PETROLEUM SCIENCE & ENGINEERING

ELSEVIER
Journal of Petroleum Science and Engineering 11 ( 1994 ) 37-50

Oil-water separation using hydrocyclones: An experimental search for optimum dimensions

G.A.B. Young, W . D . Wakley, D.L. Taggart, S.L. Andrews, J.R. Worrell
Amoco Production Company, Tulsa, OK 74133, USA (Received March 15, 1993; revised version accepted August 31, 1993 )

Abstract

This paper describes Amoco's research to evaluate existing oil/water hydrocyclone technology and to develop an optimized version of the device. The performance of hydrocyclones having dimensions as designed by Martin Thew and Derek Colman at the University of Southampton was determined. The effects of operational variables on these devices was also evaluated. Then the search for ideal hydrocyclone dimensions was conducted with an investigation of inlet size, cylindrical diameter, cone angle, straight section length/size and processing rate. The performance of the patented optimized hydrocyclone that resulted from this study is compared with the performance of a 35-mm hydrocyclone built to the Thew model. The paper also describes the test apparatus and methods used to evaluate the performance of these hydrocyclones and other oil/water separation equipment.
1. Introduction

Since produced oil is often accompanied by significant amounts of water, it is necessary to provide facilities to separate the oil and water before the oil can be sold. An initial separation is often made by a production separator which is a small baffled tank that separates most of the oil from the produced water. The small quantity of oil remaining in the water must be separated further for disposal on offshore locations or re-injection in onshore operations. In onshore operations, this separation is typically accomplished by the use of large gravity settling vessels. Offshore locations in the past have used tankagebased systems which inject a fine mist of air in combination with surfactants to float the oil droplets to the surface. In recent years, because of the platform cost to accommodate these tankage-based systems, the oil industry has been

turning to the use of liquid/liquid hydrocyclones to meet this need. The modern renaissance in de-oiling hydrocyclones was instigated by Martin Thew and Derek Colman at the University of Southampton (Woodruff, 1990). The first publication of the major gains in de-oiling hydrocyclones was at the BHRA hydrocyclone conference in October 1980 (Colman and Thew, 1980a,b). Since 1980 the use of hydrocyclones for offshore produced brine cleanup has been extensive (Medlrum, 1987; Colman and Thew, 1980b). The primary advantage of oil/water separation hydrocyclones is minimum space requirement and lack of motion sensitivity. Another important advantage is the resultant oil-rich stream readily coalesces and is not highly chemically treated (Medlrum, 1987 ). The Colman and Thew patents on de-oiling hydrocyclones cover a wide range of design ratios. The authors first conducted tests on a

0920-4105/94/$07.00 1994 Elsevier Science B.V. All fights reserved SSD10920-4105 (93) E0069-2

38
Nomenclature
Oc

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50
lcyl lu

4 do ao

leone
OL
vortex

l.
Inlet width

Dp

t"o
#vis D5o

Inside diameter of a hydrocyclone measured in the cylindrical portion Equivalent inlet diameter, the diameter of a circle with the equivalent entry area Inside diameter of the overflow (or the vortex finder) of a hydrocyclone Inside diameter of the underflow of a hydrocyclone Length of the cylindrical section of a hydrocyclone Length of the cone part of a hydrocyclone, measured as though the cone came to a point Cone included angle Length of vortex finder measured from the top of the hydrocyclone The straight portion (tail) of the hydrocyclone discharge The width of the rectangular inlet. For the experimental hydrocyclone the inlet width was half the width of the annulus formed between the cylinder section and the vortex finder outer diameter Pressure drop measured from the inlet to the underflow of the hydrocyclone in feet of head Overflow pressure measured in feet of head Plastic viscosity of the mud measured in
centipoise

] ~ - du

I
Dc = 2.75 lcyl
1

di Dc

do

- .226

Dc

.014 .028 .042

c~ = 20

d u = 1 in.

Dc
lU

13

De Fig. 1.35-mm Colman-Thew hydrocyclone.

Feed

o~,now (~lghtl ~

_~O
Off ~Reduclng ~ Core ~ ~ parallel

Underflow (heavy_)

Feed~ (ltghO ~ .

2.

Ondernow (heavy}

Invo

Underflow (heavyl

d50

Q ,ap Eff.(e)

Mean particle size of feed mud Mean particle size removed by a piece of equipment Volume flow rate ( g p m ) Density difference, pwater-oil ( g / m l ) Percent ofoil removed

3.
Overflow (light)

c>

Fig. 2. Evolution of Colman-Thew design.

35 ram, commercial implementation of this hydrocyclone having the dimensions shown in Fig, 1. The effects of operational parameters on this design were investigated to establish a baseline. Then tests were conducted on an experimental hydrocyclone in which the various dimensions of the hydrocyclone could be varied incrementally in order to determine optimum design parameters. The goal of this research was to develop an optimized oil/water separation hydrocyclone to remove oil from water at the maximum removal efficiency. Using these optimum design parameters, a prototype hydrocyclone was designed incorporating materials and manufacturing methods used in the production of existing solid/liquid hydrocyclones. The result is an economical, high

thruput hydrocyclone for use in production operations where space, weight or other considerations exclude tankage-based separation facilities.

2. Discussion Flow enters the hydrocyclone tangentially (Fig. 2), creating a spinning motion in the chamber. This spinning fluid causes the lighter oil to separate to the center where it forms a small core one to two millimeters in diameter. The bulk of the liquid reports to the underflow and the oil concentrate reports to the overflow. Typically, less than 5% of the inlet flow reports to the overflow. Before considering the performance of the hydrocyclone with respect to the various variables, let us consider the definition of underflow efficiency as given below, where the concentrations

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

39

(C) are measured in mg/liter and the flow rates (Q) in liter/minute: Oil discarded Total efficiency () - O i l presented at the feed

agency discharge limits for certain oil producing locations (Hayes et al., 1985; Cornitius, 1986; Medlrum, 1987).

(1)
Utilizing continuity:

Discharge
100 ppm

Agency/location
International Maritime Organization Tankers (Midocean) Red Sea, Mediterranean Sea U.S. Pollution Regulations Gulf of Mexico U.K. Department of Energy North Sea Victoria Petroleum Act, 1976 Victoria, Australia

Coverflowaoverflow = Cinlet Oinlet

Cunderfiowaunderflow
(2)

15 ppm 29 mg/1 (monthly avg. ) 40 mg/kg

And: CovernowQove~flow (]inlet Qinlet Cove~owQovorflow (3)

30 mg/1 (avg.)

Cov~ow+ C..do~owQ..~raow
Equation 3 can also be rewritten using continuity:

Cinlet Qinlet

CunderflowQ..d~rnow Cinlet Qinlet

-

=I-

CunderflowOunderflow
CinletQinl,t

(4)

Since a very small amount of flow is taken out the overflow, Qunde~now/Qinletis almost equal to one. This is the basis for the classic definition used by many authors when considering oil/ water separation:

A typical example for an oil/water separation system might have inlet concentrations of 1000 mg/1 and water discharge limits of no greater than 40 mg/l. Conservation of mass for this case requires an efficiency of 96% ( 1000- 40) / 1000 X 100. In other words, 96% of the oil must be removed from a feed stream of 1000 mg/1 to obtain 40 mg/1 water discharge. Figure 3 shows the required underflow efficiency versus feed oil concentration to obtain 40 mg/1 oil in the discharged water.

E=I

Cunderflow Cinlet

(5)

The efficiency (E) term in this equation is often called underflow efficiency, percent separated or simply efficiency. Equation 3, which is slightly more exact than Eq. 5, was used in all of the reported test results. As a result, Cov~mow,Qow~now, C, nd~rnowand Ounderflow were the measured variables. Cin~t and Qinlet were also measured and continuity checked. See Appendix A and B for measurement procedures. Now let us consider the implications of separator performance on the design of an oil/water separation system. This will give the reader an insight into the efficiencies necessary to clean up produced brine to the required discharge limits. The following table shows typical regulatory

Efficiency R e q u i r e d to ObtalI1 40 r a g / l i t e r of Off a t t h e Ullderflow i00_

80_

60 _

40 _

rrl fc I ~wc~ Eff = rhFcr Eff = Cr - c~ cr

20 _

~ 0 0 Feed 500 1000 mg/llter I 1500

Oil C o n c e n t r a t i o n ,

Fig. 3. Efficiency requirement.

40

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

3. Test results

For the hydrocyclone problem there are at least fourteen variables as shown in Fig. 4. These variables have been grouped into (a) dimensional variables; such as hydrocyclone diameter, inlet diameter, overflow diameter, etc., (b) operational variables; pressure drop between the feed inlet and the water discharge, and water discharge pressure, (c) feed stream variables; viscosity of the production water, oil/water density difference, oil concentration, and droplet size distribution. The result of running a test with any given set of variables is removal efficiency (total and with respect to droplet size), and water treatment rate. Initially, the effect of operational variables on the 35-mm Colman/Thew type hydrocyclonewas determined. The 35-mm hydrocyclonewas tested to determine the effects of underflow pressure, differential pressure across the hydrocyclone, overflow pressure, overflow size, inlet oil concentration, oil droplet distribution, and oil/water density differential. The 35-mm hydrocyclone was tested using Bumpass (0.85 g/cm 3) crude and later with South China Sea (0.95 g/cm 3) crude. Both were dispersed in fresh water. Both single-stage separation and double-stage separation tests were run. Effect o f flow rate or differential pressure The separation of oil droplets in the swirl chamber of the hydrocyclone is a result of the forces imposed on the oil droplets in the spin-

ning fluid and the residence time in the chamber. Lower flow rates mean longer residence times but lower acceleration forces. Conversely higher flow rates result in higher acceleration forces and smaller residence times. The 35-mm hydrocyclone's performance (Fig. 5) is independent of flow rate between 20 and 37 gallons per minute, corresponding to a pressure differential of 30 to 100 psi between the inlet and the underflow. The flow rate (thruput) of the hydrocyclone is a significant function of the pressure drop between the inlet and the underflow as shown in Fig. 6. Although not carried out in a rigorous manner because of lack of control of droplet diameter, Meldrum (1987) found the 35-mm Thew type hydrocyclone performance to be constant between 15 and 50 gpm.

35 iron Colman-Thew Hydrocyclone and Bumpass U n d e r f l o w Efficiency "as O v e r f l o w V o l u m e P e r c e n t 100_ Feed d~ = 35 ~m

LJ 0 -~

C r u d e Oil Removed

80_

~J

60_

O~ 4 0 _ E] O

Legend
DP=90 DP-70 DP=50 DP-30 psi, psi, psi, psi Q-37.4 Q-322 Q=268 Q-213 gp/n gpm gpm gpm

20_

Overflow

2 Volume

Percent,

Qo/Qi

Fig. 5. Effect of pressure drop and flow rate on separation.

35 mm Colman-Thew Hydrocyclone F l o w R a t e v s P r e s s u r e D r o p f r o t n lrllet to U n d e r l l o w

50_

Ivort~

" Con7 Angle Ct

40_

Dimensional Variables (9)

dl, do, I~, lcyi, du, lu, a, ivortex, Inlet type

~j 3 0 -

C~
20-

Operating Variables (2)

Dp, Feed Punderflow Stream Parameters

_o
I0_

vi$ production water, oil/water density difference, droplet sl~, off ~ncentraUon Results off Removal /Era, Percent Separated C u ~ e for Device, Treatment Rate (Q~)

~o

4o'
Pressure

"o
Drop, psi

do

1O 0

'

Fig. 4. Hydrocyclone dimensions and parameters.

Fig. 6. Flow rate versus pressure drop.

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

41

Effect of underflow pressure

Backpressure must be applied at the hydrocyclone underflow to force the small stream containing the oil to the overflow; otherwise, it would all come out the underflow, and no separation would be made. For a given backpressure at the underflow, if the overflow pressure is slowly increased, the core up the center of the cone elongates, ultimately resulting in part of the oil core discharging out the underflow. This results in the curves trailing abruptly downward at low volume percent overflow discharges (Figs. 7 and 8 ). The 35-mm hydrocyclone performance is independent of the underflow pressure (Fig. 7 ), provided there is enough backpressure to force enough oil/water out the overflow. This takes a minimum of 50 psi, although for field use, a set point of at least 60 psi to 75 psi is recommended. It is also critical that a constant backpressure be
35 i r o n C o l m a n - T h e w H y d r o c y c l o n e a n d B t m l p a s s C r u d e Oil U n d e r f l o w Efficiency w3 Overflow Vohllne P e r c e n t R e n l o v e d F e e d da~ = 35 ~an 100_

applied, since swings in backpressure result in the oil in the core being rapidly discharged with the cleaned water. The flow rate (thruput) of the hydrocyclone is dependent only on differential pressure between the feed and underflow and is independent of underflow pressure.

Effect of overflow diameter

Oil-water separation hydrocyclones make a separation that is independent of overflow diameter (Fig. 8). The minimum overflow flow rate to make an effective separation increases with increasing overflow diameter. M. Thew (Colman and Thew, 1980b) also demonstrated the effect of overflow diameter on separation. This minimum flow rate for each orifice opening size is a result of a minimum velocity required for the oil to report to the overflow. This minim u m velocity multiplied by the cross-sectional area of the overflow results in a minimum flow rate for effective separation for each overflow opening size. Increasing overflow size results in an increased amount of water which must be removed with the oil to obtain the same removal efficiency. This of course means that a greater flow rate of oily waste must be reprocessed. The major advantage of larger overflow openings is that it allows more oil to be removed without destroying the "clean water stream" when large slugs of oil are encountered in field operations. Furthermore, larger outlets are not as susceptible to blockage. This problem is alleviated somewhat when the units are operated in series. If the overflow opening of the first hydrocyclone becomes plugged, the oil not removed by the first unit will be treated again by the second unit.

8o_?
~ M 60__

Legend
'x Fim Pi=I50. Pi= 150, Pi=125, O Pi=100, A PI=IO0, I 2 Overflow Volume I 4 Percent, Qo/QI Pu=[00 Pu= I O0 P11=75 Pu=50 Pu=50 I 6

40_

20 _

Fig. 7. Effectof underflowpressure.

35 n u n C o l m a n - T h e w H y d r o c y c l o n e a n d B u m p a s s C r u d e Oil U n d e r i l o w Efficiency vs Overflow V o l u m e P e r c e n t R e m o v e d F e e d d ~ = 35 grrt lOO_

~
a) ~

8o~

Effect of inlet oil concentration

6o~

Legend
40~ 2o_ do = 2 mm [] do = 3 mm do = 4 r i m ,

o Overflow Volume Percent, Qo/Ql

Fig. 8. Effectof overflowdiameter.

Field reports indicate that increased oil concentrations result in improved hydrocyclone performance. The net result is that the additional oil is removed. As can be observed in Fig. 9, separation is independent of inlet oil concentration when there is adequate flow at the overflow to remove the required amount of oil. The improved separation of field installations with increasing oil content is quite likely due to larger

42
35 mm

G.A.B. Young et al. /Journal of Petroleum Science and Engineering 11 (1994) 37-50
C o h n a n - T h e w H y d r o c y c l o n e a n d B u m p a s s C r u d e Off U n d e r f l o w Efficiency vs I n l e t Off C o n c e n t l a t i o n

100__

3 5 m m C o h n a n - T h e w H y d r o e y c l o n e a n d S o u t h C h i n a S e a C r u d e Off U n d e r f l o w Efficiency ws O v e r f l o w V o h t m e P e r c e n t R e m o v e d 100_

8O__

80_

D
6O--

n"
0

60_

~, 0

40-20_

Feed

d~ o

~ o

40_
20_

35 . m [] 35 ~m

Feed d~0
[] O
0.5 Overflow 53.5 35.o 25,0 20,0 pm ~m Fm ~nl

'~

I I I l O00 2000 3000 I n l e t Off C o n c e n t r a t i o n , mg/1

40100

'

I
Volume

,'5
Percent,

'2
Qo/Q!

2'5

Fig. 9. Concentration/efficiency relationship.

Fig. 11. Effect of droplet size distribution.

3 5 m m C o l m a n - T h e w H y d r o c y c l o n e a n d B t t m p a s s C r u d e Off U n d e r f l o w Efficlency v s O v e r f l o w V o l u m e P e r c e n t R e m o v e d

I00__
El_ E]

35 rmn Colman-Thew Hydrocyclone Undea-flow E f f i c i e n c y v s M e d a a n Off D r o p l e t S i z e

0
7O A
v

g
a~

80_

100_

O" ', Feed 55 D 40 @ 35 O 25 15 ds0 ~m ~m I*m ~m ~tm

60_

O A

80_ 60_

40_
-cJ ~ 20_

40_

tq

Legend
B u m p a ~ , Single [] South China Sea, Single

o o

20_

I 0.5 Overflow

I 1 Volume

11.5
Percent, Qo/QI

]9 215
0 Median
I I 20 40 Off D r o p l e t I I 60 80 Size (microns)

Fig. 10. Efficiency/droplet size relationship.

Fig. 12. Effect ofoil droplet size.

oil droplet sizes attendant with increased oil concentrations.

cates that the oil/water density differential has a significant impact on separation.

Effect ofoil droplet distribution and oil~water differential density

The variable having the greatest impact on oil/ water separation is the oil droplet size. Figures 10 and 11 show the separation performance of 35-mm hydrocyclones for several oil distributions having the median droplet sizes shown. Figure 1 0 is for Bumpass crude (sg=0.85), whereas Fig. 11 is for South China sea crude (sg = 0.96 ). The distributions generated are quite likely similar to distributions which would be generated in the field. If the medians of these distributions are plotted against separation performance (Fig. 12), it is observed that the performance is somewhat poorer with .the South China Sea crude than with Bumpass. This indi-

0ii droplet distribution and heavier crude

Heavier, more viscous crude oil such as the South China Sea crude requires a significantly higher shear rate to break into a fine droplet size as observed in Fig. 13. As a result, in the field, oil droplet sizes for heavier, more viscous crudes will likely be coarser and hydrocyclone performance better than expected.

Two-stage separation
Two stage separation was also investigated using 35-mm hydrocyclones (Fig. 14). The total separation performance of the two units for the Bumpass and the South China Sea crudes are shown in Fig. 15.

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

43

Median Drop Size vs Homogenizer Frequency 125-100-75-50-25-

Legend
BIIIIIINRSS (TF~Id(" 11/12,87 [] [~tllillxassCrude I2'07,87
~)~l[h C h i l l a ~'~1 ( r u d e O BUlili~lss ('rll/](" 2 1 / 2 1 ; 8 7 12/ 1t3,1-17

zx
0 10
l

20 30 40 50 Homogenizer F r e q u e n c y (hz)

60

Fig. 13. Homogenizer/droplet size relationship.

tlSeeCndn~S get eOv~~sc~ar ~ 0 ~n ge

~S~dUgel~ ntgSe f a ea~' t d

presented to the piece of equipment. To partially circumvent this, analytical cut size was calculated for a number of tests. The analytical cut size is the size which would ideally split the feed droplets according to the droplet size in proportion given by the measured total efficiency. It is the size corresponding to the percent separated (e) on a cumulative feed distribution curve as shown in Fig. 16. Svarovsky published an excellent article which correlates analytical cut size to equiprobable separation (median separation, dS0 cut point) (Svarovsky, 1980). Shown in Fig. 17 is the analytical cut size versus differential density for a single-stage hydrocyclone. Knowing the differential density and the expected feed distribution of a hydrocyclone application it is possible to determine an expected efficiency (Fig. 18 ).

Feed Droplel Size I)istrillulioil

1 oo

80_

~St

Oil Con~ntrate Di~harge

Stage Overflow

Second Stage Underflow ~ Clean Water Outlet

"

60

40

t /
1o Oil Ic~o

Fig. 14. Tandem hydrocyclone operation.

20 0 35 mm Colman-Thew Hydrocyclone Underflow Efficiency vs Median Off Droplet Size Feed d~ = 35 [ml
I00_

Droplet Size (micronsl

80_

y
Legend
[] O Bumpass. Single South China Single Bumpass, Tandcm ~u~l China ~a, Ta/idcnl

Fig. 16. Definition of analytical cut, x,.

60_

4o _
20 0

Sca,

Single Slage 24-22--

80 Median Oil Droplet Size (microns)

'o

'o

~'o

'

20-O 18-16-14 004

Fig. 15. Effect of droplet size on tandem.

Effect of oil~water differential density in terms of analytical cut size The effect of oil droplet size on total efficiency has been presented. Total separation is affected to a great extent by the actual feed distribution

006

008

0i

012

0 14

0 16

I)ensity Difference, g / m l

Fig. 17. Effect of differential density on xa.

44

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (I 994) 37-50
Feed Droplet Size Dlslxlbution 100_ .

minimum loss of angular momentum. A cone angle of 6 provides the best separation over a broad flow range. Much of the separation occurs in this cone section.

60_ 40_ c~ 20_ o I Oil D r o p l e l 11o X~ Size [ m i c r o n s ) 10o

Effect of underflow length

Separation continues to occur as the fluid spins in the tail section of the hydrocyclone. Fig. 22 shows that for straight section lengths greater than 18 times the cone diameter, that there is no additional separation. This is because the fluid loses much of its spin velocity from viscous drag on the wails as it goes down the length of the tail section. Finally, the spin velocity is not great enough to provide additional separation. In other words the fluid rotation is a tight spiral initially and separation of the finer droplets occurs. This spiral widens out as it goes down the length of the tube to the point where no additional separation occurs.

Fig. 18. Using analytical cut, x~.

4. The search for optimum hydrocyclone dimension Next, the authors began to search for ideal hydrocyclone dimensions. Hydrocyclone designs having single involute entry, one cone section, and a long straight discharge were investigated. A count of the dimensional variables (Fig. 4) shows that there are nine dimensional variables. These variables were investigated utilizing hydrocyclone components which could be assembled in a variety of designs (Fig. 19). The effect of cylindrical length, cone angle, straight section length/diameter and inlet diameter are presented.

Effect of underflow size

Figure 23 shows the effect of tail section size on separation. Underflows of du/Dc=0.25 provide approximately the same separation as an underflow size of du/Dc=0.33, only at different flow rates. The dJDc= 0.33 underflow gives better separation over a broader flow range.

Effect offeed size Effect of cylindrical length

The effect of cylindrical length is shown in Fig. 20. Shorter cylindrical lengths produce better separation. This is simply because the fluid in this section is not spinning fast enough to provide appreciable separation in comparison with the separation provided in other parts of the hydrocyclone. At the same time the fluid is losing its angular m o m e n t u m by the drag of the wall of the cylindrical section. As shown in Fig. 24 the feed size di/Dc=0.25 provides the best separation. Larger feed sizes provide almost the same separation but need higher flow rates. Feed sizes smaller than this provide very poor separation. This optimum feed size probably occurs because it provides the best spin momentum transfer. That is the spin rate of the bulk fluid in the hydrocyclone is greatest for this size inlet. This is also the same size inlet which was chosen as optimum for solid liquid separation hydrocyclones (Young, 1987a,b).

Effect of cone angle

Shallow cone angles of 1.5 and 3 provide very poor separation (Fig. 21 ). This is because the fluid does not come up to speed very fast and looses its angular m o m e n t u m on the drag of the cone wall. Cone angles of approximately 6 and greater provide a rapid spin up to speed with 5, Comparison of Amoco hydrocyclone with 35mm Thew hydrocyclone Shown in Fig. 25 is a comparison of the singleand double-stage optimum design hydrocyclone

G.A.B. Young et al. /Journal of Petroleum Science and Engineering 11 (I 994) 37-50

45

Fig. 19. Modular hydrocyclone components.

De=3.0 100

Bu_mpass C r u d e O i l dl/Dc=.15 do=.078 du/Dc=.33 F e e d d ~ = 35 ~ m

alpha=3

I)c=3.0

d o / I ) c = 25

100_

B u l n p a s s C r u d e Oil do=.078 d u / D c = . 3 3 F e e d d,,, = 35 ~ml

levi / Dc =2

lu / I)c= 12

so
60_

8O _

60_

40_

4O

Legend
_

Legend
20_

]cyl=O, lu=12 [] lcyl=4, lu=12

2O_

o
0 I0 20 30 Flow Rate, gpm 40 50

0 20
Flow I

Alptm-I 5 [] Alpt~-3 Alpha-6 0 Alpha=9.5 Z~ Alpha-20 40 Rate, gpm

I

60

Fig. 20. Cylindrical length effect on separation.

Fig. 21. Effect of cone angle on separation.

with the 35-mm Thew design hydrocyclone. Within experimental error both hydrocyclones have the same performance. Shown in Fig. 26 is a set of Percent Separated curves for both single-

and double-stage Thew or Amoco hydrocyclones. The median separation for these curves can be corrected for differential density (Fig. 27).

46

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

De=3.0

dll Dc=.25

106_

B u r n p a s s C r u d e Off do=.078 du/Dc=.33 Feed d~ = 35 ~In

IcyllDe=2 alpha=6

A m o c o P r o t o t y p e ~md C o l l n a n - T h e w U n i t s C o m p a r e d

1oo. go60-

6040--

Legend
lulDc= 12 [] lulDc= 18 lulDc=24

400 a

0~ -~ 20_

Legend
Amoco Prototype C o l m a n - T l a e w "l~pe

~ U

20-

U o
0 2

I0

I 40 Flow Rate,

I 66
gpm

I 86

I 10
Feed

I 20

I 30

I 40
Size, ~m

I 50

I 60

Oil Droplet

Fig. 22. Effect ofunderflow length on separation.

Fig. 25. Efficiency of single stage units.

Dc=3.O ,~ 160_ a~

Effect of U n d e r f l o w Size o n S e p a r a t i o n Twentyflve Median Micron Feed Slurry dl/Dc=.25 do=.O78 lu/Dc=12 Icyl/Dc=2

A m o c o or C o l m a n - T h e w or Vortofl H y d r o c y c l o n e s Ap = 0. I 0 g m / c c alpha=6 I00_

Legend
du/Dc= 25 dulDc=.42 0 du/De=,50

80_
60_ "-. ql 40_

~e~ 8 0 _

g~
~ ~ 0 20 _~ 60_ 40_

20 J 0 0 _

gend
/ ~ - Single Stage Double Stage

I
Oil Droplet

I 20
Flow

I 40 Rate,

I 60
gpm

I 80

I 10 S~e

I 100
0tllcrons]

Fig. 26. Comparison of single and dual units. Fig. 23. Effect ofunderflow size.
Amoco - Cohnan-Thew Hydr ocyclones Vortofl

Dc=3.0

B u m p a s s Crude Off Feed d~ = 25 pan d0=.078 lu/Dc=18 leyl/Dc=l

28

alpha=6

24 20 16

d
S,nglc S t : : ~

12 8 4 O 0

BO '0 ;7(/0'0 ~///I

6 0 . 0 ~1/11

d2 d, d6 '. )o 12 l, ~6 )6 2'
Density Difference, g/ml

:3o.' ' III

i0 0 o lV.~ 31,

Fig. 27. Effect of A density on median separation.

Fig. 24. Effect of inlet size.

As described above, the Thew and Amoco hydrocyclones have the same performance. However, the Amoco dual-stage hydrocyclone processes 60 gpm (at 200 psi feed and 75 psi backpressure) while the dual-stage 35-mm Thew

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50
Overall Efficiency vs Median Oil Droplet Size Amoco Hydrocyclones vs Thew Type - .905 sg crude
10o 908070

47

.o_

60 50 40

30 2010 020
40

60

Legend "fflewType, 30gpm

l I

go

Oil Droplet Size, microns

Fig. 28. Typical system implementation.

hydrocyclone processes 30 gpm (at 150 psi feed pressure and 75 psi backpressure). Therefore, half the number of hydrocyclones are necessary.

6. Comparison of improved Amoco hydrocyclone with Thew type hydrocyclone

Since the original presentation of this paper at the American Filtration Conference in Baton Rouge on October 30, 1990, the performance of the Amoco oil-water hydrocyclone has been greatly improved. Shown in Fig. 28 is a comparison of the new Amoco three-inch with the old Amoco three-inch and the Thew type 35-mm hydrocyclone. Test results have shown that the new Amoco hydrocyclone is capable of a 9-/~m median separation on a 0.86 specific gravity crude at 60 gpm. An Amoco-designed one-inch hydrocyclone is also available which has a 5-/zm median separation on 0.86 specific gravity crude at 13.5 gpm. An Amoco-based two-inch hydrocyclone is also available which has substantially better performance than the Thew type 35 mm hydrocyclone.

7. Conclusions
Oil droplet size has significant impact on separation.

Oil/water density differential has a significant impact on separation. The 35-ram hydrocyclone performance is independent of the pressure drop between the inlet and the underflow from 30 to 100 psi differential. The Amoco optimum hydrocyclone performance is flowrate dependent but acceptable separation occurs over a broad range. Separation is independent of overflow diameter if adequate flow is removed. There is a minimum flow rate to make an effective separation. This minimum flow increases with increasing overflow diameter. The maximum inlet oil concentration which can be properly separated increases with increasing overflow diameter. Hydrocyclone performance is independent of the underflow pressure, provided there is enough backpressure to force enough oil/water out the overflow. This takes a minimum of 50 psi for 35mm hydrocyclones, although for field use a set point of 60 to 75 psi is recommended. Hydrocyclone performance is independent of the inlet oil concentration, assuming there is enough flow provided at the overflow to remove the required amount of oil. The effect of hydrocyclone dimensional variables on hydrocyclone performance is as follows: --Increased cylindrical length reduces separation performance. --A cone angle of approximately 6 provides the best separation over a broad flow range. --Underflow lengths up to lu/Dc= 18 improve separation. --Underflow size of dJDc = 0.33 provides a good separation over a broad flow range. --Feed size ofdi/Dc= 0.25 provides the best separation, although larger feed sizes also provide excellent separation. A patented hydrocyclone has been designed which has better separation performance than the Thew type 35-mm hydrocyclone and has twice the thruput. A high performance, economical, high thruput hydrocyclone for use in production operations is now available.

48

G.A.B. Young et al. / Journal of Petroleum Science and Engineering I 1 (1994) 37-50

Acknowledgements

The authors wish to thank the management of Amoco Production Research for the wonderful facilities and support which has made this work possible. Several individuals have contributed significantly to this project; Dick Gifford was instrumental in the design of the hydrocyclone test hardware. Eddie Vaughn, Dave Simms and Dave Hild assisted with the collection of the data. Amoco ELAF loaned Research Vortoil hydrocyclones for performance testing. Indeed, the authors are very appreciative of the expertise contributed by each member of the team.

Appendix A - - Test stand capabilities

A stand for testing all types of oil/water separation equipment has been constructed (Fig. A1 ). Water enters (lower left corner of Fig. A-1 ) and is pumped to the required pressure by a centrifugal pump. The pump discharge splits and oil is metered into one of the flow streams. The oily stream goes through a homogenizer where the desired drop size distribution is generated. The two flows join together and are fed to the piece of equipment being tested. The feed flow rate is monitored with an electromagnetic-type flow meter. After the piece of equipment has made a separation, the flow rate at the water discharge

and the oil concentrate discharge are measured. Samples from the feed, water discharge and oil concentrate discharge are analyzed for oil-inwater concentrations using a solvent extraction/ IR procedure as detailed in Appendix B. Samples from the feed and water discharge were isokinetically taken by regulating flow through the line downstream of the Malvern sample cell. Oil droplet size distributions were obtained using a Malvern forward scattering laser light size analyzer. The oil/water separation test stand has the capabilities shown in Fig. A-2. Typical size distributions shown in Fig. A-3 are easily obtained by adjusting the frequency of the homogenizer. For a given test oil, a curve of median drop size versus homogenizer frequency is obtained (Fig. A4). This allows the operator to rapidly set up the test for a given drop size. As a result of the specific arrangement of the test equipment, feed droplet distributions are independent of feed flow rate and pressure. Furthermore, the droplet size distribution is almost independent of inlet oil concentration. This type of independent control of the test stand variables allows experiments to be conducted to investigate the performance of a device with respect to one variable. This reduces the confounding of variables during testing. Several hundred usable tests have been run on this stand to date. Some of the results from these tests are reported in this paper. The stand is fully

positive Displacement Pump

~-

Disposal to Settling Tankage Discharge Sample (Underflow)

AnyUnit t

to be

Off Mass [ Flow Meter Homogel~zer

Concentratl

Off lnjecUon~ Mix Flow r RegulaUon

partlrele Size

Pump with Speed Controller

Feed and 5 Discharge Samples

Fig. A-1. Test facility.

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

49

Test

Stand

Capabilities

*Generate off concentrations u p to 3000 ppm. ,Generate various droplet size distributions. -Produce flow rates up to I00 gpm. ,Develop feed p r e s s u r e s up to 150 psi. Instrumentation Measure inlet and underflow distributions. Measure off/water concentrations. ,Measure feed and overflow rates. Collect data on microcomputer for analysis using spreadsheet and charting software.

versatile to conduct a variety of oil/water separation experiments on other pieces of equipment such as various pump types, flotation separators, centrifuges and permeable membranes.

Appendix B - - Determination of oil in water

Fig. A-2. Facility capabilities and instrumentation.

Malvern Instruments MASTER Partlele Slzer i0o-

50-

The analytical method used to determine oil concentrations in water consisted of solvent extraction to remove oil from water samples followed by infrared (IR) analysis of the extracts. Carbon tetrachloride (CCI 4) was used to extract the oil. The IR absorbance of each extract was measured at the C - H band maximum and Beer's law applied to determine the oil concentration in each water sample. A Miran 1FF Fixed Filter Laboratory Analyzer, a single-beam infrared spectrometer, was used to perform the IR analyses. Oil concentrations in water were calculated from measured absorbances and a least-squares curve fit equation of Beer's law data for the test oil (IR absorbances of known concentrations of the test oil in CC14).

Preparation of calibration curve

0 10 100 Particle size {Ixm},
Fig. A-3. Partical size distribution.

1000

Median Drop Size vs Homogenizer Frequency 125-100-75 50 25-0 I0

1 I

Legend

1t-12-87..~

Standard solutions were prepared by weighing appropriate quantities of test oil into 100-ml volumetric flasks and diluting to the mark with CCI4. IR absorbances of the standard solutions were determined and a calibration curve prepared by plotting absorbance versus concentration. Separate calibration curves were prepared for each oil tested. Examples are shown in Fig. B-1. A leastsquares curve fit of Beer's law data (absorbance versus concentration) was made for each test oil. The resulting equations were used to calculate oil concentrations in test samples.

Test procedure
Approximately 50-ml samples of the oil-inwater flow stream were collected in graduated 100-ml centrifuge tubes. Duplicate samples were collected for each determination. After exact sample volumes were recorded, the centrifuge tubes were filled to the 100 ml mark with CC14.

20

30

40

50

60

Homogenizer Frequency (hz) Fig. A-4. H o m o g e n i z e r output.

50

G.A.B. Young et al. / Journal of Petroleum Science and Engineering 11 (1994) 37-50

0.8--

0.6--

F = volume factor; F = (100-sample volume)/sample volume; and D = dilution factor (used if the CCl4extract must be diluted before IR measurement).
..// Legend
0 Bumpass
A South China Sea

0.4--

0.2--

References
I I

I
200 Oil

I
400

I
Concentration,

I
600 rag/1

I
800

I
1 OO0

Fig. B-1. Calibration curves for test oils.

The tubes were stoppered with cork stoppers and shaken vigorously for one minute to extract the oil into the CC14. Tubes were then placed in a centrifuge and spun at 3000 rpm for 10 min to separate the water/CCl4 layers. After the tubes were removed from the centrifuge, 30-ml sytinges with 14-mm pipetfing needles were used to obtain samples from the CC14 layers. Expulsion of air from the syringe while moving through the water layer prevented contamination by the water phase. Samples from the syringes were transferred to 1-cm quartz curvettes and IR absorbances were measured and recorded. Calculations Oil concentrations in water samples were calculated using the following equation:

C= (a+bXA) XFD where:

C = oil concentration (mg/1); a,b=least-squares curve fit coefficients for calibration curve of test oil, i.e., IR absorbance versus concentration of knowns; A = I R absorbance of the sample as measured (average of at least two determinations );

Anomymous, 1989. Hydrocyclone separators gain acceptance. Drilling Contractor, (Dec. 1988/Jan. 1989 ): 3940. Colman, D.A. and Thew, M.T., 1980a. Hydrocarbons for Oil/ Water Separation. Paper presented at Int. Conf. Hydrocyclones. BHRA Fluid Engineering, Cambridge. Colman, D.A. and Thew, M.T., 1980b. Hydrocyclones to Give a Highly Concentrated Sample of a Lighter Dispersed Phase. Paper presented at Int. Conf. Hydrocyclones. BHRA Fluid Engineering, Cambridge. Cornitius, T., 1986. Advances in water treating solving production problems. U.S. Specification Reference Register, July 9, 1986; Offshore, March 1988. Flanigan, D.A., Skilbeck, F., Stolhand, J.E. and Shimoda, E., 1989. Use of low shear pumps in conjunctionwith hydrocyclones for improved performance in the cleanup of low pressure water. SPE 19743, presented at SPE Annu. Tech. Conf. Hayes, J.J., Carroll, W.C., Fothergill, D.W. and Prendergast, G.J., 1985. Hydrocyclones for treating oily water: Development and field testing in Bass Strait. OTC 5-79, presented at 17th Annu. OTC Conf. Medlrum, N., 1987. Hydrocyclones: A solution to produced water treatment. OTC 5594, presented at 19th Annu. OTC Conf. Svarovsky, L., 1980. Critical Evaluation of the Simple Ways of Determining the Cut Size. Pap. 4 presented at Int. Conf. Hydrocyclones. BHRA Fluid Engineering, Cambridge. Woodruff, A., 1990. Cyclones sweep through the North Sea. PROSPECT, British Technology Group, 5 (Winter 19899O). Young, G., 1987a. An experimental investigation of the performance of a three inch hydrocyclone. SPE 16175, presented at SPE Annu. Tech. Conf. Young, G., 1987b. An experimental investigation of the dimensional and operating parameters of a hydrocyclone in a drilling mud. 3rd Int. Conf. Hydrocyclones. BHRA Fluid Engineering, Oxford.