4th International Conference on j'

.. ~

MULTI-PHASE FLOW Nice, France:

19-21 June 1989

PAPER C3

A HIGH EFFICIENCYLIQUID/LIQUID SEPARATOR

J. WOILLEZ Icb/Centre d'Etudes et de Recherches de Grenoble, France Y. LECOFFRE Bassin d'Essais

des Carenes,

Paris,

and Consultant,

France

P. SCHUMMER

Icb/Centre d'Etudes et de Recherches de Grenoble, France SUMMARY

On the basis of vortex flow enhanced by the rotation of a cylindrical chamber, a new compact oil from water separator was designed built and tested in the laboratory. Cut-size diameters lower than 20 microns are obtained at residence times of a few seconds. Experimental results and theoretical predictions are in good agreement and therefore scaling laws are available for industrial design. Limited laboratory tests show altogether a good efficiency for the treatment of water-in-oil

emulsions.

.

NOMENCLATURE

Ce Cs d cI dmax D g k L n N Q q r Ro R1 S T V Vo W IS ~ ~ p (j Cl.)

'l

"

'It T 7:J Ae 67>

1.

:

Inlet oil concentration (mg/l) Outlet oil concentration (mg/l) Droplet diameter (m) Meandroplet diameter (m) Maximumstable drop size (m)

::

Separation chamber diamet~r Gravity acceleration (m/s ) (m)

: :

:

: :

: : : : : : : : : : : : : : : : : :

:

: : : : :

~

- D = 2 Ro

Constant. Length of separation chamber (m) Exponentof vortex law

Rotation speed (Rpm) Flow rate (m3/s) Oily outlet (m3/s) Radial distance from separation chamber axis (m) Radius of separation chamber (m) Radius of water outlet orifice (m) Section (m2) Temperature (GC) Tangential velocity (m/s) Tangential velocity of wall (m/s) Droplet migration velocity (m/s) Acceleration (m/s2) Meanradial acceleration (m/s2) Energy dissipation rate (W/kg) Viscosity (kg/m.s) Interfacial.tension (N/m) Rotation speed (s-1) Effciency (%) 1.::0 1 Cs/Ce Theoretical efficiency (%) Meanresidence time (s) Droplet migration time (s) Density difference between oil and water (kg/m3) Head loss between inlet and clear water outlet (bar)

-

INTRODUCTION

In most off-shore fields, the oil is produced with a certain amount of water. The proportion of water in the oil can vary from one field to another, and durin~ the 1i fe of a gi ven well, from a few per cent to above 90%. The 011 concentrati 01 in the water produced must be reduced to less than 40 mg/l (depending on local legislation) before it can be discharged to the sea. In certain cases, special platforms must be built for this purpose, and an urgent need has arisen to reducl

the size and weight of future and existing oily water treatment installations.

2.

BACKGROUND:CYCLONES ANDCENTRIFUGES

Liquid/liquid separation can easily be performed by a gravity field uniformly distributed in all the volume of the separator. If ~ is the intensity of the acceleration field, the migration velocity of an oil droplet relative to the

water is given by STOKES' law: ~

(1)

w=:;

I::.P.3'.tI 1 8 ')J

in which

~r #

: density difference between oil and water (Kg/m3) : water viscosity (Kg/m.s)

d

: droplet

diameter

(m)

.

In ~ettling tanks to or settling parallel m/s That yields

plate separators, the 0,03 value mm/s of 0foris 20 limited velocities Wof about micronsto 10 oil droplets with an oil density of 840 Kg/m3. To capture such droplets, a parallel plate separator should have an important theoretical residence time (more than ten 15 min~tes 20m3/h mmspaced consequently a high volume (about m for for a 25 if we plates) take intoand account inlet and outlettotal volume separator) and a high specific weigh (more than 600 Kg/m3/h).

To reduce these values, it has long been thought of using gravity field enhanced by rotation of the fluid. If V is a tangential velocity given to the fluid and r the distance measured from the rotation axis, the acceleration will be ~

(2)

~=V/r

and can easily be 1000 times higher than the gravitational High values of ~ can actually hydrocyclones.

field

intensity.

be obtained either with centrifuges

or with

In centrifuges the acceleration field is generated by mechanical rotation bowl equipped with plates. The acceleration value will be z 13.1 }f = w r

of a

.

LV being the rotation

speed. The maximumvalue of ~ is reached along the walls

where r = 0/2 In hydrocylones the acceleration field is generated by tangential inlets adapted to an axisymetric chamber of separation (Fig. 1). At a current point of this chamber, the velocity is expected to follow a vortex law g~ven by :

(4)

V = VoRo/r

and hence

~=

.z 3

(VoRo)/r

The maximumvalue of ~ is reached near the axis of rotation. For having the same radius Ro and the same peripheric initial velocity following comparison can be made: I HYDROCYCLONES V.~r I CENTRIFUGES V. Vor/Ro I

-

-

v

i

I I I I I I I . I II I II

I I I I I I

I. I I I '-

R.O

no

fl1

MEANACCELERATION FIELD

I:

1

Ro

J

0

I:

~

~

~flO

I I I I I I I I II I

'"

r

'6

)'rrJ.dr

f

rrrl,dr If:..1. I flO I 0 I

I

I 1I

c

if

17.~

r

~

1\1

I

,.

I I I i

vo

RO

two separator Vo, the

~flO fl1 ..nO/3

I

. I 1.

no

trrl,dr

no

R1

I I I I I

I

I

I I

-

It can be seen that the specific nearly obtain

acceleration

field

~~JV:of

20 times higher than the value obtained for centrifuges. the same separation efficiency, a centrifuge will have

. higher and/or.

residence

high rotation

hydrocyc10n~s is This mean that

. to

time

velocities

(e.g. 5000 to 10000 rpm)

-_u-

n

--

---------------------

In addition to this three reasons have still delayed the use of-centrifuges oil from water separation in an off shore environement :

-

for -

the high rotation velocities demand a manufacturing technology which yields high specific weights (about 100 Kg/m3/h for a motor driven system);

- oilthe

acceleration field is badly fit to the treatment of light dispersions droplet must be gathered around the separator axis where ~ 0

since

=

(see eq. 3);

to

- internal

plates set to avoid turbulence must be carefully cleaned, leading to maintenance operations which are not always acceptable for an off shore environement. -

Consequently we first tried at-the CERGto develop a hydrocyc10ne in which a vortex flow was created in a conventional manner (Fig. 1). We obtained separation efficiencies equivalent to those obtained with parallel plate separators (absolute cut-size about 80 microns) at a residence time of 2 seconds. Improved results were obtained specific design (ref 111).

in such a way at Southampton University

with

a

Nevertheless we established that the separation efficiencies of conventional hydrocyc10nes were far below the performances which could be expected using a free vortex flow (eq. 4). In order to obtain a better understanding of the origin of hydrocyclone limits, L.D.A. measurementwere performed on the conventional hydrocyc10ne. Experimental shear stress

results led to the following in hydrocyc10nes.

analysis

concerning

the _effects

of

- ofReduction an oil

of mean migration velocity of droplets: the real migration velocity droplet moving toward the core is lower that the STOKES' velocity given by (1). This is due of course to the turbulent diffusion, but also to local recircu1ating flow around the drop itself (KUBIE 121)

- Swirl decay: owing to the turbulent shear stress along the walls of the separation chamber, the initial tangential velocity decreases continuously

from

Vo to a lower value V'o near the lower orifice. The decay is extremely rapid and can be only partially counterbalanced by making the separation chamber conical. If the angle of the cone is too acute, the mean residence time of the fluid inside the chamber will be too low to ensure proper separation.

- Vortex flow

.

dissipation: without turbulence or wall shear stress, the vortex flow ensuring a large acceleration field should obey eq. 4. In fact attenuated

vortex flow were measuredfollowing the law V = VoRo/rnwith n.

0,6

- 0,9.

Consequently the intensity of the meanacceleration field is reduced. At the sametime, a turbulent boundary layer on the wall creates a region of lower tangential velocity from which droplets cannot escape.

- Drop

breakup : in regions of high shear stress (tangential entry, turbulent boundary layer) an incoming droplet can be stretched by turbulent vortices and broken into two or more smaller drops which cannot be captured. The maximum

stable drop size with regard to turbulence was given by HINZE 131 as -

( 5)

_0.' -0.4dmax = O. 725 . ( (J/ fT) . (E )

.

.-

where

'

h-

---.

P = density of water (kg/m3)

v = interfacial £

tension

between oil an water (0,025 to 0,05 N/m)

= turbulent energy dissipation rate of the stretching region (W/kg).

-

-

100 W/kg can be easily obtained in a 5 10 m/s entry jet, and droplets over 100 microns will readily break up when they enter the chamber. They may be split into droplets smaller than the cut size of the cyclone, and not be separated from the water. Values of 50

- Secondary axial flows: the law of conservation of angular momentumand wall shear stress induces an irregular distribution of axial velocities in cyclones. ,These distributions discussed by Escudier 141 define flow regions in which the residence time available for separation is lower than the mean residence time. This leads to a lower efficiency for part of the inlet flow. Turbulence-phenomena are thus shown to reduce the theoretical separation efficiency of a cyclone. Another pratical consideration can limit the use of cyclones in the oil industry: The flow range of a cyclone is indeed very narrow. If a cyclone is to be fed with a flow above its nominal flow rate, unavailable head losses can appear or separation efficiency can decrease due to drop breakup. On the contrary, if the feed rate is for example half the nominal flow rate, the initial tangential velocity Vo.wi11 also be halved, and the mean acceleration field will be four times smaller.

3.

EXPERIMENTALPROCEDURE

3.1. Rotary cyclone design

Following the above analysis, the basic principle of the new cyclone design is tc overcome the limitation of conventional cyclones by rotating the separation chamber.

Fig. 2 shows the main geometrical characteristics of the device: the fluid enters a separation cylinder (I) (D = 60 mm) through an axial pipe (2), which is also the support of the cylinder. Paddles (3), move the fluid from the region of the axis to an outer annular region equipped with axisYmmetric blades (4). The blades divide the peripherica1 annular inlet section into a large number of duct~ and the fluid then enters the separation chamber axially through these ducts. The axial velocities remain sufficiently low in the ducts and separation chamber to avoid drop breakup. Relative to the walls, the flow remains axial along the separation chamber, but a rotation is superimposed on this axial flow by the motor (5) Combinedwith the effect of a downstream orifice (6) smaller than the main diameter D, the rotation induces a vortex-flow region of the same type as those found in hydrocyc10nes but with an extremely low level of turbulence. The separator is thus characterized

-- no low residence time (less inlet turbulence - no wall shear stress

- no internals

by :

than 2 s for a nominal flow of 3 m3/h)

in the separation chamber

In such conditions, an ideal free vortex flow maybe expected to appear in the separation chamber. An oily vortex core appears in the separation chamber as soon as oil droplets art introduced in the fluid (see photograph I). The oil is extracted via axial tube (l) (5 mmin diameter) at a flow rate q of about a few per cent of the main flow rate.

The specific weight obtained for a separator designed for a nominal flow of 3 m3/h is 15 kg/m3/h. 3;2. Test rig Fig. 3 gives the main characteristics of the experimental installation. Tests can be performed with temperature-controlled salt or fresh water. The range of the flow rates is from 0 to 15 m3/h for water and from 0 to 3 m3/h for oil. In order to be sure of the reproductibility of the emulsion characteristics (drop size, interfacial tension etc the fluids are not recirculated, and clean ,'products are used for each test.

...),

3.3. Drop size generation and measurement Since hydrocyclone efficiency depends on the square of drop diameter, it is of paramount importance to knowthe drop size spectrum of the emulsion for each tes~.

Inlet drop size spectra were measured by a MALVERN 2600 D analyser with a

circulation cell (Fig. 4). Circulation was controlled by a flow meter so that the velocity was sufficiently high to avoid settling and coalescence of the droplets in the junction pipes, yet sufficiently low to avoid drop breakup. For the fine dispersions used (drop size lower than 100 microns), these conditions were 1 m/s, that obtained with an 8 mminternal diameter pipe and velocities of 0,5 is,. turbulence was sufficient to avoid settling (Reynolds number = 4000 to 8000)and dmaxcalculated from eq. (5) was about 1 mm.

-

Outlet drop-size spectra where measured only if it could be considered that no change due to drop breakup appeared in the high velocity flow field at the outlet of the cyclone.

In order to check the validity of the data given by the MALVERN analyser, photographs of the emulsion were taken ~e.g photo9raph 2). It was found that the ) given by photographic measurements were about 10%larger that the mean drop size given by the analyser. It was considered that the photographic method used could not take into account droplets smaller than 10 microns and that this could explain the 10%discrepancy. The analyser was therefore considered to yield reasonably good values for the drop size spectra.

mean'drop size of the spectra (d =[ nid;/ [nidt

Typical drop size spectra is given in Fig. 5. Oil in water suspensions were generated by a series of orifices which were shown to give narrower distributions than a single orifice. Concentrations were measured from samples analysed with the tetrachlorid extraction and infrared analysis method.

4.

RESULTS

4.1. Laboratory tests Typical tests results, drawn:

are given in Table I. The following conclusions can be

- The separator

is characterized by extremely low absolute cut sizes (about 20 microns for an oil with a density of 840 kg/m3) at a residence time of a few seconds. This result is defined more precisely by Fig. 6 where it is shown that droplets smaller than the absolute cut size are still separated with good efficiency.

- The The flow field efficiency

of the separator is not affected by a shift in the flow rate. increases with decreasing flow rate (and increasing residence time) and decreases slightly with increasing flow rate as shown in Fig. 7.

- speed Increasing the rotation is related to the

speed leads to better efficiency. total pressure available.

Tfie maximumrotation

In conclusion, the performance of the separator can be adapted to the operating . asconditions (mean inlet drop size, available pressure) by design parameters such maximumflow-rate, rotation speed, and the length and diameter of the separation chamber.

4.2. Test field data A 120 millimeters in diameter and 2 m in length prototype was built and tested in offshore conditions. The purpose of the test was to verify the efficiency of the cyclone on real dispersions. Test were carried out on the outlet effluent of three phase separators. The separator pressure was 175 psig and the temperature 1650 F on platform 1, and 135 psig-135° F on platform 2. During the tests, photographs of the dispersion were taken, showing that the mean inlet drop size was centered between 30 and 50 microns on platform 1 and fluctuated between 5 and 100 microns on platform 2. Table 11 gives some of the results and demonstrates the high capacity of the separator in real life conditions. A more detailed description of the test procedure and results have been published elsewhere (ref. 151) and can be compared with results obtained with cyclones designed by Southampton University (ref. L61).

5.

THEORETICAL CORRELATIONS AND SCALE-UP

According to Fig. 8, an oil droplet a distance r of the separator axis will separated if it can reach the vortex core i.e. if it can move the distance RI) during the residence time of the fluid in the separation chamber. (r

be

-

The mean residence

time is taken to be 7:

~

= L.1l'.D 4.Q

in which: - Lis the length of separation chamber (m)

- 0 is

- Q is

the diameter of separation

chamber (m)

the volumetric flow rate (m3/s) -

All droplets

are subjected

z

to a radial

where V is given by the free-vortex

0 = vI r

acceleration

law V

VoRo I r

in which:

- Vo is the tangential velocity by the rotation speed). - Ro = 0/2

of the wall of the separation

chamber, (given

A droplet of diameter d migrates toward the core with STOKES'velocity, le: dr

~E>.)'. d

IT:

1 8. /""

t ::

l t l IJ p. d. vD . R 0 1 8.}-".

=

f (r I

r)

This equation leads to the time rJ taken by the droplet to reach the core: R1 T': E..!:... :: 1~..M . r4_R1+

Jr

f (rI

Ae. dl.v; .Rot

[

4

]

in which the vortex core radius RI is taken as the outlet orifice

radius.

Ifa'cloud of monodispersed drops is uniformly spread over the cross section

the separation

chamber. two regions

can be defined

- vortex Region I. area $1. in which z'< r core. - Region 11. are~ $2. in which r~> r the main flow. A theoretical

efficiency

(see Fig. 8) :

of

~ ~

for all for all

drops which are captured

by the

drops which are re-entrained

by

54 S" + S~ This procedure can be followed for all the inlet drop-~ize spectra available for each test point. and a theoretical efficiency for the separator will be given by : 'It:

~ (.

can thus be defined as ~UJ--

01,.' 'It (

dc.)

«,being the volumetric fraction drop-size analyser..

of droplets of diameter di given by the

Although simplifications and rough assumptions have been made as regard the flow configuration and drop behaviour. this.analytical ~odel gives a good order of magnitude for the separator efficiency as shown in Table I. and Fig. 6 and 7. Improved accuracy could be obtained with a better knowledge of the flow field but any prediction will rely essentially on the drop-size distribution, for which only rough values will be available in industrial environments. Nevertheless. the theoretical approach holds promise for the scaling-up of the separator to obtain high efficiency separators having high nominal flow rates. This was partially verified by field test of the industrial prototype described in 4.2. It can indeed be thought that the vortex flow generated by the rotation of the wall will be independent of the nominal size of the separator.

6.

CONCLUSION

A Rotary cyclone was shown to generate an almost ideal free-vortex flow, permitting high oil from water separation efficiencfes. The concept of the system seems very promising for a use as primary and casually secondary de-oiling separator. This concept should operate anyway on any liquid/liquid and/or gas/liquid flow. Recently. it was shown indeed that the rotary cyclone could perform de-waterin9 operations on crude oil. as shown on the limited results of table III (ref 171).

7.

ACKNOWLEDGMENT

The authors thank TOTAL/EXP/PROD and ALSTHOM NEYRTEC for funding this work and for technical advice. Thanks are due to P. GRISARD of TOTAL/CFP for having initiated the development program. This work was partially supported by the EEC.

8. (I)

REFERENCES:

-

D.A. COLMAN M.T. THEW

- Correlation

dispersion hydrocy1cones Vol. 61 (2) KUBIE

- July

- Settling

1983.

velocity of droplets

Chem. Eng. Sci. Vol. 35 (3) HINZE

- J.O.

- Chem. Eng. ofRes.separation Des.

- 1980.

in turbulent

results

from light

flows.

: Fundamentals of the Hydrodynamic Mechanism of Splitting n° 4 1977.

Dispersion Process. AICHEJ. Vol. 23

-

-

in

(4) M.P. ESCUDIER; J. BORNSTEIN; N ZEHNBER - Observations

.

confined turbulent

-

vortex flow. JFM vol 98

LOAmeasurements - part 1 and - 1980.

of

(5) J.C. GAY,G. TRIPONEY,C. BEZARD, P. SCHUMMER

Rotary cyclone will improve oily water treatment and reduce space requirement/weight on offshore platforms S.P.E. 1987 (6) MELDRUM N. Conoco U.K LTD

Water Treatment.

(7)

- OTt 5594 - Hydrocyclones : A Solution

to Produced

.

A. AMBLARD, R. ANSELME

- Etude-dlun cyclone

deshydrateur ~ parois tournantes.

Internal report acb/CERG1988 To be published.

Inlet oil Mean drop.1 et concentration size ( ppml (m)cronsl

Outlet oil concentration ( ppm )

Measured efficiency ( )

Theoretical efficiency ( )

Rotation speed (RPM)

Flow rate Q (m3/h)

Extract. q (Q)

1830

3,0

3,7

63

1121

20

98,2

9'J

1830

3,0

3,7

29

1038

53

94,9

93

1830

3,0

3,7

18

412

52

87,4

85

1830

3,0

3.7

10

347

111

68,0

. 60

Table I Hean residential

-

time:

Temperature of water: Fresh water density: Density

-

,

difference:

Water head

losses:

laboratory

Test

1.5 s 25'C' 1000 kg/m3 160 kg/m3

1.3 bars

Results

RotatfOt:' speed' (RPM)

Flow

rate

Q (m3/h)

Extract, q ( IQ )

Inlet 011 concentration C ppm ')

- OUtlet 011 concentration ( ppm )

5,1

1106

24

'7,8

5,3

4112

30

'19,3

5,0

1835

32

98,3

Heasured efficiency

( " )

1330

6,0

1655

,O

2073

6,0 "

'

Table

II

-

Test Field Results

Temperature of water: density: Sa1t water Density difference:

74'C 1085 kg!m3 200 kg/m3

-

Rotation speed (RPM)

Flow rate Q (m3/h)

Extract, q ( XO )

Inlet 011 concentration C ppm )

Out let 011 concentration ( ppm )

Measured efficiency ( S )

65

88,8

-

.

2073

26,5

3,8

2073

23,2

6,5

1351

80

94, 1

2013

19,9

7,0

570

24

95,8

16,6

7,0'"

591

3

-99,5

2073

-.

579.

-

Table IIb1s

Temperature of water: Salt water ~ density: Density

Rotation speed (RPM)

.

-

Platform 1 of rotation speed)

(Influence

54'c 1085 kg!m3

difference:

Flow rate Q (1II3/h)

-

Test Field Results Platform 2 (influence of flow rate)

200 kg!m3

In1et water concentration ( I )

Hean droplet size (microns)

2500

1,0

45

6,0

2500

1,0

35

2500

1,0

35

Table III

~

Outlet water concentration Cl) '

Heasured efficiency ( X )

0,6

'°,0

6,0

0,9

85,0

6,0

0,7

88,3

.

Water in 011 separation efficiency of the 60 mmrotary cyclone (outlet geometry adapted)

Photograph

1

Oily

core in the separator

......100f' Photograph

2 : Oil drops

~

distribution chamber

..

Inlet tangential InJection

stabilizing cone

011 core

water outlet

collection chamber

-El

+ 011 outlet

Fig. 1 : Conventionnal Static Hydrocyclone Developped at the CERG

7

z (,

D:W"""

~ 3