation dominates the heat flux towards the flux profiles in steam cracking furnaces as a function of two param-
reactor tubes there is a strong need for an appropriate radiation eters: the elevation at which the maximum heat flux occurs, and
model. In this study, the discrete ordinate method (DOM) is used heat flux at the floor. In turn, these two parameters are determined
to solve the radiative transfer equation (RTE) (Modest, 2003). Com- by operating conditions and furnace and burner design.
pared with other radiative heat transfer calculation methods, DOM The net rate of production Rj in the reactor model equations
is simple, reliable and especially suitable for coupled calculation of is calculated from the single event microkinetic model for steam
flow, combustion and heat transfer as shown by Habibi et al. (2007). cracking of hydrocarbons, containing thousands of elementary
The RTE of the DOM is given by: reactions of over 500 species. A detailed description of the reactor
model and the reaction network for steam cracking of hydrocar-
∇ · (I(r, s)s) + (˛ + s )I(r, s)
bons can be found in Van Geem, Marin, et al. (2008) and Van Geem,
4

T 4 s Reyniers, et al. (2008).


= ˛n2 + I(r, s )˚(s · s ) d˝ (9)

4
0
2.3. Feedstock model
Determination of the absorption coefficient of the flue gas is
crucial for the calculation of radiative heat transfer in the furnace. For industrial furnaces only rarely the detailed composition of
Calculation methods of the absorption coefficient of flue gas include the feed entering the reactor is determined by analytical techniques
the chart method, algebraic models, the weighted-sum-of-gray- such as conventional gas chromatography (1D-GC) or on-line two-
gases model (WSGGM), the wide band model, the narrow band dimensional gas chromatography (GC × GC) (Van Geem et al.,
model, the line-by-line model, etc. The WSGGM is a reasonable 2010). Generally only a limited number of global characteristics
compromise between the oversimplified gray gas model and the are measured on-line, the so-called commercial indices. These
computationally too expensive complete model which takes into indices (e.g. the average molecular weight of the mixture, some
account the different gas absorption bands. Therefore in this work points of a boiling point distillation curve, the specific density, the
the WSGGM is used to account for the radiation characteristics global PIONA weight fractions, etc.) are determined by means of
of the flue gas. In the WSGGM, the emissivity of the real gas is relatively simple and standardized analytical procedures (ASTM
expressed as the weighted sum of the emissivities of a number of procedures) (Riazi, 2005). Although they are not representative
gray gases and a transparent gas (Huang, Yang, Yu, & Zhang, 1993): of all the chemical and structural variety that such mixtures can
contain, the feedstock is generally identified and distinguished by

I these commercial indices rather than by their molecular composi-
ε= ˛ε,i (T )(1 − e−ki ps ) (10) tion (Merdrignac & Espinat, 2007). That is why over the past decade
i=0 several methods have been developed that allow to obtain the
detailed composition from these global characteristics (Klein, Hou,
where the bracketed quantity is the ith fictitious gray gas emissiv- Bertolacini, Broadbelt, & Kumar, 2006; Van Geem et al., 2007; Van
ity. Geem, Zajdlik, et al., 2007; Verstraete, Revellin, Dulot, & Hudebine,
2004). In the present work the simulation package SimCo is used for
2.2. Reactor model determining the detailed composition. For naphtha fractions either
the Shannon Entropy method or an artificial neutral network (ANN)
Currently a full CFD simulation of the reactor side with can be chosen for reconstructing the composition (Pyl, Van Geem,
detailed kinetics is not yet possible (Manca, Buzzi-Ferraris, Cuoci, Reyniers, & Marin, 2010). The ANN has been trained using the com-
& Frassoldati, 2009). Therefore the plug flow model has been used positions of over 300 reference naphtha fractions (Pyl et al., 2010)
in this work. In the 1-dimensional plug flow reactor model, a set of for determining the detailed naphtha composition. However, the
continuity equations for the process gas species is solved simulta- ANN can only be used when the considered naphtha lies within the
neously with the energy and the mechanical energy equation. The method’s application range, i.e. the ellipsoid of naphthas for which
steady-state mass balance for a component j in the process gas mix- the Mahalanobis distance is equal or smaller than 2.5 (Pyl et al.,
ture over an infinitesimal volume element with length dz leads to 2010). The inputs of this neutral network are the initial, 50% and
the following continuity equation for component j: final boiling point of the ASTM D86 boiling point curve, the global
PIONA weight fractions and the specific density. The output are the
dFj
dt2 detailed PIONA weight fractions of 28 lumped components, which
= R (11)
dz 4 j are delumped to a molecular composition of 118 naphtha feedstock
G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34 27

components based on the distributions proposed by Dente, Ranzi,


Bozzano, Faravelli, and Valkenburg (2001).

3. Simulation set-up

3.1. Furnace geometry and operating conditions

A schematic representation of the simulated industrial naphtha


cracking furnace and a top view of the simulated naphtha cracking
furnace are given in respectively Figs. 1 and 2. Because of structural
similarity, only a representative segment of one sixth of the indus-
trial cracking furnace is simulated. In the full furnace there are 24
split coils of the “U-type” vertically suspended in the center of the
furnace in a staggered arrangement in two rows as shown in Fig. 3.
The reactor is of the “4-1” type, i.e. four inlet tubes (first pass)
link with a single outlet tube (last pass). In the furnace floor, near
the walls of both sides (i.e. wall A and C in Fig. 2), 36 long-flame

Fig. 3. SL-II coil configuration.

burners are present in two rows. Each burner includes five noz-
zles, where the primary nozzles are arranged in the air inlets, and
other auxiliary nozzles are arranged around the air inlets. The fur-
nace segment is heated with 48 additional radiation burners. These
burners are positioned in one row in the front wall (wall A in Fig. 2)
and another row of burners in the rear wall (wall C in Fig. 2) of
the furnace. The furnace outlet is located at right side of the row of
reactor tubes.
Tetrahedral cells are used to discretize the physical domain
including the burner zones and the tube skin zones. The hexahe-
dral cells are used to discretize the furnace zone. The number of
grid cells and grid nodes is 338,860 and 236,667 respectively. The
furnace and coil geometry and the operating conditions are given
in Table 1.
The commercial indices of the used naphtha are given in Table 2.
First it has been verified if the considered naphtha lies within
the application range of the ANN developed by Pyl et al. (2010).
The calculated Mahalanobis distance for this naphtha is 0.42 using
the commercial indices from Table 2, and hence, the ANN is the
preferred choice for reconstructing the detailed composition of
the naphtha. A good agreement between the calculated and the
Fig. 1. Front view of the considered segment of the industrial naphtha cracking industrially used feedstock can be expected. The reconstructed
furnace. composition can be found in the supporting information.

3.2. Boundary conditions

The operating conditions mentioned in the previous paragraph


determine the mass flow and temperature of the fuel gas and air to
the long-flame and radiation burners and the mass flow and tem-
perature of the hydrocarbon/steam mixture entering the coil. The
outlet boundary condition of the furnace is determined by the pres-
sure outlet boundary, which is −50 Pa (gage pressure). The standard
wall functions are used as turbulent wall boundary conditions.
The no-slip boundary condition is imposed at the reactor and fur-
Fig. 2. Top view of the simulated furnace segment. nace walls. The furnace wall consists of refractory bricks, insulation
28 G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34

Table 1
Furnace dimension and operating conditions.

Furnace segment
Length (m) (x-direction) 18.94
Width (m) (y-direction) 3.56
Height (m) (z-direction) 13.70
Number of long-flame burners 36
Number of radiation burners 48
Firing condition
Fuel gas flow rate in bottom (kg/s) 1.07935
Fuel gas flow rate in side (kg/s) 0.1757
Air and fuel gas inlet temperature (K) 300
Furnace outlet gage pressure (Pa) −50
Oxygen excess (vol%) 2
Fuel composition (wt%)
CH4 97.68
H2 0.51
CO 0.89
C2 H4 0.89
Material properties
Emissivity of the furnace wall 0.75
Emissivity of tube skin 0.9
Reactor coils
Number of reactor tubes 24
Number of passes 2
Inlet tube diameter (outside) (10−3 m) 64
Outlet tube diameter (outside) (10−3 m) 121
Thickness of tube (10−3 m) 6.5 Fig. 4. Flow chart of coupled solution algorithm for mathematical models in both
Hydrocarbon flow rate (kg/s) 11.82 furnace and reactor.
Steam dilution (kg/kg) 0.58
Coil inlet temperature (K) 891.6
Coil outlet pressure (kPa) 95

3.3. Coupled furnace/reactor simulation


bricks, insulation cotton and a metal wall. The heat boundary on the
furnace walls is calculated based on the industrial heat loss value. In order to obtain quantitative yield data a coupled furnace (fire
The possibility of the reverse flow at the outlet boundary, during side)/reactor (process gas side) simulation is performed. The flow
the solution process is taken into account. A zero flux of all quan- chart of the complete simulation procedure is shown in Fig. 4. In
tities is assumed across the symmetry boundaries. Since the shear the simultaneous numerical solution of furnace and reactor, several
stress is zero at these boundaries, it can also be interpreted as a iterations between the two models are needed to obtain conver-
“slip” wall in viscous flow calculations. gence. The CFD furnace calculations allow to determine the flow,
For the DOM radiation model, the wall surface temperature combustion and radiative heat transfer on the furnace side. As men-
is calculated from the following equation (Modest, 2003), while tioned previously COILSIM1D is used to account for the radical
imposing qout : reactions occurring on the process gas side.
qout = (1 − εw )qin + n2 εw Tw
4
(14) Initially an external tube skin-temperature profile from indus-
trial measurement is assigned as the tube skin boundary of the
where furnace model by a user-defined function (UDF) in Fluent. Then
4
the calculation of combustion and heat transfer inside the furnace
qin = Iin s · n
 d˝ (15) begins and the first heat flux profile on the reactor tubes is calcu-
s·n
 >0 lated. This heat flux is converted to the internal heat flux according
where Iin is the intensity of the incoming ray, ˝ is the hemispherical to the difference in internal and external tube diameter. Next, this
solid angle and n  is the normal pointing out of the domain. internal heat flux is used as boundary condition for the reactor tube
model simulation. Based on the reactor model simulation a new
reactor tube skin-temperature profile can be determined, which
Table 2
is again used in the Fluent calculation. This iterative loop is con-
Commercial indices of the naphtha feedstock.
tinued until the maximal difference in tube skin-temperature of
Specific density 0.7176 two successive iterations reaches a predefined threshold value,
ASTM D86 boiling points (K)
e.g. 1 K.
IBP 318
10% 339
The governing equations for the conservation of mass, momen-
30% 361 tum, energy, turbulence, chemical species and radiation are solved
50% 387 sequentially (i.e. segregated from one another) by a control-
70% 419 volume-based method. The non-linear governing equations are
90% 459
discretisized implicitly by a second-order upwind scheme and
FBP 481
PIONA (wt%) linearized to produce a system of equations for the dependent
P 34.40 variables in every computational cell. The resulting linear system
I 35.47 with a sparse coefficient matrix is then solved to yield an updated
O 0.18
flow-field solution. The appropriate under-relaxation factors are
N 21.11
A 8.85
imposed to avoid instability in the solution. Since high values for
S (mg/kg) 236.00 pressure-correction under-relaxation lead to instability problems,
H/C (mol/mol) 2.14 the SIMPLE algorithm with a slightly more conservative under-
Average molecular weight (amu) 107.3 relaxation value (up to 0.7) is used.
G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34 29

4. Results and discussion

4.1. Furnace simulation

Fig. 5 shows the velocity vectors of the flue gas flow in the fur-
nace. Fig. 5(a) illustrates the flue gas streamlines in a section of the
furnace above a pair of long-flame burners, showing that large recir-
culation zones exist near the reactor tubes due to the entrainment
effect of the high-velocity jet of the long-flame burners. Two jets are
observed above the burners, with a strong increase in velocity in the
flame region due to the temperature rise (expansion effect). These
recirculation zones start from the furnace floor towards the reactor
tubes, and expand to the middle of the furnace, circulating the high
temperature flue-gas of the flame into the low temperature central
bottom part of the furnace, thereby increasing the residence time
of the high-temperature flue gas in the furnace. Hence, this recircu-
lation zone enhances convective heat transfer between the bottom
flue gas and the reactor tubes.
Fig. 5(a) also shows the existence of another recirculation zone
in the beginning of the convection section. When the flue gas flow
enters the convection section, it is forced to make a sharp turn, cre-
ating a third recirculation zone [zone A in Fig. 5(a)] in the convection
section inlet.
Fig. 5(b) shows the 3D velocity vector fields at three different
heights in the furnace. The lowest level is a cross section at a height
of 4.5 m in the furnace, crossing the flame of the long-flame burners.
The middle level is a cross section at a height of 9 m in the furnace
located in between two rows of the radiation burners. The upper
level is a cross section at a height of 12 m in the furnace, near the
outlet of the radiation section. From the velocity vector fields it is
clear that the velocity decreases with increasing height, and the
velocity on the right side near the rear wall is higher than that on
the left side near the front wall. It can also be seen that the vertical Fig. 6. Flue gas velocity profiles at the different heights of furnace. (a) Z velocity dis-
velocity component in between the middle reactor tubes is smaller tribution at plane x = 1.565 m along the width of furnace; (b) Z velocity distribution
at plane y = 3.445 m along the length of furnace.

than the one in between the outer reactor tubes, and is close to
zero.
Fig. 6(a) shows the flue gas velocity distribution at the differ-
ent heights in plane x = 1.565 m along the width of furnace. It can
be seen from Fig. 6(a) that the velocity distribution of the flue
gas is not completely symmetrical just above both rows of long-
flame burners. The flue gas velocity above the burners closer to the
furnace outlet is slightly higher than on the opposite side. At a dis-
tance of 0.25 m and 5 m above the burner near the front wall, the
flue gas vertical velocity is 11.32 m/s and 6.4 m/s, respectively. At
the corresponding heights, the vertical velocity above the oppo-
site long-flame burner is 10.18 m/s and 6.3 m/s, respectively. There
is a large flue gas recirculation zone in the furnace bottom near
the reactor tubes, as was already shown in Fig. 5(a). At a height
of 3 m, the recirculation velocity reaches its maximum value, i.e.
−2 m/s. The height of this recirculation zone in the middle along
the y direction is approximately 9 m. Between two rows of long-
flame burners, the flue gas velocity decreases gradually and the
velocity gradient is small due to diffusion and recirculation of flue
gas. Fig. 6(b) shows the flue gas velocity distribution at different
heights in plane y = 3.445 m along the length of furnace. As can be
seen from this figure the 2 fuel gas jets of the long-flame burners
do not affect each other when the furnace height is lower than 3 m.
However, starting from a height of 3 m, the flue gas velocity profile
flattens out due to strong turbulances and mixing caused by the 2
high-velocity jets. At a height of 7 m, the average flue gas velocity
is 4.5 m/s. Along the x direction of the furnace a recirculation zone
Fig. 5. Flow field of the flue gas in furnace. (a) Streamlines; (b) 3D vector fields at exists with a diameter of 0.25 m in between two burners and it has
different heights. a velocity of approximately −0.5 m/s.
30 G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34

Fig. 9. Average temperature distribution of flue gas along the height of furnace.

For similar furnaces run lengths of typically 100 days have been
reported by BASF, which is one of the highest reported run lengths
Fig. 7. Temperature field in a vertical cross-section, parallel to the short wall of the
furnace (a) 0.525 m; (b) 1.565 m; (c) 2.605 m. in industry for naphtha furnaces (Robinson, 2009).
Fig. 8 shows the flue gas-temperature distribution at different
sections along the height of the furnace. It shows that the flue
Fig. 7 shows the flue gas-temperature field at different sections gas-temperature at the bottom of the furnace is low. The flue gas-
along the length of the furnace. The flue gas-temperature is high temperature gradually increases from bottom to top, and reaches
near the furnace wall, the middle of the furnace and near the zone its maximum between 4 and 6 m. Above this height a gradual but
where the radiation burners are located. On the other hand the small temperature drop is observed until the radiation burners are
temperature is relatively low on the furnace floor, the top of the reached. There the flue gas-temperature begins to rise again, and
furnace and in the neighborhood of the reactor tubes. In the furnace hence, the radiation burners enlarge the zone of high-temperature
floor, the fuel gas ejected from the burners has a very high velocity flue gas at the top of furnace. They effectively complement the lack
so that it is improperly mixed with ambient air, causing incomplete of heat at the top of furnace if only long-flame burners would be
combustion and not so much heat release, hence, the flue gas- used. The latter was observed in the work of Heynderickx et al.
temperature is low. The furnace outlet is located in the rear wall (2001) for their furnace containing only long-flame burners. Note
side, and the asymmetric flow pattern caused by its positioning also that the positioning of the long-flame burners near the fur-
extends the high-temperature plane on the front wall side to the nace wall is explicitly chosen to avoid that flames would directly
rear wall side. The asymmetric design of the convection section in impinge the reactor tubes because this would result in hot spots on
combination with the positioning of the burners and the entrance the furnace wall and very high coking rates at these positions.
of the combustion air has as other advantage that the temperature Fig. 9 shows the flue gas-temperature averaged over a hori-
in the middle of furnace becomes more uniform, resulting in a zontal plane along the furnace height. Fig. 9 shows that the flue
relatively low tube skin-temperature. This is beneficial for the run gas-temperature is low at the bottom and top of the furnace, and
length of the furnace because the run length for naphtha furnaces is high and uniform at the middle of the furnace. Initially the aver-
is strongly determined by the maximal tube skin-temperature. age flue gas-temperature rises from 1406 K to 1540 K and then falls

Fig. 8. Flue-gas temperature field in a horizontal furnace cross section (a) 1 m; (b) 3 m; (c) 5 m; (d) 7 m;(e) 9 m; (f) 11 m.
G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34 31

Fig. 10. Mass fraction distribution of O2 in a vertical cross section, parallel to the
short wall of the furnace (a) 0.525 m; (b) 1.565 m; (c) 2.605 m.

to 1326.6 K, reaching its maximum value at a height of approxi-


mately 5.5 m in the furnace. The flue gas-temperature at the height
of the radiation burners does increase a little, but its influence is
significantly less pronounced than that of the long-flame burners.
Fig. 10(a)–(c) shows the O2 mass fraction contours at different
sections along the length of the furnace. In the core of the jet there
is incomplete combustion because of the high velocity of fuel gas
and air and insufficient mixing. Hence, only a small amount of O2 is
consumed resulting in a high O2 mass fraction in this zone. Outside
the jet zone, the flue gas velocity decreases sharply, resulting in
complete combustion and hence no remaining O2 . Obviously along
the furnace height, the O2 mass fraction decreases due to the com- Fig. 12. Profile of the tube skin and process gas temperature and heat flux along the
bustion of the fuel gas resulting in the O2 mass fraction being nearly reactor length (a) CFD simulation (b) using method of Colannino.
equal to zero in most of the furnace except for the jet core zone.
Fig. 11 shows the CO mass fraction profile in the plane
in a flame height of 6 m which is in good agreement with the flame
x = 1.565 m above a long-flame burner. As can be seen from Fig. 10,
lengths reported for long-flame burners.
the CO mass fraction is relatively high at height of 0–2 m above
the burners. Above a height of 2 m the CO mass fraction gradually
decreases and becomes zero (0.05% in Fig. 11) at the height of 6 m. 4.2. Reactor simulation
The height of the flame can be approximated by the end of oxidation
reaction, hence, the flame height can in this case be estimated from Fig. 12(a) shows the internal heat flux profile and the tube
the CO content profile (Han et al., 2006), that is, the place where CO skin and process gas temperature profile along the reactor length.
is completely depleted is the highest point of the flame. This results As mentioned previously, the maximum flue gas-temperature is
located in a zone at 4–6 m from the furnace floor. Hence, the tube
skin-temperature and the heat flux reach their maximum in this
zone. It can be seen in Fig. 12(a) that the tube skin-temperature in
the first pass increases from 1004 K at the entrance to a first maxi-
mum of 1202 K at an axial tube length position of about 10 m, and
falls to 1183 K at the exit of the first pass. The tube skin-temperature
of the second pass increases from 1220 K at the entrance of second
pass to the second maximum, 1269 K at tube length of about 20 m,
and falls to 1171 K at the exit of second pass. As mentioned previ-
ously the maximum external wall temperature is relatively low for
this medium severity operation because of the proper positioning
of the burners and convection section. Hence long run lengths can
be expected for these types of furnaces.
The heat flux profile is determined by the cracking reactions
in the reactor coil, the process gas temperature profile inside the
tube and the tube skin-temperature profile. The heat flux is rel-
] atively low in the beginning of the inlet tube due to low flue
Fig. 11. CO mass fraction distribution in plane x = 1.565 m above a long-flame
gas-temperature. Hence, in the first section of the reactor little
burner. to no cracking of the feed occurs. The absorbed heat makes that
32 G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34

Table 3
Comparison of calculated results with industrial data.

Items Industrial data CFD simulation Predefined heat flux


simulation

Temperature of outlet process gas (K) 1108.5 1111.1 1110.0


Maximum tube skin temperature (K) 1271.4 1269.1 1247.5
Residence time (s) 0.24 0.27 0.27
Coil inlet pressure (MPa) 0.27 0.25 0.25
Coil pressure drop (MPa) 0.07 0.05 0.05
Yield (wt%)
Hydrogen 0.86 0.65 0.70
Methane 13.81 13.82 14.01
Acetylene 0.37 0.39 0.46
Ethylene 29.69 29.56 29.40
Ethane 3.72 3.70 3.73
Propylene 15.53 15.17 15.37
Propane 0.53 0.44 0.39
1,3-Butadiene 4.20 4.49 4.39
Sum of butenes 4.21 3.59 3.74
Benzene 4.62 5.63 5.74
Toluene 3.83 3.24 3.17
Xylene 1.28 3.04 3.01

the temperature of the process gas increases rapidly. With increas- shows that using a simplified approach, based on a predefined nor-
ing process gas temperature inside the tube, the reaction rate of malized incident heat flux profile (Colannino, 2007), it is possible to
the cracking reactions and the heat absorbed by the process gas obtain a similar accuracy for the product yields requiring a signifi-
both increase. Hence, the heat flux reaches the first peak at the cantly shorter simulation time. The computation burden using this
tube length of about 8 m where the flue gas-temperature reaches a last method is reduced from a couple of days with the CFD method
maximum. With increasing conversion the rise of the process gas to a few minutes.
temperature becomes less steep because more of the absorbed heat
is consumed for maintaining the endothermic cracking reactions. 5. Conclusions
When reaching the second pass of the coil, the process gas re-enters
the high-temperature flue gas section. In the first section of the For the first time an industrial naphtha cracking furnace is sim-
second pass where the reactor diameter increases, the process gas ulated containing both radiation and long-flame burners using CFD
temperature remains almost constant. Afterwards, the increasing for the furnace side and detailed kinetics on the process gas side.
heat input from the furnace to the coil makes that the temperature Good agreement is obtained between industrially measured and
rises again. The heat flux increases until a maximum is reached at simulated light olefin yields via a coupled simulation of furnace and
a tube length of about 19 m. Fig. 12(a) also shows that the average reactor, and therefore this approach provides a proper theoretical
heat flux in the first pass is higher than that of the second pass. This basis for gaining better insight in the behavior of industrial steam
can be explained by the smaller driving force for heat transfer in cracking furnaces equipped with both long flame and radiation
the second pass because of the higher temperature of the process burners. Very high ethylene and propylene yields can be obtained in
gas in this section of the reactor. combination with long run lengths because of the specific design of
Fig. 12(b) shows the heat flux along the reactor length and tube the furnace. The optimal design of a steam cracking furnace requires
skin and process gas temperature profile when using a predefined the combination of coil design, the positioning of convection sec-
normalized incident heat flux profile in the steam cracking furnace tion and the positioning of long-flame and radiation burners, and
(Colannino, 2007). As can be seen from this figure the differences for can be addressed by coupled CFD simulations on the fire gas side
the tube skin and process gas temperature profile remain relatively with fundamental kinetic models on the process gas side. How-
small in comparison with those obtained using a full CFD simulation ever, performing such a coupled CFD simulation takes a number of
of the furnace. Hence, it can be expected that also the differences days of simulation time, while a similar degree of accuracy for the
in product yields will remain relatively small. simulated product yields can be obtained with computational less
Table 3 compares simulation results and industrial data for the demanding approaches.
converged coupled reactor/furnace simulation. High light olefin
Notation
yields are obtained because of the combination of short residence
times with the rapid heating of process gas in the split coil design.
The simulated temperature of outlet flue-gas, maximum tube skin A empirical constant
temperature, excess oxygen ratio, coil outlet temperature (COT), B empirical constant
residence time, coil inlet pressure and coil pressure drop are in bε,i,j emissivity gas temperature polynomial coefficients
good agreement with the industrial data. Also the yields of the Cj molar concentration of species j (mol/m3 )
main products ethylene and propylene agree well with the yields cpj heat capacity of component j (kJ kmol−1 K−1 )
measured in the industrial cracker. For the heavier aromatics the Di,m diffusion coefficient for species i in the mixture (m2 /s)
agreement is not perfect, but this can be caused by the way the Dij binary mass diffusion coefficient of component i in com-
yields have been determined in the industrial cracker. Generally ponent j (m2 /s)
part of the heavier products are lost in the separation section before Dt effective mass diffusion coefficient due to turbulence
they can be analyzed (Van Geem, Marin, et al., 2008; Van Geem, (m2 /s)
Reyniers, et al., 2008). This illustrates the potential of the followed dt internal tube diameter (m)
approach: it is possible to obtain quantitative results via a coupled E total energy per unit mass (J/kg)
simulation of furnace and reactor for a furnace equipped with both f Fanning friction factor
long-flame burners and radiation burners. However, Table 3 further Fj molar flow rate (kmol s−1 )
G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34 33

Gk generation of turbulent kinetic energy (J/m3 /s) Acknowledgments


Hfj0 heat of reaction j (kJ kmol−1 )
h sensible enthalpy (J/kg) This work was supported by the National Science Fund for Dis-
hj enthalpy of species j tinguished Young Scholars (No. 60625302), 973 Program of China
I radiation intensity (J/m2 /s) (No.2012CB720500), National High-Tech Research and Develop-
J diffusion flux of species j (kg/m2 s) ment Program of China (No. 2009AA04Z159), and the Shanghai
j
k turbulent kinetic energy (m2 /s2 ) Leading Academic Discipline Project (No. B504). The financial sup-
keff effective conductivity (W/m-K) port from the BOF Bilateral Scientific Cooperation (ECUST/LCT) and
kf,r forward rate constant for reaction r (1/s) the Long Term Structural Methusalem Funding by the Flemish Gov-
ki absorption coefficient of the ith gray gas (1/m) ernment (No. BOF09/01M00409) and ‘111 Project by the Chinese
kt turbulent thermal conductivity (W/m-K) Government (No. B08021) are acknowledged.
Mw,i molecular mass of species i (g/mol)
n refractive index
N number of chemical species in the system
References
p static pressure and the sum of the partial pressures of all
absorbing gases (Pa) Aspen, Plus., 2009. Ten Canal Park, Cambridge (MA) 02141: Aspen Technology, Inc.
pt total pressure (MPa) Belohlav, Z., Zamostny, P., & Herink, T. (2003). The kinetic model of thermal cracking
peff effective pressure (Pa) for olefins production. Chemical Engineering and Processing, 42, 461–473.
Bockelie, M. J., Adams, B.R., Cremer, M. A., Davis, K. A., Eddings, E. G., Valen-
q(z) heat flux along axial coordinate (kJ m−1 s−1 ) tine, J. R., et a1., (1998). Computational simulations of industrial furnaces.
r position vector Presented at the International Symposium on Computational Technologies
R ideal gas constant for Fluid/Thermal/Chemical Systems with Industrial Applications. In: Joint
ASME/JSME Pressure Vessels and Piping Conference, California: San Diego, 117-l24.
Ri net rate of production of species i by chemical reaction
Colannino, J. (2007). Mathematical models for characterizing and predicting heat
(gmol/m3 /s) flux profiles from ethylene cracking units. In AIChE spring meeting 2007 Houston,
Rj rate of reaction in which j participates (kmol m−3 s−1 ) TX, USA.
Davis, H. G., & Farrell, T. J. (1973). Relative and absolute rates of decomposition of
rb radius of the bend (m)
light paraffins under practical operating-conditions. Industrial and Engineering
s direction vector Chemistry Process Design and Development, 12, 171–181.
s scattering direction vector Dente, M., Ranzi, E., Bozzano, G., Faravelli, T., & Valkenburg, P. J. M. (2001). Heavy
s path length component description in the kinetic Modeling of hydrocarbon pyrolysis. In 13th
ethylene producers’ conference, proceedings, 10 (pp. 58–77).
Sct turbulent Schmidt number Dente, M., Ranzi, E., & Goossens, A. G. (1979). Detailed prediction of olefin yields
Sh source term in the energy equation (J/m3 /s) from hydrocarbon pyrolysis through a fundamental simulation-model (spyro).
Sϕ source term Computers and Chemical Engineering, 3, 61–75.
DeSaegher, J. J., Detemmerman, T., & Froment, G. F. (1996). Three dimensional sim-
T local temperature and process-gas temperature (K) ulation of high severity internally finned cracking coils for olefins production.
Ui , Uj ,Ul velocity component in the ith, jth or lth direction (m/s) Revue De L Institut Francais Du Petrole, 51, 245–260.
u velocity of the process gas (m/s) De Schepper Sandra, C. K., Heynderickx Geraldine, J., & Marin Guy, B. (2009). Mod-
eling the evaporation of a hydrocarbon feedstock in the convection section of a
xi , xj , xl coordinate direction in the ith, jth or lth direction (m) steam cracker. Computers and Chemical Engineering, 33(1), 122–132.
Xi mole fraction of species i Detemmerman, T., & Froment, F. (1998). Three dimensional coupled simulation of
Yj mass fraction of species j furnaces and reactor tubes for the thermal cracking of hydrocarbons. Revue De
L Institut Francais Du Petrole, 53, 181–194.
YP mass fraction of product P
Fluent. (2006). Fluent 6.3 user’s guide. Lebanon, NH, USA: Fluent Inc.
YR mass fraction of reactant R Han, Y. L., Xiao, R., & Zhang, M. Y. (2006). Combustion and pyrolysis reactions in a
z axial coordinate (m) naphtha cracking furnace. Chemical Engineering and Technology, 29(1), 112–120.
Habibi, A., Merci, B., & Heynderickx, G. J. (2007). Impact of radiation models in CFD
simulations of steam cracking furnaces. Computers and Chemical Engineering, 31,
Greek letters 1389–1406.
˛ absorption coefficient, 1/m and conversion factor Heynderickx, G. J., Oprins, A. J. M., Marin, G. B., & Dick, E. (2001). Three-dimensional
flow patterns in cracking furnaces with long-flame burners. AIChE Journal, 47,
depending on the units of pt 388–400.
˛ε,i emissivity weighting factors for the ith fictitious gray gas Hottel, H. C., & Sarofim, A. F. (1967). Radiative transfer. New York: McGraw-Hill.
ıij Kronecker delta Howell, J. R. (1968). Application of Monte Carlo to heat transfer problems. Advances
in Heat Transfer, 5(2), 1–54.
ε dissipation rate of turbulent kinetic energy (m2 /s3 and
Huang, Z. Q., Yang, G. J., Yu, Z. H., & Zhang, Z. X. (1993). Petrochemical tube fur-
emissivity) nace simulation and computer calculations. Beijing: Chemical Industry Press. (pp.
 viscosity of gas molecules (kg/m/s) 148–151).
Kang, J. W., Yoo, K.-P., Kim, H. Y., Lee, H., Yang, D. R., & Lee, C. S. (2003). Korea thermo-
eff effective viscosity (kg/m/s)
physical properties data bank (KDB). Seoul, South Korea: Department of Chemical
t turbulent viscosity (kg/m/s) Engineering, Korea University.
 gas density (kg/m3 ) Klein, M. T., Hou, G., Bertolacini, R. J., Broadbelt, L. J., & Kumar, A. (2006). Molecu-
g density of the process-gas mixture (kg/m3 ) lar modeling in heavy hydrocarbon conversions. Boca Raton: Taylor and Francis
Group.
Nekrasov factor for bends Kumar, P., & Kunzru, D. (1985). Modeling of naphtha pyrolysis. Industrial and Engi-
 net effect of third bodies on the reaction rate neering Chemistry Process Design and Development, 24, 774–782.
 generalized diffusion coefficient Lan, X. Y., Gao, J. S., Xu, C. M., & Zhang, H. M. (2007). Numerical simulation of trans-
 fer and reaction processes in ethylene furnaces. Transactions on IChemE, Part A,
i,r stoichiometric coefficient for reactant i in reaction r Chemical Engineering Research and Design, 85(A12), 1565–1579.

i,r stoichiometric coefficient for product i in reaction r Lobo, W. E., & Evans, J. E. (1939). Heat transfer in the radiant section of petroleum
heaters. Transactions of the AIChE, 35, 743.
j,r rate exponent for reactant species j in reaction r Magnussen, B. F., & Hjertager, B. H. (1976). On mathematical modeling of turbulent
j,r rate exponent for product species j in reaction r combustion with special emphasis on soot formation and combustion. In 16th
international combustion symposium Pittsburgh, USA: The Combustion Institute,
Stefan–Boltzmann constant (pp. 719–729).
S scattering coefficient (1/m) Mallard, W. G., Westley, F., Herron, J. T., Hampson, R. F. (2005). Chemical kinetics
ϕ dependent variable database, standard reference database 17, version 7.0 (web version), release 1.3.
Manca, D., Buzzi-Ferraris, G., Cuoci, A., & Frassoldati, A. (2009). The solution of very
˚ phase function
large non-linear algebraic systems. Computers and Chemical Engineering, 33(10),
˝ solid angle 1727–1734.
34 G. Hu et al. / Computers and Chemical Engineering 38 (2012) 24–34

Merdrignac, I., & Espinat, D. (2007). Physicochemical characterization of petroleum Stefanidis, G. D., Merci, B., Heynderickx, G. J., & Marin, G. B. (2007). Gray/nongray
fractions: The state of the art. Oil and Gas Science and Technology – Revue De L’ gas radiation modeling in steam cracker CFD calculations. AIChE Journal, 53,
Institut Francais Du Petrole, 62, 7–32. 1658–1669.
Modest, M. F. (2003). Radiative heat transfer (2nd ed.). New York: Academic Press. Stefanidis, G. D., Van Geem, K. M., Heynderickx, G. J., & Marin, G. B. (2008). Evaluation
Oprins, A. J. M., & Heynderickx, G. J. (2003). Calculation of three-dimensional of high-emissivity coatings in steam cracking furnaces using a non-grey gas
flow and pressure fields in cracking furnaces. Chemical Engineering Science, 58, radiation model. Chemical Engineering Journal, 137, 411–421.
4883–4893. Tao, W. Q. (2001). Numerical heat transfer (2nd ed.). Xi’an: Xi’an Jiaotong University
Oprins, A. J. M., Heynderickx, G. J., & Marin, G. B. (2001). Three-dimensional Press.
asymmetric flow and temperature fields in cracking furnaces. Industrial and Van Geem, K. M., Heynderickx, G. J., & Marin, G. B. (2004). Effect of radial temperature
Engineering Chemistry Research, 40, 5087–5094. profiles on yields in steam cracking. AIChE Journal, 50, 173–183.
Pant, K. K., & Kunzru, D. (1996). Pyrolysis of n-heptane: Kinetics and modeling. Van Geem, K. M., Hudebine, D., Reyniers, M. F., Wahl, F., Verstraete, J. J., & Marin, G.
Journal of Analytical and Applied Pyrolysis, 36, 103–120. B. (2007). Molecular reconstruction of naphtha steam cracking feedstocks based
Pyl, S. P., Van Geem, K. M., Reyniers, M. F., & Marin, G. B. (2010). Molecular reconstruc- on commercial indices. Computers and Chemical Engineering, 31(9), 1020–1034.
tion of complex hydrocarbon mixtures: An application of principal component Van Geem, K. M., Marin, G. B., Hedebouin, N., & Grootjans, J. (2008). Energy efficiency
analysis. AIChE Journal, 56(12), 3174–3188. of the cold train of an ethylene cracker. Oil Gas-European Magazine, 34, 95–99.
Riazi, M. R. (2005). Characterization and properties of petroleum fractions. West Con- Van Geem, K. M., Pyl, S. P., Beens, J., Vercammen, J., Reyniers, M. F., & Marin, G. B.
shohocken: ASTM International. (2010). On-line analysis of complex hydrocarbon mixtures using comprehensive
Rice, F. O. (1931). The thermal decomposition of organic compounds from the stand- 2D gas chromatography. Journal of Chromatography A, 1217, 6623–6633.
point of free radicals I. Saturated hydrocarbons. Journal of the American Chemical Van Geem, K. M., Reyniers, M. F., & Marin, G. B. (2008). Challenges of modeling
Society, 53, 1959–7219. steam cracking of heavy feedstocks. Oil and Gas Science and Technology – Revue
Rice, F. O. (1933). The thermal decomposition of organic compounds from De L Institut Francais Du Petrole, 63, 79–94.
the standpoint of free radicals III. The calculation of the products formed Van Geem, K. M., Zajdlik, R., Reyniers, M. F., & Marin, G. B. (2007). Dimensional anal-
from paraffin hydrocarbons. Journal of the American Chemical Society, 55, ysis for scaling up and down steam cracking coils. Chemical Engineering Journal,
3035–3040. 134, 3–10.
Rice, F. O., & Herzefeld, K. F. (1934). The thermal decomposition of organic com- van Goethem, M. W. M., Kleinendorst, F. I., van Leeuwen, C., & van Velzen, N. (2001).
pounds from the standpoint of free radicals VI. The mechanism of some chain Equation-based SPYRO (R) model and solver for the simulation of the steam
reactions. Journal of the American Chemical Society, 56, 284–289. cracking process. Computers and Chemical Engineering, 25, 905–911.
Robinson, J. (2009). Naphtha contamination and furnace coil plugging. In Ethylene Verstraete, J. J., Revellin, N., Dulot, H., Hudebine, D. (2004). Molecular reconstruction
producers conference 2009, Proceedings Tampa, FL, USA. of vacuum gasoils. Abstracts of Papers of the American Chemical Society, 227,
Roesler, F. C. (1967). Theory of radiative heat transfer in Co-Current tube furnaces. 010-FUEL.
Chemical Engineering Science, 22(l), 1325–1336. Wang, G. Q., & Zhang, L. J. (2005). Numerical simulation of gas flow and heat transfer
Shu, W. R., & Ross, L. L. (1982). Cracking severity index in pyrolysis of petroleum in cracking heater. Petrochemical Technology, 34(7), 652–655.
fractions. Industrial and Engineering Chemistry Process Design and Development, Westbrook, C. K., & Dryer, F. L. (1981). Simplified reaction-mechanisms for the oxi-
21, 371–377. dation of hydrocarbon fuels in flames. Combustion Science and Technology, 27,
Stefanidis, G. D., Heynderickx, G. J., & Marin, G. B. (2006). Development of 31–43.
reduced combustion mechanisms for premixed flame modeling in steam Zhou, H. Z., & Jia, Z. G. (2007). Three dimensional numerical simulation of flow and
cracking furnaces with emphasis on NO emission. Energy and Fuels, 20, combustion in firebox of ethylene cracking furnace. Petrochemical Technology,
103–113. 36(6), 584–590.
Stefanidis, G. D., Merci, B., Heynderickx, G. J., & Marin, G. B. (2006). CFD simulations Zhang, Z. H., Huang, G. Q., & Wu, X. (2008). Numerical simulation of combustion of
of steam cracking furnaces using detailed combustion mechanisms. Computers USC ethylene cracking furnace. Chemical Industry and Engineering Progress, 27(2),
and Chemical Engineering, 30, 635–649. 255–260.

You might also like