Sie sind auf Seite 1von 7

Ckmical Engineering Sripnrs. Vol 43. No. 6. pp. 1223-1229. 1988. OOG?-250%88 $3.00+0.

00
Printed in Great Britain. Pergamon Press plc

THE COUPLED SIMULATION OF HEAT TRANSFER AND


REACTION IN A PYROLYSIS FURNACE

M. V. RAMANA RAO,+ PATRICK M. PLEHIERS and GILBERT F. FROMENT:


Laboratorium voor Petrochemische Techniek, Rijksuniversiteit Gent, Krijgslaan 281, 9000 Gent, Belgium

(Received 15 December 1986; accepted fir publication 3 September 1987)

Abstract-The firebox of a pyrolysis furnace is divided into zones and the temperature distribution is
generated. Heat transfer by radiation and convection is considered. The location of the burners in the oven
walls is explicitly accounted for. Absorption and emission of radiation by the flue gas are calculated using a
fundamental model that considers individual band contributions for carbon dioxide and water. The
temperature distributions inside the cracking coil and in the furnace are generated simultaneously. They are
in excellent agreement with industrial results, as is the simulated product distribution.

INTRODUCTlON walls and the gas were considered to be gray. Using this
The thermal cracking of hydrocarbons for olefin model, Vercammen and Froment (1978) simulated a
production out in long tubular reactors,
is carried thermal cracking furnace with horizontal tubes.
inserted in a furnace. Usually, the simulation of the Presently, in thermal cracking as well as in many
thermal cracking coil is uncoupled from the heat other chemical processes, vertically suspended coils are
transfer phenomena in the firebox by imposing a heat used. This paper discusses the development of a
flux profile on the reactor. It is then checked a generalized and flexible furnace simulation program.
posteriori whether or not the firebox allows this heat The program can handle any type of tube lay-out in a
flux profile to be attained. In most cases this is done by rectangular firebox. Its general concept is similar to the
means of a simplified method like the Lobo and Evans previous work. For the calculation of the direct- and
(1939) approach, which assumes a uniform tempera- total-exchange areas, a fundamental approach, con-
ture throughout the firebox. sidering individual band absorption by carbon dioxide
With the help of an accurate simulation model for and water, is taken. In addition, the position of the
the radiative heat exchange in a furnace, the calculation burners in the furnace walls and the flue gas flow
of the temperature and heat flux distribution in the pattern in the firebox are explicitly accounted for. The
firebox and the reactor simulation can be coupled. The furnace model, as well as the kinetic scheme, are tested
combined furnace and reactor design can then be by the coupled simulation of heat transfer and reaction
optimized in order to obtain a favorable product in an industrial thermal cracking furnace.
distribution or to reduce unwanted side effects, such as
fouling of the heat transfer surface by carbonaceous RADIATIVE HEAT TRANSFER IN THE FIREBOX
deposits. In the zone method, as outlined by Hottel and
Vercammen and Froment (1978, 1980) further de- Sarofim (1967), the space in which radiative heat
veloped a zone method initially introduced by Hottel transfer has to be calculated is divided into a number of
and Sarofim (1967) to simulate radiation in industrial surface and volume elements which are isothermal and
heaters. Monte Carlo simulation techniques were used have uniform properties. The energy balances for the
zones form a set of non-linear algebraic equations:

z,z*
~
- EZ,Z,
__
zzz, ZZ, El Q,A*
Z,Z, z,z, - zEz,zj Z.Z, E2 = Qdz
.. (1)
..
__ __
ZIZ ZZZ z,z, - z z,z, E QA

ZiZj represents the amount of radiative energy emitted


to determine the view factors. For the calculation of
by zone Zi that is absorbed by zone Zj, both directly
direct- and total-exchange areas, absorption and reflec-
and after reflection on other zones, divided by the
tion of radiation have to be accounted for. Both the
black-body emissive power Ei of Zi. ZiZj has the
+Council of Scientific and Industrial Research, Regional dimensions of an area and is defined as the total-
Research Laboratory, Hyderabad SOOO09, India. exchange area between Zi and Z,.. In eq. (l), Qi is the
~To whom correspondence should be addressed. non-radiative heat flux leaving Zi.

1223
1224 M. V. RAMANA R.40 et al.

The calculations leading to the total-exchange areas defined. Each of them contains an entire row of tubes.
that appear in eq. (1) are performed in three steps. First, The Type-A regions are then further divided into
the view factors among surfaces in a transparent Type-B areas and empty space. The Type-B areas are
medium are calculated. In a second step, absorption by small square parallellepipeds, each enclosing one tube.
the flue gas is accounted for and the direct-exchange The path of a beam is followed by stepping from
areas among surface zones and gas zones are de- border to border in the grid. As long as no Type-B area
termined. Finally, the total-exchange areas are calcu- is entered, no incidence on a tube can occur. In a Type-
lated by considering both direct and reflected B area, incidence may occur. The beam length is
radiation. increased in small steps and after each step it is checked
The view factors in a transparent medium are whether or not the tube is hit. The position Pb of the
defined by point along the beam is expressed by the coordinates
(x, y, z) in the firebox enclosure. At the same coordi-
cosg,cospj
dAi dAj. (2) nates, a tube can be represented by means of the
7CS
equation of a cylindrical surface:
They only depend upon the furnace geometry.
P(P) = (x -x0) + (z - z,,) - R2 = 0. (3)
Analytical integration is possible in some particular
cases. For the complex geometries encountered in The tube is hit when F(P,) is negative or zero at a point
industrial furnaces, numerical integration by means of along the beam. If the derivative of F(PL) with respect
a Monte Carlo method is required. In this work, a to the beam length is positive, the beam is heading
sample of 25,000 random beams per zone was taken. away from the tube. In this case, no incidence can occur
Each beam is followed through the furnace and the and a larger step can be taken in order to save time.
location of the hit is detected. In the previous work by With the new routine, a four- to seven-fold reduc-
Vercammen and Froment (1978, 1980), extensive use tion in processing time was obtained as compared to
was made of goniometric functions for this purpose. the method based on goniometric functions.
Since the Monte Carlo procedure takes a large portion In the present furnace simulation, the location of the
of the total computer time for one furnace simulation, burners in the oven walls is explicitly accounted for.
a special effort has been made to enhance its efficiency. Because of the large number of radiant burners in a
A new routine for detecting the incidence of a beam cracking furnace, incorporating the burner cups into
avoids the use of goniometric functions, whose calcu- the simulation program as additional zones would lead
lation is very time-consuming, and replaces them by to an unjustifiable increase in computer processing
simple arithmetical and logical operations. The path of time. Therefore, the burners are considered to be point
each beam is followed in a stepwise manner. To take sources of radiant energy. This is an acceptable
the maximum controllable step, a raster is constructed assumption, since the ratio of the radiation heat
in the furnace (Fig. 1). First, Type-A regions are emitted by the burner cup to the reflected radiation is
very high. The view factors from the burners towards
all zones are calculated by means of the Monte Carlo
method that has been outlined above.
A2 In the following calculations, the burner cup tem-
- perature and a split factor y are required. y is the
fraction of the heat generated in the burner that is
effectively emitted by the burner cup. The fraction of
the combustion heat that enters the firebox with the
flue gas is then 1 - y. In the work by Vercammen and
Froment (1978), the factor y was an adjustable par-
ameter which was tuned to make the simulated
conversion match the industrial one.
The burner cup temperature T, is determined from
the energy balance for the cup. It is assumed that the
flue gas inside the burner cup has a temperature equal
to the adiabatic flame temperature, T,-. From the flue
gas, heat is transferred to the burner cup by radiation
and convection [left-hand side of eq. (4)]. The burner
cup itself emits and reflects radiation [right-hand side
of eq. (4)]:
&,oT4,+h(T,- ZJ = &$7-,4+(1 -&E,)EBOTf:. (4)

Pyrometer measurements in commercial furnaces in-


dicate a burner cup temperature of about 1200C or
1473 K. For the furnace and conditions referred to in
-
Table 1, a burner cup temperature of 1494 K is
Fig. 1. Construction of a raster in the furnace. calculated from the energy balance (4). From this
Coupled simulation of heat transfer and reaction 1225

Table 1. General furnace and reactor data


2-G
=A, =
Furnace x5 (&T _,)
Length 9720 mm
Height 9913 mm C, = hc,2
Width 1922 mm
Refractory thickness 251 mm C, = hc,/k
Insulation thickness 50 mm
Number of side wall burners 96
Side wall burner capacity 182,600 w
157,000 kcaljb
Number of bottom burners 16
Bottom burner capacity 365,200 W
314,000 kcal/h

Firing conditions
Fuel gas
Hydrogen 5 vol%
Methane 95 vol%
Air excess 15%
Reactor
x
Four reaction coils on a single row in the middle of the Fig. 2. Weighting of the individual band contributions ac-
furnace. swaged diameter design cording to Plancks law.
Total length 87.51 m
Number of passes 9
Internal diameter passes 1-5 110 mm
External diameter passes l-5 126 mm different bands to the radiative heat exchange are
Internal diameter passes 6-9 123 mm added up according to their weight in the Planck
External diameter passes 69 142 mm
2287.5 kg/h
distribution (Fig. 2). Direct-exchange areas among
Hydrocarbon flow rate per tube
Steam dilution 0.3 kg/kg HC surface zones are then obtained by multiplying Fij by
Radiant section inlet temperature 923 K the emitter area.
650C To calculate the direct-exchange area from a surface
Coil outlet pressure, not given,
to a volume zone, fictitious surface zones delimiting the
assumed to be. 174 kPa
Feedstock gas zones were introduced in the previous work. In the
Hydrogen 0.01 wt% present work, the introduction of the fictitious zones is
Methane 0.38 wt% avoided by incorporating them implicitly into the
Ethylene 0.89 wt% Monte Carlo simulations. For radiation towards a gas
Ethane 94.39 wt y0
1.98 wt y0
zone, two beam length distributions are defined: one
Propylene
Propane 1.42 wt% for the radiation entering the volume and another for
1 -Butene 0.93 wt y0 the radiation leaving the zone. The increase in CPU-
Material properties time for the Monte Carlo program is small. The new
Emissivity refractory 0.60
method proved to be very practical, especially when the
Emissivity tubes 0.95
Thermal conductivities of Wfmk oven volume is divided into a large number of zones.
Refractory 0.0913+ 118.OE-66+T(K) Since the integrated-band approach is followed, the
Insulation 0.0452+11l.lE--*T(K) direct-exchange areas from volume to surface zones
cannot be calculated by means of the principle of
reciprocity:
temperature, the split factor y is easily calculated. A
jjZ=sg. (6)
value of y = 0.0964 is found, which is in good agree-
ment with y = 0.076 proposed by Vercammen and For the calculation of the emission of radiation from a
Froment (1978). gas volume towards a surface, the effective emissive
In a real medium, a fraction of the radiation among beam length, S,, is introduced. For some furnace types,
the surface zones is absorbed. Absorption is accounted S, has been correlated as a function of the main
for by the introduction of the average transmittance dimensions of the furnace. For special geometries,
Tij. which are frequently encountered in chemical process
furnaces (e.g. thermal cracking coils with variable
Fij = T~~F$. (5)
diameter), no such correlations are available. To ensure
rij is a function of the beam length, the flue gas the general applicability of the simulation program,
composition and the temperatures of both the emitter the emission beam length distribution is generated by
and the absorbing medium. Individual band absorp- means of an additional Monte Carlo simulation. At the
tion by carbon dioxide and water is considered. Each same time, the view factors for radiation from the gas
absorption band is represented by an equivalent band zones to all surface zones are calculated. The total
which is opaque for radiation. To determine the widths amount of radiative energy that is emitted by the gas
of the equivalent bands Edwardss correlations (Siegel zone towards a surface zone is obtained by introducing
and Howell, 1972) are used. To reduce the required the effective emissive beam length into the absorption
computer storage capacity, the contributions of the model (vi& sup@. A fraction of this energy is ab-
1226 M. V. RAMANA RAO et al.

sorbed by the emitting gas itself. To account for the 2 = m


self-absorption by the gas, Siegel and Howell (1972) f
0
and Hottel and Sarofim (1967) applied a correction ZoneII
P2
factor to the emissive beam length S,. For most
geometries, this factor lies between 0.85 and 0.95. An
average value of 0.90 was taken in this work.
From the total amount of radiant energy emitted by
a gas zone, the absorption coefficient k for the gas can
be derived:

total emitted energy = 4k (volume of


the gas zone). (7)

With the method described here, values between 0.205


and 0.234 m- 1 are obtained, depending on the Rue gas
composition and temperature. These values are in
good agreement with k = 0.20 m-l proposed by
Scholand (1983) for application in a similar problem.
From the direct-exchange areas, the total-exchange
areas are calculated through a purely mathematical l v,-0
operation, for which Hottel derived a convenient p3

matrix algorithm. In principle, the method is restricted I


to systems in which the incident and the reflected
radiation have the same spectral wavelength distri- Fig. 3. Calculation of the velocity vector.
bution. Because of the introduction of the integrated-
band approach, this condition is no longer fulfihed.
The relative errors introduced in this way do not
exceed 3 %, however.

CONVECTIVE HEAT TRANSFER

Heat transfer from the flue gas to the surrounding


surfaces and to the tubes by convection is only a small
contribution to the total heat exchange. Convection
coefficients are calculated by means of the usual
correlations for immersed bodies.
Because of the flow of the flue gas through the
volume zones, considerable enthalpy flows occur in the
right-hand side of eq. (1). On the basis of some general
rules, derived from observation in the thermal cracking
pilot plant of the Laboratorium voor Petrochemische
Techniek, the flue gas flow pattern in the firebox can be
calculated with a reasonable precision. It was observed
that the flue gas leaves the radiant burners in a cone,
with the top angle depending upon the burner
construction.
The flue gas velocity induced by one burner at a
particular point in the furnace is calculated as shown in
Fig. 3. For a point located inside the cone (zone I in Fig.
3), the flue gas velocity is a vector that is directed away
from the cone top T and that has a magnitude inversely
proportional to the square of the distance between P,
and T. In zone II, above the cone, the flue gas flow is
supposed to be homogeneously distributed. In this
zone, the flue gas has an average upward velocity v,. In
zone III, below the cone, the velocity is assumed to be
zero. The vector sum of the velocities induced by ali
burners yields the final velocity vector in the point. In
Fig. 4, the final flue gas velocities in a section of the
furnace are shown. The cone top angle in this par-
ticular case was 60. The flue gas flow in the vicinity of
the oven walls and of the reactor tubes is almost 4. Flue gas velocity profile in the firebox.
Coupled simulation of heat transfer and reaction 1227

uniform. The hot flue gases leaving the burners, on the


contrary, penetrate deeply into the furnace, thus
I Initialization of 0 ond T 1
creating a risk for flame impingement on the tubes.
This is in particular the case for the bottom burners.
1 Total exchange oreo motrix CT) 1
SIMULATION RESULTS
Coupling the furnace simulation model with a
computer program that describes the phenomena
taking place inside the tubes (heat transfer and reac-
tion) allows the temperature and the heat flux distri-
bution in the furnace to be calculated. Figure 5 gives a
simplified flow chart of the calculations. A double
1 BetterestiytionforT 1
iterative scheme is used. In the inner loop, the set of
non-linear equations formed by the energy balances is
solved. Thus, a better estimation of the zone tempera-
tures is obtained. From these temperatures and the
temperatures of the process gas inside the tubes, the
heat fluxes can be re-estimated. The whole cycle is
Reactor simulation
repeated until convergence occurs in the outer iter-
ation loop.
By way of example, the simulation of a commercial
furnace for the pyrolysis of ethane is performed. The Computotlon of P

simulated results are tested against industrial data. The


main characteristics of the furnace and the reactor are
given in Table 1. Eighty-two surface zones and four
volume zones are considered. A schematical represen-
I Better estimation for 0
I
tation of the coil configuration and the subdivision

72
into zones is given in Fig. 6. For the simulation of the Yes
Iteration?
cracking process, an extended version of the detailed
radical reaction scheme proposed by Sundaram and No
Froment (1981) is used in combination with a one-
dimensional plug flow model for the reactor (Froment, STOP
198 1). The kinetic scheme contains about 150 reactions
between 14 molecular species and 12 radicals. Fig. 5. Flow chart of the calculations.
In Fig. 7, the temperatures of the furnace walls and
the flue gas are given as a function of the height in the
furnace. Because of the combined side wall and bottom
firing, the flue gas temperature gradient is quite steep. and the coil exit ethane conversion match the plant
Consequently, the heat flux along the cracking coil data excellently. The calculated product distribution is
varies considerably with the axial coordinate (Fig. 8). in good agreement with the industrial data. Notice that
The peaks correspond with the bottom of the furnace, no adjustable parameters were used in the simulation
the valleys with the top. The average heat flux is about model.
70 kW/m (16.5 kcal/m* s). The maximum value is
103 kW/m (24.6 kcal/m s). These values correspond CONCLUSION
well with the commonly accepted values in such In thermal cracking the heat flux profile along the
furnaces. The axial tube metal temperature profile, reactor determines the feedstock conversion, the olefin
which is shown in Fig. 9, follows the shape of the heat selectivities and the rate of coke deposition. An average
flux profile. The temperature peaks are important for value of the heat flux, obtained by means of simplified
the choice of the tube material. They also lead to severe furnace calculations via the Lobo and Evans approach,
and localized coke formation in the lower U-bends, cannot be used to predict the combined furnace and
causing a high pressure drop and therefore a loss in the reactor performance up to present-day standards. In
selectivity for the main cracking product, ethylene, and case the reactor simulation is uncoupled from the heat
reducing the run length of the furnace between decok- transfer calculations, the uncertainties involved are
ing periods. The process gas temperature (Fig. 10) is larger still. A detailed firebox simulation model is
relatively insensitive to the flux variations. This is due therefore a powerful tool in the design and operation
to the high mass flow rate in the reactor, which of pyrolysis furnaces and reactors.
dampens changes in heat input, and to the endothermic By means of simulation, the influences of the firing
nature of the pyrolysis process, that has a self- conditions and pattern and of the furnace and reactor
stabilizing effect on the process gas temperature. geometry on the temperature and heat flux distri-
Table 2 compares the simulated results with the bution in the firebox can be investigated. Application
industrial measurements. The calculated temperatures of the program to the simulation of a thermal cracking
1228 M. V. RAMANA R~~etd.

Fig. 6. Subdivision of the furnace in isothermal zones.

External

700
I I I I
0 25 50 75 100

Reactor coordinate (ml

Fig. 9. Axial tube wall temperature profiles.


T (-3
Fig. 7. Flue gas and oven wall temperatures.

I I I I 600
I I I I
40 100 0 25 50 75 100
0 25 50 75
Recfctor coordinate (ml Reactor coordinate (ml

Fig. 8. Axial heat flux profile. Fig. 10. Axial process gas temperature profile.
Coupled simulation of heat transfer and reaction 1229

Table 2. Simulated results vs industrial data s surface zone


T temperature, K, C
Plant Z zone
data Simulated -
ZiZj direct-exchange area from zone i to zone
Adiabatic flame j, mz
temperature 2153 K
zizj total-exchange area from zone i to zone
Burner cup temperature 1470 1494 K
Direct radiation from i, m2
burners 9.64 %
Flue gas outlet Greek letters
temperature 1300 1299 K
Outside heat loss 3 0.65 % B inclination with respect to the surface
Process gas temperature 1136 1124 K normal
at reactor outlet 863 851 C Y split factor
Coil inlet pressure 272 kPa
174
& emissivity
Coil outlet pressure kPa
0 = 5.7E-8, W/m2K4
Ethane conversion 61.51 62.07 % r transmissivity
Product yields in wt y0
Hydrogen 3.38 3.37
Subscripts
Methane 6.31 6.87
Ethylene 46.66 46.71 b beam
Propylene 3.20 1.65 burner cup
Butadiene 1.41 1.72 ; flame
9 flue gas
i,i zone indices
furnace shows that the temperature and heat flux
distributions in such a furnace are highly non-uniform.
REFERENCES

Acknowledgement-P. M. Plehiers is grateful to the Belgian Froment, G. F., 1981, Thermal cracking for olefins pro-
National Fonds voor Wetenschappelijk Onderzoek for a duction. Fundamentals and their application to industrial
Research Assistantship. problems. Chem. Engng Sci. 36, 1271-1282.
Hottel, H. C. and Sarofim, A. F., 1967, Radintiue Transfer.
McGraw-Hill, New York.
NOTATION L&o, W. E. and Evans, J. E., 1939, Trans. A_1.Ch.E. 35, 743.
A area of a zone, m2 Scholand, E., 1983, Modern procedures for the calculation of
black-body emissive power of zone i, radiant heat transfer in direct-fired tubed furnaces. Inc.
Ei
Wfm2 them. Engng 23, -10.
Siegel, R. and Howell, J. R., 1972, Thermal Radiation Heat
FijO view factor from zone i to zone i in a Transfer. McGraw-Hill, New York.
diathermal medium Vercammen, H. A. J. and Froment, G. F., 1978, Simulation of
9 gas zone thermal cracking furnaces, in ACS Symposium Series No.
absorption factor, m- 65, Chemical Reaction Engineering, Houston (Edited by
k
V. W. Weekman, Jr and D. Luss).
place vector
Vercammen. H. A. J. and Froment, G. F., 1980, An improved
; non-radiative heat flux, W/m2 zone method for the simulation of radiation in industrial
S path length, m furnaces. Int. J. Heat Mass Transfer 23, 329-337. (1980)

Das könnte Ihnen auch gefallen