Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Article

pubs.acs.org/IECR

Modeling Coil and Soaker Reactors for Visbreaking


Rodolfo A. Aguilar and Jorge Ancheyta*
Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas Norte 152, Col. San Bartolo Atepehuacan C.P. 07730, México D.F.
*
S Supporting Information

ABSTRACT: A reactor model to simulate the visbreaking process is developed. The model includes two types of visbreaking
reactors: coil and soaker. For the vaporization in the reactor to be taken into account, the vapor−liquid equilibrium (VLE) is first
determined, and then the reactor model is solved only with the liquid phase because thermal cracking reactions occurring during
visbreaking take place in this phase. The variation of liquid and vapor flow rates are also taken into consideration. A reaction
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

scheme based on nine pseudocomponents and kinetic expressions reported in the literature were used. It was found that VLE
Downloaded via UNIV INDUSTRIAL DE SANTANDER on May 14, 2019 at 14:58:15 (UTC).

affects the prediction of reactor conversion depending on the feedstock and the operating conditions. The residue conversion
decreases when VLE is considered, which is due to vaporization diminishing the reactor space available for performing the
reactions. The model predicts greater mass fractions of gas and naphtha than the experimental values. This is because the model
assumes that they are produced by all pseudocomponents heavier than them. It is anticipated that the proposed model can be
used in complex refinery planning studies with more reliability than the empirical models available to date.

1. INTRODUCTION formation, enhancing the unit run, and lowering operating


The decreasing supply of light crude oils is a matter of serious cost;2 the selectivity to the cracking of the feed in the coil-
concern for the petroleum industry because the distillation of soaker mode, due to their relatively lower temperature, allows
heavy crude oils leaves behind a significant amount of vacuum only for the cracking of higher molecules and already-cracked
residue. This vacuum residue needs to be converted into light molecules remain unaffected.
distillates to increase the profitability of the entire refinery. To Modeling of coil and soaker visbreaking, which can be
this end, there are various commercial processes based mainly broadly classified as gas−liquid systems,2 has been the subject
on hydrogen addition and carbon rejection routes.1 Visbreaking of various research studies. Mosby et al.,4 using petroleum
(VB), a carbon rejection-based technology, is a mild liquid- fractions (seven lumps) as pseudocomponents, modeled VB
phase thermal cracking process that reduces the viscosity and reactor as a CSTR (continuous stirred tank reactor).
pour point of the residue so that, with little or no addition of Castellanos et al.5 used 51 pseudocomponents to characterize
light distillates, it can meet the fuel oil quality. the feed and modeled the reactor as a two-phase plug-flow
There are two ways to perform the VB reactions: a coil reactor (PFR). Carbonell and Guirardello6 combined the zero-
reactor and a coil-soaker reactor system. The coil visbreaker is a order turbulence model with the Mosby scheme. Mittal et al.7
furnace operated at high temperature and low residence time. It modeled the coil VB as a PFR coupled with CSTR soaker in
is also termed a high-temperature, short-residence-time series but only considering the reaction of vacuum residue to
(HTST) process. The visbreaker feed from the preheater gas and gasoline. Kulkarni8 used a similar approximation with
enters the furnace at 300−330 °C and undergoes thermal more reactions and scale factors in the kinetics coefficients to
cracking in the coil in the temperature range of 460−480 °C for adjust pilot plant data. Reza and Sepehr9 modeled a VB
a duration of 2−5 min. Steam (∼1 wt %) is injected along with industrial reactor assuming the coil as a PFR and the soaker as a
the feed to generate the turbulence, control the liquid-phase CSTR. Filho and Sugaya3 modeled the visbreaking reactors
residence time, and prevent coking along the tube walls.2 The (coil and soaker) considering the VLE (vapor−liquid
cracking reactions (C-C cracking) occur in the liquid phase and equilibrium) using the Peng−Robinson equation of state, but
are slightly endothermic (192 kcal/kg of product boiling below the kinetic parameters were not reported.
204 °C).3 The conversion depends on operation conditions (residence
In the coil-soaker operation, the feed first passes through a time, temperature, etc.). Keeping constant the other conditions,
coil reactor at relatively lower temperature and residence time the longer the residence time, the greater the conversion.
(∼450 °C, 2−3 min). Then, it enters the soaker reactor−an Changing the liquid composition results in the conversion also
adiabatic hollow drum−where it is held for a longer amount of changing. At the conditions of a visbreaking reactor (temper-
time (15−25 min). The reduction in the coil temperature and ature of 450−480 °C, pressure of ∼2.5 MPa) and considering
simultaneous increase in the residence time in the soaker give the formation of light products, the VLE in the system makes
similar conversions as in the case of the coil type, it is described the gas phase increase through the reactor and, as a
as the low-temperature, high-residence-time (LTHT) route.
Some advantages of the coil-soaker process over the coil Received: June 1, 2015
process are that the soaker drum allows the heater to operate at Revised: December 11, 2015
a lower outlet temperature by providing the residence time Accepted: December 15, 2015
required to achieve the desired reaction, decreasing the coke Published: December 15, 2015

© 2015 American Chemical Society 912 DOI: 10.1021/acs.iecr.5b01985


Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

consequence, the amount of liquid decreases, reducing the


residence time and thus affecting the conversion of the reactor.
The pressure in the coil has a significant influence over the
gas formation. At constant pressure, the temperature varies
through the reactor, and the formation of gas is due to the
variation of VLE by temperature and composition originated by
the reaction. If pressure drop through the coil reactor is
considered, it also affects the VLE, increasing the vaporization
of components. If the pressure drop is high, the vaporization
could be excessive.
For modeling the visbreaking reactor including VLE, it is
necessary to account for an adequate characterization of the
feed. For modeling complex mixtures, two main methods have
been used: the lumping approach and detailed molecular
modeling.10 The lumping can be discrete or continuous. In the
discrete lumping approach, the mixture is divided in discrete
pseudocomponents (lumps) based on physical properties. The
continuous approach is based on the description of multi-
compound mixtures using a continuous distribution function.
No matter what approach is used for lumping the feed to a
visbreaking reactor, the conventional cubic EoS is a common
model for hydrocarbon−water systems.10−16 The temperature-
dependent kij of the conventional EoS can be determined using
experimental data17 or estimated with prediction methods, such
as PPR7818 or PR2SK.19
In the petroleum industry, it is a common practice to Figure 1. (a) Reaction scheme for visbreaking of 5 pseudocompo-
perform refinery planning and evaluation of several feedstock. nents,9 including S1, gas; S2, naphtha; S3, light gas oil; S4, heavy gas oil;
To do this, it is necessary to account for reliable models of and S5, residue. (b) Proposed reaction scheme for S1, gas; S2, naphtha;
different processes. The objective of this work is to develop a S3, light gas oil; S4, heavy gas oil; and S5−S9, residues.
visbreaking reactor model to simulate a coil or a coil-soaker
reactor. The model takes into consideration the VLE, the
changes of liquid and gas flow rates during the reaction, and where N is the number of pseudocomponents.
uses kinetic parameters that can be determined from feed Taking into consideration these literature reports and aiming
properties. The proposed model is based on those developed to develop an easy-to-solve model, the residue was divided into
by Kulkarni8 and Joshi et al.2 5 lumps with the following products:
S1: gas
2. MODEL DESCRIPTION S2: naphtha
This section deals with the description of the model that S3: light gas oil
includes the reaction scheme based on 9 pseudocomponents, S4: heavy gas oil
the kinetic model, the calculation of kinetic parameters, and the S5−S9: residues
reactor model (mass balance, energy balance, and VLE). The proposed reaction scheme is then illustrated in Figure
2.1. Reaction Scheme. The reaction scheme of Reza et al.9 1b.
with 5 pseudocomponents (S5: residue, S4: heavy gas oil, S3: 2.2. Kinetic Model. All of the reactions are considered to
light gas oil, S2: naphtha, S1: gas) is shown in Figure 1a, which follow first-order kinetics such that the reaction rate expressions
has been reported in the literature for modeling of visbreaking are
reactors. rij = kijCSi
This model considers the residue as only one pseudocompo- (2)
nent (S5) that can produce all of the other lumps (S1, S2, S3, where i indicates the reactant and j the product of the reaction.
and S4). When solving the mass balance equations, it results in The global rate of reaction for the pseudocomponent i is
higher conversion of residue as compared with experimental i−2 N i−2 N
values. ri = −∑ rij +
Some authors have modeled VLE in the petroleum reservoir
∑ rji = −CSi ∑ kij + ∑ kjiCSj
j=1 j=i+2 j=1 j=i+2 (3)
and have found that some information of the composition is
lost when a few pseudocomponents are used to represent the The first term on the right-hand side of eq 3 represents the
liquid phase.20−22 Joshi et al.2 suggested that the use of narrow reaction rate of the pseudocomponent as reactant, and the
boiling range residue can improve the prediction of conversion second term is the rate of reaction of the pseudocomponent as
when modeling the visbreaking reactor. Castellanos et al.5 product.
proposed to divide the residue feed into 51 pseudocomponents 2.3. Kinetic Parameters. The correlations proposed by
and that the Si reactant produces Sj products through i − 2 Castellanos et al.5 to calculate kinetic parameters from feed
parallel reactions having the products lower molecular weights properties were used.
according to the reaction scheme The temperature dependence of the rate constants is
represented by the Arrhenius equation. Aij and Bij parameters
Si → Sj i = 3, ..., N ; j = 1, ..., N − 2 (1) of the Arrhenius equation are defined in terms of products and
913 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

reactant molecular weights of the reaction i → j, according to where τ = V/ν0 is the space time in h.
the equations Substituting the reaction rate expression (eq 3) in eq 9 gives
Aij = (a0 + a1MWi + a 2 MWi2) ⎡ i−2 N ⎤
CSi0 ⎢
− αCSi = τ CSi(∑ kij) − ∑ kjiCSj⎥⎥
⎡ 2⎤ ⎢⎣
1 ⎛ a4 MWj − MWi ⎞ ⎥ ⎦
exp⎢ − ⎜
j=1 j=i+2 (10)

⎢⎣ 2 ⎝ a3a4 ⎠ ⎥⎦ (4) Dividing Eq. 10 by the total mass concentration, the mass
fraction is obtained
Bij = b0 + b1MWi + b2 MWj (5)
⎡ i−2 N ⎤
The values of the ai and bi parameters are reported in Table fSi0 − αfSi = τ ⎢fSi (∑ kij) − ∑ kjifSj ⎥
1. ⎢⎣ ⎥⎦
j=1 j=i+2 (11)

Table 1. Values of Parameters for Aij and Bij of the Solving for the pseudocomponent mass fraction I at the
Visbreaking Reactions 5 reactor exit
N
a0 1.5072 × 1012 b0 42894 fSi0 + τ(∑ j = i + 2 kjifSj )
a1 1.9 × 108 b1 −4.5 fSi = i−2
a2 2.0561 × 106 b2 3 α + τ(∑ j = 1 kij) (12)
a3 146.95
a4 11.349 b. Energy Balance. The coil reactor is a nonisothermal−
nonadiabatic reactor. The energy balance for a CSTR with
2.4. Reactor Model. Two modes of operation are found in multiple reactions is
commercial visbreaking: coil reactor and coil-soaker reactor. N
For simulating the coil visbreaking reactor, the cell model Q= ∑ ΔHRirV
i + FT CpΔT
proposed by Deans and Lapidus23,24 and McGuire and i=1 (13)
Lapidus25 are used. In this method, the coil reactor is visualized
as a set of interconnected cells in series. Each cell is an element The soaker reactor is considered to be an adiabatic reactor.
of volume of the coil reactor where the mixing is complete, The mass balance is represented by eq 12, and the energy
obtaining uniform properties. Each cell is represented by a balance is
CSTR. With this procedure, the mass and energy conservation N
expressions of the PFR are transformed into a series of 0= ∑ ΔHRirV
i + FT CpΔT
algebraic equations as CSTRs. The soaker drum reactor is an i=1 (14)
adiabatic reactor that can be considered as a CSTR. In
c. Vapor Liquid Equilibrium. Vapor−liquid equilibrium is
summary, both the coil and soaker reactors are modeled as
modeled as an isothermal flash knowing P, T, and feedstock
CSTR: the coil reactor as a series of CSTR and one CSTR for
composition.26
the soaker drum.
The computational procedure is iterative and consists first of
a. Mass Balance. The mass balance in a CSTR for
using the Rachford and Rice equation to calculate the liquid
pseudocomponent i is
composition assuming a vaporized fraction.
v0CSi0 − vCSi = −Vri (6)
N N ⎡ zj ⎤
where CSi0 and CSi are the concentrations of the reactant i in ∑ xj = ∑ ⎢ ⎥
kg/m3 at the inlet and outlet of the CSTR, respectively, ν0 and j=1
⎢ 1 + f (Kj − 1) ⎥⎦
j=1 ⎣ (15)
ν are the volumetric flow rates at the inlet and outlet of the
CSTR, in m3/h, respectively, V is the volume of the CSTR in and the vapor composition is calculated by
m3, and ri is the rate of reaction of the pseudocomponent i in yj = Kjxj (16)
kg/m3h.
During visbreaking reactions, gas is produced, which causes Initial K-values are obtained by using the Wilson equation,
changes to the mass flow of liquid. Thus, the liquid volume for such that
the visbreaking reactions to proceed is altered. To account for
this variation, a liquid volume fraction resulting after the Pci ⎡ ⎛ T ⎞⎤
Ki = exp⎢5.371 + (1 + ωi)⎜1 − ci ⎟⎥
reaction needs to be calculated. The volumetric flow rate at the P ⎣ ⎝ T ⎠⎦ (17)
reactor outlet is estimated with the liquid mass flow rate at the
reactor outlet and the densities (obtained from the VLE This is done until the sum of x’s and y’s is 1.
calculations). The liquid volume fraction is calculated as Once the x’s and y’s are determined, the K-values are
recalculated with the Peng−Robinson equation.27
v at outlet of CSTR The K-value is a function of liquid and vapor fugacity
α=
v0 (7) coefficients, such that
The difference (1 − α) is the gas volume fraction. ΦLj
Using this parameter, the mass balance gives Kj =
ΦGj (18)
v0CSi0 − αv0CSi = −Vri (8)
The fugacity coefficients are computed using the equation of
CSi0 − αCSi = −τri (9) state of Peng−Robinson.
914 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

For consideration of the vaporization in the reactor


modeling, the VLE is first determined, and then the reactor
model is solved only with the liquid phase. The vapor−liquid
product from the reactor is mixed with the vapor stream from
the VLE prior to entering the following VLE, and so forth, as
illustrated in Figure 2.

Figure 3. Modeling approaches for the different cases of study

Figure 2. Strategy for modeling of the visbreaking reactor.

The VLE and reactor model calculations were performed in


Excel spreadsheets. A program developed as a macro in Excel
Visual Basic was used to link the calculations of VLE with the
reactor model. The macro solves the VLE and determines the
temperature in the reactor using the Solver tool of Excel. This
tool uses the nonlineal optimization code GRG2, which is a
general reduced gradient method.28

3. RESULTS AND DISCUSSION


First, it is worth mentioning that experimental and commercial
information of visbreaking mass fractions in the available
literature is scarce, and the few data that have been provided
surely need rectification due to errors during the collection of Figure 4. Temperature and mass fraction profiles without VLE (case
results from plant operation, measuring, and analysis. Despite 1a). (A) Temperature profile. (B) S5−S9 mass fraction profiles. C) S1−
this, the reports are of high value and will be used here for S4 mass fraction profiles.
testing the performance of the developed model.
3.1. Case Studies. In this section, a description of the basis section are 343, 368, and 420 °C. The exit temperature of the
considered to make the simulations is described. Four cases reactor is 446 °C.
were studied, including (Case 2) An industrial plant with two furnaces (coils) in
(Case 1) An industrial plant without soaker drum (coil parallel and an adiabatic soaker (coil-soaker visbreaking).9 The
visbreaking). The feedstock is a 6.55°API vacuum residue feedstock is a 9.16° API blend of vacuum residue and vacuum
obtained by mixing two vacuum residua. The coil reactor gas oil. This coil operates with 9.5 min of residence time and
consists of three sections. The inlet temperatures to each the soaker with 36.4 min. The coil temperatures are 326.5 °C at
915 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Figure 6. Temperature, pressure, and mass fraction profiles with VLE


and pressure drop (case 1c). (A) Temperature and pressure profiles.
Figure 5. Temperature and mass fraction profiles with VLE at constant (B) S5−S9 mass fraction profiles. (C) S1−S4 mass fraction profiles.
pressure (case 1b). (A) Temperature profile. (B) S5−S9 mass fraction
profiles. (C) S1−S4 mass fraction profiles.

moles and mole fractions. With these data for each 1 vol %, the
the inlet and 439 °C at the outlet. In the simulation, one molecular weights of all pseudocomponents that constitute the
furnace is taken to account, and the soaker is simulated with feed and product are obtained.
twice the flow of the furnace. When only the specific gravity (or API gravity) and the
(Case3) A pilot plant consisting of a furnace coil with six temperature range of the cuts are reported, the molecular
independent sections followed by a soaker.29 The feedstock is a weight was calculated using the Kesler and Lee31 correlation
7.24° API vacuum residue from Bombay High crude. The inlet with arithmetic average of temperature and specific gravity.
temperature to the coil is 185 °C, and the outlet temperatures For case 1, the feed is a mixture of two residues. Data of
of each section are 290, 365, 410, 415, 420, and 430 °C. The distillation curves and specific gravity of both residues and the
residence time in the coil is 1.6 min and in the soaker is 20 min. specific gravity of the blend are available. The properties of the
(Case 4) An industrial plant with three parallel furnaces and mixture were calculated by mixing the pseudocomponents of
soakers.30 The feedstock is a 10.44° API vacuum residue. The both residues.
inlet coil temperature is 320 °C, and the outlet temperature is The distillation curves of naphtha (S2) and light gas oil (S3)
436 °C. The residence times are 2.5 min in the coil and 31 min fractions are available. The distillation curve of the heavy gas oil
in the soaker. (S4) fraction was obtained from the distillation curve of
3.2. Calculations of Properties. As mentioned earlier, the visbroken residue (S5−S9) for a cut temperature of 538 °C.
model needs the molecular weight of the feed pseudocompo- For case 2, the distillation curve of the feed is the only one
nents and of the products of the reaction. The molecular reported. The temperature range and specific gravity of the cuts
weights were calculated using the correlation of Kesler and are also reported. The molecular weights of S4 and visbroken
Lee.31 residue were obtained with the distillation curve. The molecular
When the distillation curves were available, they were fitted weights of S2 and S3 were calculated with the specific gravity
with the Kumaraswamy probability distribution function,32 and the mean value of the cut temperature reported.
which according to Sanchez et al.33 is suitable for fitting Case 3 reports the cut temperatures of the products and
distillation data, so that values of fractions for each 1 vol % can overall information of the Bombay High crude-derived residue
be generated. The API gravity of each fraction is calculated feed.29 For calculating the residue distillation curve, that of the
assuming constant KUOP.34 With the API gravity, the specific Bombay High crude was obtained from the literature,35 and
gravity is computed, and then the weight of each fraction is with the pseudocomponents characterization method, the
determined. This weight and the molecular weight of each distillation curves and the molecular weights for each cut
fraction permit calculation of the corresponding number of were determined.
916 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Figure 7. Temperature and mass fraction profiles without VLE (case Figure 8. Temperature and mass fraction profiles with VLE at constant
2a). (A) Temperature profile. (B) S5−S9 mass fraction profiles. (C) pressure (case 2b). (A) Temperature profile. (B) S5−S9 mass fraction
S1−S4 mass fraction profiles. profiles. (C) S1−S4 mass fraction profiles.

Case 4 reports the distillation curves of the feed, S2, S3, and homogeneous model for two phases of flow as reported by
the visbroken residue fractions.23 For determination of the Cooper et al.36
molecular weight of S4, the visbroken residue distillation curve Finally, for each case, the corresponding model of the
was used. reactors system was developed as shown in Figure 3. As was
For the dimensions, operating conditions of the visbreaking mentioned above, each section in a furnace is considered as a
reactors, and the properties of feed and products, refer to Table plug-flow reactor, and it is simulated with a series of 10 CSTRs,
S1 in the Supporting Information. The TBP cuts of the whereas the soaker reactor is calculated as a single CSTR. The
products for case 1 are given as data; for case 2, the cuts S2 and number of CSTRs in series was determined after developing
S3 are also given as data, and S4 and S5 were defined as in case simulations for case 1 without VLE using 5, 10, 15, 20, and 50
1; for case 3, all the data are reported; and in case 4, the TBP reactors in series. The maximum difference in the results
curves of each cut are given as data. Table S2 gives the considering more than 10 CSTRs in series with respect to that
distillation curves of feed and products. with 10 reactors was 0.22 wt % in the S2 pseudocomponent.
Table S3 shows the calculated heat capacities and the This difference is not significant and, hence, using 10 CSTRs in
molecular weights of the lumps involved in the reaction series is reliable.
scheme. The heat of reaction for the products boiling below 3.4. Results and Discussions. For each case (1−4), three
204 °C was assumed to be 800 kJ/kg.3 figures are reported, one corresponding to situation (a),
3.3. Simulations. Once the data for performing the another for situation (b) and the last one for situation (c).
simulations were either collected from reports in the literature Each figure has three other subfigures: (A) for the temperature
or calculated with proper correlations from the available and pressure profiles (pressure profile is reported when
information or, in the worst of the cases, assumed from similar pressure drop is considered), (B) for the mass fraction profile
reported cases, the following situations were simulated: (a) of residue and its pseudocomponents (S5−S9), and (C) for the
without considering VLE at constant pressure, (b) accounting mass fraction profile of pseudocomponents S1−S4. When
for VLE at constant pressure, and (c) with VLE with pressure available, data of experimental mass fractions are included as
drop. symbols.
For all of the cases and situations, the coil temperatures are 3.4.1. Case 1. Figures 4−6 show the results of the
available such that the simulations were developed by simulations for situations 1a−c, respectively.
modifying the duty of each section until the outlet temperatures Figures 4A, 5A, and 6A show the temperature profiles
were reached. Reactor pressure drop is calculated using the through the reactor. As was mentioned earlier, the inlet
917 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Figure 9. Temperature, pressure, and mass fraction profiles with VLE


and pressure drop (case 2c). (A) Temperature and pressure profiles.
(B) S5−S9 mass fraction profiles. (C) S1−S4 mass fraction profiles. Figure 10. Temperature and mass fraction profiles without VLE (case
3a). (A) Temperature profile. (B) S5−S9 mass fraction profiles. (C)
S1−S4 mass fraction profiles.

temperature of the coil was available and used to fit the energy Because of the lower residue conversion predicted, the
balance as to obtain the outlet temperatures, such that the increases of mass fraction of S1, S2, and S3 are smaller: 6.56
temperature profiles are identical. The same behavior occurs for versus 10.93% for S1; 4.79 versus 7.83% for S2, and 2.51 versus
the other cases. When only the liquid phase is considered 3.47% for S3 for situations 1b and 1a, respectively. The mass
(situation 1a), the mass fraction residue pseudocomponents fraction of S4 is practically constant (decrease from 4.64 to
(S5−S9) diminish through the coil as the temperature is 4.51%) due to the same reasons given before.
increased (Figure 4B) with a more pronounced fall of S7, S8, When the pressure drop is considered in the simulation with
and S9. This is due to that fact that pseudocomponents S8 and VLE (situation 1c), the trends of the mass fractions of
S9 are reactants and are not produced in the reaction scheme. pseudocomponents are the same as in the previous cases but
Pseudocomponent S7 is a primary product produced only from with even lower mass residue conversion (11.6%), as can be
S9; however, its rate of appearance is slower than its rate of seen in Figure 6B. Consequently, the production of S1−S4 is
disappearance, which is why it shows a decreasing profile. The also lower (Figure 6C): 5.7 versus 6.56% for S1, 4.15 versus
same behavior is observed for all of the pseudocomponents that 4.79% for S2, and 2.28 versus 2.51% for S3 for situations 1c and
constitute the residue fraction (S5−S9) for all of the cases. The 1b, respectively. The mass fraction of S4 remains the same as in
lighter the pseudocomponent (e.g., S5, S6), the smaller the drop the previous cases.
in its mass fraction. In other words, heavy pseudocomponents 3.4.2. Case 2. Figures 7−9 show the simulation results for
of the residue tend to decrease more of their mass fraction as coil and soaker of this case (situations 2a−c, respectively).
compared with light pseudocomponents. The mass fraction of If VLE is not considered in the simulation (situation 2a), the
S1−S3 increases through the reactor, whereas for S4, it is trends of the pseudocomponents behavior in the coil reactor is
maintained approximately constant between values from 4.51 the same as in case 1 but with less conversion because
to 4.66% (Figure 4C). In the case of S4, its rate of appearance is temperature and residence time are lower (Figure 7B and C).
almost equal to its rate of disappearance, causing a flat profile. The variation of the mass fractions of the S3−S9 pseudocom-
When VLE in the reactor is taken into account at constant ponents in the coil is less than 2.0%. The increase of the mass
pressure (situation 1b), the trends are the same as in situation fractions of pseudocomponents S1 and S2 is greater than 2.5%.
1a without VLE, but the reduction of mass fraction of S5 to S9 In the soaker is where most of the reactions are carried out,
pseudocomponents is lower, as can be seen in Figure 5B. The whereas in the coil, the mass fraction of the residue decreases
total mass residue conversion is lower in situation 1b (13.0%) ∼7% and in the soaker it diminishes ∼15%. In all of the
than in situation 1a (21.4%). pseudocomponents, the trend is similar. Only S4 varies in lesser
918 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Figure 12. Temperature, pressure, and mass fraction profiles with VLE
and pressure drop (case 3c). (A) Temperature and pressure profiles.
Figure 11. Temperature and mass fraction profiles with VLE at (B) S5−S9 mass fraction profiles. (C) S1−S4 mass fraction profiles.
constant pressure (case 3b). (A) Temperature profile. (B) S5−S9 mass
fraction profiles. (C) S1−S4 mass fraction profiles.
in all of the pseudocomponents except for S4, which varies only
quantity (1.06 wt %), verifying the fact that its rate of by 0.15 wt %.
appearance is practically the same as its rate of disappearance. In situation 3b (simulation considering VLE and constant
In the case of the simulation considering VLE and constant pressure), panels B and C in Figure 11 show that the mass
pressure (situation 2b, Figure 8B and C), the variation of the fraction variation of all of the pseudocomponents in the coil is
mass fractions in the coil is lower than in situation 2a. As the lower than 0.26 wt %. In the soaker, the maximum increase in
residence time in the coil is short, the conversion in this reactor the mass fraction is 3.35 wt %.
is also low. Thus, the vaporization produced by increasing the If the pressure drop is considered (situation 3c, Figure 12B
temperature is not too large, and the variation of the volume of and C), the mass fraction varies 0.25 wt % in the coil. In the
liquid in the reactor is not too high, thus causing the difference soaker, this mass fraction increases 3.28 wt %.
in volume of the coil to not be as significant as in case 1. 3.4.4. Case 4. Figures 13−15 show the simulation results of
However, in the soaker a more significant difference is case 4. In Figure 13B, it can be appreciated that the mass
observed. Although the temperature is lower, the residence fractions of the residue pseudocomponents (S5−S9) and the
time is longer, and a higher conversion is obtained. This mass fraction of the total residue are low when the reactor is
conversion affects the VLE because there are more light simulated without VLE in the coil. Most of the conversion of
pseudocomponents that vaporize and decrease the volume for the residue is obtained in the soaker, i.e., the mass fractions of
reaction in the liquid phase. In situation 2c (simulations the total residue decreases ∼15%. It can be observed in Figure
considering VLE and pressure drop), the change in the mass 13C that the mass fraction of product pseudocomponents S1−
fractions in the coil with respect to situation 1b is not S3 increases only in the soaker reactor. The mass fraction of the
significant. In the coil, the pressure drop does not cause a great S4 pseudocomponent remains almost unchanged, as in the
effect on the vaporization because there are not many light previous cases.
pseudocomponents. In the soaker, there is not a significant The mass fractions of the pseudocomponents for the
variation in the conversion. simulation of visbreaking considering VLE at constant pressure
3.4.3. Case 3. The results of the coil and soaker simulations are presented in Figure 14B and C. The trend of the behavior is
for situations 3a−c are shown in Figures 10−12. When the VLE the same as in situation 4a, but the conversion of the residue
is not considered in the simulation (situation 3a), the change in and the increase of the mass fraction of S1−S3 pseudocompo-
the mass fractions of the pseudocomponents in the coil is lesser nents are lower.
than 0.28 wt %, as can be seen in Figure 10B and C. On the It can be seen in Figure 15B and C that including the
other hand, in the soaker, the mass fractions change 3.75 wt % pressure drop with the VLE does not have an effect on the rate
919 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Figure 13. Temperature and mass fraction profiles without VLE (case
Figure 14. Temperature and mass fraction profiles with VLE at
4a). (A) Temperature profile. (B) S5−S9 mass fraction profiles. (C)
constant pressure (case 4b). (A) Temperature profile. (B) S5−S9 mass
S1−S4 mass fraction profiles.
fraction profiles. (C) S1−S4 mass fraction profiles.

From the results of the simulations of case 2 (situations 2a−


of visbreaking reactions, and the mass fraction of all of the c), it can be seen in the reactor simulations that the effect of
pseudocomponents is the same as in situation 4b. taking into account the pressure drop is not significant. For the
3.4.5. Comparison with Experimental Data. Table 2 residue product, the simulations results are 77 wt % without
summarizes the predicted mass fractions of S1−S4 and residue VLE; 82.98 wt % considering VLE and constant pressure, and
(S5−S9) and their comparison with experimental data for the 3 83.81 wt % in the simulation with VLE and a pressure drop
situations (a−c). relative to 93.0 wt % for the industrial plant. For S1−S3, the
It can be observed that situation 1c exhibits superior mass fractions of the simulations are greater than the
performance relative to situations 1a and 1b for the prediction experimental value (for S1, the simulations values ranged
of residue mass fraction seeing as though it is in general closer between 7.95 and 11.3 wt %, whereas the experimental value
to the experimental value. However, the prediction of mass was 1.85 wt %; for S2, the experimental value was 4.75 wt %,
fractions of S1−S3 is poor. Mass fractions of S1 and S2 predicted and the simulations were 6.55−8.74 wt %; and for S3, the
for situation 1c tend to approach the experimental value more simulations gave 2.15−2.96 wt %, compared with the 0.4 wt %
than situations 1a and 1b, whereas the mass fraction of S3 tends experimental value).
to move away. As expected, the vaporization at constant pressure is lower
Comparing the mass conversion of situation 1a (21.4%) with than at variable pressure in case 2. The VLE effect on the
those of situations 1b (13%) and 1c (11.2%), it can be deduced results is not as important as in case 1 because the residence
that, in this case, the inclusion of VLE does have a significant time in the coil is shorter.
effect on the residue conversion to lighter pseudocomponents. With respect to case 3, the values of the simulation results for
The light pseudocomponents and some of the heavy the different effects are similar. For the residue product, the
pseudocomponents are vaporized through the reactor, and simulation without VLE produces 84.50 wt %; considering VLE
the produced vapor diminishes the space in the reactor where at constant pressure, 83.63 and 83.81 wt % is obtained taking
the liquid phase reaction is conducted. into account the pressure drop. For S3 and S4, the mass
The difference between situations 1b and 1c shows that the fractions of the simulation are lower than the experimental ones
effect of pressure drop on the residue conversion is not (the simulation values of S3 were between 1.08 and 1.31 wt %
significant. The pressure drop diminishes the conversion 1.8%. compared with 6.69 wt % experimental value, and 7.04−7.05 wt
This is because, when the pressure decreases, the vaporized % in the simulations versus 8.94 wt % of the experimental value
fraction increases, and there is less space in the reactor for the for S4). The simulations for S1 and S2 overestimate the
liquid. experimental values (between 3.73 and 5.60 wt % versus 1.43
920 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

Table 3. Vaporization Fraction and Pressure Drop at the Exit


of the Coil and Soaker
case 1 case 2 case 3 case 4
constant P
coil 0.557 0.226 0.062 0.062
soaker 0.293 0.065 0.108
variable P
ΔP, kg/cm2 9.14 2.81 2.46 2.67
coil 0.573 0.159 0.070 0.086
soaker 0.213 0.072 0.135

Figure 16. Comparison of simulations for case 2C with variation of


volume (empty bar) and without variation of volume (filled bar).
Figure 15. Temperature, pressure, and mass fraction profiles with VLE
and pressure drop (case 4c). (A) Temperature and pressure profiles.
(B) S5−S9 mass fraction profiles. (C) S1−S4 mass fraction profiles. For case 4, it can be observed that the effect of the pressure
considering VLE is not important. For the residue product, the
simulation without VLE produces 81.58 wt %; considering VLE
at constant pressure, 82.64 and 83.04 wt % is obtained taking
wt % experimental value for S1 and 3.30−3.77 wt % compared into account the pressure drop. For S2 and S1, the mass
with 1.75 wt % for the experimental value for S2). fractions of the simulation are greater than the experimental

Table 2. Comparison of Experimental and Predicted Mass Fractions and Conversion for the Different Case Studies
residue
conversion mass fraction S4 S3 S2 S1
case 1
exp 7.8 91.32a 5.58 1.54 1.56
(a) without VLE 21.4 77.78a 3.47 7.82 10.93
(b) VLE const. P 13 86.14a 2.51 4.79 6.56
(c) VLE var. P 11.2 88.10a 2.25 4.07 5.58
case 2
exp 7 93.00a 0.40 4.75 1.85
(a) without VLE 23 77.00a 2.96 8.74 11.30
(b) VLE const. P 16.83 82.98a 2.26 6.62 7.95
(c) VLE var. P 17.1 83.81a 2.14 6.55 8.29
case 3
exp 18.81 81.19 8.94 6.69 1.75 1.43
(a) without VLE 15.50 84.50 7.05 1.19 3.53 3.73
(b) VLE const. P 16.37 83.63 7.04 1.08 3.30 5.60
(c) VLE var. P 16.19 83.81 7.04 1.31 3.77 4.07
case 4
exp 10.00 90.00a 5.20 2.20 2.60
(a) without VLE 18.42 81.58a 2.55 7.36 8.51
(b) VLE const. P 17.36 82.64a 2.37 7.04 7.95
(c) VLE var. P 16.96 83.04a 2.31 6.12 7.65

a
S4−S9 pseudocomponents.

921 DOI: 10.1021/acs.iecr.5b01985


Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

ones (6.87−7.36 wt % for simulations, compared with 2.2 wt % 3.4.6. On the Effect of Variation of Liquid and Gas Flow
experimental value for S2, and 7.75−8.51 wt % in the Rates. As has been stated before, including the change of
simulations versus 2.6 wt % of the experimental value for S1). volume of the reacting system indeed modifies the flow rates of
The case of S3 is the opposite, being higher than the liquid and gas during the reaction. For this to be verified, the
experimental value (5.2 wt %). The simulation values decrease reactor model was solved by considering or not considering the
when both effects are considered (approximately 2.34−2.55 wt change of volume. The results for case 2C are shown in Figure
%). 16 as an example. It is clearly seen that the inclusion of volume
Finally, from Table 2, the mass fraction of the residue in the reactor model strongly affects the predictions. Not
obtained by simulation is always lower than the experimental considering this change causes the residue mass fraction at the
one. Only for case 1c are all case 3 values comparable to the outlet of the reactor to be approximateyl 6 wt % lower than
taking into account the change of volume.


experimental values with variation in mass fraction less than 4%.
In cases 2 and 4, the residue mass fractions are significantly
lower than those reported experimentally. In case 2, the CONCLUSIONS
calculated mass fractions are between 16 and 9.19% lower than The following conclusions were reached: A reaction scheme
the experimental value. In case 4, the calculated mass fractions with 9 pseudocomponents using kinetic expressions reported in
are found to be between 6.96 and 8.42% lower than the the literature is proposed. The vapor−liquid equilibrium affects
experimental value. the reactor conversion prediction only at high temperatures and
The simulation for products S1 and S2 overpredicts the space-time. The volume of liquid in the reactor decreases when
experimental values in all of the cases. The values of considering VLE. Pressure drop does not have great effect over
pseudocomponent S3 do not follow a trend. In cases 1, 3, the residue conversion. The inclusion of the change of volume
and 4, the experimental value is underpredicted, and for case 2, in the reactor model strongly affects the residue conversion and
it is overpredicted. product yields. For similar situations, the reaction extent in the
When VLE is considered, the residue mass fraction is greater, soaker is greater than that in the coil. The temperature in the
which means that the residue conversion diminishes because soaker is lower than that in the coil, but the residence time is
when the light components are vaporized, the reactor space much larger. The effect of residence time is greater than the
effect of temperature.


decreases, letting less volume for the liquid reactions to occur
(cases 1a and 1b). If pressure drop is considered, the
ASSOCIATED CONTENT
vaporization increases when the pressure decreases, and the
residue conversion is lower (cases 1b and 1c). *
S Supporting Information

The high estimates of mass fractions of S1−S3 are due to the The Supporting Information is available free of charge on the
reaction pathway used for developing the kinetic parameters ACS Publications website at DOI: 10.1021/acs.iecr.5b01985.
reported in Castellanos et al.5 The reaction scheme used for Simulation data for the different cases, distillation curves
parameter estimation involves the cracking and addition of feed and products, and heat capacity and molecular
reactions. In this work, those reactions are not considered. weight of lumps (PDF)
The values of mass fraction of residue in simulations in the
first two cases without considering VLE are significantly lower
than those in which VLE is taken into account. In cases 3 and 4,
■ AUTHOR INFORMATION
Corresponding Author
the effect is minor. *E-mail: jancheyt@imp.
When constant pressure is considered in the simulations, or Notes
there is pressure drop, the differences in the mass fractions are The authors declare no competing financial interest.


not relevant. The maximum difference is 1.27% for case 1.
From the experimental data, considering the products S1, S2, NOMENCLATURE
and S3 in cases 1, 3, and 4, S3 is the product with the greatest Aij = Arrhenius parameter defined in eq 4
mass fraction; S1 and S2 have similar values. In case 2, the order ai = constants of the Arrhenius parameter Aij (i = 0−4)
of mass fractions is S2 > S1 > S3. Bij = Arrhenius parameter defined in eq 5
The vaporization at the exit of the coil and the soaker in each bi = constants of the Arrhenius parameter Bij (i = 0−2)
case is presented in Table 3. As can be seen, the vaporization of CSi = final molar concentration of pseudocomponent i
cases 3 and 4 is almost negligible, and the mass fractions of the CSi0 = initial molar concentration of pseudocomponent i
residua are close. In cases 1 and 2, the difference between Cp = heat capacity (kJ/kg-°C)
considering VLE and not results in a greater residue mass FT = liquid mass flow
fraction due to the vaporization. This is because the space of f Si = mass fraction of pseudocomponent Si (i = 1−5)
the reactor to perform the reaction decreases as the quantity of f Si0 = initial mass fraction of pseudocomponent Si
vapor increases. f = vaporized fraction
Case 1 reports the highest temperature at the coil inlet and Kj = equilibrium constant of pseudocomponent j
outlet. The pilot plant of case 3 has the lower temperature at ΔHRi = reaction enthalpy (kJ/kg)
the entrance. The temperatures in cases 2 and 4 are similar kij = rate constant for the reaction i → j (i,j = 1−5)
(326 and 300 °C at the entrance of the coil, respectively). MWi = molecular weight of pseudocomponent “i”
In addition to the temperature effect, the residence time is N = number of pseudocomponents
also important. Case 4 exhibits the longest residence time in the P = pressure of the system
soaker reactor, where the reaction is mainly performed. The Pci = critical pressure of pseudocomponent I
pilot plant is the case that has the shorter residence time. Q = heat transfer rate from CSTR
Therefore, the residue conversion is the least of the four cases. ri = rate of reaction i (kg/h-m3)
922 DOI: 10.1021/acs.iecr.5b01985
Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

rij = rate reaction of pseudocomponent i to produce Phase Envelope Calculations by Use of the Soave-Redlich-Kwong
pseudocomponent j Equation of State. Ind. Eng. Chem. Process Des. Dev. 1984, 23, 163−170.
Si = pseudocomponent i (i = 1−9) (13) Manafi, H.; Mansoori, G. A.; Ghotbi, S. Phase behavior
S1 = gas prediction of petroleum fluids with minimum characterization data. J.
Pet. Sci. Eng. 1999, 22, 67−93.
S2 = naphtha
(14) de Hemptinne, J. C.; Béhar, E. Thermodynamic Modelling of
S3 = light gas oil Petroleum Fluids. Oil Gas Sci. Technol. 2006, 61, 303−317.
S4 = heavy gas oil (15) Riazi, M. R. Improved Phase Behavior Calculations of Resevoir
S5−S9 = vacuum residue pseudocomponents Fluids Using a New Distribution Model. Sientia Iranica 2003, 10, 341−
T = reaction temperature (°C) 345.
ΔT = temperature difference between inlet and outlet of the (16) Gani, R.; Fredenslund, A. Thermodynamics of Petroleum
CSTR Mixtures Containing Heavy Hydrocarbons: An Expert Tuning System.
Tci = critical temperature of pseudocomponent I Ind. Eng. Chem. Res. 1987, 26, 1304−1312.
V = reactor volume (17) Chen, K. Vapor-liquid equilibrium and its effects on trickle bed
xj = molar fraction of liquid phase in VLE hydrotreating reactors. Canmet ENERGY Leadership in ecoInnova-
yj = molar fraction of gas phase in VLE tion, Natural Resources Canada. One Oil Patch Drive, Devon, AB
T9G 1A8, Canada, Seminar Series in Tianjin University, October 20,
zj = feed molar fraction composition
2010.
Greek Letters (18) Jaubert, J. N.; Privat, R.; Mutelet, F. Predicting the Phase
α = relation between volumetric flow rate at inlet and outlet, Equilibria of Synthetic Petroleum Fluids with the PPR78 Approach.
eq 7 AIChE J. 2010, 56, 3225−3235.
ν0 = volumetric flow rate at the inlet of the CSTR (m3/h) (19) Jaubert, J. N.; Privat, R. Relationship between the binary
ν = volumetric flow rate at the outlet of the CSTR (m3/h) interaction parameters (kij) of the Peng−Robinson and those of the
Soave−Redlich−Kwong equations of state: Application to the
ρ = density of reaction mixture (kg/m3)
definition of the PR2SRK model. Fluid Phase Equilib. 2010, 295,
τ = residence time (h) 26−37.
ΦLj = fugacity coefficient of pseudocomponent i in liquid (20) Li, Z., RERI, Firoozabadi, A. General strategy for stability testing
phase and phase-split calculations in two and three phase. Presented at the
ΦGj = fugacity coefficient of pseudocomponent i in gas phase 2010 SPE improved oil recovery Symposium. Tulsa, Oklahoma, USA,
ωi = acentric factor of pseudocomponent i April 24−28, 2010.


(21) Mohebbinia, S.; Sepehrnoori, K.; Russell, T.-J. Four-phase
REFERENCES equilibrium calculations of CO2/hydrocarbon/water systems using a
reduced method. Presented at the 18th SPE improved oil recovery
(1) Rana, M. S.; Samano, V.; Ancheyta, J. A review of recent advances Symposium. Tulsa, Oklahoma, USA, April 14−18, 2012.
on process technologies for upgrading of heavy oils and residua. Fuel (22) Firoozabadi, A.; Pan, H. Fast and robust algorithm for
2007, 86, 1216−1231. compositional modeling: part I − stability analysis testing. SPE J.
(2) Joshi, J.; Pandit, B. A. B.; Kataria, K. L.; Kulkarni, R. P.; Sawarkar, 2002, 78−89.
A. N.; Tandon, D.; Ram, Y.; Kumar, M. M. Petroleum residue (23) Deans, H. A.; Lapidus, L. A computational model for predicting
upgradation via visbreaking: A review. Ind. Eng. Chem. Res. 2008, 47, and correlating the behavior of fixed-bed reactors: I. Derivation of
8960−8988. model for nonreactive systems. AIChE J. 1960, 6, 657−663.
(3) Filho, R. M.; Sugaya, M. F. A computer aided tool for heavy oil (24) Deans, H. A.; Lapidus, L. A computational model for predicting
thermal cracking process simulation. Comp. Comput. Chem. Eng. 2001, and correlating the behavior of fixed-bed reactors: II. Extension to
25, 683−692. chemically reactive systems. AIChE J. 1960, 6, 663−668.
(4) Mosby, J. F.; Buttke, R. D.; Cox, J. A.; Nokolaides, C. Process (25) McGuire, M. L.; Lapidus, L. On the stability of a detailed
characterization of expanded-bed reactors in series. Chem. Eng. Sci. packed-bed reactor. AIChE J. 1965, 11, 85−95.
1986, 41, 989−995. (26) Seader, J. D.; Henley, E. J.; Roper, D. K. Separation process
(5) Castellanos, J.; Cano, J. L.; del Rosal, R.; Briones, V. M.; Mancilla, principles. Chemical and biochemical operations, 3rd ed.; John Wiley &
R. L. Kinetic model predicts visbreaker mass fractions. Oil Gas J. 1991, Sons, Inc., 2011
11, 76−82. (27) McCain, W. D., Jr. The properties of petroleum fluids, 2nd ed.;
(6) Carbonell, M. M.; Guirardello, R. Modelling of a slurry bubble Penn Well Publishing Company: Tulsa, OK, 1990
column reactor applied to the hydroconversionof heavy oils. Chem. (28) Lasdon, L. S.; Waren, A. D.; Jain, A. Ratner, M. Design and
Eng. Sci. 1997, 52, 4179−4185. testing of a GRG code for nonlinear optimization. ACM Trans. Math.
(7) Mittal, S., Maiti, R. N., Lahiri, R. N., Ram-Babu, D. Modeling and Software 1978, 4, 34.
simulation of visbreaker. In Proceedings of Third International Petroleum (29) Kulkarni, R. P.; Pandit, A. B.; Joshi, J. B.; Kataria, K. L.; Tandon,
Conference and Exploration; PETROTECH-99: New Delhi, India, D.; Kumar, M. M. Visbreaking studies in the presence of soaker
1999; p 349. internals. Ind. Eng. Chem. Res. 2010, 49, 11221−11231.
(8) Kulkarni, R. P. Modelling of multiphase reactors: visbreaking. (30) Stratiev, D.; Nikolaev, N. Dependence of visbreaker residue
Ph.D. Thesis, University of Mumbai, Mumbai, India, 2005. properties on unit operation severity and the residual fuel oil
(9) Reza, S. M.; Sepehr, S. Simulation and kinetic modeling of specificacion. Pet. Coal 2009, 51, 140−145.
vacuum residue soaker-visbreaking. Pet. Coal 2011, 53, 26−34. (31) Kesler, M. G.; Lee, B. I. Improve prediction of enthalpy of
(10) Petitfrere, M.; Nichita, D. V.; Montel, F. Multiphase equilibrium fractions. Hydrocarbon Process. 1969, 55, 153−158.
calculations using the semi-continuous thermodynamics of hydro- (32) Kumaraswamy, P. A generalized probability density function for
carbon mixtures. Fluid Phase Equilib. 2014, 362, 365−378. double-bounded random processes. J. Hydrol. 1980, 46, 79−88.
(11) Liang, X.; Yan, W.; Thomsen, K.; Kontogeorgis, G. M. Modeling (33) Sanchez, S.; Ancheyta, J.; McCaffrey, W. C. Comparison of
the liquid−liquid equilibrium of petroleum fluid and polar compounds probability distribution functions for fitting distillation curves of
containing systems with the PC-SAFT equation of state. Fluid Phase petroleum. Energy Fuels 2007, 21, 2955−2963.
Equilib. 2015, 406, 147−155. (34) Hariu, O. H.; Sage, R. C. Crude split figured by computer.
(12) Pedersen, K. S.; Thomassen, P.; Fredenslund, A. Thermody- Hydrocarbon Process. 1969, 48, 143−148.
namics of Petroleum Mixtures Containing Heavy Hydrocarbons. 1. (35) Aalund, L. R. New Assay, Bombay high. Oil Gas J. 1983, 78.

923 DOI: 10.1021/acs.iecr.5b01985


Ind. Eng. Chem. Res. 2016, 55, 912−924
Industrial & Engineering Chemistry Research Article

(36) Cooper, J. R.; Penney, W. R.; Fair, J. R.; Walas, S. M. Chemical


Process Equipment Selection and Design, 2nd ed.; Elsevier Inc., 2005.

924 DOI: 10.1021/acs.iecr.5b01985


Ind. Eng. Chem. Res. 2016, 55, 912−924

You might also like