Sunteți pe pagina 1din 4

Numerical Simulation of a Gas/Liquid/Solid Wringing Separator

A. Bennardo1), F. Podenzani1), L. Ciccarelli2), U. Spinelli3)


1)
ENI Division R&M, Via Maritano 26, 20097 S. Donato Mil (MI), Italy
2)
ENI Division E&P, Via Emilia 1, 20097 S. Donato Mil (MI), Italy
3)
Micronsep, Via Cavour 3, 20040 Briosco (MI), Italy

ABSTRACT

Offshore gas platforms badly need compact gas/liquid/solid separators with very high separation efficiency (>
98%), in order to reduce the size of the offshore production platform, potentially saving millions of euros in
construction and operating costs. Those requirements are fulfilled by the Wringing separator, where a very
complicated fluid dynamics situation is established and can be visualised only by CFD simulations.
CFD simulations have been helpful in describing the physical phenomena taking place in the separator and thus
improving the understanding of the separation mechanisms. These studies created the basis for the present work
in which the effect of geometrical changes in the Wringing separator such as the changing of the deflector or the
adding of one or two partition walls have been investigated both in terms of the fluid dynamics behaviour and
separation efficiency. The CFD approach allowed an easier way to evaluate the effect of those variations on the
apparatus performances compared to the cumbersome and expensive experimental testing.

1 INTRODUCTION
Gas and oil separators are essential in oil extraction..
They are used to separate gas, oil / water and
sometime sand before any other device on the
platform. The separator efficiency is very important for
the well productivity and has to be maintained in a
large range of flow rates.
The Wringing separator (Figure 1) has been designed
to fulfil all these requirements.
In a previous paper (1) the basic separation
mechanisms in the Wringing apparatus have been
highlighted by the use of the CFD. In particular the
combined effects of curved ducts (creating Taylor's
eddies) and boundary layer (where separation takes
place) has been proved to be crucial for the separation.
Starting from this point, the effects of geometrical
modifications of the apparatus has been studied and
presented in this paper both in terms of the fluid
dynamics behaviour and separation efficiency.

2 APPARATUS DESCRIPTION
The Wringing separator is a vertical separator divided
in two main parts: the upper part (Micronsep
proprietary), where the separation process is
performed, consisting of a cylindrical spiral made of
sheet metal, a holed disc and a hopper; and the lower Figure 1: the Wringing separator.
part acting as a reservoir.
apparatus). A part of the main flow entering the spiral
The mixture enters in the separator from the inlet (see
follows the spiral itself and exits directly from the gas
Figure 1) and is subdivided into two flows through a
outlet while a set of little secondary flows exits
deflector (not shown in the figure); the main flow (about
throughout the holes of the initial rounds of the spiral
75-80%) goes into the spiral while the other descends
into the hopper. Also the descending flow coming from
into the bottom reservoir through the small gap
the inlet comes up through the stump into the hopper.
between the skirt and the case (the outer wall of the
These two flows re-enters into the spiral throughout the
holes of the last rounds and exit from the gas outlet as simulations of the whole separator to evaluate the
well. boundary conditions for the round.
The particles follow the gas flows and are separated All the calculations were performed with a commercial
initially into the spiral and than definitively into the code (Fluent 6.1). For the simulations of the whole
hopper from where they go down, throughout the separator, we used the RNG k-ε model (4) for the
stump, into the bottom reservoir. How much gas turbulence description with the “standard wall
follows each flow and, among the first flow, how much functions” for the wall treatment. For the study of the
gas goes into the hopper throughout the holes, has a gas flow rates in the apparatus, the use of the wall
very strong impact on the separator performances and functions is adequate, as discussed in a previous
is related to the different pressure drops associated to paper (1). For each geometrical configuration, the
the different flow paths. Pressure drops are than number of cells required to simulate the whole
related to the geometry (e.g. the spiral dimensions and apparatus was about 1.5x106 and the cell dimension
number of rounds or the type of deflector adopted) and near the walls was about 1 mm.
the physical conditions (mainly density) in the To evaluate correctly the separator efficiency, a fine
separator. description of the boundary layer is needed, but this
We focused our attention on a particular configuration, means tens of millions of cells for the whole separator..
with a fixed geometry of the apparatus and a specific The CFD analysis was therefore concentrated on the
mixture: air at 4.7 bar mixed with toner particulate (1-5 first round of the spiral, with the very high cell
μm diameter range, about 0.03% in weight of the gas refinement near the walls (with a cell adjacent the wall
and about 2000 kg/m3 of density). of 20 μm) necessary to correctly simulate, by a
Due to the their small size (1-5 μm), the particles follow numerical point of view, the boundary layer. To
the gas flow that undergoes to a progressively higher evaluate the separation efficiency, an Eulerian (gas) /
centrifugal force (the spiral has a growing bend radius), Lagrangian (particles) approach was used. In the
creating a set of eddies perpendicular to the flow Lagrangian approach the motion equation of the single
direction (2, 3), the Taylor's eddies. The particles move particle is solved (4). The Eulerian / Lagrangian
towards and concentrate at the external wall of the approach allows to calculate separately the gas motion
spiral because of the centrifugal force and descend first and the particles motion after. Besides the
along the wall because of their weight (Figure 2). Walls calculation can be performed as a stationary one
presence is essential to the separation, because it is instead of a transient one (much heavier for the CPU).
right into the walls boundary layer that the separation is It's also very easy to evaluate the effect of particles
actually performed due to the very low gas velocities. diameter on their behavior, introducing different particle
groups with different diameter. On the other hand this
approach has some shortcomings: the particle is
considered of zero dimension so that gas and solid
coexist in the same volume; the particle-particle
interaction is not considered. These limitations are not
severe for the conditions considered because the
amount of particulate is well below 1% in volume.
For both configurations we first checked both coupled
and uncoupled calculations for the Eulerian /
Lagrangian approach, but the differences were
negligible, allowing us to use the uncoupled simulation,
reducing considerably the CPU time to obtain a
solution.

3.1 Deflectors
A better flux distribution in the separator could allow a
more compact and/or more efficient apparatus to be
realized. We studied the following configurations:
Figure 2: the Taylor's eddies.
1. actual configuration (plane deflector, Figure 3.1);
2. actual configuration with a cylindrical deflector
3 CFD SIMULATIONS coherent with the coiling of the spiral (counter
We have studied the effect of changing of the deflector clockwise, Figure 3.2);
or the adding of one or two partition walls on the gas 3. actual configuration with a cylindrical deflector
fluxes and the separation efficiency of the first round of opposite to the coiling of the spiral (clockwise,
the spiral. Figure 3.3).
Two different set of simulations were performed, the The configurations with the different deflectors are
first on the whole separator, the second focused on the almost undistinguishable from the point of view of the
first round of the spiral, using the results of the general flow rates: the flow rates calculated into the
Stump

(1) (2) (3)


Spiral inlet
Figure 3: Geometrical details of (1) the actual
configuration; (2) a cylindrical deflector coherent with
the coiling of the spiral (counter clockwise); (3) a Reservoir
cylindrical deflector opposite to the coiling of the spiral
(clockwise). (a)
hopper through the stump and through all the holes of
the spiral are respectively about 23% and 29-30% of
the inlet flow rate. Nonetheless the three deflectors
create quite different fluxes into the separator. In fact Hopper
the device equipped with the cylindrical deflector
coherent with the spiral has the lowest total pressure
drops, followed by that with the plane deflector and
eventually by the configuration with the cylindrical
deflector opposite to the coiling of the spiral. An
experimental work carried out with the three different
configurations in a scaled separator confirm this trend.
In the actual configuration the gas enters
perpendicularly into the device (see Figure 1). A part of
the entering flow rate (about 60-70%) is re-directed in
the same direction of rotation of the spiral (counter-
clockwise) by a plane deflector, inclined by 45° respect Stump
to the initial flux direction (see Figure 3.1). The main
portion of this deflected flux climbs rotating (counter-
clockwise) and enters into the spiral, while the (b)
remaining portion descends throughout the gap
between the skirt and the external wall into the Figure 4: Stump, actual configuration. Gas velocity
reservoir. The part of the entering flux which is not vectors: the intensity is proportional to the module of
deviated by the deflector passes up, down and on the the velocity, while the colour is a function of the
other side of the deflector, originating a flux rotating velocity component along the “X” axis [m/s].
clockwise (opposite to the coiling of the spiral), the
main portion of which descends into the reservoir pressure drops. If the cylindrical deflector directs the
through the gap, while the remaining portion climbs flux in the opposite direction, all the flux rotates
towards the spiral inlet. This flux “collides” with the clockwise. The flux ascending is forced at the spiral
main flux (rotating counter-clockwise) and causes a big inlet to change the direction of rotation (high pressure
turbulent zone right at the spiral inlet. The gas present drops), while the descending flux involves both the
into the spiral comes out into the hopper through the reservoir and the stump in its clockwise rotation.
holes of the initial rounds of the spiral and, given the Table 1. Particles/gas exited from the holes (% in
counter-clockwise coiling of the spiral, it rotates weight) of the 1st round of the spiral.
counter-clockwise. The gas descends towards the Type of deflector gas 1 μm 5 μm 10 μm
stump along the wall of the hopper and climbs (again
rotating counter-clockwise) at the center, re-entering (a) planar 14 14 18 23
the spiral through the holes of the last rounds. (b) coherent (c.clockwise) 14 13 15 26
On the other hand, the gas present into the reservoir (c) opposite (clockwise) 14 14 17 31
rotates clockwise, as in the stump (see Figure 4a, flux
into the stump), through which the gas climbs into the
hopper. It is right here, at the end of the stump, that the In table 1 are shown the separation efficiency for the
direction of rotation changes (see Figure 4b, flux out of three deflector configurations. In this contest, for
the stump), conforming to the main rotation in the separation efficiency we intend the whole mass of
hopper (and in the spiral, counter-clockwise). particles separated through the holes respect to the
total mass of particles entering the first round of the
When a cylindrical deflector coherent with the spiral is spiral. The 1 μm particles are easily dragged by the
used, the flux is totally directed counter-clockwise so fluid, therefore the amount of them separated
that both the ascending and descending portion of the throughout the holes corresponds to the gas entering
gas rotate counter-clockwise, thus allowing the lowest
into the hopper from the 1st round of the spiral, around the 10 μm particles. This effect is not physical but
14%. The 5 and 10 μm particles have progressively numerical. In the case of the single or double duct, the
higher inertia, tend to accumulate into the wall grid is very fine near the wall (and very similar for both
boundary layer and are more effectively separated. geometry), while in the case of the three ducts, due to
Particles separation is maximum for the configuration the computer memory constriction, some compromises
in which the cylindrical deflector is opposite to the had to be made. The Lagrangian approach used for
coiling of the spiral; at the contrary it is minimum when tracking the particles is, unfortunately, quite sensible to
the cylindrical deflector is coherent to the coiling of the the grid. The gas, on the contrary, is not influenced by
spiral. The actual deflector (planar) is in some way in the kind of refinement adopted near the wall as
between, more similar to the opposite cylindrical confirmed by simulations performed with much coarser
deflector. In fact the flow distribution into the meshes (with the standard wall functions). As the 1 μm
spiral/hopper is rather similar for the two cases. Semi- particles follow the gas, their behaviour is very little
quantitative evaluations of the separation efficiency influenced by the grid refinement adopted near the
with the three deflectors on a scaled Wringing wall.
separator gave similar trends. Table 2. Particles/gas exited from the holes (% in
weight) of the 1st round of the spiral.
3.2 Partition walls Simulation cells gas 1 μm 10 μm
The spiral is the key element of the separator, as it is 1 duct 1076284 13 14 55
in the boundary layer of its external wall (due to the 2 duct 1911968 25 25 76
centrifugal force) where the particles are separated.
Keeping fixed the internal section of the spiral (thus 3 duct 2175024 36 36 48
maintaining the same gas flowrate and velocity), a
higher extension of this surface allows a greater
amount of particles to be trapped in it. For this reason 4 CONCLUSIONS
the coils are tall and narrow. Nevertheless a coil too tall The CFD has been used to optimize the actual
and narrow means a reduction of the Taylor eddies geometrical configuration, investigating the effects of
which re-supply the wall of particles. The ratio between different geometrical variations such as changing the
the height and the width of the spiral is a balance deflector type and orientation or putting partition walls
between two contrasting needs, to trap and separate in order to increase the separation surface.
the particles in the boundary layer and to supply it with The three deflectors investigated create quite different
the particles coming from the main flux. Besides, a fluxes into the separator and consequentially different
higher spiral (with the same section) causes an total pressure drop (as confirmed by experimental
increase in the particle residence time in the boundary work). Also separation is affected by the deflector
layer (the particles have to be disposed of throughout configuration and the simulation have shown that the
the holes at the base of the spiral) thus increasing the clockwise configuration has a slight better separation
time needed to separate the particles, meaning that than the counter clockwise.
the particles have to travel a greater portion of the
A modification which greatly affect the separator both
spiral before to be disposed of.
in terms of gas flow rate into the hopper and separator
The insertion of a vertical partition wall in the spiral, efficiency into the spiral, is the adding of one or more
subdivides the duct into two or more sub-ducts, partition walls. Adding a partition wall increases the
increasing the useful surface (actually doubling it in the particles separation, and a second partition wall, i.e.
case of a single insertion wall), avoiding a spiral too tall creating three ducts, increases even more the
and thus avoiding an increase of the particulate separation efficiency.
residence time.
Unfortunately, with the computer available, it was not
possible to simulate the boundary layer for more than REFERENCES:
one duct at full height. So we simulated the first round
of the spiral at half height. The results are shown on [1] A. Bennardo, F. Podenzani, L. Ciccarelli, U.
Tab.2. Spinelli: CFD Simulation of a Gas/Solid Wringing
Separator, Proceedings of Partec 2004.
Adding a partition wall increases the particles
separation. As before, the 1 μm particles are easily [2] S. V. Patankar, V. S. Pratap et al.: “Prediction of
dragged by the gas and come out of the holes in more turbulent flow in curved pipes”, Journal Fluid
or less the same quantity than the gas; while the 10 Mechanics, Vol.67, part 3 (1975), pp. 583-595.
μm particles have a much higher inertia and are more [3] P. A. Mees, K. Nandakumar et al.: “Secondary
easily accumulated on the external walls of each duct instability of flow in a curved duct of square cross-
and then disposed of through the holes. Adding a section”, Journal of Fluid Mechanics, Vol.323 (1996),
partition wall has a dramatic effect on the separation pp.387-409.
efficiency. Adding a second partition wall, i.e. creating [4] Fluent User's Guide, Vol 2 (1998).
three ducts, increases even more the separation
efficiency for the 1 μm particles, while reducing that for

S-ar putea să vă placă și