Documentos de Académico
Documentos de Profesional
Documentos de Cultura
A R T I C L E I N F O A BS T RAC T
Keywords: Downhole oil-water separator is the most important component of downhole oil-water separation system. Its
Downhole oil-water separator structure has great eects on oil concentration of underow (OCU) thus requiring further optimization. This
Response surface methodology work aims to propose an optimization method for the structural optimization of downhole oil-water separator
Particle swarm optimization that consists of two series de-oiling hydrocyclones. Firstly, the signicant factors inuencing OCU are identied
Numerical simulation
by two-level Plackett-Burman Design (PBD) with twelve factors. Then the central levels of the ve signicant
Structural optimization
factors identied from PBD are determined by steepest ascent design. Secondly, response surface methodology
(RSM) is used to establish the second order model between the OCU and ve signicant factors for the
implementation of particle swarm optimization (PSO). Finally, the optimal structural parameters are obtained
by PSO algorithm. Computational uid dynamics (CFD) is employed to calculate the OCU for each particular
case and analyze the separation performance variations before and after optimization. The simulation results
show that compared with the original geometry and the best geometry in CCF design, the OCU of the optimized
decreases. And the separation performance of the optimized geometry is remarkably improved for the oil
droplets whose particle sizes are smaller than 35.78 m. Furthermore, laboratory experiments have been
conducted to validate the proposed optimization method. The experimental results conrm that the OCU also
reduces following the proposed optimization method. It can be summarized that PSO algorithm combined with
PBD, steepest ascent design and RSM can be an eective method for the structural optimization of the downhole
oil-water separator.
1. Introduction with higher water concentration (Amini et al., 2012). Compared with
the solid-liquid hydrocyclone (SLHC), it is more dicult to improve the
Downhole oil-water separation is an ecient, environmental- separation eciency of DOHC due to low density dierences between
friendly and energy-saving technology, which can realize a complete two liquids in the DOHC. The DOHC generally follow a conventional
production process including oil-water recovery, separation, oil lift and design and may have some limitations on separation performance
water injection in the same wellbore (Ogunsina and Wiggins, 2005; (Ghodrat et al., 2014a). Therefore, many previous works have been
Amini et al., 2012). Downhole oil-water separator is the most im- done to optimize the geometry of DOHC.
portant device of downhole oil-water separation system, and can Early, experiments and theoretical analysis are the main method to
determine the oil-water separation eect. In order to preventing the study hydrocyclones. In the1960s, the hydrocyclone used for oil-water
injection layer (waste layer or formations with high permeability) from separation was rstly studied by a group of researchers from
being blocked, the oil concentration of the liquid injected to the Southampton University (Thew, 1986). They designed a typical
injection layer should be as small as possible (Scaramuzza et al., DOHC with two cone sections and obtained some optimal parameters.
2001). In other words, the higher the separation eciency of downhole In particular, the ratio of the inlet diameter to the reference diameter
oil-water separator, the smaller the OCU. To improve the separation was 0.35; the ratio of the tailpipe diameter to the reference diameter
performance, the scheme of two de-oiling hydrocyclones in series is was 20. They also found that the large cone section with an angle of 20
often adopted in the design of downhole oil-water separator. Typically, and the small cone section with an angle of 1.5 could enhance
a de-oiling hydrocyclone (DOHC) is composed of one inlet stream separation eciency. Young et al. (1994) proposed a modied design
(tangential inlet in the cylindrical section) and two outlet streams, that called Amoco DOHC with a single cone section. They found that the
is overow stream with higher oil concentration and underow stream optimal ratio of the inlet diameter to the reference diameter was 0.25,
Corresponding author. Tel: +86 0546 6383140.
E-mail address: zcw860204@163.com (C. Zhao).
http://dx.doi.org/10.1016/j.petrol.2016.09.033
Received 21 June 2016; Received in revised form 14 September 2016; Accepted 19 September 2016
Available online 28 September 2016
0920-4105/ 2016 Elsevier B.V. All rights reserved.
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
and the optimal cone angle was 6. And a very slightly cone tailpipe was eciency could be improved approximately 10% when using a helical
better than a cylindrical tailpipe. For the droplet whose particle size form of inlet. Noroozi and Hashemabadi (2011) studied the eects of
smaller than 20 m, the Amoco DOHC could raise separation eciency dierent inlet chamber designs (exponential, conical, quadratic poly-
by 20% compared with the DOHC with two cone sections. Thew (2000) nomial body prole and standard design) on eciency. The simulation
proposed the rectangular tangential inlet introduced through a gentle results indicated that the separation eciency could be improved about
volute. Li et al. (2011a) found that the average separation eciency of 8% using exponential body shape. Saidi et al. (2013) investigated the
the DOHC using rectangular tangential inlet could be up to 99.4% in eects of dierent cone angles (6, 10 and 20) on the ow eld and
their experimental studies. Li et al. (2011b) also conducted experi- separation eciency of the DOHC using large eddy simulation. As a
ments and found that the DOHC with cylinder-cone-inverted cone result, the design with 6 cone angle showed the best separation
overow pipe reduced the short-circuit ow, and the average separa- performance.
tion eciency of the DOHC could be up to 98.1%. Belaidi and Thew However, no specic optimization techniques are used for systema-
(2003) proposed a new modied design to increase the separation of tic optimization of DOHC in the above works. Chu et al. (2000)
free gas based on the typical DOHC. In this design, a tapered swirl investigated the inuence of body and inlet chamber design of SLHC
chamber was used to separate gas at entry, a twin-tangential inlet was on separation eciency using orthogonal design method. Vieira et al.
employed to provide a stable and axisymmetric ow and a total angle of (2011) studying the eect of geometric parameters on the performance
2 was adopted to accelerate the swirling stream. of SLHC by CFD and RSM, found a better design leading to Euler
In recent years, CFD has been widely used in the analysis of internal number and cut size in minimum levels. Silva et al. (2012) used the
ow eld and structural optimization of DOHC (Ghodrat et al., 2014b). dierential evolution algorithm to improve the performance of SLHC.
Noroozi et al. (2009) investigated the eects of four dierent inlet The optimized conguration of hydrocyclone presented a high e-
designs on the DOHC eciency using an Eulerian-Eulerian three- ciency and a small reduced cut size. Silva et al., (2015) obtained an
dimensional CFD model. Their simulations showed that the separation optimized geometry for a ltering hydrocyclone with maximum total
Fig. 1. Structure diagram of the downhole oil-water separator: 1tubing; 2rod; 3tubing joint; 4screw pump; 5upper cross-channel nipple; 6the rst DOHC; 7shell; 8
central cross-channel nipple; 9the second DOHC; 10under cross-channel nipple; 11inner pipe; 12outer pipe; akow channel.
116
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
117
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
to determine the central level of the ve signicant factors obtained by tion problems developed by Eberhart and Kennedy in 1995 (Dong
PBD. et al., 2005). It is an evolutionary computation technique that starts
with a crowd of grain called as the swarm and mainly based on social
2.2.3. Response surface methodology models, such as bird ocking, sh schooling and the swarm theory (Rao
To attain the optimal solution using PSO algorithm, the mathema- and Patel, 2010). The basic equations of PSO algorithm (Rao and Patel,
tical relationship of the OCU to the ve signicant factors must be 2010) are as follows:
determined. RSM is a statistical method of multifactorial analysis of Vid (t + 1) = Vid (t ) + c1rand ()( pid xid (t )) + c2rand ()( pgd xid (t ))
experimental data which provides a better understanding of the
(2)
process than the standard experimental methods, because it is able
to predict how the inputs inuence the outputs in a complex process xi (t + 1) = xi (t ) + Vi (t + 1) (3)
where dierent factors can interact among themselves (ere et al.,
2016). Central composite face-centered (CCF) design is frequently used Where Vid is the velocity of particle i in d-dimensional space, xi(t)
in RSM because of its advantages in optimizing multifactor problems =(xi1, xi2, xi3, xin) is a vector with length n, and represents the
with optimum number of experimental runs (Mangili et al., 2015). position of particle i in ith generation; Vi(t)=(Vi1, Vi2, Vi3,Vin) is
Therefore, the RSM based on CCF design is chosen to build the also a vector with length n, and represents the velocity of particle i in
mathematical model between the OCU and ve signicant factors ith generation; represents inertia coecient;c1andc2 are the accelera-
and analyze the eect of each factor and their interactions. tion constants with positive values, and usually, c1=c2 =2 (Bergh and
Due to the desirable properties of second order polynomials, such Engelbrecht, 2006; Dong et al., 2005; Rao and Patel, 2010); rand() is a
as high predictability, robustness and simplicity, they are often used for random number in [0,1]; pid is the individual extreme value, pgd is the
global extreme value.
estimating response value and determining the size of eects (Karami
At present, a widely adopted dynamic is the linear decreasing
et al., 2016). The general form of a second order model is as below:
inertia weight proposed by Shi and Eberhart (1998), which can be
k k k
expressed as:
y = 0 + i xi + ii xi2 + ij xi xj +
i =1 i =1 i<j (1) (Tmax t )
(t ) = (ini end ) + end
Tmax (4)
Where y is the purposed response, here, it is OCU; xi values are the ith
independent variables of the function; value shows random error; 0 Where Tmax is the maximum iteration number; ini is the initial inertia
is a real constant of regression; i, j, ii and ij values are regression weight; end is the inertia weight when t=Tmax; in general, ini=0.9,
coecients of the main, interaction and quadratic terms. end=0.4.
118
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
number. positive eect on OCU, and the other factors have negative eect on it.
The physical parameters of materials under normal temperature The signicance of each parameter is evaluated by standardized main
and pressure are shown in Table 2. In addition, Table 3 shows the eect. The eects of the twelve factors on reducing OCU is in the
setting of boundary condition in this study. According to the actual following order: J > D > C > I > K > L > F > H > G > A > E > B. The small
working conditions, the boundary conditions are set as follows. cone angle 2 (J), small cone angle 1(D), large cone angle 1(C), large
The boundary style of inlet is set to velocity-inlet. According to the cone angle 2 (I) and tail pipe diameter Du2 (K) are considered as
rated ow rate and the cross-sectional area of the inlet (Fig. 3), velocity signicant factors and included for the next stage of optimization,
at the inlet can be calculated. The outlets including the rst overow, because their standardized main eect are not less than 2.365.
second overow and underow are set to pressure-outlet. The other
boundaries are set to wall. And no-slip boundary condition is selected 3.2. Determination of the central level for CCF design by steepest
at the wall. ascent design
In the following procedures including PBD, steepest ascent design
and CCF design, the OCU is obtained by using this numerical The steepest ascent path of the signicant factor was determined by
simulation method. The calculation work is performed on a work- the PBD results. Due to the large cone angle 2 and the tail pipe
station (Gistom W510) whose CPU is Xeon E5645 with speed of diameter Du2 having positive eects on OCU, their values should be
2.40 GHz. increased. While the small cone angle 2, small cone angle 1 and large
cone angle 1 having negative eects on OCU, their values should be
2.5. Laboratory experiment reduced. The values of other factors keep the same as listed in Table 1.
The steepest ascent design and results are shown in Table 6. It shows
The downhole oil-water separation process consists of extraction, that the OCU gradually decreases from experiment 1 to experiment 3;
separation, lifting and injection. In this experiment, only the separation the OCU tends to gradually increase in experiment 4. According to the
process is considered to test the separation performance of downhole principle of steepest ascent (Joyce and Leung, 2003; Yao et al., 2003),
oil-water separator. Therefore, an experimental joint was specially the minimum OCU may be between experiment 3 and experiment 4.
designed for this experiment (Fig. 4a). The ow diagram of experiment Thus, the parameters of experiment 3 are used as the central level of
is shown in Fig.4 b.The work ow of the system is as follows: the liquid the next response surface design.
from the pump passes through ow channel g (Fig. 1) into the rst
DOHC. After the rst separation, the oil-rich liquid ows back to the oil
tank through the rst overow channel and pipes; the water-rich liquid 3.3. Establishment of the equation for OCU using RSM
passes through channel h into the second DOHC. After the second
separation, the oil-rich liquid passes through the second overow In the CCF design, the eect of each factor was evaluated at three
channel and pipes into the oil tank; the liquid with trace oil ows back levels, low level (1), central level (0) and high level (+1). The central
to the oil tank. level of the ve factors is from the parameters of experiment 3 in
Globe valve V2, V3, V4 are used to control the outlet pressures of the steepest ascent design, the low level and high level are shown in
rst overow, second overow and underow for the adjustment of Table 7. The factors and response results of the CCF design are shown
split ratio. Globe valve V1 is served to control the inlet ow rate. Sample in Table 8.
valve L1, L2, L3, L4 are used for the collection of samples, including the The equation for OCU (Y) was obtained by performing multiple
inlet, rst overow, second overow and underow. P1, P2, P3 and P4 regression analysis, which could be written as:
represent pressure gauge; Q1, Q2, Q3 and Q4 represent turbine Y (OCU) = 2136.820 287.955X1 401.283X2 43.765X3 29.417X 4
owmeter. The centrifugal pump is adopted to drive the liquid in this
210.282X5
experiment, and its ow rate can be controlled by variable frequency
technique. + 87.295X12 + 111.295X22 + 1.456X32 + 1.706X42 + 23.122X52 + 52.125X1 X2
The following steps are proposed to prepare experimental media. + 7.406X1 X3 5.156X1 X 4 12.604X1 X5 + 4.406X2 X3 3.906X2 X 4
Firstly, prepare the oil-water mixture with oil (diesel) concentration of 12.188X2 X5 1.133X3 X 4 1.172X3 X5 1.536X 4 X5
50,000 mg/L. Secondly, add OP-12 emulsier to this mixture until its (5)
concentration is up to 15106 mg/L. Finally, keep the blender
Where X1, X2, X3, X4 and X5 are given by Table 7.
running 70 s at the speed of 940r/min. The oil droplets size distribu-
The adequacy of the regression model was checked using ANOVA,
tion is measured by the Mastersizer-S laser particle size analyzer. The
as shown in Table 9. Signicance of the ve factors and their
oil concentration of the sample is detected by spectrophotometric
interactions in the observed model were checked using the F-test.
method with the ultraviolet spectrophotometer (752 C).
The Model F-value of 224.66 implies the model is signicant. There is
only a 0.01% chance that a model F-value this large can occur due to
3. Results and discussion
noise. Values of Prob > F less than 0.05 indicate model terms are
signicant. In this case, linear terms X1, X2, X3, X4, X5, and quadratic
3.1. Screening of signicant factors using PBD
terms X12 and X22 are signicant for OCU. Interactive terms are also
signicant for OCU except X3X5 and X4X5.
In this study, a 20 runs PBD was performed to evaluate the twelve
The R2 value of the regression equation is 0.993 indicating that the
factors at 95% condence level. The eects of each factor were
model is t to explain OCU as a function of the ve factors. The value of
evaluated at two levels, i.e. high level (+1) and low level (1). Based
adjusted R2 is 0.989, which indicates that only1.1% of the variations
on the results obtained by our research group in previous studies, high
can not be explained by this model.
level should be 1.252 times larger than low level (Zhao et al., 2014b).
The level of each factor is shown in Table 4. The factors and response
Table 2
(OCU) results of the PBD are shown in Table 5. Physical parameters of materials.
The standardized Pareto charts (P < 0.05) of main eects is
depicted in Fig. 5. The critical values representing the statistically Material Density (kg/m3) Viscosity (Pa s) Median particle size (m)
signicant eect of factors at 95% condence level is 2.365 for the 4
Water 997 8.9410
response. It illustrates that overow diameter Do1 (A), large cone angle Oil 868 3.46103 30
2 (I), tail pipe diameter Du2 (K) and tail pipe length Lu2 (L) have
119
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
Table 3 Table 4
Setting of boundary condition. Factors and levels in PBD.
Fig. 4. Flow diagram of experiment and experimental joint: (a) experimental joint and (b) ow diagram of experiment.
120
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
Table 8
Factors and response results of CCF design.
X1 X2 X3 X4 X5
1 0 1 0 0 0 217
2 0 0 0 0 0 155
3 0 0 1 0 0 174
4 1 1 1 1 1 132
5 1 1 1 1 1 177
6 1 1 1 1 1 233
7 1 1 1 1 1 113
8 1 1 1 1 1 150
9 0 1 0 0 0 147
10 1 1 1 1 1 241
11 1 1 1 1 1 246
12 1 1 1 1 1 147
13 0 0 0 0 0 155
14 1 1 1 1 1 367
15 1 1 1 1 1 332
16 1 1 1 1 1 134
17 0 0 0 0 0 155
18 0 0 0 1 0 148
19 1 1 1 1 1 395
Fig. 5. Standardized (P < 0.05) main eect Pareto charts of the PBD for OCU. 20 0 0 0 0 0 155
21 0 0 0 0 0 155
22 1 1 1 1 1 212
Table 6
23 0 0 0 1 0 174
Steepest ascent design and results.
24 0 0 0 0 0 155
25 0 0 1 0 0 146
Run Small Small Large Large Tail pipe OCU
26 1 1 1 1 1 435
order cone cone cone cone diameter (mg/L)
27 0 0 0 0 0 155
angle angle angle angle Du2 (mm)
28 1 1 1 1 1 175
2 () 1 () 1 () 2 ()
29 1 0 0 0 0 230
30 1 1 1 1 1 200
1 3 3 25 17 5.2 264
31 1 1 1 1 1 154
2 2.5 2.5 23 19 5.8 203
32 1 1 1 1 1 266
3 2 2 21 21 6.4 151
33 1 1 1 1 1 214
4 1.5 1.5 19 23 7 184
34 0 0 0 0 0 155
5 1 1 17 25 7.6 219
35 0 0 0 0 0 155
6 0.5 0.5 15 27 8.2 257
36 1 1 1 1 1 158
37 1 1 1 1 1 230
38 1 1 1 1 1 304
Table 7 39 0 0 0 0 1 155
Factors and levels of CCF design. 40 1 1 1 1 1 186
41 1 0 0 0 0 122
Factors Symbol level 42 1 1 1 1 1 317
43 1 1 1 1 1 121
Central (0) Low (1) High (+1) 44 1 1 1 1 1 195
45 1 1 1 1 1 301
Small cone angle 2() X1 2 1.5 2.5 46 0 0 0 0 0 155
Small cone angle 1() X2 2 1.5 2.5 47 1 1 1 1 1 350
Large cone angle 1() X3 21 19 23 48 1 1 1 1 1 264
Large cone angle 2() X4 21 19 23 49 0 0 0 0 1 170
Tail pipe diameter Du2(mm) X5 6.4 5.8 7 50 1 1 1 1 1 129
51 1 1 1 1 1 106
52 1 1 1 1 1 225
3.5. Comparative analysis of structural parameters before and after
optimization
small cone section will become very long, which can lead to the
Compared with initial structure parameters, ve structural para- reduction of centrifugal eld and the increase of OCU. However, due
meters have changed after optimization. As shown in Fig. 7, the small to the increase of large cone angle 2, the large cone section becomes
cone angles 1, 2 and the large cone angle 1 decrease, while the large steeper. And the rotary speed of the uid entering into the small section
cone angle 2 and the tail pipe diameter Du2 increase. will be accelerated, which makes up for the reduction of centrifugal
The reduction of the large cone angle 1 and small cone angles 1 of eld intensity in the small section and enhances the chance of
the rst DOHC can increase the residence time of oil droplets without separating more droplets. Moreover, it can be noticed from Fig. 9 that
damping the swirling stream. Therefore, more oil droplets will be the optimal OCU will be observed when the small cone angle 2 is small
separated with full radial migration time thus reducing the OCU. and the large cone angle 2 is large. Besides, the increase of tail pipe
Furthermore, it can be seen from the response surface plot and diameter Du2 avoids a too long small cone section in the second DOHC,
corresponding contour plot as shown in Fig. 8 that in a certain range, and ensures the centrifugal eld intensity in the small cone section is
the smaller the large cone angle 1 and the smaller the cone angle 1, strong enough. And as shown in Fig. 10, with the reduction of the small
the lower is the OCU, as expected. cone angle 2 and the increase of the tail pipe diameter Du2, the OCU
The reductive amplitude of the small cone angle 2 is larger than gradually decreases in a certain range.
that of the small cone angle 1. If the small cone angle is too small, the
121
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
Table 9 Table 11
ANOVA for multiple regression model. Optimized results of PSO algorithm.
122
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
Fig. 8. Surface and contour plot showing interactions between small cone angle 1 and large cone angle 1 on OCU while the other three factors at central level.
Fig. 9. Surface and contour plot showing interactions between small cone angle 2 and large cone angle 2 on OCU while the other three factors at central level.
Fig. 10. Surface and contour plot showing interactions between small cone angle 2 and tail pipe diameter Du2 on OCU while the other three factors at central level.
obvious oil core can be found in the second DOHC. whose particle sizes are not smaller than 35.78 m, the OCU decreases
The OCU of each particle size for the original and optimized slightly after optimization. However, for the oil droplets whose particle
geometries is shown in Fig. 13. It can be seen that for the oil droplets sizes are smaller than 35.78 m, the OCU reduces greatly. Therefore,
123
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
Fig. 11. Oil droplet size distribution in inlet: (a) actual measured distribution and (b) simulation model distribution.
the separation performance of the optimized geometry is remarkably geometry) and 121 mg/L (the best geometry in CCF design), respec-
improved for the oil droplets with small particle size. tively. Compared with original geometry and the best geometry in CCF
design, the OCU of the optimized geometry is reduced by 54.7% and
3.7. Validation of laboratory experiment 16.5%, respectively.
On the basis of the above analyses, it can be summarized that the
In order to validate the eectiveness of the proposed optimization OCU of the optimized downhole oil-water separator signicantly
method, laboratory experiments were carried out. The physical para- reduces. At the same time, a general agreement between the numerical
meters of the experimental media can be in agreement with that of the result and experimental result is obtained. Therefore, the PSO algo-
numerical simulation by using the method in Section 2.5 to prepare rithm combined with PBD, steepest ascent design and RSM can be an
experimental media. The outlet pressures of overow and underow eective method for structural optimization of the downhole oil-water
can be in accordance with that of the numerical simulation by adjusting separator.
the globe valve V1, V2, V3 and V4. The OCUs obtained by the laboratory
experiments are 223 mg/L (original geometry), 101 mg/L (optimized
124
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
References
Amini, S., Mowla, D., Golkar, M., Esmaeilzadeh, F., 2012. Mathematical modelling of a
hydrocyclone for the down-hole oil-water separation (dows). Chem. Eng. Res. Des.
90, 21862195.
Belaidi, A., Thew, M.T., 2003. The eect of oil and gas content on the controllability and
separation in a de-oiling hydrocyclone. Chem. Eng. Res. Des. 81, 305314.
Bergh, F.V.D., Engelbrecht, A.P., 2006. A study of particle swarm optimization particle
trajectories. Inf. Sci. 176, 937971.
Brennan, M., 2006. CFD simulations of hydrocyclones with an air core: comparison
between large eddy simulations and a second moment closure. Chem. Eng. Res. Des.
84, 495505.
Chu, L.Y., Chen, W.M., Lee, X.Z., 2000. Eect of structural modication on hydrocyclone
performance. Sep. Purif. Technol. 21, 7186.
Delfos, R., Murphy, S., Stanbridge, D., Oluji, ., Jansens, P.J., 2004. A design tool for
optimising axial liquidliquid hydrocyclones. Miner. Eng. 17, 721731.
Dong, Y., Tang, J., Xu, B., Wang, D., 2005. An application of swarm optimization to
nonlinear programming. Comput. Math. Appl. 49, 16551668.
Grady, S.A., Wesson, G.D., Abdullah, M., Kalu, E.E., 2003. Prediction of 10-mm
hydrocyclone separation eciency using computational uid dynamics. Filtr. Sep.
40, 4146.
Ghodrat, M., Kuang, S.B., Yu, A.B., Vince, A., Barnett, G.D., Barnett, P.J., 2014a.
Numerical analysis of hydrocyclones with dierent conical section designs. Miner.
Eng. 62, 7484.
Ghodrat, M., Kuang, S.B., Yu, A.B., Vince, A., Barnett, G.D., Barnett, P.J., 2014b.
Numerical analysis of hydrocyclones with dierent vortex nder congurations.
Miner. Eng. 63, 125138.
Fig. 13. Oil concentrations in underow of pre-optimization and post-optimization. Joyce, A.P., Leung, S.S., 2013. Use of response surface methods and path of steepest
ascent to optimize ligand-binding assay sensitivity. J. Immunol. Methods 392,
1213.
4. Conclusions Karami, H.R., Keyhani, M., Mowla, D., 2016. Experimental analysis of drag reduction in
the pipelines with response surface methodology. J. Pet. Sci. Eng. 138, 104112.
Kharoua, N., Khezzar, L., Nemouchi, Z., 2010. Hydrocyclones for de-oiling applications
An optimization study was conducted in the downhole oil-water a review. Petrol Sci. Technol. 28, 735755.
separator by using PSO algorithm combined with PBD, steepest ascent Li, D., Zhou, X., Zhou, H., 2011a. Experimental study on inlet structure of the rod pump
design and RSM in order to obtain a geometry that could minimize with down-hole oil-water hydrocyclone. Procedia Eng. 18, 369374.
Li, D., Zhou, X., Chen, Y., 2011b. Experimental study on overow pipe structure of the
OCU of this separator. Before and after optimization, the variations of rod pump with down-hole oil-water hydrocyclone. Procedia Eng. 18, 387391.
separation performance were investigated by CFD simulation. Mangili, I., Lasagni, M., Huang, K., Isayev, A.I., 2015. Modeling and optimization of
Laboratory experiments were conducted to validate the optimization ultrasonic devulcanization using the response surface methodology based on central
composite face-centered design. Chemom. Intell. Lab. Syst. 144, 110.
method. Based on the above study, the following conclusions can be
Noroozi, S., Hassan, S., Hashemabadi, S.H., 2009. CFD simulation of inlet design eect
achieved: on de-oiling hydrocyclone separation eciency. Chem. Eng. Technol. 32,
18851893.
Noroozi, S., Hashemabadi, S.H., 2011. CFD analysis of inlet chamber body prole eects
(1) Small cone angle 2, small cone angle 1 and large cone angle 1
on de-oiling hydrocyclone eciency. Chem. Eng. Res. Des. 89, 968977.
are signicant factors having negative eect on OCU, while large Ogunsina, O.O., Wiggins, M.L., 2005. A review of downhole separation technology. In:
cone angle 2 and tail pipe diameter Du2 are signicant factors Proceedings of SPE Production and Operations Symposium. Society of Petroleum
having positive eect on OCU. Engineers.
Priyadharshini, S.D., Bakthavatsalam, A.K., 2016. Optimization of phenol degradation by
(2) The optimal structural parameters obtained are: the small cone the microalga Chlorella pyrenoidosa using placket-burman design and response
angle 2 is 1.5, the small cone angle 1 is 1.8, the large cone angle surface methodology. Bioresour. Technol. 207, 150156.
1 is 19.9, the large cone angle 2 is 22.6 and the tail pipe Rao, R.V., Patel, V.K., 2010. Thermodynamic optimization of cross ow plate-n heat
exchanger using a particle swarm optimization algorithm. Int. J. Therm. Sci. 49,
diameter Du2 is 6.7 mm. Compared with the original structural 17121721.
parameters, it is found that the variations of the structural Saidi, M., Maddahian, R., Farhanieh, B., 2013. Numerical investigation of cone angle
parameters tend to reduce OCU. eect on the ow eld and separation eciency of deoiling hydrocyclones. Heat.
Mass Transf. 49, 247260.
(3) The CFD simulation results show that the separation performance Scaramuzza, J.L., Fischetti, H., Strappa, L., Figliuolo, S., 2001. Downhole oil/water
of the optimized geometry is remarkably improved for the oil separation system-eld pilot-secondary recovery application project. In: Proceedings
droplets whose particle sizes are smaller than 35.78 m in of SPE Latin American and Caribbean Petroleum Engineering Conference. Society of
Petroleum Engineers.
comparison with that of the original geometry.
Shi, Y., Eberhart, R.C., 1998. Parameter selection in particle swarm optimization. In:
(4) The experimental results conrm that compared with the original Proceedings of the 7th International Conference on Evolutionary Programming VII.
geometry and the best geometry in CCF design, the OCU of the Springer-Verlag.
Silva, D.O., Vieira, L.G.M., Lobato, F.S., Barrozo, M.A.S., 2012. Optimization of the
optimized geometry is reduced by 54.7% and 16.5% respectively,
design and performance of hydrocyclones by dierential evolution technique. Chem.
which are close to the numerical results 52.4% and 16%. Therefore, Eng. Process. 61, 17.
the separation performance of the downhole oil-water separator Silva, N.K.G., Silva, D.O., Vieira, L.G.M., Barrozo, M.A.S., 2015. Eects of underow
can be improved by using PSO algorithm combined with PBD, diameter and vortex nder length on the performance of a newly designed ltering
hydrocyclone. Powder Technol. 286, 305310.
steepest ascent design and RSM. ere, Z., Maravi, N., Takai, A., Nikoli, I., oronja-Simovi, D., Joki, A., Hodur, C.,
2016. Treatment of vegetable oil renery wastewater using alumina ceramic
membrane: optimization using response surface methodology. J. Clean. Prod. 112,
31323137.
Thew, M.T., 1986. Hydrocyclone redesign for liquid-liquid separation. Chem. Eng. 427,
1723.
Acknowledgement Thew, M.T., 2000. Flotation | cyclones for oil/water separations. In: Wilson, I.D. (Ed.),
Encylopedia of Separation Science. Academic Press, 14801490.
Vieira, L.G.M., Silvrio, B.C., Damasceno, J.J.R., Barrozo, M.A.S., 2011. Performance of
The author would like to thank the nance supported by the hydrocyclones with dierent geometries. Can. J. Chem. Eng. 89, 655662.
National High Technology Research and Development Program of Yao, C., Liu, C., Chu, I., Hsieh, T., Hwang, S., 2003. Factorial designs combined with the
steepest ascent method to optimize serum-free media for ex vivo expansion of
China (No. 2006AA06Z224) and the Fundamental Research Funds for
human hematopoietic progenitor cells. Comput. Enzym. Micro Tech. 33, 343352.
the Central Universities (No.10CX05009A and No. 13CX06077A). Young, G.A.B., Wakley, W.D., Taggart, D.L., Andrews, S.L., Worrell, J.R., 1994. Oil-water
125
C. Zhao et al. Journal of Petroleum Science and Engineering 148 (2017) 115126
separation using hydrocyclones an experimental search for optimum dimensions. system with screw pump. Zhongguo Shiyou Daxue Xuebao 37,
J. Petrol. Sci. Eng. 11, 3750. 129133, (in Chinese).
Zhao, C., Li, Z., 2011. Design of down-hole oil-water separation and reinjection in the Zhao, C., Li, Z., Dong, X., Zhang, X., 2014a. Numerical simulation and experiment of
same well system with surface driving single screw pump. Adv. Mater. Res. 339, downhole two-stage tandem hydrocyclone. Acta Pet. Sin. 3, 551557, (in Chinese).
595602. Zhao, C., Li, Z., Xu, W., Wu, H., 2014b. Structural parameter optimization of downhole
Zhao, C., Li, Z., Dong, X., 2013. Design and ground test of downhole oil-water separation dual-stage tandem hydrocyclone. Chem. Eng. 42, 4853, (in Chinese).
126