Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

DE GRUYTER International Journal of Chemical Reactor Engineering.

2018; 20170243

Amir Barza1 / Behrouz Mehri1 / Vahid Pirouzfar2

Mathematical Modeling of Ethane Cracking


Furnace of Olefin Plant with Coke Formation
Approach
1 Department of Process engineering, Kavian Petrochemical Company, PSEEZ, Asaluyeh, Iran
2 Department of Chemical Engineering, Central Tehran Branch, Islamic Azad University, Tehran, Iran, E-mail:

v.pirouzfar@iauctb.ac.ir

Abstract:
In this study, ethylene furnace modeling is carried out by ethane pyrolysis (thermal cracking or hydrocracking)
method in Arya Sasol Petrochemical Company (ninth olefin unit, Assaluyeh, Iran), which includes the solution
of kinetic equations and transfer phenomena, by the forward finite difference method in the MATLAB. Due to
study and compare coke formation, a specific time period has been selected in equal segments and equations
have been solved. It means that in a length segment of coil (Δz), momentum, energy as well as mass equations
are solved, then the amount of precipitated coke in each length segment is achieved. With new efficient coil
diameter calculating all mentioned approach will be repeated for next time segment. The results of this model
have been compared with actual data and deviation has been reported. It was found that modeling approach is
more capable to define the parameters of coke formation equations. The model has a good agreement between
the values of prediction and experimental of in most cases.
Keywords: thermal cracking of ethane, pyrolysis mathematical modeling, coke formation, transfer phenomena,
radiation in furnace, hydrocracking
DOI: 10.1515/ijcre-2017-0243
Received: December 13, 2017; Revised: April 4, 2018; Accepted: July 7, 2018

1 Introduction
Olefins production are the third capital after Ammonium and oil refinery. Pyrolysis is a reaction in gas phase
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

with high temperature and because of this, reactions accomplish in the coil with direct radiation at more than
800°C (840–890). The furnaces (F101 ~ 109) are designed to crack gaseous fresh ethane feed and recycle ethane
feed at an ethane conversion of 65 % with a steam to hydrocarbon ratio of 0.3. A mix of fresh ethane and recycle
ethane feedstock is delivered to the furnace area at a pressure of 7.1 Bara and a temperature of 55°C (Ghasemi,
Gilani, and Daryan 2016; Karamullaoglu and Dogu 2007; Nabavi et al. 2011). At the furnace the feedstock is split
into two passes through pass flow control valves and routed to the first row of the feed preheater (FPH) in the
furnace convection section. Superheated dilution steam coming from the hot section is supplied to the furnace
area at a pressure of 7.0 Bara and a temperature of 190°C. At the furnace, the dilution steam flow is split into
two passes through pass flow control valves. The signals from each of the individual feed and dilution steam
passes are fed to a steam/feed ratio and lead control system (Albright and Marek 2015). This system controls
the total hydrocarbon throughput and the steam to hydrocarbon weight ratio in such a way that dilution steam
is always in excess of the ratio specified whenever the hydrocarbon throughput is changed. Additionally, it
equally distributes the individual flows of hydrocarbons and dilution steam over the two passes. The ethane
feed is pre-heated in six passes of the FPH bank. After the FPH bank, the six passes are combined to two passes
and each pass is mixed individually with one dilution steam pass (Technip Lay-Out Radiant Coil).
The dilution steam has the dual function of lowering the hydrocarbon partial pressure, and reducing the
coking rate in the radiant coils. The steam and hydrocarbon mixture is split into six passes and further super-
heated to the required radiant coil inlet temperature in the first high temperature coil (HTC-I) and the second
high temperature coil (HTC-II). This temperature has been carefully selected to suit the ethane feed, in order to
obtain maximum heat recovery without reaching cracking conditions. Note that the flow after the HTC-I is first
combined to two passes and then split again to six passes before entering the HTC-II. The hydrocarbon/steam
mixture leaving HTC-II is split into 24 radiant coil inlet streams before entering the radiant section. The radiant

Vahid Pirouzfar is the corresponding author.


© 2018 Walter de Gruyter GmbH, Berlin/Boston.

Brought to you by | University of California - Santa Barbara


Authenticated 1
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

section of ASPC’s furnace is symmetrical with respect to a plane through the center line of the furnace and at
an angle of 90° with the side walls. There are a total of twenty-four (24) vertical radiant coils located centrally
in the firebox (Index Furnace Performance Sheets).
The radiant coil is of the Swaged Multiple Diameter Kinetics (SMK) with straight coil type. The inlet and
outlet tube of each coil (total of four tubes) enters the radiant section through the roof section. The last two
tubes of each radiant coil have a larger diameter than the first two tubes. Using a larger diameter tube in the
outlet part, the region of the coil where feed conversion and the related coking rate is the highest, the radiant
coil is less susceptible to fouling and plugging. This increases both run length and operability.
Due to descent any undesired progress; pyrolysis outlet shall be cooled-down immediately. Therefore trans-
fer line exchangers (TLE) have been set in the furnace outlet (Technip Process Description). Figure 1 illustrates
the furnace and TLEs flow diagram and arragement.
Figure 2 shown the furnace zones and coil arrangement. Also for better view about type of segment, Figure
3 illustrates Δz segment.

Figure 1: Furnace & TLEs.


Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 2: Coil arrangement (top view).

Hydrocarbons pyrolysis usually is accompanied with coke formation as a negative component which settle
down on the internal side of coil’s wall that it leads to pressure and heat transfer drop as well as hot spots that
will eventuate to furnace run-time and selectivity decrease.

1.2 Coke formation mechanisms

Because of thermal condition (more than 700°C) as well as feed phase (gaseous without Aromatic composition)
catalytic coke formation mechanism is happened not pyrolysis or radical mechanism. Also droplet Conden-
sation mechanism is not happened when feed is gas or temperature is low but this mechanism in TLEs and
downstream equipment is dominant (Albright and Marek 1988; Cai, Krzywicki, and Oballa 2002; Kopinke,
Zimmermann, and Nowak 1988).
In the catalytic coke formation mechanism, polished coil’s surface, provide the first layer of coke with ab-
sorbing hydrocarbons molecules and as consequence convert them to carbon atoms by metal components of
coil’s surface which are usually Nickel, Chrome and Iron until component’s surface is saturated by carbon
atoms. In the next step saturated crystals will work as an active site or catalyst for following carbon layer for-
mation and at the end of layer forming, metal component will transfer on the top of formed layer therefore the
coke forming progress will be being continued (Alizadeh, Towfâighi, and Karimzadeh 2008).
It goes without saying that, for reduction of coke formation rate, vendors recommend sulphur-based ad-
ditives such as DiMethyl DiSulfid (DMDS) because these are cracked in the vicinity of heat, settle down on

Brought to you by | University of California - Santa Barbara


2 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

the coil’s surface as an interface layer so that it prevents coke formation. Although coke formation cannot be
devastated entirely while the process is going ahead, the rate of forming can be controlled and finally with a
longer run-time, decoke operation will be needed.
The developed model and the findings of this research are unique in their nature. The results presented
here represent an interesting contribution to ethane cracking technology or olefin plant development. This
model can be used for simulation and optimization of the performance in ethane cracker modules and coke
formation prediction as a model which represents the performance in the olefin furnace. The developed model,
in particular, was proposed as an important factor to estimate the product flow rate. this valuable work has been
done to calculate the amount of coke and compare it with the amount of coke formed in the operating state
(actual data). So far, the information from coke modeling in a furnace Run-Time with the actual amount of
coke has not been announced together for a gas feed unit that can be used to estimate the model error in coke
formation. The most crucial point of this study is the comparison between actual and simulated amount of coke
which is formed during a furnace run time that has not been reported yet. Employed technique and developed
models can be used as a useful tool for design and optimization of ethane cracking furnace with coke formation
approach for various industrial applications.

2 Theoretical
2.1 Selection of reactions

After a wide and accurate research, it is understood that Sundaram & Forment model (Froment 1992; Sundaram
and Froment 1977; Sundaram and Froment) shall be opted because of akin operation condition with ASPC’s
furnaces such as dilution factor and Coil Outlet Temperature (COT). Actual operation & selected model condi-
tions are included in Table 1.

Table 1: Actual operation & selected model conditions.


Parameter equation Froment Operation
Outlet pressure 1.5–2 2–3
(atm.)
Dilution Factor 0.4 0.3
COT Rang (°C) 780–850 780–840
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

According to actual data and analysis 8 basic reactions have been presented which are dominant in the
reactor among more than 1000 reactions (Norinaga and Deutschmann 2007; Norinaga et al. 2009; Norinaga, Ja-
nardhanan, and Deutschmann 2007; Rahimpour et al. 2013). The rate coefficient of model reactions are included
in table 2.

C2 H6 ↔ C2 H4 + H2 ΔH = 136330000 (J/Kmol) (1)

2C2 H6 ↔ C3 H8 + CH4 ΔH = −11560000 (J/Kmol) (2)

C3 H8 ↔ C3 H6 + H2 ΔH = 124910000 (J/Kmol) (3)

C3 H8 → C2 H4 + CH4 Δ𝐻 = 82670000 (J/Kmol) (4)

C3 H6 ↔ C2 H2 + CH4 ΔH = 133450000 (J/Kmol) (5)

C2 H2 + C2 H4 → C4 H6 Δ𝐻 = −171470000 (J/Kmol) (6)

2C2 H6 ↔ C2 H4 + 2CH4 Δ𝐻 = 71102000 (J/Kmol) (7)

C2 H4 + C2 H4 → C3 H6 + CH4 Δ𝐻 = − 22980000 (J/Kmol) (8)

Brought to you by | University of California - Santa Barbara


Authenticated 3
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

With considering mentioned reactions, rate of them will be declared as following:

𝐹𝐶2 𝐻6 𝑃𝑡 𝐹𝐶 𝐻 𝐹𝐻 𝑃 2
𝑟1 = 𝑘1 ⎡
⎢ ( ) − 22 4 2 ( 𝑡 ) ⎤
⎥ (9)
⎣ 𝐹𝑡 𝑅𝑇 𝐹𝑡 𝐾𝐶1 𝑅𝑇 ⎦

𝐹𝐶2 𝐻6 𝑃𝑡
𝑟2 = 𝑘 2 [ ( )] (10)
𝐹𝑡 𝑅𝑇

𝐹𝐶3 𝐻8 𝑃𝑡
𝑟3 = 𝑘 3 [ ( )] (11)
𝐹𝑡 𝑅𝑇

𝐹𝐶3 𝐻8 𝑃𝑡
𝑟4 = 𝑘 4 [ ( )] (12)
𝐹𝑡 𝑅𝑇

𝐹𝐶3 𝐻6 𝑃𝑡 𝐹𝐶2 𝐻2 𝐹𝐶𝐻4 𝑃𝑡 2


𝑟5 = 𝑘 5 ⎡
⎢ ( ) − 2𝐾
( )⎤
⎥ (13)
⎣ 𝐹 𝑡 𝑅𝑇 𝐹 𝑡 𝐶5 𝑅𝑇 ⎦

2
𝐹𝐶2 𝐻2 𝐹𝐶2 𝐻4 𝑃𝑡 ⎤
𝑟6 = 𝑘 6 ⎡
⎢ ( )⎥ (14)
⎣ 𝐹𝑡2 𝑅𝑇 ⎦

𝐹𝐶2 𝐻6 𝑃𝑡
𝑟7 = 𝑘 7 [ ( )] (15)
𝐹𝑡 𝑅𝑇

2
𝐹𝐶2 𝐻6 𝐹𝐶2 𝐻4 𝑃𝑡 ⎤
𝑟8 = 𝑘 8 ⎡
⎢ ( )⎥ (16)
⎣ 𝐹𝑡 𝑅𝑇 ⎦
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Table 2: Rate coefficient of model reactions. (Sundaram and Froment 1977).


Rate coefficient A(sec-1orlmole-1sec-1) E(J/mole)
K1 4/65E+ 13 2/73E+ 05
K2 3/85E+ 11 2/73E+ 05
K3 5/89E+ 10 2/15E+ 05
K4 4.69E+ 10 2/12E+ 05
K5 6/81E+ 08 1/54E+ 05
K6 1/03E+ 09 1/73E+ 05
K7 6/37E+ 23 5/30E+ 05
K8 7/08E+ 10 2/53E+ 05

Where, T is temperature in K, R is constant of gas, P is pressure in Pa, and F is molar flow rate of in mole s−1 .
Due to gaseous feed and spacious rang of produced component in the length of coil, Ethylene (C2 H4 ) propylene
(C3 H6 ) and 1,2butadine (C4 H6 ) are in majority so they have been selected as infrastructure of coke formation
(Rahimpour et al. 2013; Zarinabadi et al. 2010; Zou et al. 1993). The rate coefficient of coke forming reaction is
mentioned in Table 3.

𝐹𝐶2 𝐻4 𝑃𝑡 1.34

𝑟1 = 𝑘 1 ⎢ ( ) ⎤ ⎥ (17)
⎣ 𝐹𝑡 𝑅𝑇 ⎦

𝐹𝐶3 𝐻6 𝑃𝑡 2.89
𝑟2 = 𝑘 2 ⎡
⎢ ( ) ⎤ ⎥ (18)
⎣ 𝐹𝑡 𝑅𝑇 ⎦
Brought to you by | University of California - Santa Barbara
4 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

𝐹𝐶4 𝐻6 𝑃𝑡 2.17
𝑟3 = 𝑘 3 ⎡
⎢ ( ) ⎤ ⎥ (19)
⎣ 𝐹𝑡 𝑅𝑇 ⎦

Table 3: Rate coefficient of coke forming reaction (Nabavi et al. 2011).


Rate coefficient A(sec-1orlmole-1sec-1) E(kJ/mole)
K9 5/00E+ 14 2/24E+ 02
K10 2/77E+ 09 1/16E+ 02
K11 5/61E+ 18 2/74E+ 02

Likewise, slip mode has been defined as eq. 9 ~ 15.


For simplification some assumption has been applied in simulator that are mentioned in following items:

1. Plug reactor because of high length to diameter ratio


2. Ideal gas because of low pressure, high temperature of gas mixture
3. No temperature, speed and concentration in axial direction because of short coil diameter
4. No thermal resistance in reactor because of high stream’s speed so thermal transfer coefficient
5. Constant fire box temperature

2.2 Basic equations

Since, feed reforming and cracking (mass transfer), thermal energy for cracking (thermal transfer) and even-
tually gas movement (momentum transfer) are provided in length of the coil during the radiation section,
thus, all transfer phenomena equations shall be solved simultaneously. It is worth noting that although trans-
fer phenomena equations just describe ideal condition in the coil, coke settling is inevitable so coke thickness
changes in the run-time, shall be declared and solved with other three phenomena equations (Froment 1992,
2000; Masoumi et al. 2006; Sundaram and Froment 1977; Towfighi, Modarres, and Omidkhah 2004; Sundaram
and Froment).
For simulation and parameter analysis, Δz is provided as following figure, then all transfer phenomena
and coke formation equation can be written for this control volume. Afterward, to the mass and momentum
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

equations shall be expanded as well as simplified for each of the eight materials contained in the model. Thus,
in simultaneous solving in different length steps, the amount of production or consumption of each material
can be obtained and an adequate analysis of the changes in the amount of material with the pressure changes
which is caused by the coke formation or passing through the knees. It is also possible to reduce the amount of
unsatisfactory products by changing each of these materials in the feed and simulating the process. However,
the change in feed is largely unavailable, but the impact of auxiliary feeds, such as the addition of some propane,
can be considered. In this way, the optimal amount of propane in the feed can be obtained in such a way that
the increase in the production of the admixture C3 + does not cause a loss in the amount of ethylene production.

𝑑 𝜋𝑑𝑡2
𝑓𝑖 = . ∑ 𝜗𝑖 .𝑟𝑖 (20)
𝑑𝑧 4

𝑑𝑇 𝜋𝑑𝑡2
∑ 𝑓𝑗.𝐶𝑝 . = 𝑄 (𝑧) .𝜋𝑑𝑡 + ∑ 𝑟 (−Δ𝐻𝑖 ) (21)
𝑗
𝑗 𝑑𝑧 4 𝑖 𝑖

𝑑𝑃𝑡
𝑑
𝑑𝑧
( 𝑀1 ) + 1
𝑀𝑚
( 𝑇1 𝑑𝑇
𝑑𝑧
+ 𝐹𝑟 )
𝑚
= (22)
𝑑𝑧 1
𝑀𝑚 𝑃𝑡
− 𝐺2𝑃𝑅𝑇
𝑡

Where, ʋ is stoichiometric coefficient, r is rate of reaction in mole m−2 s−1 , Cp is heat capacity in J mole-1 K-1, Q
is heat flux in W m−2 , d is coil diameter in m, ΔH is heat of reaction in J kmol−1 , M is average molecular weight
kg mole−1 , Fr is friction factor and finally G is total mass flow rate in Kg m−2 s−1 .

Brought to you by | University of California - Santa Barbara


Authenticated 5
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

The first improvement of this study is providing an equation in cylindrical form for convectional heat trans-
fer Q (z), basic parameter for the proportionate fluid temperature profile that has not been published yet. Also
in this study a thrive has happened that coke thickness changes (tc) has been linked to coil internal diameter
(dt) with appropriate dimension balance rather previous studies (Alizadeh, Towfâighi, and Karimzadeh 2008;
Keyvanloo, Sedighi, and Towfâighi 2012). To shed light on its significance there is no doubt that a junior change
in fire box temperature, transferred thermal energy and therefore fluid temperature profile will revolute. This
will lead the distribution of cracked product composition to undesired or by-products by using Q (z) in the
next step to solve mass transfer equation. Also kwall has been extracted according to coils material. The material
of the first three tubes is centrifugally-cast 25 Cr-35 NiNb and the material of the outlet tube is 35Cr-45NiNb
to allow high tube skin temperatures up to the required 1080°C for the first two tubes of each radiant coil and
1110°C for the last two tubes. This maximizes the run-time between successive decoke operations.

𝑇𝑤𝑎𝑙𝑙 − 𝑇𝑓 𝑙𝑢𝑖𝑑
𝑄 (𝑧) = (23)
𝑟1
𝑘𝑐𝑜𝑘𝑒
𝑙𝑛 𝑟𝑟21 + 𝑟1
𝑘𝑤𝑎𝑙𝑙
𝑙𝑛 𝑟𝑟32

2
𝑘𝑤𝑎𝑙𝑙 = 6.6667 × 10−6 + 0.009𝑇 + 12.0333 (24)

Where, Kwall is convectional heat coefficient in W/m.K, (Belohlav, Zamostny, and Herink, 2003; Gal and Lakatos,
2014; and TechnipRadiant Coil Assemblies)
There are some subsidiary equations that are needed to calculate internal parameters in momentum transfer
equations: (Froment 1992; Niaei et al. 2004; Sadrameli and Green 2005)

Re−0.2
Fr = 0.092 (25)
dt

Re−0.2 𝜖
Fr = 0.092 + (26)
dt 𝜋Rb
Fr is different and varies in the knees, depending on the movement of the fluid in the straight area of the coil.
Where eq. (25) is for straight parts and eq. (26) is for knees:

Λ dt
𝜖 = [0.7 + 0.35 ] [0.051 + 0.19 ] (27)
90∘ Rb
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

∑ 𝐹𝑖
𝑀𝑚 = 𝑀𝑖 (28)
𝐹𝑡

𝐺 = ∑ 𝐹 𝑖 𝑀𝑖 (29)

𝐺𝑑𝑡
𝑅𝑒 = (30)
𝜇𝑚
In order to calculate the viscosity of hydrocarbons at low pressures, Bromley & Wilke provided an equation as
follows (Perry and Green 1997):

8
𝜇𝑖
𝜇𝑚 = ∑ 8 𝑥 (31)
𝑖=1 1 + ∑ 𝑗 = 1 (𝑄𝑖𝑗 𝑥𝑗𝑖 )
𝑗≠𝑖

1/2 1/4 2
𝜇 𝑀𝑗
1 + [( 𝜇𝑖𝑗 ) (𝑀 ) ]
𝑖
𝑄𝑖𝑗 = 1/2
(32)
√8[1 + 𝑀𝑖
𝑀𝑗
]

Brought to you by | University of California - Santa Barbara


6 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

𝐹𝑖
𝑥𝑖 = (33)
𝐹𝑡

Where, µ is mixture viscosity in Cp that for hydrocarbons in low pressures, Bromley & Wilke provided men-
tioned equation, ɛ is elbow parameter, ʌ is elbow parameter and G total mass flow rate (Kg m−2 s−1 ) and x is
mole fraction.
In the case of Ft , it should be noted that this amount is the sum of all components and dilution steam injected
with feed ratio of 0.3 (Technip Operating Manual 2003).

Ft = ∑ Fi + 𝐹𝑠𝑡𝑒𝑎𝑚 (34)

In the simultaneous equations solving for transfer phenomena and coke formation alike, in each segment inner
diameter will restrict by coke forming so it shall be recalculated and set in next step computation.in this study
a thrive has happened that coke thickness changes(tc ) has been linked to coil internal diameter(dt ) with appro-
priate dimension balance rather previous studies(Manafzadeh, Sadrameli, and Towfâighi 2003; Rahimpour et
al. 2013) .moreover, in these equations rc is rate of coking in mg/m2 .s, ρc is coke specific gravity in Kg m−3 , α is
coking factor and r2 is coil internal radius in m.

𝑑tc r 𝛼 × 10−6
= c (35)
dt 𝜌c

dt = 2 ( r2 − tc ) (36)

2.3 Numerical solution of equations

As mentioned in the previous sections, after writing the equations of the mass balance, the balance of energy
and momentum, and simplifying them as far as possible, it is necessary to solve these three sets of equations
simultaneously. By using forward finite difference method, the corresponding equations were discriminated
and compiled by MATLAB software. To determine and compare the formation time of the coke, a time period
selected and divided into equal intervals then the set of equations obtained for the total time period and solved.
In this way, in a longitudinal step( Δz) the transfer phenomena’s equations are solved for the entire length of
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

the coil, and then the amount of coke deposited on the coil wall, in each longitudinal interval is achieved. Then,
by calculating a new effective diameter for the coil, all calculations will be repeated for the next time interval.

3 Result and discussion


In this section all hydrocracking reactor’s simulation outputs will be presented and discussed. The results of
gas analysis and COT and wall temperatures are specified in Tables 4 and 5, respectively.

Table 4: gas analysis.


Gas Analyze ( Wt % )
Component C2H6 C2H4 H2 C3H8 C3H6 CH4 C4H6 C2H2
Inlet 99.33 0.18 0 0.06 0.14 0.12 0.17 0
Outlet 34 54 3.88 0.02 1.2 4.8 1.4 0.7

Table 5: COT and wall temperature.


Time Pass1 Pass2 Pyrometer
Day Outlet Temp (�C) Outlet Temp (�C) Wall Temp (�C)

Brought to you by | University of California - Santa Barbara


Authenticated 7
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

2 839.1 841.2 1034


4 84,036 840.1 1035
6 837.5 839.6 1037
8 838.7 840.8 1039
10 839.3 838.4 1041
12 839.4 841.5 1043
14 840.7 838.8 1045
16 837.2 839.3 1046
18 838.9 839.0 1048
20 838.1 840.2 1050
22 838.5 839.6 1052
24 840.2 842.3 1054
26 839.1 839.2 1056
28 840.4 840.5 1058
30 838.2 838.3 1059
32 839.8 839.9 1061
34 840.2 838.3 1063
36 841.9 839.8 1065
38 839.9 838.0 1067
40 838.4 839.5 1069
42 839.6 840.7 1071
44 840.9 840.0 1072
46 839.4 838.5 1074
48 839.5 841.6 1076
50 838.7 840.8 1078
52 839.1 839.2 1080
54 838.8 838.9 1082
56 840.1 839.2 1083
58 840.2 838.3 1085
60 839.9 839.2 1087

The temperature of cracked gas is always fixed about 840°C, however by increasing in the coke formation;
the fuel gas flow will increase to reach the proper temperature.In the third column of Table 10 actual coils’ wall
temperature has been measured and noted. Technip® has considered 1100°C as maximum coil temperature
according to its material and has called it End of Run (EOR) temperature that will be reached in the case of
coke accumulation. The wall temperatures of the coil will change between 1030 to 1090°C so decoking operation
shall be started before this point. The figure 4 illustrates coke and coil radial sketch. Algorithm for numerical
solution of equations is described in Figure 5.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 6(a) shows that the crack gas temperature is rising sharply at arrival and after preheating up to
the 0.1 reactor’s length. Then it reached to the reaction temperature, and refers to endothermic reactions the
temperature of the fluid raises continuously but with low slope

Brought to you by | University of California - Santa Barbara


8 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

Figure 3: Δ𝑧 segment.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 4: Coke and coil radial sketch.

Brought to you by | University of California - Santa Barbara


Authenticated 9
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

Figure 5: Algorithm for numerical solution of equations.


Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 6: (a) Temperature profile; (b) Pressure profile vs. reactor length for first day of furnace run.

AS the Figure 6(b) is quite clear, the pressure, temperature, and coke formation diagram are perfectly
matched. Regarding the cleanliness of the input coil, the input pressure is 2.2 bar absolute, which is charac-
terized by the development of coke reactions and tensile pressure dropping to reach the outlet pressure. Of
course, due to resident time adjustment in the reactor, the output pressure will be set by compressor speed.
This does not mean that the output pressure changes in different capacities but the compressor by controlling
its anti-surge valves can adjust its inlet pressure, which is the outlet pressure of the furnace, therefore keep
the resident time constant. According to the momentum equation and the involvement of the friction factor,
although it was expected that the pressure diagram in the distance between the knees would be dropped, with
the case study of this equation, it figured out that the second term in the numerator 𝑀1 ( 𝑇1 𝜕𝑇
𝜕𝑧
+ 𝐹𝑟 ), no matter
𝑚
𝜕
in straight length or knees in flat areas of the tube, or in the knees, is much less than the first term 𝜕𝑧 ( 𝑀1 ).
𝑚
For this reason, the pressure diagram will be non-volatile, and consequently the product and the coke graphs,
where the pressure is involved, will not decrease.
Total flow rates of components versus reactor lengths is shown in Figure 7. Figure 7 shows that all compo-
nent, reactants and product distribution will change along the reactor. It is interesting to note that the diagram
of flow vs. reactor length follows different trends. These trends are continuously descending for a ethane and

Brought to you by | University of California - Santa Barbara


10 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

propane. Ethane conversion is low in view of low temperature through the initial 0.1 of reactor length. With
regard to the temperature profile, ethane conversion curve soared gradually from 1068 K to the extent that the
amount of ethane reaches its lowest level and leaves the reactor. In this case, the furnace conversion rate is 65 %
relative to ethane. All these trends are properly predicted by the model. Therefore, it seems that the model is
quite satisfactory and useful.
The resident time in the reactor will be between 0.2 and 0.3 seconds, depending on the compressor suc-
tion. Therefore, the conversion rate will change with respect to the unit capacity and compressor speed, which
straightly effect on resident the time.
The rate of ethylene and hydrogen is low through the initial 0.1of the reactor length, regarding to the low
ethane conversion that is caused by lack of temperature required. After this distance, the slope is approximately
identical to the conversion rate of ethane. The small difference between the slope of the ethylene and hydrogen
production chart is due to the consideration of ethylene as a source of coke in the reactor.
It must be pointed out the produced methane is almost zero through the initial 0.1 of reactor length. As
result, the Propane and Propylene cracking reactions which are the main source for creating Methane didn’t
have any progress or have not been produced sufficiently, or the heat that is needed to crack them and producing
methane is not provided that the third hypothesis is much more likely. Since the propane source is ethane that
is available from the beginning. From this point onwards, methane will produce with raising slope.
Besides, consumption of propylene for coke formation, with decrease in propane in the last 0.2of the reactor
length less Propylene will be produced so this causes a downward slope in consumes.
Given the great similarity between the propylene and acetylene graphs, it can be concluded from equilib-
rium reaction No. 5, which is the source of acetylene production from propylene, acetylene production would
be more than consumption during the initial 0.8 of the reactor length.. Because acetylene is produced up to here
will be used regarding to reaction No. 6.
It would be worth mentioning that the produced 1, 3 Butadiene is only produced according to reaction No.
6, but as shown in the graph, after 0.8 of the reactor length is being consumed. Given that this product is also
one of the coke sources, it is also consuming throughout the reactor, but due to the prevalence of production,
the slope of the graph is ascending. From that point on, the production of this 1, 3 butadiene has dropped
according reaction 6 and since it does not any consumption except coke formation, 1 and 3 butadiene will only
be consumed.
Finally, Propane is produced in reaction No. 2 and is consumed in reactions 3 and 4. As shown in the figure,
as stated in the preference for ethylene and hydrogen graphs, in the initial 0.1 times the reactor, the amount of
propane conversion is not much improved due to the lack of temperature required, but from this point on, the
conversion rate continues with a decreasing slope. It should be noted that the reason for this descending slop is
propane consumption more than its production. It is because of the two main products from ethane cracking,
which are ethylene and hydrogen generation in propane consumption reactions, so propane consumption will
be higher than its production.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 8 shows the thickness of cock at the end of run that would be very useful to predict thickness of coke
according to pressure drop in various days.

Figure 7: Flow rates of components vs. reactor length for first day of furnace run.

Brought to you by | University of California - Santa Barbara


Authenticated 11
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

Figure 8: Coke thickness vs. reactor length for different day of furnace run.

In the cracking process, the amount of coke will increases sharply at the final 0.3 of coil length because of
the fact that the high temperature of the fire box to maintain the required cracking temperature. By estimating
the area under coke formation curves, amount of produced coke can be achieved. For each decoking operation
500–600 Kg coke has been extracted from decoke pot and TLEs actually that can be a criterion for comparison
with simulation output.

3.1 How to calculate the amount of coke produced by the model

The modeling coke formation can be estimated by the calculation of the area under the curve during the sixtieth
days. During the operation, the actual amount of coke formed from the decoke Pot and TLEs is depleted for
each time between 500 and 600 kg, which can be basis to make comparison between actual and modeling to
calculate the error rate. The performed calculations are as follows:

1
∫ 12.577𝑥4 − 13.76𝑥3 7.1286𝑥2 − 0.916𝑥 + 0.0269 = 1.025
0

1.0205 × 0.001 × 55 (𝑙𝑒𝑛𝑔𝑡ℎ) = 0.5613 𝑚2 𝐴𝑟𝑒𝑎 𝑢𝑛𝑑𝑒𝑟 𝑡ℎ𝑒 𝑐𝑢𝑟𝑣𝑒

0.5613 × 3.14 × 0.1 (𝐼.𝐷) = 0.0176 𝑚3 𝑉𝑜𝑙𝑢𝑚𝑒 𝑜𝑓 𝑓 𝑜𝑟𝑚𝑒𝑑 𝑐𝑜𝑘𝑒

0.0176𝑚3 × 1600 𝑘𝑔/𝑚3 (𝑐𝑜𝑘𝑒 𝑑𝑒𝑛𝑠𝑖𝑡𝑦) = 28.198𝑘𝑔 𝑊𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑓 𝑜𝑟𝑚𝑒𝑑 𝑐𝑜𝑘𝑒

3.2 Model validation

The comparison between experimental and predicted values of model for various parameters are summarized
in Table 6. According to data from this table, there is a good agreement between the values of prediction and
experimental parameters in most cases. After modeling and obtaining results and comparing them with actual
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

conditions, it is possible to have accurate study by changing the conditions on the furnace to achieve higher
efficiency, either in terms of purity of products or in terms of the amount of energy needed to perform the entire
cracking operation. And also, all responses indicate tolerable coordination between experiments and estimated
data.

Table 6: The error of model results with actual data.


Parameter Real Modeling Error%
Total coke (kg) 25 28.198 12.8
COT(�C) 842 877 4.2
C2H6 (wt %) 34 32.787 3.6
C2H4 (wt %) 54 54.67 1.2
H2 (wt %) 3.88 4.196 8.1
C3H8 (wt %) 0.02 0.0272 36.0
C3H6 (wt %) 1.2 1.274 6.2
CH4 (wt %) 4.8 5.056 5.3
C4H6 (wt %) 1.4 1.424 1.7
C2H2 (wt %) 0.7 0.563 19.6

Brought to you by | University of California - Santa Barbara


12 Authenticated
Download Date | 7/22/18 8:58 AM
DE GRUYTER Barza et al.

4 Conclusion
In this research, deviation of the model in actual condition can be obtained by actual data and modeling results.
These data are compared in Table (11) briefly. Changing in cracking products can be seen foe instant by adding
propane or reduction in the amount of ethane which is injected into the feed. This sort of changes are expected
during periodic maintenance (overhauls) that usually occurs in the summer for gas refinery as feed supplier in
the Assaluyeh region so as consequence, received ethane by downstream units such olefins decrease. Therefore
significant amount of propane (up to 25 % by weight) is added to the feed. Because there is no C3 + separation
section in the gas cracking unit, and the size of the storage tanks of this cut is limited, so it is difficult to add
uncounted propane in the feed. Another optimization that can be done on the system is to get the best COT
for the furnace. By doing this, the maximum conversion rate of ethane to the desired product in the gas unit
(ethylene) will be achieved. Also, according to the actual amount of coke produced during a runtime, the best
temperature can be obtained to reduce the amount of coke formation and prevent over-cracking of the feed.
In the modeling was assumed that the temperature of the fire box was fixed every day, but to bring the model
closer to reality, a radiant heat transfer profile could be considered for that area, and this make the error between
the assumption and reality clear and it can be concluded whether this assumption is true or not.

4.1 Advantages of the model

In the equation presented for the momentum, depending on the position of the fluid in the coil, two modes of
bending and straight coil are involved that different friction coefficient is chosen for each of them. Due to these
coefficients, the pressure graph must have ups and downs which indicate the effect of fluid passing through the
knees and the straight parts. But due to the type of coil structure and the angles of the knees, with the case study
of this equation, it became clear that the second term in the numerator, no matter in straight length or knees
𝜕
in flat areas of the tube, or in the knees, is much less than the first term 𝜕𝑧 ( 𝑀1 ). For this reason, the pressure
𝑚
diagram will be non-volatile, and consequently the product graphs, which are affected by the pressure, will not
fluctuate.
The valuable work has been done in this study is to calculate the amount of coke and compare it with the
amount of coke formed in the operating state. So far, the information from coke modeling in a furnace Run-
Time with the actual amount of coke has not been announced together for a gas feed unit that can be used to
estimate the model error in coke formation.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Nomenclature

z axial reactor coordinate (m) T temperature (K) r2 coil internal radius (m)
F molar flow rate (mole s−1 ) Q heat flux (W m−2 ) G total mass flow rate (Kg m−2
s−1 )
dt coil diameter (m) ΔH heat of reaction (J kmol−1 ) tc coke thickness (m)
ʋ stoichiometric coefficient P pressure (Pa) rc rate of coking (mg/m2 .s)
r rate of reaction (mole m−2 M molecular weight (kg mole−1 ) ρc coke specific gravity (Kg m−3 )
s−1 )
Cp heat capacity (J mole-1 K-1) Fr friction factor Rb elbow radius (m)
α coking factor R universal gas constant (J r1 coil internal radius with
mole−1 K−1 ) coke(m)
k Rate coefficient(1/S or Re Reynolds number ʌ elbow angel
1/mole.s )
r3 coil external radius(m) x mole fraction ε elbow parameter
α coking factor µ mixture viscosity(Cp)
Subscript
i component t total m average

References
Albright, L.F., and J.C. Marek. 1988. “Coke formation during pyrolysis: roles of residence time, reactor geometry, and time of operation.” In-
dustrial Engineering Chemical Researcher 27 (5): 743–51.

Brought to you by | University of California - Santa Barbara


Authenticated 13
Download Date | 7/22/18 8:58 AM
Barza et al. DE GRUYTER

Belohlav, Zdenek, Petr Zamostny, and Tomas Herink. 2003. “The Kinetic Model of Thermal Cracking for Olefâins Production.” Chemical Engi-
neering and Processing 42: 461–73.
Cai, H., A. Krzywicki, and M.C. Oballa. 2002. “Coke Formation in Steam Crackers for Ethylene Production.” Chemical Engineering and Processing
41: 199–214.
Froment, G.F.. 1992. “Kinetics and Reactor Designer the Thermal Cracking for Olefins Production.” Chemical Engineering Science 47 (9-l i): 2163–
77.
Gal, T., and B.G. Lakatos. 2004. “Modeling and Simulation of Steam Crackers.” European Symposium on Computer-Aided Process Engineering-
14–2004
Ghasemi, H., N. Gilani, and J. Daryan. 2016. “CFD Simulation of Propane Thermal Cracking Furnace and Reactor: A Case Study.” International
Journal of Chemical Reactor Engineering 15 (3) Retrieved December 13, 2017. doi:10.1515/ijcre-2016-0125.
Karamullaoglu, G., and T. Dogu. 2007. “Dynamic and Steady State Analysis of Low Temperature Ethane Oxidative Dehydrogenation over
Chromia and Chromia-Vanadia Catalysts.” International Journal of Chemical Reactor Engineering 5 (1): Retrieved December 13, 2017.
doi:10.2202/1542-6580.1513.
Keyvanloo, K., M. Sedighi, and J. Towfâighi. 2012. “Genetic Algorithm Model Development for Prediction of Main Products in Thermal Crack-
ing of Naphtha: Comparison with Kinetic Modeling.” Chemical Engineering Journal 209: 255–62.
Kopinke, F.D., G. Zimmermann, and S. Nowak. 1988. “On the Mechanism of Coke Formation in Steam Cracking-Conclusions from Results
Obtained by Tracer Experiments.” Carbon 26 (2): 117–124.
Lahaye, J., P. Badie, and J. Ducret. 1977. “Mechanism of carbon formation during steamcracking of hydrocarbons.” Carbon 15: 87.
Mackenzie, Allan l., Philip D. Pacey, and Jayantha H. Wimalase. 1983. “Induction Periods in the Formation of Ethane from the Pyrolysis of
Ethylene.” Canada Journal Chemical 61 (9): 2033–2036.
Manafzadeh, H., S.M. Sadrameli, and J. Towfâighi. 2003. “Coke Deposition by Physical Condensation of Poly-Cyclic Hydrocarbons in the
Transfer Line exchanger(TLX) of Olefâin Plant.” Applied Thermal Engineering 23: 1347–58.
Masoumi, M.E., S.M. Sadrameli, J. Towfâighi, and A. Niaei. 2006. “Simulation, Optimization and Control of a Thermal Cracking Furnace.”
Energy 31: 516–27.
Alizadeh, Mohammad, J. Towfâighi, and R. Karimzadeh. 2008. “Modeling of Catalytic Coke Formation in Thermal Cracking Reactors.” Jour-
nal Analysis Applications Pyrolysis 82: 134–39.
Nabavi, R., Rangaiah, G.P., Niaei, A., Salari, D. 2011 (December) 13. “Design Optimization of an LPG Thermal Cracker for Multiple Objectives.”
International Journal of Chemical Reactor Engineering 9 (1). Retrieved December132017. DOI: 10.1515/1542-6580.2507.
Niaei, A., J. Towfâighi, S.M. Sadrameli, and R. Karimzadeh. 2004. “The Combined Simulation of Heat Transfer and Pyrolysis Reactions in
Industrial Cracking Furnaces.” Applied Thermal Engineering 24: 2251–65.
Norinaga, K., and O. Deutschmann. 2007. “Detailed Kinetic Modeling of Gas-Phase Reactions in the Chemical Vapor Deposition of Carbon
from Light Hydrocarbons.” Industrial Engineering Chemical Researcher 46: 3547–57.
Norinaga, K., O. Deutschmann, N. Saegusa, and J. Hayashi. 2009. “Analysis of Pyrolysis Products from Light Hydrocarbons and Kinetic Mod-
eling for Growth of Polycyclic Aromatic Hydrocarbons with Detailed Chemistry.” Journal Analysis Applications Pyrolysis 86: 148–60.
Norinaga, K., V. Janardhanan, and O. Deutschmann. 2007. “Detailed Chemical Kinetic Modeling of Pyrolysis of Ethylene, Acetylene, and
Propylene at 1073–1373 K with a Plug-Flow Reactor Model.” Wiley Interscience. doi:10.1002/kin.20302.
Perry, R.H., and D.W. Green. 1997. Perry’s Chemical Engineering Handbook„ 411. New York: McGraw-Hill
Rahimpour, M.R., O. Dehghani, M.R. Gholipour, M.S. ShokrollahiYancheshmeh, S.S. Haghighi, and S. Raeissi. 2013. “Modeling of Ethane
Pyrolysis Process: A Study on Effects of Steam and Carbon Dioxide on Ethylene and Hydrogen Productions.” Chemical Engineering Journal
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

215–216: 550–60.
Ramanarao, M. V., P.M. Plehiers, and G.F. Froment. 1988. “The Coupled Simulation of Heat Transfer and Reaction in a Pyrolysis Furnace.”
Chemical Engineering Science 43 (6): 1223–29.
Rase, H.F. 1977. Chemical Reactor Design for Process Plants: Case Study 102,. Vol. 2. New York: John Wiley and Sons
Sadrameli, S.M., and A.E.S. Green. 2005. “Systematics and Modeling Representations of Naphtha Thermal Cracking for Olefâin Production.”
Journal Analysis Applications Pyrolysis 73: 305–13.
Simon, M., and M.H. Back. 1969. “Kinetics of the Thermal Reactions of Ethylene.” Canadian Journal of Chemistry 47: 251.
Sundaram, K.M., and G.F. Froment. 1977. “Modeling of Thermal Cracking kinetics—I: Thermal Cracking of Ethane, Propane and Their Mix-
tures.” Chemical Engineering Sciences 32: 601–08.
Sundaram, K.M., and G.F. Froment. 1980. “Two Dimensional Model for the Simulation of Tubular Reactors for Thermal Cracking.” Chemical
Engineering Science 35: 364–71.
Technip Operating Manual. July 2003. Rev. : 1, Chapter : 3, Page : 6.
Towfighi, J., J. Modarres, and M. Omidkhah. 2004. “Estimation of Kinetic Parameters of Coking Reaction Rate in Pyrolysis of Naphtha.” IJE
Transactions B:Applications 17, no. 4 (December2004): 319.
Zarinabadi, S., E. Ziarifar, M. SadeghMarouf, and A. Samimi. “Modeling and Simulation for Olefin Production in Amir Kabir Petrochemical.”
In Proceedings of the World Congress on Engineering and Computer Science 2010, Vol. II
Zou, R., Q. Lou, S. Mo, and S. Feng. 1993. “Study on a Kinetic Model of Atmospheric Gas Oil Pyrolysis and Coke Deposition.” Industrial Engi-
neering Chemical Researcher 32: 843–47.

Brought to you by | University of California - Santa Barbara


14 Authenticated
Download Date | 7/22/18 8:58 AM

You might also like