Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Fuel 235 (2019) 113–129

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Viscosity modeling of reservoir fluids using the Friction Theory with T


PC-SAFT crude oil characterization

Mohammed I.L. Abutaqiya, Jieyi Zhang, Francisco M. Vargas
Department of Chemical and Biomolecular Engineering, Rice University, Houston, TX 77005, USA

A R T I C LE I N FO A B S T R A C T

Keywords: The Friction Theory (FT) for viscosity modeling is applied to a characterization method based on the Saturates-
PC-SAFT Aromatics-Resins-Asphaltenes content (SARA) which makes use of the Perturbed-Chain version of the Statistical
Peng-Robinson Associating Fluid Theory equation of state (PC-SAFT EoS). The viscosity modeling results are compared to Chi-
Friction Theory Square (CS) characterization method that uses Peng-Robinson (PR) EoS. In the Friction Theory, viscosity is
Live oil viscosity
calculated using repulsive and attractive pressures from an EoS. In this work, the FT parameters for pseudo-
Crude oil characterization
components are obtained by tuning to a single value of viscosity at saturation condition. A new fitting approach
for PC-SAFT FT model is proposed where the number of fitting parameters for each pseudo-component is re-
duced from two to one. The simulation results for 10 light crude oils from the Middle East region show that using
PC-SAFT EoS to represent the phase behavior of the crude oil provides viscosity predictions that are at least as
good as the CS method that uses PR EoS, even though a lower number of tuning parameters and pseudo-fractions
are used. However, the use of PC-SAFT to calculate repulsive and attractive pressures in the FT model does not
significantly improve the accuracy in viscosity modeling as compared to the use of PR EoS.

1. Introduction theoretical description of the liquid phase viscosity is arduous. Because


of these challenges, viscosity of liquid is modeled with either empirical
The determination of the viscosity of hydrocarbon fluids is needed or semi-empirical models. Empirical models [7–9] for modeling visc-
to simulate the flow dynamics inside pipes and reservoirs, which is of osity of hydrocarbons have not found much popularity in the industry
utmost importance for the development of strategies for hydrocarbon because of the limited applicability in terms of temperature and pres-
production, transportation, and processing. Throughout these processes sure range. Semi-empirical models [10–13] generally perform better
the reservoir fluids experience temperature and pressure changes that but still require a proper crude oil characterization method that accu-
can even induce phase separation and alter composition. The viscosity rately captures the phase behavior and chemical structure.
of reservoir fluids can change multiple orders of magnitude due to The Friction Theory (FT) for viscosity modeling [14,15] is a rela-
temperature, pressure, and composition variations. tively recent semi-empirical model that has gained a wide popularity in
Several models for viscosity have been developed on empirical, the oil industry. The viscosity of a fluid is modeled as a sum of kine-
semi-empirical, or theoretical bases. Generally, prediction of viscosity matic and mechanical contributions. The theory relates the friction
for crude oils is highly challenging due to the complex structure of viscosity to repulsive and attractive pressures, which can be calculated
reservoir fluids. Many authors have reviewed the viscosity models in from an equation of state. The Friction Theory has been widely used for
the literature [1–5]. Mehortra et al. [2] and Vesovic et al. [3] have accurate modeling of pure component viscosities [14,16,17]. Further-
discussed in depth models that are applicable to hydrocarbon systems. more, it has proven successful in modeling different hydrocarbon
Most viscosity models are applicable to either gas or liquid phases mixtures such as natural gases [18,19], natural gas liquid [20], light
within a specific range of temperature and pressure. Models for gases crude oils [13,20,21] heavy oils [22–24] and CO2 + hydrocarbons
are more developed mainly because the momentum transfer in a dilute [25]. While the viscosity modeling of pure components and their mix-
gas is dominated by random collisions between molecules, which can tures can be purely predictive, the crude oil modeling generally re-
be described by the kinetic theory and Chapman-Enskog theory accu- quires the tuning of Friction Theory parameters to available experi-
rately [2,6]. On the other hand, since the momentum transfer me- mental data.
chanisms involved in liquids are far more complex than for gases, a As a compositional model, the implementation of the Friction


Corresponding author.
E-mail addresses: abutaqiya@rice.edu (M.I.L. Abutaqiya), jieyi.zhang@rice.edu (J. Zhang), fvargas@rice.edu (F.M. Vargas).

https://doi.org/10.1016/j.fuel.2018.06.062
Received 28 March 2018; Received in revised form 9 June 2018; Accepted 15 June 2018
0016-2361/ © 2018 Elsevier Ltd. All rights reserved.
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Nomenclature Kz Compressibility correction factor in PR FT


m number of segments per molecule
Acronyms Pa Attractive pressure
Pr Repulsive pressure
A+R Aromatics + Resins Rm Molar refractivity
A + R+A Aromatics + Resins + Asphaltenes TB Normal boiling point
AAPD Average Absolute Percent Deviation TBr Reduced normal boiling point
CS (p) Chi Square distribution with p degrees of freedom zi Mole fraction of component i
EoS Equation of State
MW Molecular Weight Greek symbols
PC-SAFT Perturbed-Chain Statistical Associating Fluid Theory
PNA Polynuclear Aromatics ε/k Segment-segment interaction energy
PR Peng Robinson I Gamma distribution function
PVT Pressure Volume Temperature γ Aromaticity factor
SARA Saturates, Aromatics, Resins, Asphaltenes ηa Attractive viscosity term in PC-SAFT FT
SLF Single Liquid Fraction ηc Critical viscosity parameters in PR FT
ηf Friction viscosity contribution
Symbols ηo Dilute gas viscosity contribution
ηr Repulsive viscosity term in PC-SAFT FT
c Peneloux correction ρ20 Density at 20 °C
fc Scaling parameter for critical pressure σ Segment diameter
KcPR Scaling parameter for critical viscosity in PR FT ζ Volume shift implemented into repulsive and attractive
KcPC-SAFT Scaling parameter for critical viscosity in PC-SAFT FT pressures

Theory for simulating crude oil viscosity requires the use of a char- pressure and high temperature viscosity data of crude oils from the
acterization procedure. Several crude oil characterization methods are Middle East region. In addition, viscosity of oil is an important factor in
available in the literature such as: Whitson method [26], Pedersen the modeling of flow dynamics and asphaltene deposition in pipelines.
method [27], Chi-Square method [20], and SARA-based method The SARA-based method using PC-SAFT EoS has proven reliable for
[28–30]. The Whitson, Pedersen, and Chi-square methods use cubic modeling asphaltene precipitation. Therefore, a methodology for cou-
equations of state to represent the pressure-volume-temperature rela- pling the SARA-based method with a viscosity model is of great im-
tion of crude oils. The SARA-based method, originally developed by portance for flow assurance applications. Such a model can be used to
Ting et al. [28], uses the Perturbed-Chain form of the Statistical Asso- capture the effect of pressure, temperature, and composition on oil
ciating Fluid Theory (PC-SAFT) equation of state. It has been shown viscosity in asphaltene deposition simulators [34–36] that implement
that the SARA-based characterization method can accurately model the SARA-based method. These deposition models currently assume
asphaltene precipitation and PVT properties of crude oils [28–32]. constant oil viscosity in solving the flow equations.
Furthermore, Abutaqiya et al. [33] proposed a simplification to the
SARA-based method called the Single Liquid Fraction (SLF) that does 2. Background on the implemented modeling methods
not require SARA analysis as an input. The authors showed that the SLF
model yields good results for modeling a wide range of PVT experi- This section provides a review for all the models and equations used
ments for oils from the Middle East region and concluded that the SLF in this work. First, a description of the characterization methods is
method can be used in case SARA analysis is not readily available. provided for the SARA-based, SLF, and Chi-Square methods along with
Previous work using Friction Theory mainly investigated viscosity a description of the most common technique to assign EOS parameters
modeling of crude oil using Chi-Square characterization method with for the pseudo-components in each characterization method. Then, a
PR EoS [13,20–24]. The main purpose of this work is to formulate a brief description of the Friction Theory using PR and PC-SAFT EoS is
procedure for the application of the Friction Theory to the SARA-based presented. Finally, important equations used for estimating the critical
method (which uses PC-SAFT EoS) for the modeling of high-tempera- properties for the cubic EoS are discussed.
ture and high-pressure viscosity of live oil and compare the results to
the Chi-Square method. In case SARA analysis is not available for the
crude of interest, the SLF method is used. Furthermore, this work in- 2.1. Crude oil characterization and EOS parameter estimation methods
vestigates the use of PC-SAFT and PR EoS to calculate the repulsive and
attractive pressures in the Friction Theory model. The work aims to 2.1.1. SARA-based characterization method using PC-SAFT EoS
In the SARA-based method [28–32] the stock tank oil is character-
compare three scenarios in modeling the viscosity of crude oils at high-
pressure and high-temperature: ized based on the content of saturates, aromatics, resins, and asphal-
tenes. The model uses the perturbed chain version of the Statistical
1- SARA-based characterization method with PC-SAFT Friction Theory Associating Fluid Theory EoS (PC-SAFT) [37]. PC-SAFT is a variation of
the original SAFT EoS developed by Chapman et al. [38,39] and has
2- SARA-based characterization method with PR Friction Theory
3- Chi-Square characterization method with PR Friction Theory proven to accurately model complex systems such as polymers and
asphaltenes [28–32,40,41].
The experimental data used for the comparison include liquid In the SARA-based characterization method, the gas and liquid
viscosity of 10 light crude oils from the Middle East region. The study phases are characterized separately and then combined based on the
aims to investigate which scenario would best reproduce the experi- zero-flash gas-to-oil ratio (GOR) to form the live oil composition. The
mental data after fitting the Friction Theory simulation parameters to a gas phase is represented by as seven components, namely: Nitrogen
single value of viscosity at saturation. This investigation helps in de- (N2), Carbon Dioxide (CO2), Hydrogen Sulfide (H2S), Methane (C1),
ciding which scenario is most suitable for quality-control of high Ethane (C2), Propane (C3), and a “Heavy Gas” pseudo-component (C4+
fractions in the flashed gas). The liquid phase is characterized by three

114
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

pseudo-components: Saturates, Aromatics + Resins (A + R), and where Γ represents the gamma function [26,45] and s can be inter-
Asphaltenes. preted as a molecular-weight-scaled variable that can have values equal
In order to use the PC-SAFT EoS, three parameters are needed for or greater than zero. The mass fraction corresponded to s range from
each non-associating component, namely: the number of segments per zero to s0 is regarded as the light-component mass fraction up to the C6
molecule (m), the segment diameter (σ), and the segment-segment in- fraction (M6 ), which should satisfy Eq. (5).
teraction energy (ε/k). The three simulation parameters for pure com- s0
ponents are adapted from the work of Gross and Sadowski [37]. For the ∫ fdis ds = M6
pseudo-components, the correlations from Gonzalez et al. [42] form the 0 (5)
basis for the parameter estimation technique. Gonzalez et al. [42]
For each component i in the heavy fraction, the mass fraction ( fmi )
correlated the PC-SAFT parameters to molecular weight for different
is represented by the area under CS ( p ) function between si and si − 1 that
homologous series of hydrocarbons such as alkanes, benzene deriva-
satisfies Eq. (6).
tives, and polynuclear aromatics (PNA). Punnapala and Vargas [30]
si
defined an aromaticity factor γ that ranges from 0 to 1, where γ = 0
represents n-alkanes and γ = 1 represents polynuclear aromatics (PNA).
fmi = ∫ fdis ds
si − 1 (6)
The final expressions for the PC-SAFT parameters for the pseudo-com-
ponents are [30]: and the molecular weight (MWi ) is calculated using Eq. (7).

m = (1−γA + R )(0.0257MW + 0.8444) + γA + R (0.101MW + 1.7296) (1) MWi = MW si ̂ (7)


where the molecular weight scaling parameter MW is calculated using
ln(MW ) ⎞ 93.98 Eq. (8) and the center of mass si ̂ is calculated using Eq. (9).
σ (Å) = (1−γA + R ) ⎛4.047−4.8013 + γA + R ⎛4.6169− ⎞
⎝ MW ⎠ ⎝ MW ⎠ (2) m
MW + fmi
9.523 ⎞ 234100
MW =
(1−M6)
∑ si ̂
ε / k (K ) = (1−γA + R ) ⎛exp ⎛5.5769− ⎞ + γA + R ⎛508− ⎞ i=1 (8)
⎝ ⎝ MW ⎠ ⎠ ⎝ MW 1.5 ⎠ (3)
si
1
The aromaticity of “Heavy Gas” and “Saturates” pseudo-fractions is si ̂ =
fmi
∫ sfdis ds
assumed to be that of n-alkanes (γ = 0) . The aromaticity of the “A + R” si − 1 (9)
pseudo-fraction (γA + R ) is regressed to match the bubble pressure and where MW +
is the molecular weight of the C7+ fraction, m is the total
the live oil density at saturation. The molecular weight for the “A + R” number of components reported for the liquid fraction.
and “Saturates” fractions are chosen such that the average molecular To find the best CS ( p ) function that describes the heavy fraction
weight of the STO is satisfied. For the “Asphaltene” pseudo-fraction, the mass distribution, the degree of freedom p is optimized to match the
aromaticity and the molecular weight (MWasphal ) are fit to match ex- molecular weight of the first few fractions as they are typically mea-
perimental data for asphaltene onset of precipitation. sured with higher accuracy. The average molecular weight of the heavy
fraction reported from experiment can be matched by adjusting the
2.1.2. Single Liquid Fraction (SLF) characterization method using PC-SAFT molecular weight of the last fraction. Any number of pseudo fractions
Abutaqiya et al. [33] proposed the SLF method to eliminate the necessary for the modeling calculations can be generated from the
need of SARA analysis as an input to the model since SARA may be optimized CS ( p ) function through Eqs. (7)–(9), where equal mass
prone to experimental uncertainty and is not readily available. The SLF fraction fmi can be set for each pseudo-fraction and molecular weights
differs from the SARA-analysis only in the liquid phase lumping pro- can then be calculated.
cedure. The liquid phase in the SLF method is characterized as a single
pseudo-component called the Single Liquid Fraction “SLF” instead of 2.2. The Friction Theory for viscosity modeling
three pseudo-components; “Saturates”, “A + R”, and “Asphaltene”. The
gas phase lumping procedure and the parameter estimation method The Friction Theory was proposed by Quiñones-Cisneros and cow-
described in the previous section are the same. The aromaticity para- orkers who approached viscosity as a mechanical property instead of a
meter for the “SLF” pseudo-fraction (γSLF ) is tuned to match experi- transport property [14,15]. The friction theory approach separates
mental bubble pressure and density of reservoir fluid at saturation. viscosity η into two parts as expressed in Eq. (10).
Despite its simplicity, the model has shown good match to experimental η = η0 + ηf (10)
data from a wide range of PVT experiments for light crude oils from the
Middle-East [33]. where η0 represents the dilute gas viscosity contribution as a results of
collisions of molecules in dilute gas state; the friction viscosity con-
2.1.3. Chi-Square characterization method using cubic EoS tribution ηf represents the energy dissipation associated to motion due
Quiñones-Cisneros and coworkers proposed using the Chi-Square to external stress (analogous to the Amontons-Coulomb friction law).
(CS) characterization of crude oil to describe the molar mass distribu- The dilute viscosity term η0 can be calculated from the model by Chung
tion of the heavy fractions (C7+) [20,22,23,43]. In essence, the CS et al. [46] which is based on Chapman-Enskog theory. The friction
characterization procedure first finds the best CS distribution with p viscosity term ηf is correlated to the repulsive and attractive pressure
degrees of freedom (CS( p )) that describes the C7+ molar mass dis- terms [14,15]. Therefore, different Friction Theory models may be
tribution, then generate any desired number of pseudo fractions from constructed depending on the EoS that is used to calculate the repulsive
the CS( p ) function for modeling work. The ability to conveniently and attractive pressure terms. Two Friction Theory models are dis-
generate any number of pseudo fractions with equal mass fraction is an cussed in Sections 2.2.1 and 2.2.2: Peng-Robinson Friction Theory (PR
important advantage of this procedure, since weight-based procedure FT), and PC-SAFT Friction Theory (PC-SAFT FT).
gives all hydrocarbon fractions in the C7+ equal importance [44].
In general, a Chi-Square probability distribution function fdis with p 2.2.1. Peng-Robinson Friction Theory model
degrees of freedom has the form of Eq. (4). For FT models that are based on cubic EoSs, ηf has the general form
p
of Eq. (11) [15]:
2− 2 p
e− 2 s ( 2 )
s
fdis =
−1 ηf
= ηf ̂ = kr Pr + ka Pa + krr Pr2
Γ ()
p
2 (4) ηc (11)

115
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

where ηc is the characteristic critical viscosity, ηf ̂ is the dimensionless temperature and critical pressure based on the One-Third Rule pro-
friction viscosity contribution, Pr is the repulsive pressure, Pa is the at- posed by Vargas et al. [51]. The One-Third Rule was derived based on
tractive pressure, k is the viscous friction coefficient whose subscript r the observation that the ratio of molar refractivity Rm (cm3/mol) to the
means repulsive, subscript a means attractive, and rr means second MW of a given component is approximately equal to one-third for a
order repulsive. In the case of PR FT, the repulsive and attractive wide range of hydrocarbons at different temperatures and pressures.
pressure can be readily obtained from the pressure expression of PR EoS This correlation is described by Eq. (16).
[47]. The viscous friction coefficients kr , ka , krr are functions of tem- 2
perature, critical properties, and 16 universal constants that are de- ⎛ n −1 ⎞ 1 = Rm ≈ 1
⎜ ⎟
2
pendent on the cubic EoS used. For PR FT, the universal constants can ⎝n + 2⎠ ρ MW 3 (16)
be found in ref [15]. where n is the sodium D-line refractive index and ρ is the mass density
The characteristic critical viscosity ηc for CO2, N2, and n-alkanes up (g/cm3). Panuganti et al. [50] extended one-third rule to predict critical
to n-octadecane are reported in Ref. [15]. Quiñones-Cisneros et al. [13] temperature Tc and critical pressure Pc from easily measurable proper-
suggested estimating the characteristic critical viscosity ηc for pseudo- ties such as the normal boiling temperature TB , density ρ , and molecular
components from a modified correlation from Uyehara and Watson weight MW . Using fundamental thermodynamic property relations and
[48] as shown in Eq. (12). the one-third rule to relate the Van Der Waals attractive term a to TB ,
MW, and ρ20 , Panuganti et al. [50] derived the following equations for
MWi Pc2/3
,i
ηc, i = K c Tc and Pc :
Tc1/6
,i (12)
2
where the constant K c is 7.95 for n-alkanes. Quiñones-Cisneros et al. (
⎛ 0.613 MW +
Tc = TB ⎜
0.1674MW
ρ20 ) + 24.85 ⎞⎟
[13] fitted the K c value for all the pseudo-fractions generated by the
Chi-Square distribution to match experimental values of viscosity of ⎝
(
⎜ 0.577 MW + 0.1674MW
ρ20 ) + 11.12 ⎟⎠ (17)
live oil above bubble pressure.
2

2.2.2. PC-SAFT Friction Theory


(
⎛0.613 MW +
Pc = TB2 ⎝
0.1674MW
ρ20 ) + 24.85⎞⎠
4
Quiñones-Cisneros et al. proposed the PC-SAFT FT model that aimed
to improve compressibility prediction at high pressure [49]. The PC-
(
⎛0.577 MW +

0.1674MW
ρ20 ) + 11.12⎞⎠ (18)
SAFT FT is a quadratic model that has the form shown in Eq. (13). where temperature is in units of K, pressure in atm, and density in g/
2 2 cm3. ρ20 is the density measured at 20 °C.
∼ P ∼ P ∼ P ∼ P
ηf = ηa ⎛⎜ka ⎛ a ⎞ + kaa ⎛ a ⎞ ⎞⎟ + ηr ⎛⎜kr ⎛ r ⎞ + +krr ⎛ r ⎞ ⎞⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟

⎝ ⎝ ⎠ Pc P
⎝ ⎠⎠
c ⎝ ⎝ ⎠ Pc ⎝ Pc ⎠ ⎠ (13) 2.3.2. Lee-Kesler correlations for ω
where ηa and ηr are the scaling parameters for the attractive and re- Kesler and Lee [52] proposed that the acentric factor can be esti-
∼ ∼ ∼ ∼ mated from critical temperature and critical pressure using Eq. (19):
pulsive term, respectively, and ka , kaa , kr , and krr are the reduced fric-
tion coefficients. The scaling parameters for pure components and the s
ln(PBr )−5.92714 +
6.09649 6
+ 1.28862ln(TBr )−0.169347TBr
TBr
38 universal constants for the model can be found in ref [49]. Quiñones- ω= 15.6875 6
Cisneros et al. [49] suggested an empirical correlation for estimating ηa 15.2518− T −13.4721ln(TBr ) + 0.43577TBr (19)
Br
and ηr for n-alkanes from Eqs. (14) and (15): s
where PBr is the ratio of atmospheric pressure to Pc , TBr is the ratio of TB
ηa = −124.491 + 5.88015/ m + 1023.46/ m2 (14) to Tc .

ηr = 160.376 + 44.6249/ m−852.173/ m2 (15) 2.3.3. Quiñones-Cisneros et al. [22] correlations Tc, Pc, and ω
Quiñones-Cisneros et al. [22] suggested the following correlations
2.3. Estimation of critical properties of crude oil components for the calculation of critical properties of pseudo-components in crude
oils:
Cubic EoS calculations require input of the critical properties (Tc , Pc , Tc = −423.587 + 210.152ln(MW ) (20)
ω ) of crude oil components. The critical properties of pure components
are available in literature, while the critical properties of the pseudo- Pc = fc exp (9.67283−4.05288MW 0.1) (21)
components can be estimated from correlations. In this section, three
15.1665
models are discussed for the estimation of critical properties. ω = exp ⎛8.50471− ⎞
⎝ MW 0.1 ⎠ (22)
2.3.1. The One-Third Rule correlations for Tc and PC where MW is in g/mol, Tc is in K, Pc is in bar. The perturbation factor fc
Panuganti et al. [50] presented a new method to estimate critical reflects the deviation of the critical pressure from that of an n-alkane

Table 1
Properties of crude oils investigated in this study.
Property Crude A Crude B Crude C Crude D Crude E Crude F Crude G Crude H Crude I Crude J

STO MW (g/mol) 201.5 212.4 201.9 204.0 193.5 208.0 177.5 181.2 193.0 189.7
STO API gravity 37.4 31.7 36.9 36.9 40.5 38.0 40.2 39.2 38.4 39.7
GOR (Sm3/m3) 83.5 38.1 147.8 192.4 189.7 163.7 135.4 63.4 171.4 159.1
ρsat (kg/m3) 692.0 764.0 654.7 636.2 624.2 647.6 643.0 692.0 632.4 635.5
Reservoir T (oC) 126.7 100.0 121.1 126.7 120.6 121.1 121.1 126.7 123.9 123.9
Psat (MPa) 8.79 5.76 16.39 19.79 15.79 17.35 15.41 7.63 15.88 14.81
Saturates, wt% 68.6 59.4 70.42 70.2 49.5 – 68.9 68.6 – –
Aromatics, wt% 22.3 22.57 22.64 22.9 40.2 – 21.8 22.3 – –
Resins, wt% 6.5 13.97 6.24 4.9 7.2 – 7.1 6.5 – –
Asphaltene wt% 0.5 1.73 0.18 0.1 3.1 – 0.5 0.5 – –

116
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

with the same molecular weight. The perturbation factor is considered a 3.1. SARA-based + PC-SAFT FT
fitting parameter which is fit to match experimental saturation pressure
for the crude oil of interest. The characterization using the SARA-based method is performed as
described in Section 2.1.1, with the exception that the Asphaltene
pseudo-fraction is combined with the Aromatics + Resins fraction,
3. Methodology forming a pseudo-fraction here called A + R + A. The amount of as-
phaltenes in the modeled crudes is small and is not expected to have a
In this study, 10 light crude oils from the Middle East region with significant effect on the phase behavior or viscosity of crude oil, even if
low asphaltene content are investigated. The properties of the crude oils neglected. The estimation of the PC-SAFT EoS parameters for the
are shown in Table 1. Crudes B-F and I are taken from the work of pseudo-components is performed using Eqs. (1)–(3). For “Heavy Gas”
Abutaqiya et al. [33] with the same crude designation. The composition and “Saturates”, aromaticity (γ) in Eqs. (1)–(3) is set equal to zero
of the crudes can be found in Appendix A. The asphaltene content in the (which corresponds to n-alkanes). For the “Aro-
oils in Table 1 is measured using a gravimetric technique based on IP matics + Resins + Asphaltenes” fraction (A + R + A), the aromaticity
143 standard. The determination of the amounts of Saturates, Aro- (γA+R+A) is fit to match the saturation pressure and density. In case the
matics and Resins for Crude A was originally performed using a clay-gel SARA analysis is not reported for the crude of interest, the SLF method
adsorption column chromatography (ASTM D-2007 m). However, the is used (Section 2.1.2). The PC-SAFT EoS parameter estimation in the
service laboratory reported a 30 wt% loss of volatiles from the oil SLF method is similar to the SARA-based method. The main difference
sample. Therefore, the SARA analysis from Crude H is used to represent between the SLF and SARA-based methods is the number of pseudo-
that of Crude A since both crudes are produced from the same field and components in the liquid phase, i.e. one pseudo-component in the SLF
formation, and both have similar asphaltenes and resins content. A method, and two pseudo components in the SARA-based method.
sensitivity analysis will be performed at the end of this investigation to The PC-SAFT FT model requires two critical viscosity parameters for
study the effect of SARA analysis on the viscosity modeling results of each component (ηa and ηr) in addition to the three PC-SAFT EoS
Crude A. For Crudes B, C, D, E, G, and H a chromatographic technique is parameters (m, σ, ε/k). For pure components (H2S, N2, CO2, C1, C2, and
used that includes a column filled with activated alumina and various C3), these two critical viscosity parameters are taken from the litera-
solvents to elute each fraction separately. Note that the SARA analyses ture.[49] The “Heavy Gas” and “Saturates” pseudo-components are
in Table 1 do not necessarily add to 100% due to loss of some volatiles assumed to consist of n-alkanes. Therefore, the correlations in Eqs. (14)
and presence of inorganics. This small difference (0.5–2.3 wt%) is and (15) can be used to obtain ηa and ηr. For the “A + R + A” pseudo-
added to the Aromatics + Resins + Asphaltenes pseudo-fraction as per component, a correction factor (KcPC-SAFT) is introduced to Eqs. (14) and
the characterization method used in this work (see Section 3.1). (15) to take into account the deviation from n-alkanes behavior. The
In order to model the viscosity of a crude oil using the Friction correlations to obtain PC-SAFT FT parameters for “A + R + A” be-
Theory, the crude oil must be characterized first to obtain the compo- comes:
sition and molecular weight of the pseudo-fractions as well as the EoS
5.88015 1023.46
parameters that match the phase behavior (saturation pressure and ηa = K cPC − SAFT ⎛−124.491 + + ⎞
⎝ m m2 ⎠ (23)
density). Then, the Friction Theory (FT) parameters are optimized de-
pending on available experimental data of viscosity. In this work, the 44.6249 852.173
ηr = K cPC − SAFT ⎛160.376 + − ⎞
viscosity of live oil at saturation is used to tune the FT parameters. ⎝ m m2 ⎠ (24)
As mentioned earlier, the Friction Theory requires the use of an EoS
where K cPC − SAFT is considered a fitting parameter. The superscript “PC-
(e.g. PR or PC-SAFT) to calculate repulsive and attractive pressures.
SAFT” is used to distinguish the K c for the PC-SAFT FT from that used
This implies that several combinations of crude oil characterization and
by Quiñones-Cisneros et al. [13] for the PR FT. When the SLF method is
Friction Theory models may be formulated for modeling crude oil
used, a similar approach is followed where the correlations in Eqs. (23)
viscosity. In this work, three combinations are investigated:
and (24) are used for the SLF pseudo-component.
1- SARA-based characterization method with PC-SAFT Friction Theory
3.2. SARA-based + PR FT
2- SARA-based characterization method with PR Friction Theory
3- Chi-Square characterization method with PR Friction Theory
The characterization results from the previous section are used in
this scenario, i.e. the crude oil components, mole fractions, molecular
A brief description of each scenario can be found in Table 2. A
weight, and PC-SAFT parameters. Here, the PR EoS is used for the
detailed description is given in the following sections.
Friction Theory model. Therefore, the PR EoS parameters (Tc, Pc, ω) are

Table 2
Description of the investigated scenarios for viscosity modeling in this work.
SARA-based + PC-SAFT FTa SARA-based + PR FTa CS + PR FT

EoS used for PVT behavior of crude oil PC-SAFT PC-SAFT PR


PVT data used to fit EoS parameters - Saturation pressure - Saturation pressure - Saturation pressure
- Saturation density - Saturation density - Density above saturation
Number of fitted parameters to match PVT behavior of crude oil 1b 1b 2
(γA+R) (γA+R) (fc and Peneloux correction)
EoS used to calculate repulsive and attractive pressures in the Friction Theory PC-SAFT PR PR
Viscosity data used to fit FT parameters Viscosity at saturation
Number of fitted parameters for viscosity modeling 1 1 1
(KcPC-SAFT) (KcPR) (KcPR)
References for the crude oil characterization method [28–33] [28–33] [20,22,23,40]

a
In case SARA analysis is not available for the crude of interest, SLF method is used for the crude characterization.
b
The Asphaltene pseudo-fraction is lumped with the A + R pseudo-fraction. Therefore, there is no need to fit the two asphaltene parameters γasphal. and MWasphal.
(see Section 2.1.1).

117
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

required for each component. The PR EoS parameters are readily 3.3. CS + PR FT
available for pure components (H2S, N2, CO2, C1, C2, and C3). For
pseudo-components in the SARA-based method (“Heavy Gas”, Quiñones-Cisneros et al. used the Chi-Square characterization of
“Saturates”, and “A + R + A”) or in the SLF method (“Heavy gas” and crude oil with the PR EoS for both the PVT behavior and the FT to
“SLF”), the critical pressure and critical temperature are calculated model the viscosity of crude oil [20,22,23,43]. Here, the same proce-
from the One-Third rule using Eqs. (17) and (18) (Section 2.3.1). The dure is followed as suggested by the authors.
normal boiling point (Tb) and density at 20 °C (ρ20) for each pseudo- The heavy fraction in the crude oil is characterized as four pseudo-
component, which are required in Eqs. (17) and (18), are calculated components following a Chi-Square distribution (Section 2.1.3). The PR
using PC-SAFT with EoS parameters that were already obtained from EoS is used to describe the phase behavior of the crude oil. For pure
the characterization step. Therefore, PC-SAFT is used as a source of components (H2S, N2, CO2, C1, C2, C3, i-C4, n-C4, i-C5, and n-C5, and C6),
“experimental” values for Tb and ρ20 for the pseudo-components. The the PR EoS parameters (Tc, Pc, ω) are available from the literature. For
acentric factor (ω) for each pseudo-component is calculated from the the four pseudo-fractions, the PR EoS parameters are obtained from the
Lee-Kesler correlations Eq. (19), with Tc and Pc calculated from the One- correlations of Quiñones-Cisneros et al. [22] shown in Eqs. (20)–(22)
Third rule. (Section 2.3.3). The perturbation parameter fc in Eq. (21) is treated as
The PR FT model requires knowledge of the critical viscosity (ηc) for an adjustable parameter to match the saturation pressure for the crude
each component. For pure components, the critical viscosity values can oil of interest [22]. Furthermore, to match the density of the crude oil,
be found in the literature [15]. The “Heavy Gas” and “Saturates” Quiñones-Cisneros et al. [22] suggested a Peneloux shift correction
pseudo-fractions are assumed to consist of n-alkanes. Therefore, the from the actual volume v to a corrected volume v′,
correlations for ηc in Eq. (12) can be used with Kc value of 7.95 which
v′ = v−c (25)
corresponds to n-alkanes [13]. For “A + R + A” or “SLF” pseudo-frac-
tions, the Kc in Eq. (12) is treated as a fitting parameter. In this work, where the volume shift c is estimated with:
the KcPR of “Saturates” or “SLF” pseudo-fractions is fitted to match the
viscosity at saturation. Here we introduce the superscript PR to dis- c = Kv ∑ x i MWi
tinguish between the Kc used for PR from that used in PC-SAFT. HF (26)

with the summation performed over the heavy fractions, and Kv is the
adjustable parameter for the volume correction.

Fig. 1. Description of the scenarios used for crude oil viscosity modeling in this work (FT: Friction Theory, CS: Chi-Square).

118
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

The PR FT requires the critical viscosity ηc for each component. For Table 4. The PR EoS parameters, along with the density at 20 °C (ρ20)
pure components, the critical viscosity values can be found in the lit- and normal boiling point (Tb) are shown in Table 5 for the pseudo-
erature [15]. For the four pseudo-fractions, the correlations for ηc in Eq. components in the SARA-based method. As mentioned earlier, the
(12) can be used with KcPR treated as a single fitting parameter for all density at 20 °C (ρ20) and normal boiling point (Tb) are obtained from
pseudo-fractions [22]. In this work, KcPR is tuned to match the viscosity the PC-SAFT EoS.
at saturation. The characterized crudes and the optimized EoS parameters using
Fig. 1 demonstrates a flowchart for the description of the three the Chi-Square characterization method are shown in Table 6. The
scenarios studied in this work. Example calculations for two scenarios characterization results shown in Table 6 were obtained using the
(SARA-based + PR FT and SARA-based + PC-SAFT FT) can be found in commercial software VLXE|BLEND® from VLXE Aps. The optimized PR
the supplementary information document attached to this manuscript, FT parameters are shown in Table 7.
as well as tabulated experimental data for viscosity and density of each
crude oil.
4.2. Viscosity modeling results

4. Results After fitting the Friction Theory parameters to viscosity at satura-


tion, predictions are made for viscosity values on the entire pressure
4.1. Characterization results and FT simulation parameters range. The value of viscosity at saturation is chosen in order to in-
vestigate the capability of the Friction Theory to predict viscosity of live
The characterized crudes and the optimized EoS parameters using oil at high-pressure and high-temperature from minimum available
the SARA-based method (or SLF for crudes F, I, and J) are shown in experimental data. Fig. 2 shows the simulation results of the viscosity of
Table 3. The simulation parameters for crudes B-F and I in Table 3 are the live oil for the 10 crudes studied. Table 8 provides the average
taken from Abutaqiya et al. [33] The binary interaction parameters absolute percent deviation (AAPD) of the viscosity predictions for the
(BIPs) used for the SARA-based method are taken from Punnapala and three scenarios. The AAPD of the viscosity predictions is reported for
Vargas [30], while those for the SLF method are taken from Abutaqiya the under-saturated oil (above bubble point), saturated oil (below
et al. [33] The sets of BIPs used in this work can be found in Appendix bubble point), and for the entire pressure range. The error associated
B. The optimized Friction Theory parameters for the PC-SAFT FT and with the viscosity point at ambient pressure is not included in the
the PR FT using the SARA-based characterization method are shown in AAPD. This data point is subject to the highest experimental

Table 3
Simulation parameters for crude oils using the SARA-based and SLF characterization methods.
Component MW (g/mol) γ zi (mol%) m σ (Å) ε/k (K) Component Component MW (g/mol) γ zi (mol%) m σ (Å) ε/k (K)

Crude A N2 28.01 – 0.22 1.21 3.31 90.96 N2 28.01 – 0.10 1.21 3.31 90.96 Crude B
CO2 44.01 – 2.38 2.07 2.79 169.21 CO2 44.01 – 0.90 2.07 2.79 169.21
C1 16.04 – 18.12 1.00 3.70 150.03 C1 16.04 – 12.51 1.00 3.70 150.03
C2 30.07 – 5.51 1.61 3.52 191.42 C2 30.07 – 4.60 1.61 3.52 191.42
C3 44.10 – 6.24 2.00 3.62 208.11 C3 44.10 – 5.11 2.00 3.62 208.11
Heavy gas 71.80 0.00 13.68 2.69 3.76 231.43 Heavy gas 65.42 0.00 5.11 2.53 3.74 228.45
Saturates 183.30 0.00 40.62 5.56 3.91 250.87 Saturates 177.90 0.00 52.96 5.42 3.91 250.48
A + R+A 257.40 0.60 13.23 5.58 4.13 372.65 A + R+A 310.40 0.69 18.62 6.09 4.20 400.43

Crude C N2 28.01 – 0.12 1.21 3.31 90.96 N2 28.01 – 0.12 1.21 3.31 90.96 Crude D
CO2 44.01 – 2.77 2.07 2.79 169.21 CO2 44.01 – 2.66 2.07 2.79 169.21
C1 16.04 – 32.03 1.00 3.70 150.03 C1 16.04 – 37.87 1.00 3.70 150.03
C2 30.07 – 6.43 1.61 3.52 191.42 C2 30.07 – 6.83 1.61 3.52 191.42
C3 44.10 – 6.05 2.00 3.62 208.11 C3 44.10 – 5.92 2.00 3.62 208.11
Heavy gas 72.84 0.00 12.66 2.72 3.76 231.87 Heavy gas 73.92 0.00 13.34 2.74 3.77 232.31
Saturates 187.62 0.00 30.19 5.67 3.91 251.17 Saturates 199.87 0.00 21.93 5.98 3.92 251.96
A + R+A 242.65 0.63 9.74 5.25 4.12 375.03 A + R+A 211.45 0.60 11.33 4.83 4.07 360.16

Crude E N2 28.01 – 0.16 1.21 3.31 90.96 N2 28.01 – 0.12 1.21 3.31 90.96 Crude F
CO2 44.01 – 1.66 2.07 2.79 169.21 CO2 44.01 – 2.59 2.07 2.79 169.21
C1 16.04 – 32.57 1.00 3.70 150.03 C1 16.04 – 34.15 1.00 3.70 150.03
C2 30.07 – 7.82 1.61 3.52 191.42 C2 30.07 – 6.72 1.61 3.52 191.42
C3 44.10 – 7.71 2.00 3.62 208.11 C3 44.10 – 6.32 2.00 3.62 208.11
Heavy gas 72.16 0.00 15.48 2.70 3.76 231.58 Heavy gas 72.33 0.00 13.41 2.70 3.76 231.65
Saturates 160.61 0.00 20.62 4.97 3.90 249.04 SLF 208.00 0.210 36.68 5.66 3.97 289.48
A + R+A 228.30 0.37 13.89 5.72 4.03 322.53

Crude G N2 28.01 – 0.15 1.21 3.31 90.96 N2 28.01 – 0.91 1.21 3.31 90.96 Crude H
CO2 44.01 – 1.91 2.07 2.79 169.21 CO2 44.01 – 2.34 2.07 2.79 169.21
C1 16.04 – 31.23 1.00 3.70 150.03 C1 16.04 – 16.84 1.00 3.70 150.03
C2 30.07 – 7.08 1.61 3.52 191.42 C2 30.07 – 4.99 1.61 3.52 191.42
C3 44.10 – 6.49 2.00 3.62 208.11 C3 44.10 – 5.14 2.00 3.62 208.11
Heavy gas 66.06 0.00 8.24 2.54 3.74 228.78 Heavy gas 71.77 0.00 6.60 2.69 3.76 231.42
Saturates 162.42 0.00 33.80 5.02 3.90 249.20 Saturates 151.95 0.00 51.57 4.75 3.89 248.20
A + R+A 220.00 0.52 11.09 4.89 4.09 368.48 A + R+A 307.53 0.60 11.46 6.40 4.17 381.23

Crude I N2 28.01 – 0.26 1.21 3.31 90.96 N2 28.01 – 0.18 1.21 3.31 90.96 Crude J
CO2 44.01 – 1.85 2.07 2.79 169.21 CO2 44.01 – 1.85 2.07 2.79 169.21
C1 16.04 – 31.10 1.00 3.70 150.03 C1 16.04 – 29.85 1.00 3.70 150.03
C2 30.07 – 7.11 1.61 3.52 191.42 C2 30.07 – 7.18 1.61 3.52 191.42
C3 44.10 – 6.84 2.00 3.62 208.11 C3 44.10 – 7.04 2.00 3.62 208.11
Heavy gas 74.35 0.00 15.52 2.76 3.77 232.48 Heavy gas 72.54 0.00 14.63 2.71 3.76 231.74
SLF 193.00 0.200 37.31 5.38 3.96 285.34 SLF 189.70 0.204 39.26 5.30 3.96 285.41

119
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table 4
Optimized PR and PC-SAFT Friction Theory parameters using the SARA-based (or SLF) characterization method.
Component PR FT PC-SAFT FT Component PR FT PC-SAFT FT

Kc PR
ηc (μP) Kc PC-SAFT
ηa (μP) ηr (μP) KcPR ηc (μP) KcPC-SAFT ηa (μP) ηr (μP)

Crude A N2 – 174.2 – 2.3 57.8 N2 – 174.2 – 2.3 57.8 Crude B


CO2 – 376.9 – −16.1 119.3 CO2 – 376.9 – −16.1 119.3
C1 – 152.9 – 8.2 48.7 C1 – 152.9 – 8.2 48.7
C2 – 217.6 – 9.0 49.9 C2 – 217.6 – 9.0 49.9
C3 – 249.7 – 1.3 65.0 C3 – 249.7 – 1.3 65.0
Heavy gas 7.95 269.5 – 19.2 59.2 Heavy gas 7.95 269.3 – 37.7 44.9
Saturates 7.95 243.2 – −90.3 140.8 Saturates 7.95 247.1 – −88.6 139.6
A + R + A (γ = 0.60) 7.02 269.0 0.428 −38.7 60.3 A + R + A (γ = 0.69) 5.50 146.1 1.14 −109.6 165.3

Crude C N2 – 174.2 – 2.3 57.8 N2 – 174.2 – 2.3 57.8 Crude D


CO2 – 376.9 – −16.1 119.3 CO2 – 376.9 – −16.1 119.3
C1 – 152.9 – 8.2 48.7 C1 – 152.9 – 8.2 48.7
C2 – 217.6 – 9.0 49.9 C2 – 217.6 – 9.0 49.9
C3 – 249.7 – 1.3 65.0 C3 – 249.7 – 1.3 65.0
Heavy gas 7.95 269.5 – 16.0 61.6 Heavy gas 7.95 269.4 – 14.0 63.2
Saturates 7.95 242.0 – −91.6 141.7 Saturates 7.95 238.7 – −94.9 144.0
A + R + A (γ = 0.63) 11.46 447.8 0.556 −48.0 76.7 A + R + A (γ = 0.60) 13.74 540.4 0.516 −41.0 68.7

Crude E N2 – 174.2 – 2.3 57.8 N2 – 174.2 – 2.3 57.8 Crude F


CO2 – 376.9 – −16.1 119.3 CO2 – 376.9 – −16.1 119.3
C1 – 152.9 – 8.2 48.7 C1 – 152.9 – 8.2 48.7
C2 – 217.6 – 9.0 49.9 C2 – 217.6 – 9.0 49.9
C3 – 249.7 – 1.3 65.0 C3 – 249.7 – 1.3 65.0
Heavy gas 7.95 269.5 – 18.1 60.0 Heavy gas 7.95 269.2 – 17.7 60.3
Saturates 7.95 249.4 – −81.9 134.9 SLF (γ = 0.210) 7.55 253.8 0.280 −25.6 39.7
A + R + A (γ = 0.37) 8.71 311.1 0.272 −25.1 38.7

Crude G N2 – 174.2 – 2.3 57.8 N2 – 174.2 – 2.3 57.8 Crude H


CO2 – 376.9 – −16.1 119.3 CO2 – 376.9 – −16.1 119.3
C1 – 152.9 – 8.2 48.7 C1 – 152.9 – 8.2 48.7
C2 – 217.6 – 9.0 49.9 C2 – 217.6 – 9.0 49.9
C3 – 249.7 – 1.3 65.0 C3 – 249.7 – 1.3 65.0
Heavy gas 7.95 269.3 – 36.5 45.9 Heavy gas 7.95 269.4 – 19.2 59.1
Saturates 7.95 248.9 – −82.7 135.4 Saturates 7.95 251.8 – −77.9 132.0
A + R + A (γ = 0.52) 12.02 475.3 0.275 −22.1 36.8 A + R + A (γ = 0.60) 5.04 186.8 0.258 −25.5 37.8

Crude I N2 – 174.2 – 2.3 57.8 N2 – 174.2 – 2.3 57.8 Crude J


CO2 – 376.9 – −16.1 119.3 CO2 – 376.9 – −16.1 119.3
C1 – 152.9 – 8.2 48.7 C1 – 152.9 – 8.2 48.7
C2 – 217.6 – 9.0 49.9 C2 – 217.6 – 9.0 49.9
C3 – 249.7 – 1.3 65.0 C3 – 249.7 – 1.3 65.0
Heavy gas 7.95 269.4 – 12.5 64.3 Heavy gas 7.95 269.5 – 17.2 60.7
SLF (γ = 0.200) 7.96 261.3 0.277 −24.4 38.5 SLF (γ = 0.204) 7.14 242.3 0.242 −21.0 33.5

Table 5
PR EoS parameters for the pseudo-components in the SARA-based (or SLF) method.
Component ρ20 (g/cm3) Tb (K) Pc (bar) Tc (K) ω Component ρ20 (g/cm3) Tb (K) Pc (bar) Tc (K) ω

Crude A Heavy gas 620 309.20 37.78 496.30 0.11 Heavy gas 0.60 293.57 40.18 482.45 0.07 Crude B
Saturates 757 505.15 17.31 674.26 0.61 Saturates 0.75 497.87 17.82 667.45 0.59
A + R + A (γ = 0.60) 996 737.60 20.49 946.85 1.04 A + R + A (γ = 0.69) 1.07 826.98 18.06 1041.2 1.15

Crude C Heavy gas 620 311.69 37.41 498.51 0.12 Heavy gas 0.63 314.24 37.03 500.76 0.13 Crude D
Saturates 760 510.86 16.92 679.62 0.62 Saturates 0.77 526.56 15.90 694.38 0.64
A + R + A (γ = 0.63) 1000 718.87 21.93 929.45 1.02 A + R + A (γ = 0.60) 0.97 660.87 24.15 869.32 0.92

Crude E Heavy gas 620 310.1 37.65 497.08 0.11 Heavy gas 0.62 310.48 37.59 497.44 0.12 Crude F
Saturates 740 473.4 19.63 644.68 0.55 SLF (γ = 0.210) 0.83 583.84 18.82 767.91 0.77
A + R + A (γ = 0.37) 890 650.5 19.59 845.71 0.89

Crude G Heavy gas 600 295.19 39.92 483.89 0.07 Heavy gas 0.62 309.15 37.84 496.35 0.11 Crude H
Saturates 740 476.07 19.43 647.14 0.55 Saturates 0.74 460.49 20.66 632.71 0.52
A + R + A (γ = 0.52) 990 680.35 23.77 890.49 0.95 A + R + A (γ = 0.60) 1.02 808.35 17.37 1018.12 1.12

Crude I Heavy gas 630 315.25 36.88 501.67 0.13 Heavy gas 0.62 310.98 37.51 497.88 0.12 Crude J
SLF (γ = 0.215) 820 563.66 20.06 748.41 0.73 SLF (γ = 0.204) 0.82 556.46 20.16 740.50 0.72

uncertainty and its inclusion in the overall AAPD may cause misleading predictions that are within 10% of the experimental data. The “SARA-
interpretations of the results. based + PC-SAFT FT” model shows the lowest deviation from experi-
mental data with an AAPD of 3.3%, while the CS + PR FT shows the
5. Discussion highest deviation with an AAPD of 6.4%. As observed from the results
in Table 8, the error in the predictions for the three scenarios below
In general, the three scenarios provide acceptable overall saturation pressure is higher than the error above saturation pressure.

120
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table 6
Simulation parameters for crude oils using the PR Chi-Square characterization method.
Component MW (g/mol) zi (mol%) Tc (K) Pc (bar) ω c Component MW (g/mol) zi (mol%) Tc (K) Pc (bar) ω c

Crude A N2 28.01 0.22 126.20 33.98 0.04 −0.15 N2 28.01 0.12 126.20 33.98 0.037 −0.15 Crude B
CO2 44.01 2.37 304.12 73.74 0.23 −0.04 CO2 44.01 0.92 304.12 73.74 0.225 −0.04
H2S 34.08 0.00 373.40 89.63 0.09 −0.09 H2S 34.08 0.00 373.40 89.63 0.090 −0.09
C1 16.04 18.10 190.56 45.99 0.01 −0.32 C1 16.04 12.33 190.56 45.99 0.011 −0.32
C2 30.07 5.56 305.32 48.72 0.10 −0.19 C2 30.07 4.74 305.32 48.72 0.099 −0.19
C3 44.10 6.53 369.83 42.48 0.15 −0.14 C3 44.10 6.25 369.83 42.48 0.152 −0.14
i-C4 58.12 2.03 407.85 36.40 0.19 −0.12 i-C4 58.12 1.25 407.85 36.40 0.186 −0.12
n-C4 58.12 5.00 425.12 37.96 0.20 −0.11 n-C4 58.12 4.63 425.12 37.96 0.200 −0.11
i-C5 72.15 2.60 460.39 33.81 0.23 −0.09 i-C5 72.15 2.07 460.39 33.81 0.229 −0.09
n-C5 72.15 3.43 469.70 33.70 0.25 −0.07 n-C5 72.15 3.35 469.70 33.70 0.252 −0.07
C6 84.00 5.10 509.84 28.61 0.29 0.02 C6 84.00 4.86 509.84 28.61 0.295 0.02
Heavy 1 124.36 20.99 594.68 29.77 0.41 −0.09 Heavy 1 131.01 27.49 631.13 32.34 0.387 −0.14
Heavy 2 193.32 13.51 688.12 22.14 0.62 −0.09 Heavy 2 227.49 15.83 752.92 22.24 0.638 −0.14
Heavy 3 283.46 9.21 769.18 16.94 0.87 −0.09 Heavy 3 348.15 10.34 846.84 16.42 0.921 −0.14
Heavy 4 488.85 5.34 884.62 11.37 1.37 −0.09 Heavy 4 618.19 5.83 973.56 10.68 1.475 −0.14

Crude C N2 28.01 0.12 126.20 33.98 0.04 −0.15 N2 28.01 0.13 126.20 33.98 0.04 −0.15 Crude D
CO2 44.01 2.78 304.12 73.74 0.23 −0.04 CO2 44.01 2.64 304.12 73.74 0.23 −0.04
H2S 34.08 0.01 373.40 89.63 0.09 −0.09 H2S 34.08 0.00 373.40 89.63 0.09 −0.09
C1 16.04 32.18 190.56 45.99 0.01 −0.32 C1 16.04 37.47 190.56 45.99 0.01 −0.32
C2 30.07 6.47 305.32 48.72 0.10 −0.19 C2 30.07 6.78 305.32 48.72 0.10 −0.19
C3 44.10 6.19 369.83 42.48 0.15 −0.14 C3 44.10 5.94 369.83 42.48 0.15 −0.14
i-C4 58.12 1.54 407.85 36.40 0.19 −0.12 i-C4 58.12 1.47 407.85 36.40 0.19 −0.12
n-C4 58.12 4.02 425.12 37.96 0.20 −0.11 n-C4 58.12 3.76 425.12 37.96 0.20 −0.11
i-C5 72.15 1.89 460.39 33.81 0.23 −0.09 i-C5 72.15 1.73 460.39 33.81 0.23 −0.09
n-C5 72.15 2.59 469.70 33.70 0.25 −0.07 n-C5 72.15 2.32 469.70 33.70 0.25 −0.07
C6 84.00 3.74 509.84 28.61 0.29 0.02 C6 84.00 3.35 509.84 28.61 0.29 0.02
Heavy 1 117.13 17.02 594.29 32.90 0.37 −0.14 Heavy 1 111.90 15.64 583.09 33.68 0.36 −0.12
Heavy 2 190.64 10.46 699.65 23.75 0.58 −0.14 Heavy 2 183.80 9.52 690.19 24.20 0.56 −0.12
Heavy 3 284.96 6.99 786.59 17.93 0.82 −0.14 Heavy 3 288.12 6.07 787.20 17.68 0.83 −0.12
Heavy 4 498.15 4.00 907.40 11.91 1.31 −0.14 Heavy 4 549.14 3.19 926.39 10.99 1.43 −0.12

Crude E N2 28.01 0.12 126.20 33.98 0.04 −0.15 N2 28.01 0.14 126.20 33.98 0.037 −0.15 Crude F
CO2 44.01 2.59 304.12 73.74 0.23 −0.04 CO2 44.01 2.59 304.12 73.74 0.225 −0.04
H2S 34.08 0.00 373.40 89.63 0.09 −0.09 H2S 34.08 0.00 373.40 89.63 0.09 −0.09
C1 16.04 34.16 190.56 45.99 0.01 −0.32 C1 16.04 34.01 190.56 45.99 0.011 −0.32
C2 30.07 6.72 305.32 48.72 0.10 −0.19 C2 30.07 6.69 305.32 48.72 0.099 −0.19
C3 44.10 6.36 369.83 42.48 0.15 −0.14 C3 44.10 6.34 369.83 42.48 0.152 −0.14
i-C4 58.12 1.54 407.85 36.40 0.19 −0.12 i-C4 58.12 1.53 407.85 36.40 0.186 −0.12
n-C4 58.12 4.05 425.12 37.96 0.20 −0.11 n-C4 58.12 4.03 425.12 37.96 0.2 −0.11
i-C5 72.15 1.99 460.39 33.81 0.23 −0.09 i-C5 72.15 1.98 460.39 33.81 0.229 −0.09
n-C5 72.15 2.66 469.70 33.70 0.25 −0.07 n-C5 72.15 2.65 469.70 33.70 0.252 −0.07
C6 84.00 3.56 509.84 28.61 0.29 0.02 C6 84.00 3.65 509.84 28.61 0.2945 0.02
Heavy 1 114.32 16.72 565.71 29.66 0.41 −0.10 Heavy 1 113.34 16.68 584.76 33.22 0.361 −0.13
Heavy 2 191.86 9.96 673.26 20.99 0.65 −0.10 Heavy 2 188.51 10.03 694.34 23.66 0.577 −0.13
Heavy 3 304.02 6.29 768.88 15.20 0.98 −0.10 Heavy 3 297.38 6.36 792.53 17.19 0.862 −0.13
Heavy 4 584.24 3.27 904.57 9.36 1.68 −0.10 Heavy 4 569.54 3.32 932.50 10.63 1.479 −0.13

Crude G N2 28.01 0.16 126.20 33.98 0.04 −0.15 N2 28.01 0.90 126.2 33.98 0.04 −0.15 Crude H
CO2 44.01 1.90 304.12 73.74 0.23 −0.04 CO2 44.01 2.33 304.1 73.74 0.23 −0.04
H2S 34.08 0.00 373.40 89.63 0.09 −0.09 H2S 34.08 0.07 373.4 89.63 0.09 −0.09
C1 16.04 31.12 190.56 45.99 0.01 −0.32 C1 16.04 16.81 190.6 45.99 0.01 −0.32
C2 30.07 7.16 305.32 48.72 0.10 −0.19 C2 30.07 5.14 305.3 48.72 0.10 −0.19
C3 44.10 6.90 369.83 42.48 0.15 −0.14 C3 44.10 5.83 369.8 42.48 0.15 −0.14
i-C4 58.12 1.81 407.85 36.40 0.19 −0.12 i-C4 58.12 1.79 407.9 36.40 0.19 −0.12
n-C4 58.12 4.48 425.12 37.96 0.20 −0.11 n-C4 58.12 4.34 425.1 37.96 0.20 −0.11
i-C5 72.15 2.14 460.39 33.81 0.23 −0.09 i-C5 72.15 2.33 460.4 33.81 0.23 −0.09
n-C5 72.15 2.90 469.70 33.70 0.25 −0.07 n-C5 72.15 3.04 469.7 33.70 0.25 −0.07
C6 84.00 4.24 509.84 28.61 0.29 0.02 C6 84.00 4.76 509.8 28.61 0.29 0.02
Heavy 1 89.60 20.76 531.77 38.15 0.29 −0.11 Heavy 1 119.69 22.36 562.0 26.75 0.46 0.02
Heavy 2 203.54 9.14 707.74 22.16 0.63 −0.11 Heavy 2 185.43 14.43 650.9 19.97 0.68 0.02
Heavy 3 372.91 4.99 837.58 14.41 1.06 −0.11 Heavy 3 268.72 9.96 726.1 15.42 0.95 0.02
Heavy 4 809.56 2.30 1003.8 7.98 1.98 −0.11 Heavy 4 453.94 5.90 832.5 10.53 1.47 0.02

(continued on next page)

121
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table 6 (continued)

Component MW (g/mol) zi (mol%) Tc (K) Pc (bar) ω c Component MW (g/mol) zi (mol%) Tc (K) Pc (bar) ω c

Crude I N2 28.01 0.27 126.20 33.98 0.04 −0.15 N2 28.01 0.17 126.20 33.98 0.04 −0.15 Crude J
CO2 44.01 1.86 304.12 73.74 0.23 −0.04 CO2 44.01 1.87 304.12 73.74 0.23 −0.04
H2S 34.08 0.00 373.40 89.63 0.09 −0.09 H2S 34.08 0.00 373.40 89.63 0.09 −0.09
C1 16.04 31.24 190.56 45.99 0.01 −0.32 C1 16.04 29.93 190.56 45.99 0.01 −0.32
C2 30.07 7.16 305.32 48.72 0.10 −0.19 C2 30.07 7.24 305.32 48.72 0.10 −0.19
C3 44.10 6.95 369.83 42.48 0.15 −0.14 C3 44.10 7.23 369.83 42.48 0.15 −0.14
i-C4 58.12 1.84 407.85 36.40 0.19 −0.12 i-C4 58.12 1.95 407.85 36.40 0.19 −0.12
n-C4 58.12 4.56 425.12 37.96 0.20 −0.11 n-C4 58.12 4.81 425.12 37.96 0.20 −0.11
i-C5 72.15 2.31 460.39 33.81 0.23 −0.09 i-C5 72.15 2.38 460.39 33.81 0.23 −0.09
n-C5 72.15 3.09 469.70 33.70 0.25 −0.07 n-C5 72.15 3.16 469.70 33.70 0.25 −0.07
C6 86.20 4.26 509.84 28.61 0.29 0.02 C6 84.00 4.45 509.84 28.61 0.29 0.02
Heavy 1 114.95 15.57 589.63 33.21 0.36 −0.15 Heavy 1 116.45 15.62 587.92 32.22 0.38 −0.13
Heavy 2 177.04 10.11 682.94 24.91 0.54 −0.15 Heavy 2 177.84 10.22 678.72 24.30 0.56 −0.13
Heavy 3 260.76 6.86 766.61 19.04 0.76 −0.15 Heavy 3 260.85 6.97 760.86 18.63 0.78 −0.13
Heavy 4 456.12 3.92 887.43 12.68 1.22 −0.15 Heavy 4 454.85 4.00 880.09 12.43 1.25 −0.13

Table 7
Critical viscosities (ηc) and fitted KcPR values for the PR Friction Theory model after fitting to viscosity at saturation using the Chi-Square characterization method.
Crude ηc (μP)

A B C D E F G H I J

Component
N2 174.2 174.2 174.2 174.2 174.2 174.2 174.2 174.2 174.2 174.2
CO2 376.9 376.9 376.9 376.9 376.9 376.9 376.9 376.9 376.9 376.9
H2S 318.0 318.0 318.0 318.0 318.0 318.0 318.0 318.0 318.0 318.0
C1 152.9 152.9 152.9 152.9 152.9 152.9 152.9 152.9 152.9 152.9
C2 224.1 224.1 224.1 224.1 224.1 224.1 224.1 224.1 224.1 224.1
C3 239.9 239.9 239.9 239.9 239.9 239.9 239.9 239.9 239.9 239.9
i-C4 244.4 244.4 244.4 244.4 244.4 244.4 244.4 244.4 244.4 244.4
n-C4 249.6 249.6 249.6 249.6 249.6 249.6 249.6 249.6 249.6 249.6
i-C5 254.0 254.0 254.0 254.0 254.0 254.0 254.0 254.0 254.0 254.0
n-C5 252.6 252.6 252.6 252.6 252.6 252.6 252.6 252.6 252.6 252.6
C6 242.4 242.4 242.4 242.4 242.4 242.4 242.4 242.4 242.4 242.4
Heavy 1 321.0 291.2 351.3 344.3 246.6 273.2 189.0 381.4 330.4 303.8
Heavy 2 320.7 290.3 351.0 344.2 246.4 273.1 189.1 381.1 330.4 303.7
Heavy 3 318.9 287.7 348.9 342.0 244.6 271.2 186.8 379.3 328.8 302.3
Heavy 4 313.5 281.2 342.8 334.6 238.9 265.1 180.1 373.6 323.7 297.6

KcPR 7.71 7.34 9.71 9.02 6.92 7.18 5.02 11.20 8.64 8.05

The experimental data above saturation pressure (single-phase fluid) is The following mixing rule for the volume shift was proposed by the
expected to be more accurate than the data below saturation pressure authors [22]:
(in the two-phase region). This is mainly because other factors of ex-
1
perimental uncertainty arise when measuring viscosity in the two-phase ζ = Kz ∑ x i MWi3
region such as sample equilibration time. The modeling results in i (27)
Table 8 show trends that are consistent with the expected experimental
uncertainty. The CS + PR FT scenario shows a larger deviation to ex- where the summation applies only to pseudo-components and K z is
perimental data below the bubble point than the other two scenarios. tuned over viscosity data in the entire pressure range. However, since in
However, given the high experimental uncertainty in the measurements this work we are studying the predictive capability of the models after
below bubble point, these high deviations are still acceptable. tuning to only one data point at saturation, it is not possible to in-
The modeling results for crude G using CS + PR FT show deviation corporate the K z correction factor into the modeling results.
from experimental data in the regions above and below bubble pres- The accuracy of viscosity predictions are expected to depend on how
sure. Deviations are also observed for Crude B above the bubble point accurately the model captures the density. In search for a relation be-
using SARA-based + PR FT. Quiñones-Cisneros et al. recommended tween the viscosity and density predictions, modeling of oil density is
using a correction factor (K z ) to correct for the compressibility effects performed using the two characterization procedures: SARA-based and
which are usually inadequately described by a regular cubic EOS for CS. The AAPD for the density predictions are shown in Table 9. Similar
dense fluids [22]. The correction factor is implemented by in- to viscosity, the modeling results for density show a consistent trend to
corporating a volume shift (ζ ) to the volume used in the calculation of the expected experimental uncertainty. The AAPD for the predictions
the repulsive and attractive pressure. This volume shift is different than below bubble point (two phase region) is higher than AAPD above
the one used to match the experimental density (Eqs. (25) and (26)). bubble point (single phase region). Also, the CS method shows a larger

122
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Fig. 2. Modeling results for viscosity of crude oils at reservoir temperature (FT: Friction Theory, PR: Peng Robinson, CS: Chi-Square).

deviation for density predictions below saturation pressure than the observed which indicates that good match to viscosity data is expected
SARA-based method. This can explain the larger deviation in the visc- if the density predictions of the EoS model are accurate. Therefore, the
osity predictions below bubble point for the CS method. In order to improved predictions of viscosity from the SARA-based method can be
analyze the relation between density and viscosity predictions, the attributed to the more powerful characterization method and EoS used
AAPD of viscosity predictions in each region is plotted against AAPD of for the phase behavior (PC-SAFT). The improved performance of the
density predictions as shown in Fig. 3. A positive relationship is SARA-based method can be justified by the more physical nature of the

123
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table 8
Average absolute percent deviation for viscosity predictions after fitting to viscosity at saturation.
SARA-baseda + PC-SAFT FT SARA-baseda + PR-FT CS + PR FT

sat sat sat sat


Above P Below P Overall Above P Below P Overall Above Psat Below Psat Overall

Crude A 1.4% 3.1% 2.1% 0.1% 11.1% 4.6% 1.1% 21.3% 10.1%
Crude B 1.4% 7.1% 3.8% 2.6% 4.0% 3.1% 0.5% 2.9% 1.5%
Crude C 0.7% 5.4% 2.5% 1.7% 7.0% 3.8% 0.6% 19.0% 8.0%
Crude D 1.6% 5.5% 2.8% 1.1% 8.3% 3.2% 0.8% 22.5% 7.1%
Crude E 3.0% 9.4% 5.2% 2.1% 3.8% 2.6% 0.8% 9.7% 4.2%
Crude F 5.4% 9.6% 7.1% 1.7% 12.6% 6.2% 1.1% 13.3% 6.1%
Crude G 3.1% 2.4% 2.9% 3.8% 3.4% 3.7% 7.4% 19.6% 11.7%
Crude H 0.8% 4.9% 2.2% 1.0% 0.5% 0.9% 2.2% 4.3% 2.9%
Crude I 1.2% 0.7% 1.0% 0.5% 10.8% 4.4% 2.2% 17.9% 8.1%
Crude J 2.1% 3.9% 2.8% 2.6% 6.5% 4.1% 0.2% 12.1% 4.6%

Overall 2.1% 5.2% 3.2% 1.7% 6.8% 3.7% 1.7% 14.3% 6.4%

a
For crudes F, I, and J the SLF method is used for the crude characterization.

Table 9 The density predictions above saturation pressure using CS (AAPD


Average absolute percent deviation for density predictions using the SARA- 0.22%) show a slightly better match to experimental data than the
based characterization and the Chi-Square (CS) characterization. SARA-based method (AAPD 0.41%). This is because the Peneloux
SARA-baseda CS correction for PR EoS used for the CS method was fit to match the
entire set of density data above saturation pressure, unlike the SARA-
sat sat
Above P Below P Overall Above Psat Below Psat Overall based method which is tuned only to one density data point at sa-
turation. Although both characterization procedures show similar
Crude A 0.47% 0.39% 0.44% 0.16% 1.80% 0.69%
Crude B 0.14% 1.43% 0.57% 0.60% 0.45% 0.55% overall predictions for density, the SARA-based method should be
Crude C 0.10% 0.94% 0.38% 0.11% 1.83% 0.67% considered more superior to CS method for three reasons: 1) lower
Crude D 0.51% 1.01% 0.65% 0.10% 2.36% 0.67% number of data points required for the fitting (one density value),
Crude E 0.58% 0.40% 0.52% 0.09% 1.38% 0.46%
lower number of fitting parameters (one parameter), and lower
Crude F 0.52% 0.72% 0.59% 0.08% 2.05% 0.71%
Crude G 0.37% 1.54% 0.85% 0.24% 0.53% 0.33% number of pseudo-components used in the characterization. In fact,
Crude H 0.21% 1.55% 0.58% 0.55% 0.76% 0.61% even when the SLF method is used where the heavy fraction is char-
Crude I 0.61% 0.98% 0.73% 0.17% 1.55% 0.61% acterized as a single liquid pseudo-component (Crudes F, I, and J) the
Crude J 0.64% 0.81% 0.69% 0.08% 2.15% 0.70% viscosity predictions using PR-FT are at least as good as the predic-
Overall 0.41% 0.98% 0.60% 0.22% 1.49% 0.60% tions from the CS + PR-FT.
Comparing the SARA-based + PR FT and the SARA-based + PC-
a
For crudes F, I, and J the SLF method is used for the crude characterization. SAFT FT, it can be seen that using PC-SAFT FT yields slightly better
match to experimental data. Furthermore, the reduction of the FT
parameters for the pseudo-components proposed in this work from two
(ηa and ηr) to one (KcPC-SAFT) seems to provide acceptable results.
However, given the increased complexity in the PC-SAFT FT model, the
improvements in the viscosity are too slight to justify its use. The PR FT
model seems to provide sufficient description of the viscosity of the
crude oils at varying pressure and composition. The EoS used to capture
the phase behavior of the crude oil has a more pronounced effect on the
viscosity modeling results than the EoS used for the FT model.

6. Sensitivity to SARA analysis

The SARA-based method studied in this work requires SARA ana-


lysis as an input to characterize the heavy fraction. It is well-known,
however, that SARA analysis is subject to significant experimental un-
Fig. 3. The relation between error in viscosity predictions and error in density certainty. Different service laboratories follow different experimental
predictions for all crudes. The three points for each scenario correspond to procedures to estimate the SARA fractions and these techniques can
AAPD for data above Psat, overall, and below Psat (in increasing order). show significant differences [53,54]. Panuganti et al. briefly discussed
the sensitivity of SARA analysis on the predictive capability of the
SARA-based model to capture the effect of gas injection on asphaltene
characterization procedure. The SARA-based method splits the STO into precipitation [32]. The authors showed an example where two service
defined groups of pseudo-components (Saturates, Aro- laboratories provided significantly different results of SARA analysis for
matics + Resins + Asphaltenes) unlike the CS method which assumes the same crude, one of which was unreliable due to significant loss of
that the pseudo-components follow a specific mass distribution. volatiles. The authors showed that when the unreliable SARA analysis is

124
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table 10 three scenarios: SARA-based method + PR FT, SARA-based


Results of the sensitivity analysis of Saturates fraction on the density and method + PC-SAFT FT, and CS + PR FT. By fitting the FT simulation
viscosity modeling of Crude A using SARA-based + PR FT method. parameters to a single value of viscosity at saturation point, the visc-
Case Saturates wt% γ* KcPR** AAPD (overall) osity is predicted for the entire pressure range. The modeling results for
10 light crude oils from the Middle East region show that the predic-
Density Viscosity tions from the SARA-based + PC-SAFT FT model show the best match
to experimental data with an AAPD of 3.2%, while the CS + PR FT
+20 wt% 88.6 1.57 25.79 0.51% 4.87%
+15 wt% 83.6 1.02 11.87 0.53% 4.87% show the least match with AAPD of 6.4%. The viscosity predictions
+10 wt% 78.6 0.90 8.84 0.50% 4.81% from the three scenarios show consistent trends to the expected ex-
+5 wt% 73.6 0.72 7.59 0.47% 4.69% perimental uncertainty, since the AAPD above saturation pressure
Base 68.6 0.60 7.02 0.44% 4.60%
(single-phase region) is lower than AAPD below saturation pressure
−5 wt% 63.6 0.53 6.57 0.59% 4.81%
−10 wt% 58.6 0.46 6.44 0.52% 4.97% (two-phase region). The difference in modeling accuracy between the
−15 wt% 53.6 0.41 6.31 0.52% 5.13% three scenarios mainly occurs over data below the bubble point, which
−20 wt% 48.6 0.36 6.22 0.52% 5.28% are subject to significant experimental uncertainty due to the presence
of two phases. Therefore, in effect, the three modeling scenarios argu-
* Aromaticity is optimized in each case to match bubble point and density.
ably provide the same accuracy.
** Kc factor is optimized in each case to match viscosity at saturation.
The capability of the characterization procedure to capture the
phase behavior of the crude oil seems to have a pronounced effect on
used as an input to the model, the predictive capability of asphaltene the viscosity predictions using the FT model. It is found that there exists
phase behavior is compromised. a positive relationship between the AAPD in the viscosity predictions
In this investigation, SARA analysis for Crude A was originally re- and the AAPD in the density predictions. The highest deviation in
ported with 30 wt% loss of volatiles. As a result, the SARA analysis for viscosity predictions corresponds to the CS + PR FT scenario below
Crude H is used to represent Crude A since both are being produced saturation pressure (14.3%), which also shows the highest deviation in
from the same field and formation and have similar resins and as- density predictions (1.49%) in the same region.
phaltenes content. No significant advantage, for viscosity modeling purposes, is ob-
Under the current modeling framework using the SARA-based served from using PC-SAFT FT model as compared to PR FT. The overall
method, the Aromatics, Resins, and Asphaltenes are lumped into a AAPD for the SARA-based + PC-SAFT FT (3.2%) is not significantly
single pseudo-fraction (A + R + A). Therefore, it is only possible to different from the SARA-based + PR FT (3.7%). The PR FT model is
investigate the effect of Saturates (or A + R + A) on the modeling re- sufficient for the modeling of crude oil viscosity when coupled with the
sults, and not each fraction of SARA separately. The sensitivity analysis SARA-based characterization (which uses PC-SAFT for modeling the
is performed by changing the Saturates wt% then re-optimizing the phase behavior of the crude). The EoS used to capture the phase be-
aromaticity of A + R + A fraction (to match bubble point and density) havior has a more pronounced effect on the viscosity modeling than the
and KcPR (to match viscosity at saturation). Results of the sensitivity EoS used for the FT model.
analysis on predictions of density and viscosity are shown in Table 10. Previous work has shown the accuracy of the SARA-based method in
The results in Table 10 indicate that the predictive capability for modeling a wide range of PVT properties and asphaltene precipitation
live oil density and viscosity are not significantly compromised by in crude oils. The successful implementation of the PR FT to the SARA-
changing SARA. The aromaticity and KcPR are re-tuned to match the based characterization method expands the range of applicability of the
phase behavior and viscosity of the live oil. Re-tuning aromaticity, in SARA-based method to the modeling of live oil viscosity. In fact, visc-
effect, maintains the unique total aromatic character of the crude oil. As osity modeling can be now implemented in asphaltene deposition si-
observed from Table 10, when the Saturates fraction increases, a higher mulators that utilize the SARA-based method for the thermodynamic
aromaticity of the A + R + A fraction is needed to match the phase modeling of asphaltene precipitation. This work has shown that even
behavior of the oil and maintain the total aromatic character. One can with a lower number of fitting parameters and lower number of pseudo-
think of the total aromatic character of the STO as the aromaticity that components, the SARA-based characterization method yields results
would be obtained had the crude been characterized using the SLF that are at least as good as (was better) the CS characterization method.
approach. Since Saturates are assumed to consist of n-alkanes (γ = 0), The superior performance is attributed in part to the more physical
increasing the amount of Saturates leads to a higher aromaticity of the nature of the SARA-based method as compared to the CS method, and
A + R + A fraction. In fact, for Saturates fraction higher than 83 wt%, to the more powerful EoS used to capture the density and phase be-
an aromaticity of A + R + A that is greater than 1 is needed to match havior (PC-SAFT). The proposed method has been verified for light
the phase behavior. Aromaticity is defined to be zero for n-alkanes and crude oils and extension to heavy oils is the topic of future investiga-
one for PNA, which is the definition proposed by Punnapala and Vargas tion. Light oils, nonetheless, tend to be more problematic with regard to
[30]. An aromaticity greater than 1 means that the pseudo-fraction is asphaltene deposition since asphaltenes are immiscible in light com-
more aromatic than PNA as defined above. Although this is not ne- ponents. Therefore, the SARA-based method proposed in this work
cessarily unphysical, it is highly unlikely and is not reported in the provides a comprehensive and promising tool for flow assurance ap-
literature as per the definition of Punnapala and Vargas. The base case plications in crude oil systems.
aromaticity and KcPR values for Crude A show reasonable values since
they are very similar to those of Crude H. This supports our assumption Acknowledgments
of using SARA analysis of Crude H to represent Crude A.
This work was undertaken with the generous support of Abu Dhabi
7. Conclusion National Oil Company (ADNOC) Oil Subcommittee. The authors are
grateful to Dalia S. Abdallah and Sameer Punnapala from ADCO for
This work investigates the modeling of crude oil viscosity using fruitful discussions.

125
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Appendix A. Compositional analysis of crude oils

See Tables Table A.1–A.3

Table A.1
Compositional analysis of flashed gas for the crudes modeled in this work.
Flashed gas composition (mol%)

MW Crude A Crude B Crude C Crude D Crude E Crude F Crude G Crude H Crude I Crude J

N2 28.01 0.47 0.34 0.20 0.18 0.25 0.19 0.37 2.47 0.42 0.29
CO2 44.01 5.15 3.29 4.61 3.96 2.54 4.09 3.44 6.33 2.95 3.05
H2S 34.08 0 0.00 0.00 0.50 0.00 0.00 0.00 0.22 0.00 0
C1 16.04 39.27 44.04 53.32 56.45 49.79 53.94 56.23 45.65 49.62 49.15
C2 30.07 11.94 16.23 10.71 10.18 11.96 10.61 12.80 13.52 11.34 11.83
C3 44.10 13.52 18 10.07 8.83 11.79 9.99 11.71 13.93 10.91 11.59
iC4 58.12 3.93 2.84 2.42 2.14 2.79 2.36 2.77 3.52 2.81 3.01
nC4 58.12 9.12 8.23 6.08 5.32 6.97 6.05 6.17 7.23 6.75 7.11
iC5 72.15 3.87 2.14 2.50 2.30 2.94 2.67 2.01 2.3 3.06 3.08
nC5 72.15 4.64 2.69 3.16 2.90 3.72 3.36 2.27 2.38 3.85 3.79
C6 86.18 4.31 1.55 3.28 3.19 3.68 3.33 1.46 1.49 4.03 3.77
M-c-C5 84.16 0.00 0.13 0.40 0.41 0.38 0.37 0.00 0.00
Benzene 78.11 0.00 0.02 0.15 0.16 0.13 0.14 0.00 0.00
c-C6 84.16 0.00 0.05 0.50 0.53 0.52 0.42 0.00 0.00
C7 95.00 2.45 0.30 1.15 1.23 1.20 1.20 0.56 0.64
M-c-C6 98.19 0.00 0.03 0.24 0.28 0.23 0.24 0.00 0.00
Toluene 92.14 0.00 0.02 0.21 0.25 0.19 0.24 0.00 0.00
C8 108.8 0.93 0.06 0.55 0.64 0.54 0.48 0.17 0.22
C2-Benz. 106.17 0.00 0.00 0.04 0.05 0.04 0.05 0.00 0.00
m&p-Xyl. 106.17 0.00 0.00 0.05 0.07 0.05 0.07 0.00 0.00
o-Xyl. 106.17 0.00 0.00 0.02 0.03 0.03 0.01 0.00 0.00
C9 120.80 0.28 0.01 0.19 0.24 0.18 0.07 0.04 0.07
C10 140.20 0.03 0.01 0.02
C11 147.00 0.00 0.00 0.01

Plus fraction C10+ C12+ C10+ C10+ C10+ C10+ C12+ C12+ C7+ C7+
Plus fraction mol% 0.12 0.00 0.11 0.14 0.09 0.12 0.00 0.00 4.26 3.33
Plus fraction MW 142.46 – 143.1 136.8 136.9 134.0 – – 100.0 98.0

Table A.2
Compositional analysis of flashed liquid for the crudes modeled in this work.
Flashed liquid composition (mol%)

MW Crude A Crude B Crude C Crude D Crude E Crude F Crude G Crude H Crude I Crude J

N2 28.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
CO2 44.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
H2S 34.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
C1 16.04 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.01
C2 30.07 0.02 0.22 0.03 0.05 0.05 0.00 0.13 0.25 0.02 0.08
C3 44.10 0.27 1.65 0.28 0.24 0.35 0.10 0.86 1.10 0.21 0.45
iC4 58.12 0.27 0.62 0.21 0.15 0.27 0.13 0.63 0.77 0.20 0.33
nC4 58.12 1.08 3.20 0.90 0.69 1.08 0.63 2.39 2.63 0.85 1.24
iC5 72.15 1.22 2.05 0.98 0.65 1.17 0.81 2.32 2.34 1.03 1.3
nC5 72.15 2.04 3.62 1.71 1.22 2.06 1.45 3.74 3.40 1.80 2.2
C6 86.18 4.81 6.15 4.48 3.43 5.19 3.96 7.76 6.66 4.67 5.25
M-c-C5 84.16 0.00 0.87 0.87 0.98 0.93 0.89 0.00 0.00
Benzene 78.11 0.00 0.20 0.35 0.36 0.26 0.30 0.00 0.00
c-C6 84.16 0.00 0.46 0.59 0.53 0.66 0.53 0.00 0.00
C7 95.00 7.76 5.57 5.48 4.94 6.22 5.15 9.02 8.38
M-c-C6 98.19 0.00 0.75 1.27 1.29 1.36 1.19 0.00 0.00
Toluene 92.14 0.00 1.02 1.37 1.33 1.09 1.16 0.00 0.00
C8 108.80 9.34 5.93 6.52 7.02 7.43 6.54 9.4 9.00
C2-Benz. 106.17 0.00 0.59 0.77 1.53 0.74 0.74 0.00 0.00
m&p-Xyl. 106.17 0.00 0.53 1.64 1.12 1.64 1.57 0.00 0.00
o-Xyl. 106.17 0.00 0.31 0.76 0.66 0.79 0.74 0.00 0.00
C9 120.80 8.54 5.31 5.78 6.18 6.72 5.91 8.32 7.72
1,2,4TMB 120.19 0.00 0.00 0.96 0.95 0.67 1.05 0.00 0.00
C10 140.20 6.36 7.60 7.23
C11 147.00 5.40 6.00 5.83

Plus fraction C10+ C12+ C10+ C10+ C10+ C10+ C12+ C12+ C7+ C7+
Plus fraction mol% 64.65 49.19 65.05 66.70 61.32 67.15 41.83 44.7 91.22 89.14
Plus fraction MW 257.4 327.4 256.3 257.7 254.6 260.9 278.8 275.4 204 204.0

126
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table A.3
Compositional analysis of combined sample for the crudes modeled in this work.
Combined sample composition (mol%)

MW Crude A Crude B Crude C Crude D Crude E Crude F Crude G Crude H Crude I Crude J

N2 28.01 0.21 0.10 0.12 0.13 0.16 0.14 0.20 0.90 0.27 0.17
CO2 44.01 2.36 0.93 2.78 2.64 1.67 2.59 1.88 2.33 1.86 1.87
H2S 34.08 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.07 0.00 0.00
C1 16.04 18.02 12.39 32.18 37.47 32.63 34.01 30.78 16.81 31.24 29.93
C2 30.07 5.52 4.73 6.47 6.78 7.86 6.69 7.07 5.14 7.16 7.24
C3 44.10 6.49 6.25 6.19 5.94 7.85 6.34 6.80 5.83 6.95 7.23
iC4 58.12 2.02 1.24 1.54 1.47 1.92 1.53 1.80 1.79 1.84 1.95
nC4 58.12 4.97 4.62 4.02 3.76 4.92 4.03 4.46 4.34 4.56 4.81
iC5 72.15 2.57 2.08 1.89 1.73 2.33 1.98 2.15 2.33 2.31 2.38
nC5 72.15 3.41 3.36 2.59 2.32 3.15 2.65 2.93 3.04 3.09 3.16
C6 86.18 4.95 4.85 3.74 3.35 4.20 3.65 4.31 4.76 4.26 4.45
M-c-C5 84.16 0.00 0.66 0.58 0.6 0.532 0.480 0.00 0.00
Benzene 78.11 0.00 0.15 0.23 0.23 0.153 0.160 0.00 0.00
c-C6 84.16 0.00 0.34 0.53 0.53 0.533 0.400 0.00 0.00
C7 95.00 5.48 4.09 2.87 2.47 3.265 2.740 4.39 5.53
M-c-C6 98.19 0.00 0.55 0.65 0.62 0.677 0.590 0.00 0.00
Toluene 92.14 0.00 0.74 0.67 0.61 0.514 0.550 0.00 0.00
C8 108.80 5.40 4.27 2.91 2.77 3.702 3.180 4.35 5.77
C2-Benz. 106.17 0.00 0.42 0.33 0.54 0.331 0.320 0.00 0.00
m&p-Xyl. 106.17 0.00 0.38 0.68 0.42 0.709 0.670 0.00 0.00
o-Xyl. 106.17 0.00 0.22 0.32 0.24 0.341 0.310 0.00 0.00
C9 120.80 4.78 3.82 2.41 2.23 3.474 2.910 3.79 4.90
1,2,4TMB 120.19 0.00 0.00 0.38 0.32 0.309 0.460 0.00 0.00
C10 140.20 4.58 3.45 4.58
C11 147.00 3.88 2.71 3.96

Plus fraction C10+ C12+ C10+ C10+ C10+ C10+ C12+ C12+ C7+ C7+
Plus fraction mol% 33.82 35.35 25.9 22.37 21.19 24.72 18.94 28.24 36.43 36.78
Plus fraction MW 257.4 327.4 256.0 257.1 254.3 258.7 278.9 275.4 197.0 198.0

Appendix B. Binary interaction parameters

See Tables Table B.1–B.3

Table B.1
Binary interaction parameters for the SARA-based method (PC-SAFT EoS).
Component N2 CO2 H2S C1 C2 C3 Heavy gas Saturates A+R+A

N2 – 0 0.090 0.030 0.040 0.060 0.075 0.14 0.158


CO2 – 0.0678 0.05 0.097 0.1 0.12 0.13 0.1
H2S – 0.062 0.058 0.05 0.07 0.09 0.015
C1 – 0 0 0.01 0.03 0.029
C2 – 0 0.02 0.012 0.025
C3 – 0.015 0.01 0.01
Heavy gas – 0.005 0.012
Saturates – 0.007
A + R+A –

Table B.2
Binary interaction parameters for the SLF method (PC-SAFT EoS).
Component N2 CO2 H2S C1 C2 C3 Heavy gas SLF

N2 – 0 0.090 0.030 0.040 0.060 0.075 0.100


CO2 – 0.0678 0.057 0.097 0.107 0.090 0.100
H2S – 0.062 0.058 0.053 0.080 0.050
C1 – 0 0 0.080 0
C2 – 0 0.050 0
C3 – 0.030 0
Heavy gas – 0.067
SLF –

127
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

Table B.3
Binary interaction parameters for the Chi-Square method (PR EoS).
Component N2 CO2 H2S C1 C2 C3 i-C4 n-C4 i-C5 n-C5 C6 Heavy 1 Heavy 2 Heavy 3 Heavy 4

N2 – −0.02 0.165 0.036 0.05 0.085 0.095 0.095 0.095 0.095 0.1 0.08 0.08 0.08 0.08
CO2 – 0.097 0.1 0.13 0.135 0.13 0.13 0.125 0.125 0.125 0.15 0.15 0.15 0.15
H2S – 0.085 0.075 0.075 0.06 0.06 0.06 0.06 0.05 0.1 0.1 0.1 0.1
C1 – 0.002 0.007 0.012 0.012 0.017 0.018 0.024 0 0 0 0
C2 – 0.001 0.003 0.003 0.004 0.005 0.007 0 0 0 0
C3 – 0 0 0.001 0.002 0.003 0 0 0 0
i-C4 – 0 0 0 0.001 0 0 0 0
n-C4 – 0 0 0.001 0 0 0 0
i-C5 – 0 0 0 0 0 0
n-C5 – 0 0 0 0 0
C6 – 0 0 0 0
Heavy 1 – 0 0 0
Heavy 2 – 0 0
Heavy 3 – 0
Heavy 4 –

Appendix C. Supplementary data

A detailed step-by-step characterization is presented in the supplementary information document attached to this article. The example de-
monstrates how to obtain the EoS simulation parameters and the Friction Theory parameters. Also, the supplementary document contains tabulated
values of experimental data for viscosity and density at reservoir conditions for all crudes studied.
Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.fuel.2018.06.062.

References a north sea crude oil, natural gas liquid, and their mixtures †. Energy Fuels
2005;19(4):1303–13.
[21] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Stenby EH. Friction theory prediction
[1] Reid RC, Prausnitz JM, Poling BE. The properties of gases & liquids. fourth ed. New of crude oil viscosity at reservoir conditions based on dead oil properties. Fluid
York: McGraw-Hill Inc; 1987. Phase Equilib 2003;212(1–2):233–43.
[2] Mehrotra AK, Monnery WD, Svrcek WY. A review of practical calculation methods [22] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Baylaucq A, Boned C. Viscosity
for the viscosity of liquid hydrocarbons and their mixtures. Fluid Phase Equilib modeling and prediction of reservoir fluids: from natural gas to heavy oils. Int J
1996;117:344–55. Thermophys 2004;25(5):1353–66.
[3] Vesovic V, Trusler JPM, Assael MJ, Riesco N, Quiñones-Cisneros SE. Dense fluids: [23] Quiñones-Cisneros SE, Andersen SI, Creek J. Density and viscosity modeling and
viscosity. Assael MJ, Goodwin ARH, Vesovic V, Wakeham WA, editors. characterization of heavy oils. Energy Fuels 2005;19(4):1314–8.
Experimental thermodynamics volume IX : advances in transport properties of fluids [24] Kumar A, Henni A, Shirif E. Heavy oil viscosity modeling with friction theory.
Cambridge, UK: The Royal Society of Chemistry; 2014. Energy Fuels 2011;25(2):493–8.
[4] Viswanath DS. Viscosity of liquids: theory, estimation, experiment, and data. [25] Zéberg-Mikkelsen CK, Quiñones-Cisneros SE, Stenby EH. Viscosity prediction of
Dordrecht: Springer; 2007. carbon dioxide + hydrocarbon mixtures using the friction theory. Pet Sci Technol
[5] Monnery WD, Svrcek WY, Mehrotra AK. A critical review of practical predictive and 2002;20(1–2):27–42.
correlative methods. Canad J Chem Eng 1995;73(1):3–40. [26] Whitson CH. Characterizing hydrocarbon plus fractions. SPE 1983;23(04):683–94.
[6] Hirschfelder JO, Curtiss CF, Bird RB. Molecular theory of gases and liquids. Wiley; [27] Pedersen KS, Fredenslund A, Thomassen P. The properties of oils and natural gases.
1954. Gulf Publishing Company; 1989.
[7] Allan JM, Teja AS. Correlation and prediction of the viscosity of defined and un- [28] Ting PD, Hirasaki GJ, Chapman WG. Modeling of Asphaltene Phase Behavior with
defined hydrocarbon liquids. Canad J Chem Eng 1991;69(4):986–91. the SAFT Equation of State. Pet Sci Technol 2003;21(3–4):647–61.
[8] Orbey H, Sandler SI. The prediction of the viscosity of liquid hydrocarbons and their [29] Gonzalez DL, Hirasaki GJ, Creek J, Chapman WG. Modeling of asphaltene pre-
mixtures as a function of temperature and pressure. Canad J Chem Eng cipitation due to changes in composition using the perturbed chain statistical as-
1993;71(3):437–46. sociating fluid theory equation of state. Energy Fuels May 2007;21(3):1231–42.
[9] Twu CH. An internally consistent correlation for predicting the critical properties [30] Punnapala S, Vargas FM. Revisiting the PC-SAFT characterization procedure for an
and molecular weights of petroleum and coal-tar liquids. Fluid Phase Equilib improved asphaltene precipitation prediction. Fuel 2013;108:417–29.
1984;16(2):137–50. [31] Vargas FM, Gonzalez DL, Hirasaki GJ, Chapman WG. Modeling asphaltene phase
[10] Ely JF, Hanley HJM. Prediction of transport properties. 1. Viscosity of fluids and behavior in crude oil systems using the perturbed chain form of the statistical as-
mixtures. Ind Eng Chem Fundam 1981;20(4):323–32. sociating fluid theory (PC-SAFT) equation of state. Energy Fuels 2009;23:1140–6.
[11] Pedersen K, Fredenslund A, Christensen P, Thomassen P. Viscosity of crude oils. [32] Panuganti SR, Vargas FM, Gonzalez DL, Kurup AS, Chapman WG. PC-SAFT char-
Chem Eng Sci 1984;39(6):1011–6. acterization of crude oils and modeling of asphaltene phase behavior. Fuel
[12] Assael MJ, Dymond JH, Papadaki M, Patterson PM. Correlation and prediction of 2012;93:658–69.
dense fluid transport coefficients. I. n-Alkanes. Int J Thermophys [33] Abutaqiya MIL, Panuganti SR, Vargas FM. Efficient algorithm for the prediction of
1992;13(2):269–81. pressure–volume–temperature properties of crude oils using the perturbed-chain
[13] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Stenby EH. The friction theory for statistical associating fluid theory equation of state. Ind Eng Chem Res May
viscosity modeling: extension to crude oil systems. Chem Eng Sci 2001;56:7007–15. 2017;56(20):6088–102.
[14] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Stenby EH. The friction theory (f- [34] Vargas FM, Creek JL, Chapman WG. On the development of an asphaltene de-
theory) for viscosity modeling. Fluid Phase Equilib 2000;169(2):249–76. position simulator. Energy Fuels 2010;24(4):2294–9.
[15] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Stenby EH. One parameter friction [35] Kurup AS, et al. Development and application of an asphaltene deposition tool
theory models for viscosity. Fluid Phase Equilib 2001;178(1–2):1–16. (ADEPT) for well bores. Energy Fuels 2011;25(10):4506–16.
[16] N. US Department of Commerce, “nist23.” [Online]. Available: http://www.nist. [36] Kurup AS, Wang J, Subramani HJ, Buckley J, Creek JL, Chapman WG. Revisiting
gov/srd/nist23.cfm [Accessed: 16-Feb-2016]. asphaltene deposition tool (ADEPT): field application. Energy Fuels
[17] Schmidt KA, Carroll JJ, Quiñones-Cisneros SE, Kvamme B. Hydrogen Sulphide 2012;26(9):5702–10.
Viscosity Modelling. 86th Annual Convention of the Gas Processors Association, vol. [37] Gross J, Sadowski G. Perturbed-chain SAFT: an equation of state based on a per-
36. TX: San Antonio; 2007. turbation theory for chain molecules. Ind Eng Chem Res 2001;40(4):1244–60.
[18] er Zeberg-Mikkelsen CK, Quiñones-Cisneros SE, Stenby EH. Viscosity prediction of [38] Chapman WG, Jackson G, Gubbins KE. Chain molecules with multiple bonding
natural gas using the friction theory. Int J Thermophys 2002;23(2):437–54. sites. Mol Phys 1988;65(5):1057–79.
[19] Zéberg-Mikkelsen CK, Quiñones-Cisneros SE, Stenby EH. Viscosity prediction of [39] Chapman WG, Gubbins KE, Jackson G, Radosz M. New reference equation of state
hydrogen + natural gas mixtures (hythane). Ind Eng Chem Res for associating liquids. Ind Eng Chem Res 1990;29(8):1709–21.
2001;40(13):2966–70. [40] Dominik A, Chapman WG. Thermodynamic model for branched polyolefins using
[20] Schmidt KAG, Quiñones-Cisneros SE, Kvamme B. Density and viscosity behavior of the PC-SAFT equation of state. Macromolecules 2005;38(26):10836–43.

128
M.I.L. Abutaqiya et al. Fuel 235 (2019) 113–129

[41] Tumakaka F, Sadowski G. Application of the Perturbed-Chain SAFT equation of News, vol. 36. 1944. p. R-714-722.
state to polar systems. Fluid Phase Equilib 2004;217(2):233–9. [49] Quiñones-Cisneros SE, Zéberg-Mikkelsen CK, Fernández J, García J. General friction
[42] Gonzalez DL. Modeling of asphaltene precipitation and deposition tendency using theory viscosity model for the PC-SAFT equation of state. AIChE J
the PC-SAFT equation of state. Texas: Rice University; 2008. 2006;52(4):1600–10.
[43] Quiñones-Cisneros SE, Dalberg A, Stenby EH. PVT characterization and viscosity [50] Panuganti SR, Vargas FM, Chapman WG. Property scaling relations for nonpolar
modeling and prediction of crude oils. Petrol Sci Technol 2004;22(9):1309–25. hydrocarbons. Ind Eng Chem Res 2013;52(23):8009–20.
[44] Pedersen KS, Christensen PL. Phase behavior of petroleum reservoir fluids. Boca [51] Vargas FM, Chapman WG. Application of the One-Third rule in hydrocarbon and
Raton, FL: Taylor & Francis Group; 2007. crude oil systems. Fluid Phase Equilib 2010;290(1–2):103–8.
[45] Casella G, Berger RL. Statistical inference. second ed. Cengage Learning; 2002. [52] Kesler MG, Lee BI. Improve prediction of enthalpy of fractions. Hydrocarbon Proces
[46] Chung TH, Ajlan M, Lee LL, Starling KE. Generalized multiparameter correlation for 1976;55(3):153–8.
nonpolar and polar fluid transport properties. Ind Eng Chem Res 1988;27(4):671–9. [53] Fan T, Buckley JS. Rapid and accurate SARA analysis of medium gravity crude oils.
[47] Peng D-Y, Robinson DB. A new two-constant equation of state. Ind Eng Chem Energy Fuels 2002;16(6):1571–5.
Fundam 1976;15(1):59–64. [54] Kharrat AM, Zacharia J, Cherian VJ, Anyatonwu A. Issues with comparing SARA
[48] Uyehara OA, Watson KM. A Universal Viscosity Correlation. National Petroleum methodologies. Energy Fuels 2007;21(6):3618–21.

129

You might also like