Application of Multiphase Desander Technology to Oil and

Gas Production
C. Hank Rawlins
Kvaerner Process Systems US

Presented at the 3rd International Conference on Multiphase Technology in Banff, Alberta,

Canada: 3-5 June 2002.

1 ABSTRACT

Solid- liquid hydrocyclone technology has matured to encompass multiphase flow

applications. Multiphase desanders were first employed by the oil industry in 1995 to
remove solids upstream of the wellhead choke1 . The technology has primarily advanced
based on field applications, with little testing or modeling. This paper will provide a
background of multiphase desanding and its applications, and summarize the work
accomplished to date on modeling pressure drop and separation efficiency. An outline of an
empirical model for scaling multiphase desanders from liquid based applications is presented
as work to be undertaken in a multiphase flow loop. Comparison to the existing hydraulic
and pneumatic cyclone mechanistic models is also discussed.

2 INTRODUCTION

2.1 Oil and Gas Solids Control

All oil and gas wells produce solids in varying concentration with the produced fluids. These
solids originate either from the reservoir, such as sand, clay, or silt, or may be corrosion
byproducts from the downhole equipment. Produced solids are normally smaller than 250
micron in diameter and are found in varying amounts, with a typical concentration range of 5
to 250 PPM2 .

Production of sand from oil & gas wells is a common difficulty for processing facilities, and
is often caused by unconsolidated reservoirs, high water production rates, or the failure of
gravel packs and other sand control measures. Detrimental effects of sand production include
mechanical damage (erosion) to chokes, flow lines, control valves, pumps and other high
velocity areas, reduced residence time (in the case of separators) or partial blockage of
pipelines, and high environmental discharge costs.
Page 1 of 19
The main sand control solutions to date have been to cut back fluid production rates to
reduce the rate of sand, work over of wells to repair or install gravel packing or screens, or
increased equipment maintenance which leads to reduced plant availability. The need for an
alternative solution that enhances rather than restricts production was recognized by
operators, which lead to the introduction of multiphase desanders in the mid 1990’s. This
device enables a producing operation to remove solids on- line at the wellhead choke, which
both maintained production while protecting process equipment.

Selection of the proper solids removal location is primarily a matter of economics. The four
principal solids removal locations identified for a common separation train are shown in
Figure 1. Solids removed upstream of the wellhead choke will protect all downstream
process equipment, while removal after the choke will protect the separation processes but
not the choke. The determining factor on desander location is the cost balance of an
upstream (high pressure) separating device with the cost of choke maintenance. Post-choke
remova l has a lower design pressure (lower cost), but the equipment may be larger due to the
increased actual volume of gas treated. Alternatively the production separator (2 or 3 phase)
makes a great separation device for solids, however solids build up decreases residence time.
A sand jet system can be employed to remove collected solids in the vessel, but does not
benefit downstream oil or water treating equipment. Many standard (liquid only) desander
systems installed on water and oil treating lines do an excellent job of protecting treaters and
flotation cells from filling up or preventing injection wells from plugging. The aim of this
paper is to discuss the removal of solids in multiphase applications, before or after the
production choke.

A serendipitous benefit of removing solids early in the process scheme is the cleanliness of
the separated sand. Upon commissioning of the first commercial wellhead desander, one of
the major concerns of the operator pertained to solids handling. This concern was placated
when the sand dumped from the accumulator was found to be relatively free of hydrocarbons
(clean enough for direct discharge). The reasons for the low hydrocarbon concentration on
sand are twofold. First, sand is typically (and preferably) water wet in the reservoir, and will
remain so unless allowed to "cook" in production separators for extended periods. During
retention in process vessels the sand is subjected to interfacial modifying chemicals and
processes that allow the solids to become oil wet, thereby creating an oil layer on the sand
particle. Removing the sand ahead of the separators prevents oil adsorption. Second, proper
operation of a multiphase desander ensures that the collection vessel (accumulator) is pre-
filled with clean water after each purge cycle. Therefore the separated sand is allowed to sit
in a clean water environment as hydrocarbon well fluids are prevented from entering the
accumulator to discharge with the sand. In field applications typical total hydrocarbon
concentration on sand has measured less than 0.5% by weight (kg oil per kg dry sand). This
is analogous to oil on drill cuttings, which can be directly discharged in many regions,
providing the oil on solids concentration is less than 1% by weight.

3 DESANDING HYDROCYCLONES

3.1 Desanders as Subset of Hydrocyclones

Solid- liquid cyclones (hydrocyclones) have been used for many decades in various industrial
applications ranging from classification of mineral solids to removing grit from hot tomato

Page 2 of 19
catsup. In the oil and gas industry solid- liquid cyclones are used in drilling operations, while
only recently has this technology found significant use in oil and gas production operations2 .

By definition, a “desander” is a solid- liquid hydrocyclone that operates with an enclosed

underflow and an accumulation section (integral or separate) with a batch discharge, as
shown in Figure 2. They are also called water-sealed hydrocyclones or grit separators. Solids
are centrifugally separated from the liquid as per a standard hydrocyclone, however the y are
not continuously discharged from the apex (spigot). Solids exiting the apex collect into a
closed accumulation section for batch discharge. The rate of discharge is directly
proportional to the rate of solids recovery and the size of the accumulation vessel. Because
of batch discharge, desanders should be limited to cases where solids are 1 vol.% or less.
When the concentration of the solids is significantly larger than this amount the application
would be better suited for a standard hydrocyclone wit h open underflow.

Closing the underflow allows use in high-pressure operations and prevents air from being
drawing into the vortex center. With a desander all fluids entering the unit, are discharged
from the overflow, as no liquid split occurs. Because the underflow is closed, the pressure
drop is controlled from inlet-overflow. Throttling of this stream controls the pressure drop,
and in effect the desander acts as an orifice in the line, regardless of the operating pressure.
Desanders have been built for 10,000# API rating (68,900 kPag), while standard
hydrocyclones have been generally limited to 100 psig (689 kPag). Additionally, since the
underflow is closed no air core is formed in the center of the vortex. In open underflow
hydrocyclones the vortex flow pattern creates a low-pressure region at the cyclone center,
which draws air through the apex up into the vortex finder. Air mixed with hydrocarbons is a
hazardous mixture, and closing the underflow prevents this from occurring. Closing the
underflow does have the effect of coarsening the separation size but increasing the sharpness
of separation. This occurs by limiting the settling velocity of small solids, but this effect has
not been quantified and will be investigated in future analysis 3 .

Desanders are available in different sizes and styles. The vessel style and liner style are
shown in Figure 2. These units vary in size range and construction, and selection of each is
dependent upon the capacity, separation size, and turndown requirements. The vessel size
ranges from 2-30 inch nominal diameter (50-750 mm), with the vessel itself acting as the
desander. The liner style ranges from 0.5-4.0 inch nominal diameter (10-100 mm), and
consists of a pressure vessel with several “liners” contained within. The vessel style units are
used in applications requiring large flow rates and coarse separation size (25-200 microns),
while the liner style is used in applications at most any flow rate with fine separation size
(10-50 microns).

Like all cyclonic devices, desanders offer the highest throughput-to-size ratio compared to
conventional separation equipment, such as plate interceptors or filters. Their weight and
size is usually 25% of standard equipment, and the capital cost is likewise much lower.
Desanders require little, if any, operator interaction for operation, and due to the fact they
have no moving parts maintenance is minimal.

3.2 Desander Models

Solid- liquid hydrocyclones have been studied extensively because of their applicability in
separation processes, however desanders specifically have received little attention.
Researchers such as Yoshioka & Hotta4 , Bradley5 , Rietema6 , Holland-Batt7 , Trawinksi8 ,
Page 3 of 19
Plitt9 , and Svarovsky3 have provided a good base of mechanistic and empirical models to
draw upon, but their work is primarily based on open underflow hydrocyclones used in
classification applications. Some work has been added by each in terms of underflow
control, however only Quian10 and Witbeck11 have published experimental work on closed
underflow (water sealed) hydrocyclones. This work did not provide a model to design or size
a desander, only experimental comparison for specific applications. Multiphase desanders
models have not been addressed, which will be the subsequent work from this paper.

4 MULTIPHASE DESANDERS

4.1 History of Multiphase Desanders

A joint industry project (JIP) initiated in 1994 was set up to investigate the viability of using
specially designed desanders for solids removal from multiphase production operations. The
operators involved in the initial study were BP, Chevron, LASMO, Shell Expro, and Statoil1 .

The comprehensive development, from concept to commercial product, of the first wellhead
desander was carried out in three stages. Internal testing at a cyclone vendor laboratory
performed the initial phase of testing. This involved the investigation of air and water
systems in controlled laboratory conditions focusing on the hydrodynamic relationship
between pressure drop, flow rate, and gas void fraction (GVF) in a desander. Additionally,
the sand removal efficiency at optimum flow parameters was tested to establish the design of
a pilot unit. The second phase of work was JIP funded, and involved the construction and
testing of a pilot plant at BP Wytch Farm onshore production facility. A third, parallel
exercise was undertaken on a 5,000 psig (34,450 kPag) design pressure prototype wellhead
desander to prove the mechanical viability of a commercial unit. Using successful laboratory
and field data, the first commercial unit was manufactured and installed on the Shell Expro
Brent Field, 18 months after first concept.

Since the installation of the first unit in 1995, at least 40 more units have been installed in
the North Sea, Gulf of Mexico, Venezuela, Brazil, Saudi Arabia, Indonesia, and Canada.
These units are split between wellhead and wellstream use, with the wellhead units in use as
a service tool and the wellstream units in use for standard production. These units have
ranged in size from 3-20 inch (75-500 mm) nominal diameter and from 150-10,000# rating
(1033-68900 kPag). Vessel material of construction is carbon steel, 316 stainless steel, and
standard/super duplex stainless steel. Internal wear components are typically made of boride
hardened 410 stainless steel. Very high wear applications have added alumina (Al2 O3 ) or
silicon carbide (SiC) ceramic lining as tile or cast components.

4.2 Application and Use in Oil & Gas Industry

The first multiphase desanders units were installed upstream of the production choke and
became termed “wellhead desanders”, as shown in Figure 3. As application base grew
specific designs were made to operate downstream of the choke. These units were
correspondingly termed “wellstream desanders”. Both units operate in multiphase flow and
are of similar design. The main difference in each is due to mechanical design requirements.
A wellhead desander may be designed for 5,000-15,000 psi API rating, while a wellstream
desander is typically designed for 150#-600# ASME. Benefits of pre-choke installation
include protection of the choke and all downstream process equipment, lower actual volume
of fluids to be treated (smaller diameter desander), possible elimination of PSV as unit
Page 4 of 19
designed for full shut- in pressure. Well service companies have used wellhead desanders as a
work-over tool. Benefits of post-choke installation include lower design pressure and
potential to treat several wells through single unit. Primary wellstream desander installations
are with oil companies to protect downstream process equipment (e.g. separators).

Multiphase desander technology can be a valuable tool for use by both oil producers as well
as service companies. Generally the equipment is provided as permanent installation on
operating systems, or as a portable skid for service company work. The initial design role
was to remove solids from fracture flow back operation. Due to the flexible robust design,
this scope has greatly increased to include many applications.

1. Removal of naturally produced formation sand from wells in which a gravel pack
was never installed or has failed. A 30% increase in sand free production rate has
been achieved in some applications.
2. Removal of sand flow back from coiled tubing fracture jobs, well bore flush out,
or well acidizing.
3. Removal of solids from well fluids in underbalanced drilling operations.
4. Removal of solids from gas and condensate wellstream to prevent plugging of
printed circuit heat exchanger technology used in gas well cooling applications.
5. Removal of sand from oil & gas well streams to protect sensitive downstream
equipment, and avoid choke, manifold, and flow line erosion.
6. Monitoring and measurement of sand production rates (i.e. test unit).

Both the vessel and liner type desander configurations have been used in multiphase
applications. Theoretically both configurations are applicable, however to date most
installations have used the vessel style because it handles a higher solids loading, is less
susceptible to plugging by large particles, has a smaller vessel diameter, and does not suffer
from possible maldistribution of gas- liquid flow to cyclone internals. The liner style does
have the advantage of finer separation size and modularized design that allows for simpler
maintenance.

5 MECHANISTIC MODELS

5.1 Operating Principles

A multiphase desander operates on the same principles as a standard liquid only desander,
except that it is designed to treat combined gas- liquid flow. The unit uses pressure energy
from the wellstream fluid to achieve cyclonic separation of solids. The feed stream from the
wellhead, a mixture of liquids (oil, drilling fluid, water, etc.), gas (natural, nitrogen, etc.),
and solids (sand, scale, rust, etc.) enters the cyclone insert tangentially through the inlet. The
change in flow direction forces the mixture to spin in a vortex pattern. This vortex flow is
accelerated as the internal diameter is reduced over the length of the cone. Due to angular
acceleration of the flow pattern, centrifugal forces are imparted on the solid particle forcing
them toward the wall of the cone. The solids continue to spin in a radial vortex motion, down
the length of the cone, to discharge through the apex creating the underflow stream. Due to
the cone convergence, the multiphase fluid flow is reversed, and sent upward through the
vortex finder to create the overflow stream. The solids that exit through the apex collect into
an accumulation chamber, where they are periodically purged, while the overflow discharges
continually. The presence of free gas in the feed stream has the beneficial effect of
Page 5 of 19
increasing inlet fluid velocity, decreasing the fluid mixture density and viscosity, and
creating an annular liquid flow patch (gas core) that decreases the particle migration
distance. These effects work together to increase the solids recovery efficiency.

The internal geometry of the cyclone insert is extremely important, and is driven by detailed
knowledge of the fluid properties, multiphase flow regime, allowable pressure drop, and
separation efficiency requirements. Conventions used in this paper for desander geometry
are shown in Figure 4. Multiphase desanders have been applied to well fluids from 0-100%
GVF, with the geometry customized for the process conditions, solids characteristics, and
site space requirements.

5.2 Flow Regime

The methodology for determining pressure drop and separation efficiency of a multiphase
desander has stemmed primarily from empirical scaling relationships based hydraulic and
pneumatic cyclone models. To determine which base and scale model to apply, the inlet flow
regime must first be determined. As a simplistic approach, low GVF applications are treated
as hydraulic desanders with gas in the fluid (liquid dominant), and high GVF applications
have been treated as pneumatic cyclones with a heavy wet gas inlet (gas dominant). The
boundary line for which model to use is based on the determination of the inlet flow regime.

Desanders are normally operated in the vertical position, so the inlet line runs horizontally
into the unit. Flow regime determination for two-phase flow in a horizontal line can be
determined by several methods (Beggs-Brill12 , Gregory-Mandhane-Aziz13 , Taitel-Dukler14 ,
etc.). The first generation multiphase desander model uses the Gregory model because of the
relative simplicity in plotting the flow regime map13 , as shown in Figure 5. Pipe geometry
and fluid data determine the flow regime. Annular mist flow is deemed gas dominant and all
other regimes (slug, dispersed, churn, etc.) are deemed liquid dominant. Liquid dominant
flows are treated with the hydraulic cyclone model and gas dominant flows are treated with
the pneumatic cyclone model.

5.3 Pressure drop

Pressure drop is the energy source for separation in any cyclone. It is also the only practical
operating parameter that can be changed on- line to affect separation efficiency. The pressure
drop is directly related to a cyclone flow rate, thus is controlled by the amount of fluid put
through the unit. Increasing or decreasing the throughput, will have a directly proportional
effect on the pressure drop.

A multiphase desander can operate over a wide pressure drop range providing an inherent
turndown ratio. The recommended minimum operating pressure drop for a multiphase
desander is 10 psi (69 kPa). Below this value, the fluid does not contain enough energy to
form the proper vortex flow pattern. No theoretical maximum pressure drop exists, but 75-
100 psi drop (517-689 kPa) is offered as a practical maximum. Above this value the wear
rate continues to increase as velocity to the fourth power, while the added recovery is
minimal. Liquid dominant flow regimes will generally produce a pressure drop in the 25-75
psi range (172-517 kPa), while gas dominant flow regimes will generally produce a pressure
drop in the 15-40 psi range (103-276 kPa). Very dry gas flows may operate at 2-5 psi
pressure drop (14-34 kPa), however this occurs in low-pressure gas only applications.

Page 6 of 19
5.3.1 Liquid Dominant Flow
Hydrocyclones will have a specific pressure drop versus capacity relationship exhibited by
the capacity power law. This law is shown in Equation 1 below to express the flow rate (Q)
in terms of the pressure drop (∆P).

Q = a ( ∆P) b Eq. 1

The terms a and b are correlating factors to fit the mathematical relationship. For the
desander geometry in which these correlations were developed (termed “standard
geometry”), the power factor (b) in this relationship is found to average 0.5, which is the
slope of the curve as plotted on an ln- ln graph. The equation then reduces to the simple
relationship shown in Equation 2.

Q = k ∆P Eq. 2

The k term is called the “k- factor” and is specific for each hydrocyclone size and geometry.
In effect it is similar to a listed Cv value for a valve. Cyclone diameter has the greatest
directly proportional effect on k- factor, with inlet area, vortex finder diameter, and cyclone
length exhibiting proportional lesser effects. The capacity relationship for a range of
hydrocyclones is shown in Figure 6. These curves are determined by direct flow laboratory
measurements.

The Hadfield-Rawlins15 relationship is a modification of the capacity power law corrected

for multiphase flow, and is shown in Equation 3.
2
 Q
∆P =   Eq. 3
 Gk 

This is the same equation as previously presented, only factored for pressure drop. The k-
factor is present, as in the standard desander rela tionship, and has the same function. A
further “G-factor” (dimensionless) is present as an increase to the pressure drop due to the
presence of gas. This increase in pressure drop is due to the actual fluid volume the free gas
will occupy, that both increases the inlet velocity and increases the effective fluid volume
present in the hydrocyclone. The G- factor was determined experimentally both in lab and
subsequent field installations. The term was found to provide a distinct increase in pressure
drop with increase in gas liquid ratio (v/v), as shown in Figure 7. This is a very simple
approximation found valid for most field applications, although field data has been difficult
to obtain. Further work will be undertaken to develop a mechanistic model from first
principles to validate and extend the model.

This increase in pressure drop is measured for gas liquid ratios (v/v), up to approximately
20:1, above which it is assumed the flow regime is gas dominant (mist flow). Once the k and
G factors are applied to the flow rate, the resultant pressure drop can be input into the
separation size calculation.

5.3.2 Gas Dominant Flow

A pneumatic cyclone model is used for gas dominant flow in determination of pressure drop.
The Stairmand relationship was selected both for its simplicity of calculation and
Page 7 of 19
applicability to the cyclone geometry tested16 . The mist flow acts as a heavy gas and follows
this curve closely. The formula for pressure drop calculation is shown in Equation 4.

N H ρ f Q2
∆P = Eq. 4
2K a 2 Kb 2 Dc 4

a
Ka = Eq. 5
Dc

b
Kb = Eq. 6
Dc

The term NH is a cyclone configuration constant, dependent upon the geometry. This value
has been determined for standard geometry function. The Dukler mixing density14 must be
used for the fluid density (ρf), and the actual total liquid flow rate (gas plus liquid) is used as
the capacity. The terms Ka and Kb relate the inlet height (a) and width (b) for a rectangular
inlet design to the cyclone diameter (Dc). These terms can be calculated for a circular inlet
using the hydraulic radius. This equation has been found to provide good correlation to
liquid laden produced gas in field applications.

5.4 Separation Efficiency

Similar to pressure drop calculations, the algorithm for multiphase desander separation
efficiency starts with determination of inlet flow regime. The gas or liquid dominant nature
present at the multiphase desander inlet will determine the hydrodynamic and separation
efficiency nature of the unit. A multiphase desander operating in the annular mist flow
regime (gas dominant) will follow the recovery efficiency correlation of Leith-Licht for gas
cyclones. Units operating in all other flow regimes (liquid dominant) follow the Lynch-Rao
efficiency model for hydrocyclones. Once the inlet flow regime has been determined the
corresponding correlation can be used to designate the proper multiphase desander
configuration and model the separation efficiency and size.

5.4.1 Liquid Dominant Flow

The Lynch-Rao relationship modified for multiphase flow can be used for all liquid
dominant flow regimes17 . A thorough review of the basic relationship has been detailed in
previous publications 2 , and only the basics are presented here. The reduced recovery curve,
base d50 , and actual d50 formulas are followed as the basis. The reduced recovery curve, also
called graded efficiency cur ve, plots the recovery percentage for each particle size. It is
called reduced because the flow split correction taken for a open underflow is removed to
reduce the recovery to 0% for a particle of 0 microns. The base d50 is the particle size with a
50% chance of capture, based only on cyclone diameter. The actual d50 corrects the base d50
for cyclone geometry and fluid properties. The actual d50 is what the desander will exhibit
when installed. Corrections for inlet concentration, inlet area, vortex finder diameter, body
length, and cone angle remain the same, while the corrections for pressure drop, fluid
viscosity, and fluid density are modified for multiphase fluids. The net result is to produce
the actual d50 .

Page 8 of 19
The pressure drop for multiphase flow is much higher than hydraulic flow for a given liquid
rate as presented in the previous section. The correction to the base d50 can still be utilized,
once the resultant (higher) pressure drop is determined.

Both the fluid density and viscosity are greatly changed by the presence of free gas. In a
hydraulic desander, a solid particle must migrate through the carrier liquid, and is impeded
by the liquid viscosity and density. The presence of gas has the benefit of changing the
carrier fluid from a liquid to vapor- liquid mixture. The resultant mixture has a viscosity and
density between the base constituents, as shown in the Dukler mixing relationships of
Equations 7-9. First the liquid fraction (λ) is determined from the actual gas (QL) and liquid
(QG) flow rates. The resultant mixture density (ρM) and viscosity (µM) are determined by rule
of mixture relationships to the gas and liquid density (ρL, ρG) and viscosity (µL, µG)
respectively.

QL
λ= Eq. 7
QL + QG

ρ M = ρL λ + ρG (1 − λ ) Eq. 8

µ M = µ L λ + µ G (1 − λ ) Eq. 9

These equations must be calculated in-situ to account for the actual volume of gas at the
process pressure and temperature. Once the mixture density and viscosity are known, the
standard correction to base d50 can be used.

Using actual d50 for the specific hydrocyclone, the separation efficiency can be obtained. The
actual d50 and the inlet feed particle size distribution are used with the recovery curve to
calculate the percent recovery for each particle size. The shape of a recovery curve for a
standard geometry desander is defined mathematically by the Lynch-Rao relationship shown
in Equation 10, and illustrated in Figure 8. This equation can be used to calculate the
recovery probability for each particle size in a range.
d
α
d e d 50
−1
Y( )= d × 100 Eq. 10
d 50 α
e d 50
+ eα − 2

In this equation the actual d50 is used, along with each particle size (d) and the sharpness
slope (α). The sharpness slope varies with fluid and solid properties. The recovery (Y) for
each particle is then summed up for the complete range to give a total recovery. The
recovery percentage (E) is the relationship of the recovered amount- inlet solids content (f)
minus overflow solids content (o)-to the inlet solids content, which is shown in more general
terms for a hydrocyclone in Equation 11.

f −o
E= x100 Eq. 11
f

Page 9 of 19
The calculated cut size can further be used to generate the separation size. Using the Lynch-
Rao relationship, with α of 4.0 (sand in water), the separation size (d98 ) is equal to 1.97 times
the cut size (d50 ).

A relationship determined by Plitt can be used to calculate α for the Lynch-Rao equation3 .
Plitt uses Rosin- Rammler as an alternative to the Lynch-Rao equation, which defines a
sharpness value m. The value for m can be calculated with Equation 12, and converted to α
with Equation 13.

D 2 ( L − l ) 0. 15
m = 2.96 exp( −1.58R f )( ) Eq. 12
Q

α = 1.54m − 0.47 Eq. 13

Where Rf is the flow split to underflow (0 for desanders), L is the cyclone internal length,
and l is the vortex finder length.

For general comparison, in a liquid dominant application a 4.0 inch (100 mm) multiphase
desander will have a separation size of 25 microns, while a 20 inch (500 mm) desander will
have a separation size of 100 microns. The practical limit for sand separation from a
hydrocyclone is 10 microns.

5.4.2 Gas Dominant Flow

For applications involving annular mist flow, particularly gas wells, a different approach to
separation efficiency is taken. A practical gas cyclone model developed by Leith-Licht is
used for the cyclone design18 . This model is based upon the fluid tangential velocity profile
in Equation 14, where the product of the tangential velocity (v θ) and the radius (r) remains
constant.

vθ r m = cons tan t Eq. 14

The exponent (m) was experimentally derived, and is given by Equation 15. It is shown to be
a function of cyclone diameter (Dc ) and temperature (T).

T 0.3
m = 1 − (1 − 0.67 D c 0.14 )( ) Eq. 15
283

The collection efficiency (η) per particle size (Dp ) is further determined by Leith-Licht
model given by Equation 16.

η ( Dp ) = 1 − exp( −ψDp M ) Eq. 16

In which Equations 17 and 18 give the main parameters ψ and M.

1
M= Eq. 17
( m + 1)

Page 10 of 19
 M
 KQρ p Cc ( m + 1)  2
ψ = 2  Eq. 18
 18µDc 3 

The coefficients K and Cc are called the dimensionless geometric configuration parameter
and the Cunningham correction factor respectively. Both values are specific to the
hydrocyclone geometry, and are experimentally determined for a set geometry ratio. The
particle density (ρp ), gas viscosity (µ), and cyclone diameter are the other main factors.

In the equations above, the fluid viscosity used must be corrected both for actual gas
conditions, as well as the vapor- liquid mixture. This is easily taken into account using the
Dukler mixture viscosity calculated at operating conditions. Cyclone geometric parameters
such as inlet area, vortex finder, cylinder length, and cone angle are configured into the K
and Cc values. Particle specific gravity is corrected to account for fluid mixture density. The
flow rate instead of the pressure drop is used in the efficiency calculation, which is
acceptable as they are directly proportional by a square power factor. The net result is similar
to the actual d50 calculation for hydraulic desanders.

The resultant particle recovery curve is similar in style to the Lynch-Rao curve, but exhibits
a slightly differing shape. Total solids recovery is calculated by summing each individual
particle recovery fraction. The cut size and separation size values are determined from the
above equations, calculated for the 50% and 98% sizes respectively.

For general comparison, in a gas dominant application a 4.0 inch (100 mm) multiphase
desander will have a separation size of 15 microns, while a 20 inch (500 mm) desander will
have a separation size of 50 microns. The practical limit for sand separation from a
pneumatic cyclone is 10 microns.

5.5 Particle Trajectory

It is worthwhile to discuss the trajectory of solid particles in a multiphase desander. Since
the utility of a multiphase desander is to remove sand particles from a mixed fluid stream,
every effort in design and operation must be undertaken to separate and capture these
particles. Transition from axial pipe flow to cyclone vortex flow is undertaken by the
desander inlet transition. The transition both increases the fluid velocity (from 20-30 m/s to
75-100 m/s) and changes the flow direction (axial to centrifugal). Of key importance is to
direct the randomly moving solids in the pipe to the cyclone wall as quick as possible. This
is done with a tangential or volute inlet. Volute inlets, from 180-360º, provide a more
efficient transition but are much more complicated to manufacture on pressure vessels.
Typically tangential inlets are used, as the difference in efficiency is sma ll compared to the
lowered cost.

Particles that successfully translate to the cyclone wall are considered captured and follow
the vortex flow along the inner wall of the cyclone cylinder and cone sections, as shown in
Figure 9. As the radius decreases down the length of the cone, tangential velocity increases,
which captures more and smaller particles. The boundary layer at the wall has a downward
axial velocity that pushes particles towards the apex. If the boundary layer becomes
overloaded with particles, some of the boundary layer flow may remix into the main vortex
flow and leave with the overflow3 . In an open underflow removal of boundary layer fluid is
regulated the apex diameter, which controls the liquid split. On a desander an accumulating
Page 11 of 19
chamber seals the underflow reducing this flow rate to zero. For this reason the amount of
solids a desander can handle in the inlet flow is limited to prevent overcrowding at the
boundary layer and apex. Experimentally this value has been found to be 1 vol.%.

Once the particles are fully captured and report through the apex, they collect in the
accumulating chamber, as illustrated in Figure 9. The solid particles have tangential, radial,
and axial velocity components as they leave the apex, however because the accumulator is a
static liquid filled chamber, the velocity is immediately arrested. The particles then settle by
gravity through the liquid medium into the bottom of the accumulator. Most desanders are
oriented vertically to allow solids to collect near the accumulator discharge orifice.
Horizontal operation of a desander is possible (centrifugal separation forces negate gravity
effects), however the apex must be designed to discharge solids at 90º to drop into the
discharge orifice. The velocity arrest by static liquid coarsens the desander separation size
compared to open underflow operation as small particles are unable to settle quickly and
may be recaptured into the vortex flow. This elutriation effect of small particles does
increase the sharpness of separation (increases α).

The effect of captured particles settling into the static accumulator is called “particle- fluid
swapping”. As a sand particle of given volume settles into the accumulator it displaces an
equivalent volume liquid droplet upward into the desander, as illustrated in Figure 9. Gravity
settling is a function of density therefore sand will settle in a water filled chamber however
oil will not. A beneficial effect arising from this is that solids are washed while settling into
the accumulator. To negate the coarsening effect of separation size by a closed chamber a net
removal of liquid out of the chamber through an orifice other than the apex must be imposed.
Liquid can be removed through the solids discharge orifice, however throttling of slurry in
high-pressure applications leads to premature wear of valves. An alternative is to remove
liquid from the top of the accumulation chamber. As long as the volume of liquid removed
from the accumulator is greater than the volume of solids entering the chamber, no
coarsening effect will occur. The simplest method is to provide a balance line from the top of
the accumulator to the overflow line as shown in Figure 9. Controlling this flow with a
spring balanced flow valve is an economic method of ensur ing sufficient liquid removal
without substantial support equipment. Connecting the balance line to the overflow is a
convenient location as it keeps all liquids within the desander package, and it minimizes the
pressure drop on the flow valve (it will be equivalent to the inlet-overflow pressure drop).
This effect has been exhibited in field applications but no comprehensive study has been
undertaken to quantify the liquid removal rate versus increase in separation efficiency.

6 CONCLUSIONS AND FUTURE WORK

The purpose of this paper is to provide an overview of the history, application, and function
of multiphase desanders. A wide variety of material is presented to form the basis of a
multiphase desander model using hydraulic and pneumatic cyclone models. A
comprehensive understanding of the hydrodynamics and separation theory has been
leapfrogged by the industry with little negative effect on equipment deployment and use.
The simplicity of cyclone operation has enabled these theoretical gaps to be covered by
adjustment in operation, as most units installed in the field have provided excellent
separation characteristics. Variation in unit life has been exhibited, which is mainly
attributed to improper material selection or misapplication of the technology. Overall the
Page 12 of 19
multiphase desander has proven to be a robust addition to the process engineer tool kit in
solving solids separation problems in oil and gas production operations.

Work has recently begun to produce a comprehensive hydrodynamic model for

determination of pressure drop and separation efficiency at the various flow regimes. This
work will be undertaken with a transparent multiphase cyclone test loop using air-water-
sand. Further work includes visualization and modeling of desander-accumulator interaction
(fluid-particle swapping), solids settling and stacking mechanisms in accumulators, fluid
partitioning in liner type vessels, desander orientation effect on separation, and pressure
recovery with tangential overflow.

7 REFERENCES

1. Barrett R. Experiences in solids exclusion operations in support of production operations.

Proceedings of IBC Global Conferences on Production Separation Systems; 2001
December 10-11; Houston, TX.
2. Rawlins, Charles H.; Wang, Isaac I. Design and installation of a sand separation and
handling system for a Gulf of Mexico oil production facility. Paper 63041 Presented at
the Society of Petroleum Engineers Annual Technical Conference; 2000 October 1-4;
Dallas, TX.
3. Svarovsky, L. Hydrocyclones. Lancaster, PA: Technomic Publishing Co., Inc.; 1984.
4. Yoshioka, N. and Hotta, Y. Liquid cyclone as a hydraulic classifier. The Society of
Chemical Engineers, Japan. Vol. 19, N. 12, Dec. 1955.
5. Bradley, D. Cyclone liquid separator. The Hydrocyclone, Pergamon Press, London 1965.
6. Rietema, K. Performance and design of hydrocyclones I – general considerations.
Chemical Engineering Science. Vol. 15, 1961. 298-302.
7. Holland-Batt, A.B. A bulk model for separation in hydrocyclones. Trans. Instn. Min.
Metall. Section C: Mineral Process Extr. Metall.: 91, 1980.
8. Trawinski, H.F. Practical aspects of the design and industrial applications of the
hydrocyclone. Filtration and Separation, Vol. 6. Jul/ Aug and Nov/Dec 1969.
9. Plitt, L.R. A mathematical model of the hydrocyclone classifier. CIM Bulletin, Dec.
1976. 114-123.
10. Quian, L.; Changlie, D.; Jirun, X. Comparison of the performance of water-sealed and
commercial hydrocyclones. Proceedings of the International Journal of Mineral
Processing. Volume 25; 1989. 297-310.
11. Witbeck, W.O.; Woods, D.R. Pressure drop and separation efficiency in a flooded
hydrocyclone. The Canadian Journal of Chemical Engineering. Volume 62; February
1984. 91-98.
12. Beggs, H.D. and Brill, J.P. A study of two-phase flow in inclined pipes. Trans. AIME,
1973. p. 607.
13. Gregory, G.A., Mandhane, J.M., Aziz, K. Some design considerations for two-phase
flow in pipes. Presented at the 25th Annual Tech. Mtg. of Petroleum Society of CIM.
Calgary, Alberta; 1974 May 7-10. Paper No. 374020.
14. Taitel, Y. and Dukler, A.E. A model for predicting flow regime transitions in horizontal
and near horizontal gas- liquid flow. AIChE Journal, Vol. 22, No. 1. Janua ry 1976, 47-55.
15. Rawlins, C.H. Solid Liquid Separation Technical Manual, Kvaerner Process Systems,
Houston, TX (1998), 45.
16. Stairmand, C.J. Trans. Inst. Chem. Engrs. London, Vol. 29. 1951. 356-383.
Page 13 of 19
17. Lynch, A.J. and Rao, T.C. Modeling and scale-up of hydrocyclone classifiers. 11th Int.
Mineral Processing Congress. 1975. 245-269.
18. Benitez, J. Process engineering and design for air pollution: cyclonic devices. 1993. 333-
369.

Page 14 of 19
 Wellstream
Desander
3 Phase Production Separator

Well Wellhead
Tree Desander Sand Jet System

Standard
Desander

Figure 1: Solids removal options in typical production scenario.

Vessel Style
Liner Style

Cyclone
Element

Integral
Accumulation
Chamber

Figure 2: Desander geometry and types - vessel style (left) and liner style (right).

Page 15 of 19
 Well Fluids

Desander

Wellhead

Accumulator

Sand Discharge

Figure 3: Application layout of wellhead desander.

Overflow

b
a
Inlet l
D

Θ L

Accumulator D

Underflow

Figure 4: Desander geometry conventions use in mechanistic models.

Page 16 of 19
 Liquid Gas
Dominant Dominant

Figure 5: Gregory -Mandhane -Aziz flow regime map for horizontal pipes.

100000

20”
Volumetric Flow Rate, BPD

16”
10000

10”

6”

4”
1000
1 10 100
Pressure Drop, psi

Figure 6: Hydraulic (water) capacity relationship for various desander sizes.

Page 17 of 19
 30

25

20
G Factor

15

10

0
0 10 20 30 40 50 60 70 80 90 100 110
GLR (vol./vol.) in-situ

Figure 7: Hadfield-Rawlins relationship for gas effect on pressure drop (G-factor).

100.0
D98=Separation Size
90.0
Particle Recovery (% to Underflow)

80.0

70.0

60.0

50.0
D50=Cut Point
40.0

30.0

20.0

10.0

0.0
0.1 1 1.97 10
Particle Diameter/D50 (µm/µm)

Figure 8: Reduced recovery curve shown as Lynch-Rao relationship.

Page 18 of 19
 Overflow

Inlet Balance
Line

Sand Water Droplet

Particle

Accumulator

Underflow

Figure 9: Particles trajectory in desander (vortex flow and “particle-fluid swapping”).

Page 19 of 19