Andresen, Arntzen, Sjøblom - 2000 - Stability of model emulsions and determination of droplet size distributions in a gravity separator

7/30/2019 Andresen, Arntzen, Sjblom - 2000 - Stability of model emulsions and determination of droplet size distributions in

1/12

Colloids and Surfaces

A: Physicochemical and Engineering Aspects 170 (2000) 3344

Stability of model emulsions and determination of dropletsize distributions in a gravity separator with different inlet

characteristics

Per Arild Kjlseth Andresen a,* , Richard Arntzen b, Johan Sjblom c

a Department of Chemistry, Uni6ersity of Bergen, Allegt. 41, N-5007 Bergen, Norwayb K6rner Process Systems a.s, R&D Group, S.P. Andersens 6ei 7, N-7465 Trondheim, Norway

c Statoil A/S, R&D Centre, Rot6oll, N-7005 Trondheim, Norway

Received 4 June 1999; accepted 23 November 1999

Abstract

A model system consisting of an aliphatic oil (Exxsol D60), a commercial surfactant (nonyl-phenol-ethoxylat

Berol 26) and water was examined in a gravity separator loop system. By using a surfactant, we tried to control th

stability of the dispersion and to extract the influence of some of the separator characteristics. The parameters varie

were water cut, pressure drop, volumetric flow rate and inlet device. Initial droplet size distributions (DSDs) wer

obtained and examined for both water- and oil-continuous systems. It was observed that under these experiment

conditions and for these surfactant concentrations (5330 ppm) the oil-continuous dispersion was very unstable an

consequently the DSD measurements were not representative for the whole population of droplets. For thwater-continuous emulsions, variations were found to be dependent on pressure drop, water cut and flow rate. In th

case all the DSD data seemed reliable and accurate. 2000 Elsevier Science B.V. All rights reserved.

Keywords: Gravity separator; Initial droplet size distribution; Model oil; Surfactant and separator characteristics

Nomenclature

dP pressure drop over choke (bar)

NIL normal interface level, as measured by pressure transmitters (m)

total liquid flow rate (m3 h1)Qt

WC water cutp separator efficiency

www.elsevier.nl/locate/colsur

* Corresponding author. Tel.: +47-55-583382; fax: +47-55-589490.

E-mail address: [email protected] (P.A.K. Andresen)

0927-7757/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved.

P I I : S 0 9 2 7 - 7 7 5 7 ( 9 9 ) 0 0 5 1 8 - X


2/12

P.A.K. Andresen et al. /Colloids and Surfaces A: Physicochem. Eng. Aspects 170 (2000) 334434

1. Introduction

New trends will emerge during the next 3 5

years in the petroleum production on the Norwe-

gian Continental Shelf. First of all the amount of

water produced from topside platform separators

will increase mainly due to ageing fields with

water-break-through and a concomitant co-pro-duction of injection water together with the oil.

This high production rate of water will place high

demands on separator efficiency and treatment of

wastewater. In addition to this, many new fields

to be explored in the future will be complicated to

develop since the crude oil produced will contain

large amounts of heavy components like asphalte-

nes and resins. These components will strongly

increase the capability of the crude oil to bind

water, which will increase the retention time in the

separator. These new types of crude oils will also

most likely necessitate an increase in the use of

production chemicals in the separator and in the

transport process.

The effects of these two trends have to be

implemented into the design tools used to opti-

mize topside gravity separators. Tools in use at

present do not have a coalescence model for the

dispersion entering the separator and use only

modified versions of Stokes law [1] of settling

when describing the settling/creaming of droplets

and the subsequent separation of phases. The

influence of higher water cuts and more stabilizingsurface active components enhance the need for a

coalescence model.

There are literature reports on break-up and

coalescence of droplets in oil/water systems, but

to our knowledge these reports are mainly based

on low dispersed fractions and are specific for the

instrumentation used. The apparatus is usually a

vessel with a stirrer implemented as an energy

dissipating device [26] and for obvious reasons,

it is difficult to convert these relations to a large-

scale continuously flowing separator system.The overall coalescence rate of the dispersion

band in a separator is the most important design

criteria. Unfortunately, this rate is a product of

several complex mechanisms like binary coales-

cence, interfacial coalescence and settling/cream-

ing. Each of these mechanisms is further related

to other even more complex processes/factors lik

hydrodynamic micro and macro motions, dropl

size distribution and interfacial components. I

order to understand the overall coalescence ra

one must also understand the interactions b

tween these mechanisms. This makes it difficult t

separate the overall rate into a sum of distin

rates, and is probably the reason why there is ngeneralized coalescence model for concentrate

dispersion with a sound theoretical foundation.

The aim of this work was to carry out exper

mental work in a pilot-scale separator and obta

empirical correlations between separator chara

teristics, initial droplet size distribution (DSD

and upstream conditions (choke pressure drop

By using some of the typical geometric features o

a full-scale separator, the experimental data w

be much more suitable for performing scale-u

and implementation into design tools.

Another weakness with most design tools

that they assume an initial DSD. Based on liter

ture the reason for this is the lack of experiment

data connecting initial DSD and upstream cond

tions. Since the droplet size entering the separato

determines the settling velocity and hence rete

tion time, correct DSD is crucial.

Most authors have examined water/oil system

that coalesce completely within seconds after th

energy-dissipating device is stopped. Alternative

they have examined extraction columns [7,

which have different design and dispersing deviccompared to gravity separators. The experimen

conducted are usually small-scale with regard t

mass flow.

With the objective to carry out tests in a contin

uously flowing separator and to examine the di

persion band along the separator length, it w

desirable to create a dispersion that did not coa

lesce completely within the retention time of th

separator (24 min). It was, however, imperativ

for the system to separate during the model loop

total retention time (

15 min). Outside theboundaries, the result would be no dispersio

band at all or circulation of a stable dispersion

emulsion. Using a pure, classical model oil in

large-scale apparatus will fail to create a dispe

sion band that can be examined without usin

flow rates that are too large with regard to th


3/12


rate region of interest. The reasons why one can-

not use crude oil directly are obvious. First, it is

very cost effective and easy to use an open system

since environmental issues can be maintained.

This is not the case when a crude oil is used.

Secondly, many crude oils may generate stable

dispersions within the model loops retention time

and hence are of no use in a continuous modelloop. A third factor is that one will lose the

advantages of a transparent dispersion. Finally

one should keep in mind that crude oil batches

are not reproducible due to ageing effects.

When using an aliphatic oil containing a surfac-

tant (ppm range) one should get dispersions with

characteristics between stable and unstable. The

final level of stability is also determined by the

flow conditions. By controlling the amount of

surfactant, it is reasonable that one also would

control the separation characteristics, particularly

the dispersion band. Additions of a suitable com-

mercial surfactant will not remove the advantages

of a transparent system or make the system envi-

ronmentally hostile. The surfactant used in this

study is a non-ionic ethoxylated nonyl phenol,

with approximately six EO groups. The commer-

cial name is Berol 26. The reason for choosing

this surfactant is that the group in Bergen has

collected a lot of data on emulsions (w/o) stabi-

lized by this chemical.

2. Experimental

This section is divided into four parts. The first

two describe the chemical system and some pre-

liminary tests performed to establish a concentra-

tion range with regard to the surfactant. The third

part is related to the separator system and how

separation characteristics are extracted. The last

part describes the method used for sampling the

DSD.

2.1. Chemical system

The dispersions were prepared by using Exxsol

D-60 model oil (mixture of aliphatic hydrocar-

bons with chain lengths from C10H22 to C13H28,

transparent) and water with a salinity of 2.2 wt.%.

The inversion point between water- and oil-co

tinuous systems is in the range of WC=0.38

0.40 for flow conditions in our tests. Hence a

tests at [WC=0.16, 0.25 and 0.35] are oil-contin

uous, while tests at [WC=0.5 and 0.84] are w

ter-continuous. Berol 26 (Berol Nobel Industrie

Sweden) was used as w/o surfactant in order t

control the stability. Berol 26 is a commercinonyl-phenol-ethoxylate with approximately s

EO groups (in reality a distribution of EOs).

2.2. Preliminary bottle tests

The objective of the bottle tests was to establis

an emulsion stability range with regard to th

concentration of commercial surfactant. The tes

were very simple and were performed by shakin

bottles with different Berol 26 concentrations an

visually observing the evolution of the wat

phase. With concentrations lower than 100 ppm

no visual differences were observed, and for in

stance with a concentration of 1% a very stab

emulsion was formed.

2.3. Separator system

2.3.1. Separator efficiency

The separator tests were carried out under th

prerequisite that the normal interface level (NIL

should be at fixed values, contrary to standar

operation conditions (for example, keeping a

ceptable water quality for downstream proces

ing). As the model separator has limite

dimensions, the water quality will normally e

ceed standard downstream specifications. It w

also desirable to do both water- and oil-continu

ous runs within the same parameter ranges, whe

one would expect different separator behavio

The system will generally behave more robustly

this prerequisite is disregarded. This howev

made it necessary to define an efficiency based o

combined outlet qualities (Eq. (1)) as the phaqualities are generally interdependent. In norm

separator operation the first term (1WCwa

outlet) is close to 1, and the efficiency is hen

based on the oil quality only.

p= (1WCwater outlet)WCoil outlet (


4/12


Fig. 1. P&ID of the multiphase separator loop.

2.3.2. SystemThe separator system consists of a multiphase

flow loop with a bulk feed gravity separator, two

positive displacement pumps, and a model separa-

tor made in acrylic plastic. Total liquid volume in

the flow loop is approx. 2.4 m3. Fig. 1 shows the

Process and Instrument Diagram (P&ID) for the

loop. Geometry data and explanation of the

figure indices for the model separator are given in

Tables 1 and 2.

The bulk gravity separator is several times

larger than the model, and has consequently alarger operational window with respect to load-

ing. This will give stable feed conditions to the

model. The positive displacement pumps deliver a

stable pressure to the choke valve within their

capacity, and facilitate the experimental work.

The model separator is a pilot-scale first-stage

separator, with a flow distributor (perforated

plate) to remove unwanted channelling and un-

even flow distribution. The levels are controlled

by differential pressure (DP) measurements up-

and downstream of the weir, connected to but-

terfly control valves at the outlets.

In order to investigate the effect of shear within

a cyclone inlet on separation and also the effect of

the liquid outlet height, four inlets were tested as

shown in Fig. 2. Inlet B is a simple 2 bend and

tube, designed for low shear. Inlets C (and D) are

Table 1

Geometry data for model separator in Fig. 1

Symbol/unitParameter Value

Length tan-tan LTT [m] 2.80

ID [m]Inner diameter 0.63

LTW [m] 2.53Length tan-weir

0.55LTP [m]Length tan-perforated plate

HW [m]Height weir 0.25

Table 2

Explanation of indices in Fig. 1

Symbol in Fig. Explanation

1

LCVO NOL [oil] level control valve

LCVW NIL [water] level control valve

Qw Electromagnetic water flow meter

Qo Oil turbine flow meter

P Pressure gauge

CV Choke valve (manual ball valve)1 Weir, height 250 mm from vessel bottom

Perforated distribution plate, 500 mm2

from vessel bottom

3 Inlet device

4 Feed separator, tan-tan length 3 m, inner

diameter 1 m


5/12


Fig. 2. Inlet device geometries used in the tests.

dresen et al. [9]. The technique was developed i

order to examine unstable dispersions with a hig

internal phase in flowing systems. The basic prin

ciple is a fast dilution of the dispersion with th

continuous phase, typically with a dilution rat

of 1:100. Fig. 3 shows the set-up of the measur

ment apparatus. By inserting a sampling tube int

the inlet pipeline, one can withdraw a sampiso-kinetically without subjecting the dispersion t

any dissipating force. Using two magnetic valv

(MV1, MV2) connected to a timer, a controlle

injection of a sample into the dilution tank can b

obtained. Depending on the continuous phase o

the dispersion, the droplets will settle on the bo

tom plate (oil-continuous) or cream onto the to

plate (water-continuous). To prevent the drople

from wetting the plates, the bottom plate w

made of a hydrophobic material (acrylic plasti

and the top plate was made of a hydrophilmaterial (glass). A video camera was mounte

beneath or over the dilution tank and sever

images of droplets were captured and analyzed i

order to extract the DSD. The image analysis too

Image Pro Plus was used to measure the size an

generate droplet size distributions. The smalle

droplets possible to measure are about a few m

since the Brownian force will prevent them from

settling/creaming.

In the experiments, DSDs were extracted from

points 1 and 2 described above. The samplintubes were placed in the middle of the pipelin

and inlet liquid diffuser.

The major weakness of the technique is th

large amount of time consumed in extracting th

DSD. The analysis vessel has to be draine

cleaned and filled up between each experimen

The measurement of the droplet sizes on th

images is also rather time-consuming. The dilu

tion behavior can be performed in two differen

ways, i.e. either in a pure oil phase or alternativel

in an oil phase with the same amount of surfac

tant as used in the experiments. The advantage o

using a continuous phase containing the stabiliz

is that there will be no drainage of surfactant

the interface of the droplets. Such effects migh

accelerate the coalescence of especially larg

droplets.

conventional twin CCI as manufactured by

Kvrner Process Systems. Inlet F is the same twin

cyclone, but with one cyclone blocked at the gas-

and liquid outlets. Thus inlet B represents a low

shear inlet, C and D are reference cases, and F is

a high shear inlet (double load vs. inlet C/D). The

other parameter investigated is the outlet height

of the liquid diffuser; inlets B, D, and F have aliquid exit above NIL, while inlet C has a liquid

exit below NIL.

2.3.3. Sampling

Seven sample points were chosen for character-

izing the separator performance. These were lo-

cated at: (1) 0.4 m upstream from the inlet, (2)

inside the inlet liquid diffuser, (3 5) 0.2 m up-

stream from the weir plate at heights of 0.12, 0.15

and 0.19 m (measured from the bottom of the

vessel), (6) at the water outlet, and (7) at the oil

outlet. These are indicated in Fig. 1.

2.4. Droplet size measurement technique

The technique used for extracting the DSD has

been developed and described elsewhere by An-


6/12


3. Results and discussion

This section is divided into one part concerning

the separator performance and one related to the

droplet size distribution.

3.1. Separator performance

The results showed very high separation effi-

ciency and little variation in the oil-continuous

systems, which indicates that the separator load-

ing was far from the limit for this system, and all

variations were of the same order as the measure-

ment resolution. Fig. 5 shows results typically

found during the experiments. As can be seen, the

major influence in the tests was the effect of Berol

26 in the water-continuous regime.

No significant difference was found between the

various inlets, as shown in Fig. 4. This indicates

that the impact of shear and liquid diffuser height

is less than the resolution of the measurement

technique for the given system. Neither was the

effect of choke pressure drop found significant

within the varied interval. Higher water cuts gave

an increase in separation efficiency, as shown in

Fig. 5. This is in accordance with Refs. [1,10], a

the dispersed fraction decreases with increasin

water cut for water-continuous systems.

The separator performed well for all water-co

tinuous flow rates at Qt=12 m3 h1, but faile

for [Qt=18 m3 h1, WC=50%, 330 ppm Ber

26] and performed poorly (p=0.84) for [Qt=1

m3

h1

, WC=83%, 330 ppm Berol 26].

3.1.1. Separator efficiency

Typical results from the water-continuous sep

rator efficiency tests are shown in Fig. 5. As ca

be seen from the figure, the only significan

parameters found were water cut (amount of di

persed phase) and concentration of Berol 2

Also, the water cut was only a significant param

ter for runs with 330 ppm of Berol 26. The effec

of Berol 26 is obviously to stabilize the oil-in-wa

ter droplets. This surfactant effect is somewh

surprising and will be discussed in more deta

below. The effect of water cut is attributed to th

amount of internal phase required to be tran

ported through the coalescing interface, as su

gested by Refs. [1,10].

Fig. 3. Iso-kinetic injection system for measuring DSD.


7/12


Fig. 4. Water-continuous tests of different inlet types, efficiency versus inlet type. Constant Qt=12 m3 h1.

3.2. Dispersion layer measurements

The sample points inside the separator (points

3, 4, and 5) showed pure phases in at least two

sample points for all oil-continuous tests. This is

caused by the low stability of the system, being

unable to produce a dispersion layer thicker than

the resolution of the sample point (0.03 m). These

results are therefore not discussed further. The

only experiments where the separator performed

satisfactorily, i.e. having all the three sample

points within the dispersion band are shown in

Fig. 6. The linear concentration gradients for the

different systems are shown in Table 3. As can be

seen from the table, the difference in dispersion

layer gradient between systems [water cut 50%,

no Berol 26] and [water cut 83%, 330 ppm Berol

26] is 1.8 versus 2.6 WC% cm1. In Fig. 5,

the difference in separator efficiency for these

cases is 0.99 and 0.84, which probably is the

result of a difference in dispersion layer thickness.

However, the dispersion concentration gradients

suggest the opposite effect; the thickness of the

dispersion layer should be larger with decreasing

slope. This implies that the concentration gradi-

ent of the dispersion layer is non-linear, at least

for the [water cut 83%, 330 ppm Berol 26]-sys-

tem.

The reason for the variance within the calcu

lated concentrations at Normal Interface Lev

(NIL) origins is difficulties with the pressu

transmitter calibrations between runs. Assuming

linear gradient and neglecting all other forces bu

gravity, these values should coincide at WC=0

which is the definition of NIL when using pre

sure transmitters as controllers. The gradient ma

indeed be non-linear and/or discontinuous (ha

ing a discontinuity at a maximum dispersed pha

Fig. 5. Tests of different Berol 26 concentrations, efficien

versus two-way interactions between WC and Berol 26 conce

tration. Constant inlet type D, flow rate Qt=18 m3 h

Pressure drop variations disregarded. This figure also includ

the oil-continuous regime, for comparison.


8/12


Fig. 6. Dispersion layer gradients.

Table 3

Dispersion layer gradientsa

CorrelationBerol 26 (ppm)Water cut Concentration gradient Calculated interception (water fraction at

measured NIL)(WC% cm1)coefficient

0.992 1.8 0.3650% 0

0.996 12.60 0.4484%

0.96684% 2.6330 0.31

a All flow rates are Qt=18 m3 h1.

value), but this cannot explain the large variance

found. The slopes should however be unaffected

by this absolute value at calculated NIL.

3.3. Droplet size distribution

Table 4 shows the experimental set-up and val-

ues for the DSD experiments. The set-up consist

of five parts: Two 23 factorial design (w/o, o/w)

with 330 ppm Berol 26, extended data points forevaluating the pressure drop effect, effect of dif-

ferent inlet devices and the effect of no surfactant

added. The DSDs are in all cases represented as

linear average diameter, which is defined as the

sum of measured diameters divided with the num-

ber of measurements. Although many authors

prefer volumetric or maximal diameter [11], it wa

found that in these experiments a weighting of th

measured values would respond to a large erro

due to the broadness of the distribution and th

limiting number of droplets counted (1000 p

experiment). The volumetric weighting of th

largest droplet would typically represent 30% o

the total volume.

3.3.1. 23 factorial design oil-continuous

The combined effects expressed as interactioterms are relatively small and by neglecting them

one will get a model (Eq. (2)) which explain

89.9% of the total variation in the experiment

data.

d(lin=63+9WC+7Qt7.3 dP (


9/12


The predicted values from the model versus

measured values are given in Fig. 7. The effect of

higher water cut and total flow rate in Eq. (2) is to

increase the linear average drop size while higher

pressure drops will create smaller linear drop size.

The water cut and the pressure drop show effects

that are consistent with theory [12], i.e. higher

internal phase fraction gives larger droplets andhigher shears give smaller droplets. A higher total

flow rate should result in smaller droplets since it

gives rise to a higher shear and the retention time

between the choke and the measuring point (inle

would be less, hence shorter time for coalescenc

to occur.

The reason for this is probably the fact that a

the oil-continuous experiments suffered to som

degree from coalescence in the dilution tank an

this would further increase the experimental erro

to the value of the data variation. This coalecence gave rise to a couple of very large, non-me

surable droplets, and numerous small satelli

droplets. Measurements of these satellite drople

Table 4

Experimental results from the DSD experiments

Qt (m3 h1) DP (bar) Measuring pointWC Berol 26 (ppm) Linear average diameter (mm)

Oil-continuous, 23 factorial design

0.25 33018 787 1

0.25 33018 863 1

5833010.25 7120.25 33012 663 1

18 70.16 1 330 46

7033010.16 318

12 70.16 1 330 41

330 590.16 12 3 1

Water-continuous, 23 factorial design

7 1 330 820.83 18

3 118 3300.83 95

12133017120.83

12 30.83 1 330 102

10133010.5 718

18 30.5 1 330 99

3300.5 12412 7 1330 1220.5 12 3 1

Pressure drop effect

10.5 68330120.16

873300.25 112 0.5

10.5 206120.5 330

1060.83 12 3300.5 1

Inlet de6ice effect

180.16 487 3302, device D

630.16 18 7 2, device B 330

2, device F18 330 4970.16

7 1 0 470.16 18

Effect of no surfactant0.25 5701718

01 2757180.5

7 1 0 212180.83

18 70.83 2, device D 0 227

22902, device D70.83 18

18 7 1 0 2070.83


10/12


Fig. 7. Linear regression line, oil-continuous factorial design.

The combined effects and the pressure dro

effect are relatively small and by neglecting the

one will get a model (Eq. (3)) which explain

83.1% of the total variation in the experiment

data.

d(lin=105.85.8WC11.5Qt (

The predicted values from the model versumeasured values are given in Fig. 8. The effect o

higher water cut, in this case lower disperse pha

fraction, and higher total flow rate sugges

smaller droplets. Both these trends are in acco

dance with common theory [12].

3.3.3. Pressure drop effect

In order to evaluate the pressure drop effe

more precisely, four additional experiments with

pressure drop of 0.5 bar at a flow of 12 m 3 h

were carried out. Fig. 9 displays the results an

the effect of larger pressure drops and low

internal phase fraction is definitely small

droplets. Theory [12] based on the viscosity an

density of dispersed and continuous phase pr

dicts that water droplets in general should b

larger than the oil droplets and this is the opposi

of these results. The coalescence of the large wate

droplets observed for the oil-continuous system

is one possible explanation.

3.3.4. Inlet de6ice effect

Unfortunately these experiments were carrieout with the oil-continuous system and one mu

have the experimental drawbacks in mind. Th

different inlet devices tested are shown in Fig.

and their effects on the linear droplet size a

shown in Table 4. The table indicates very litt

difference in the linear droplet size due to diffe

ent inlet devices. Hence the shear arising from th

different configuration of the inlets does not seem

to be large enough to alter the average diamete

of the droplets.

3.3.5. Effect of no surfactant

Fig. 10 displays the effect of surfactant on th

linear average drop size at a total flow rate of 1

m3 h1 and at 7 bar pressure drop. For oil-con

tinuous systems and experimental condition

given above it seems that addition of surfacta

Fig. 8. Linear regression line, water-continuous factorial de-

sign.

would then give a smaller average diameter than

the actual one entering the separator. The coales-

cence of water droplets is obviously affecting the

results for the oil-continuous emulsions.

3.3.2. 23 factorial design water-continuous

The water-continuous experiments did not suf-

fer from coalescence in the dilution tank and gave

much more accurate values than the correspond-ing oil-continuous experiments discussed above.

This is probably due to the fact that Berol 26

stabilized the oil droplets and this is also in

accordance with Fig. 5 which clearly shows that

the dispersions are more stable in the water-con-

tinuous regime.


11/12


Fig. 9. Average linear diameter versus pressure drop at differ-

ent water cuts.

droplets will increase at the expense of wat

droplets. However at higher surfactant concentr

tions the w/o stability will predominate, which

also experimentally observed.

The experiments with inlet device D and repl

cates without surfactant confirm that the wate

continuous system gives reliable results (227 an

229 mm).Another important factor concerning the use o

surfactants are the ageing of the system due to th

different processes. Especially when working wit

large apparatus and flowing systems open to a

mosphere one cannot discard that the concentra

tion of the surfactant in the two phases w

change with time. The surfactant molecules ca

for instance interact with impurities, biologic

agents degrading the oil phase, forming unwante

by-products or be accumulated in the syste

thereby affecting the bulk concentration an

hence the equilibrium. In the separator system

is possible that a small layer of stable emulsion

formed and accumulated in the feed tank. I

order to create this emulsion one must have

relatively large amount of surfactant; in this ca

]330 ppm since this gave an unstable dispersion

Local regions with higher surfactant concentr

tion than in the bulk can be created when dense

packed droplets coalesce and the interfacial are

is greatly reduced. If these local regions are sub

jected to turbulence, droplets with enough su

gives rise to somewhat larger droplets. On the

other hand the water-continuous systems clearly

show that the surfactant alters the droplets tosmaller sizes. In other words Berol 26 seems to

stabilize the oil droplets. There might be a

straightforward explanation for the findings with

regard to emulsion stabilization at these surfac-

tant concentrations and flow rates. Since the com-

mercially available surfactant will have a

polydispersity in the number of EO units, one can

presume that the molecules with the highest

amount of EO groups will possess the highest

surface activity. A direct consequence of this is

that at low concentrations the stability of oil

Fig. 10. Average linear diameter versus water cut at different Berol 26 concentrations.


12/12


face-active components to create a stable emul-

sion can be formed. The amount of surfactant

consumed in these processes will increase with

time given ageing of the system. Therefore it is

important when using these kinds of systems to

carry out the experiments in a rapid sequence and

hopefully without errors.

4. Conclusions

The main objectives of these experiments were

to establish a model system that performed with

some of the same characteristics as a crude oil

system. It was especially important to achieve

some stability with regard to the oil-continuous

experiments. The use of a commercial surfactant

did not come up with the anticipated results,

which is attributed to the dual and effective na-

ture of the commercial surfactant. In the concen-

tration range selected for the experiments, the

opposite effects occurred. The dispersions in the

oil-continuous experiments were unstable, but the

dispersions in the water-continuous experiments

performed well with the desired stability. The

results from the water-continuous experiments

with regard to the influence of instrument

parameters and flow showed the same trends as

found in earlier work.

In order to obtain more reliable results, a new

type of model oil system with promising resultsregarding stability of the oil-continuous disper-

sions has been established and will be published

soon.

Acknowledgements

Per Arild Kjlseth Andresen would like to ac-

knowledge the technology program Flucha

financed by the Norwegian Research Council

(NFR) and the industry for a Ph.D. gran

Kvrner Process Systems is especially acknow

edged for allowing the experimental work to b

carried out on their gravity separator and wil

ingly sharing their knowledge regarding separato

systems. Richard Arntzen would like to acknow

edge the Mobility Program financed by the No

wegian Research Council (NFR) and Kvrner foa Ph.D. grant.

References

[1] S.A.K. Jeelani, S. Hartland, Effect of dispersion prope

ties on the separation of batch liquidliquid dispersion

Ind. Eng. Chem. Res. 37 (1998) 547554.

[2] A. Bhardwaj, S. Hartland, Kinetics of coalescence

water droplets in water-in-crude oil-emulsions, J. Dispe

sion Sci. Technol. 15 (2) (1994) 133146.

[3] W.J. Howarth, Measurment of coalescence frequency

an agitated tank, AIChE J. 13 (5) (1967) 10071013.[4] R. Shinnar, On the behaviour of liquid dispersions

mixing vessels, J. Fluid Mech. 10 (1961) 259275.

[5] B. Weinstein, R.E. Treybal, Liquid liquid contacting

unbaffled, agitated vessels, AIChE J. 19 (2) (1973) 304

312.

[6] F.B. Sprow, Drop size distributions in strongly coalesci

agitated liquid liquid systems, AIChE J. 13 (5) (196

995997.

[7] A. Kumar, S. Hartland, A unified correlation for t

prediction of dispersed-phase hold-up in liquidliquid-e

traction columns, Ind. Eng. Chem. Res. 34 (1995) 3925

3940.

[8] A. Kumar, S. Hartland, Unified correlations for the pr

diction of drop size in liquidliquid extraction column

Ind. Eng. Chem. Res. 35 (1996) 26822695.

[9] P.A.K. Andresen, X. Yang, J. Sjblom, H. Linga, F.

Nilsen, A new method for determining droplet size distr

bution of unstable dispersions, J. Dispersion Sci. Techno

20 (1&2) (1999) 187197.

[10] H. Polderman, Field tests on Draugen. SPE Proceeding

1998 Annual Technical Conference and Exhibition.

[11] A.J. Karabelas, Droplet size spectra generated in turb

lent pipe flow of dilute liquid/liquid dispersions, AIChE

24 (2) (1978) 170180.

[12] P. Walstra, Principles of emulsion formation, Chem. En

Sci. 48 (2) (1993) 333349.

.

Documents

Andresen, Arntzen, Sjøblom - 2000 - Stability of model emulsions and determination of droplet size distributions in a gravity separator