Proceedings World Geothermal Congress 2005

Antalya, Turkey, 24-29 April 2005

Design Concept for a More Efficient Steam-Water Separator

K.C. Foong
21 Merriefield Avenue, Forrest Hill, North Shore City, New Zealand
[email protected]

Keywords: Separator surface to separate two fluids of different densities (brine

and steam) and then use gravity to drop out the heavier
ABSTRACT brine from the steam. This led to the design of the U bend
separator (Figure 2) in 1951, where the two phase fluid
The Webre steam-water separator design is widely used in
(mixture of brine and steam) is made to go around a U bend
New Zealand, Philippines, Indonesia, Kenya and elsewhere
in a vertical loop to concentrate the brine on the outside
in the world. It has been very successful with a separation
wall. As the two separated fluids come down from the loop,
efficiency claimed to be as high as 99.97%. While this
gravity accelerates the brine downwards. The lighter phase
seems adequate, 0.03 % brine carry over equates to about
steam is then extracted from the inside wall.
0.24 TPH for a 100MW plant. The paper looks at designs
and the operation of new and old separators and proposes a
new equation to calculate separator efficiency. It then
presents a conceptual design to improve the separation
efficiency by trapping creep water. The design also reduces
the pressure drop across the separator and improves access
for internal vessel inspection.

1. INTRODUCTION
The current Webre type separator design dates back to the
1950’s. It was first used in New Zealand at the Wairakei
Geothermal Power Station. Since then, it has been used in
many parts of the world to separate steam from brine
(geothermal water) in geothermal power projects. Except
for improvements in inlet design and internal construction,
its design has changed very little.

2. SEPARATOR DEVELOPMENT
Water and steam can be separated by flowing the mixture
into a large drum. The lighter steam will rise up while the
heavier water will fall to the bottom of the drum. Indeed
such a knock out drum (Figure1) is still in use today and it
is still being built for new geothermal projects.

Figure 2 – U Bend Separator

The separation efficiency of the U bend was reported to be
only about 80%. A centrifugal separator was later added
downstream to improve separation efficiency. Further
experimentation showed that the centrifugal separator could
handle wet fluid without the U bend and the latter was
discarded.

One of the first centrifugal separators used at Wairakei has

steam discharge from the top of a vertical vessel and brine
discharge from the bottom of the vessel (Figure 3).

The diagram in Bangma’s paper showed internal baffles at

the top, middle and bottom of the separator. It is speculated
that the middle baffle forces the incoming fluid against the
wall and spin it at high velocity to separate steam from
brine. The top baffle is probably used to prolong the
spinning action and to prevent up flowing steam from
moving to the centre of the separator early. The bottom
baffle is probably an anti vortex plate for the brine.
Figure 1 – Knock Out Drum
Later, the separator design was changed to a Webre type
Faced with the problem of separating brine from steam with the steam pipe running inside the vessel and coming
more efficiently, the early designers would have thought of out at the bottom. This concept is still the most widely used
using the centrifugal effects of fluid going around a curve
1
Foong

separation method employed in New Zealand geothermal probably obstruct and interrupt flow, causing brine to
projects and many parts of the world. splash inside the separating chamber (Figure 5).

Figure 3 – An Early Separator Design

3. THE WEBRE TYPE SEPARATOR Figure 5 – Separator with Guide Vanes

In 1961, Bangma ran a series of experiments with separator The second one has the two-phase inlet at the top, the steam
configuration and came to the conclusion that a separator tube in the middle and the brine pipe at the bottom. (Figure
with a spiral inlet is more efficient than a tangential inlet. 6). The baffle plate at the top of the steam tube presumably
By progressively increasing the inlet velocity of the “cuts” the steam from the brine, which continues
separator, he also demonstrated that the separation downwards to the bottom of the vessel. Such a
efficiency increases until a breakdown velocity (Figure 4) is configuration would assume that the steam/brine ratio is
reached. Above this velocity, the efficiency deteriorates fixed for the life of the plant and the size of the baffle plate
rapidly. He noted that the spiral inlet separator has a higher sized accordingly
inlet breakdown velocity than the tangential one (about
60% higher). His spiral inlet separator achieves the highest
efficiency when the steam inlet velocity is between 30 and
40 m/s. The breakdown velocity is approximately 45 m/s.
Current separator design is based on this 1961 model.

Figure 4 – Break Down Velocity

There are some variations to the Webre separator design.
Following are three such variations. Figure 6 – Separator with Baffle Plate
The first one has plates at the two-phase inlet to presumably The third one has a spiral inlet with a constant curvature
direct flow downwards at a fixed pitch for the spiraling and a small downward angle on the inlet to encourage brine
fluid. This does not happen in practice and these plates will to spiral downwards (Figure 7). Once inside the vessel, the

2
 Foong

fluid is left to find its own equilibrium unobstructed. It is likely to cause turbulence leading to entrainment of fine
very similar to the one tested by Bangma in 1961. water droplets.

In later design, a pipe reducer was installed at the steam

inlet. This makes the inlet more rigid and it also acts like a
cheap “flared” inlet. This innovation worked and the
pressure drop across the separator shows an improvement.

Figure 7 – Separator with Spiral Inlet

The design familiar to the author has a rectangular spiral
inlet with a constant change in curvature over a 90 degrees
flow path. The scroll like inlet has the fluid subjected to an
increasing centrifugal force and “lays” the water film on the
vessel wall as it enters to minimize the “splash” that high
velocity brine makes when it comes in contact with a curve
surface. It also has a sloped bottom that directs the brine in
a downwards spiral. Very high efficiency is achieved with Figure 8 – Whirling Steam Inlet
this design. Mechanical separation efficiency is probably as
good as it can be. Further improvements may be possible by
optimizing the aspect ratio of the inlet, slope and inlet
velocity. But these probably require extensive laboratory
testing and would vary with fluid conditions. But fluid
conditions vary with the life of a geothermal resource so it
is impossible to define what these are.

4. CURRENT SEPARATOR DESIGN EXPERIENCE

The sharp inlet coupled with 180 degrees reversal of flow
and very high angular velocity contributed much to the
pressure drop across the separator. This varies from 0.3 to
0.5 bars depending on the design.

One separator inspected showed failure in the steam tube at

two areas. The point where the steam tube is welded to a
baffle plate inside the separator and the area around the
inlet to the steam tube. Both were thought to have been
caused by the forces generated by spiraling fluid.

The fatigue cracks on the welds attaching the tube to the

vessel baffle plate suggested that the swirling flow of steam
had caused the steam tube to behave like a whirling shaft
(Figure 8). It is thought that similar forces also cause the
inlet to “ovalate” (Figure 9), going from round to oval Figure 9 – “Ovalation”
shape and vice versa. This is inspite of the inlet being
reinforced with a steel collar. 5. SEPARATOR EFFICIENCY
In 1984, Lazalde-Crabtree presented a paper on the Design
To overcome the problem, the inlet was reinforced to make Approach of Steam Water Separators. The paper stated that
it more rigid and the tube was supported by struts welded to Separator efficiency is measured by the amount of brine
the vessel wall. The latter obstructed steam flow. The carry over into the steam. This definition has been
effects of this obstruction cannot be quantified but it is “universally accepted” as the way to measure separator
efficiency. Lazalde-Crabtree also proposed that the
3
Foong

efficiency of separator is a product of mechanical and The following picture (Figure 10) shows blocked holes of a
annular efficiency. diffuser connected to the steam line downstream of a Webre
separator with efficiency of the high 99%! Turbines with

η η η
steam scrubbers and steam wash facilities are reported to
ef = m x A also have silica deposition problems.

η m = Centrifugal efficiency

η A = Annular efficiency

The former is function of centrifugal forces and the latter

deals with entrainment of water droplets in the annular
space between the outer wall and the inner steam tube. It is
related to the terminal velocity of the water droplets and the
vertical steam velocity inside this annular space. “Settling
efficiency” is probably a better term to use than annular
Figure 10 – Blocked Diffuser Holes
efficiency.
Therefore, whatever efficiency can be gained, no matter
I believe it is more appropriate to express the efficiency, how small, will improve the steam quality leaving the
using the same terminology as follows:- separator and will cause fewer problems in the steam
system and the turbines in the power plant.

η =ηef m+ η )η
x(1- m A 6. FLUID BEHAVIOUR INSIDE THE SEPARATOR
Two fluid behaviors are considered in this section.
+ y(1-η )η m B
6.1 Wall Creep

+ z(1-η )η
The steam moving up to the inlet of the steam tube picks up
m C +… tiny water droplets. As it spirals upwards, the centrifugal
force moves these water droplets radially from the center of
where (x + y + z +..) =1 and holds true only if “parallel” the vessel to the outside wall. If the straight section is tall
separation mechanisms occur after the initial separation. enough, there maybe enough time for the droplets to reach
the vessel wall and coalesce with others to form a water
In a liquid dominated system, most of the brine is separated film. If this film is too thick, water will peel off the film and
by centrifugal forces when the fluid enters the separator. drops back into the vessel to rejoin the rest of the brine
This could be as high as 95%. The balance of the brine, flowing out of the vessel. However a small thin film of
small amount, is separated in a different part of the vessel water will continue to flow in the general direction of the
by different mechanism(s). The second term represents the steam and moves up the vessel (Figure 11). When it reaches
efficiency contributed by the annular or settling efficiency the top, the roof of the vessel, it will coalesce and fall from
described in Lezalde’s equation. If this is 100% and x = 1, the underside of the roof into the steam outlet pipe.
the separator efficiency is 100%.
A similar film will cling to the outside wall of the steam
If only the first two terms are considered and x=1, high tube, move up and flow into the steam tube.
efficiency can be achieved in a knock out drum when
mechanical efficiency is low but settling separation This mechanism where water creeps up a wall in a spinning
efficiency is high. However, a very large knock out drum is fluid is quite well known especially in industries where
required. centrifuges are used.

The third and subsequent terms are left in the equation to Pollak showed in his experiments with cyclones that the
allow for other mechanisms of separation after initial water in the outlet steam is mainly due to entrainment and
separation. This equation needs further investigation as to the creep of water film up the cyclone walls. Ter Linden
sequential separation mechanisms require it to be written tried to overcome the entrainment problem due to wall
quite differently. creep by introducing a skirt, a concentric annular ring, at
the top of the vessel to trap the water (Figure 12). This
The separation efficiency achieved by the Bangma type probably did not improve the situation, as the water would
Webre separator is in the order of 99.9% and would be still drop into the steam flowing out of the cyclone
quite sufficient for most processes. However in geothermal separator.
power application, the cumulative effects of small brine
carry over with associated dissolve chemicals can be quite The water really needs to be trapped and led away from the
large. For example in a 100 MW plant, the steam flow main steam flow.
required is in the order of 800 TPH. Even with separator
operating at 99.97% efficiency, the amount of brine carried
over into the steam line is 0.24 TPH. A 0.01%
improvement will reduce brine carry over by 0.08 TPH.

4
 Foong

moved out from the center of the Webre separator only to

be to be recycled again.

Figure 13 – Conservation of Momentum

7. PROPOSED MODIFICATIONS
With separator efficiency of 99.7% and some claimed
99.9%, it would be very hard to improve on it. However, it
is felt that the wall creep mechanism described above is not
adequately addressed by the current Webre design and it is
believed that significant improvement can be made to
minimize the problem.
Figure 11 – Wall Creep
A modification can be made to the Webre Separator by
installing a plate under the roof of the separator and
draining the collected fluid down the center of the steam
tube to the main brine stream (Figure 14). The inverted
“chinaman” hat could also improve the dynamics in the
inlet and could possibly improve the pressure drop.

Figure 12 – Annular Ring to Trap Water

6.2 Conservation of Momentum

The tangential velocity of a spinning fluid near the wall of
the vessel is the product of radius and angular velocity (V =
ωr ). By the conservation of momentum, ω increases as the
steam moves from the wall to the center where the steam
outlet is located. Therefore the fluid spins faster near the
steam outlet. The centrifugal force being a function of the
radius and the square of the angular velocity (rω2) is large
near the center of the steam outlet (Figure 13). A drop of
water falling into the center of this spinning fluid will be Figure 14 – Modification with Drip Tray
shattered. Some fluid particles are flung radially outwards However a more radical change is proposed.
to the wall. One can imagine the many fluid particles being

5
Foong

Back in 1961, Bangma gave two reasons for choosing the 7.2 Pressure Loss
bottom outlet cyclone (BOC) separator over the top outlet The current Webre design has a sharp entry to the steam
cyclone (TOC) separator tube. The flow also does a very sharp 180 degrees turn
before traveling down a straight steam tube to connect to a
1. No interior baffles and fittings steam main.
2. Steam removal from the bottom make simpler Looking at just the entry to the point where steam starts
pipe work. traveling down a straight tube, it can be said that the pipe
system in this short section (Figure 16) consist of:
The first reason was given probably because the original
TOC separator was a complicated design and had a lot of
• 1 x sharp inlet
internal baffles. The second reason is debatable.
• 2 x miter bends
Changes proposed unfortunately take the conceptual design
of the separator back to a top outlet unit.

7.1 Trapping Water

Central to the conceptual separator is a bell mouth inlet at
the top of the separator (Figure 15). It is hung from the roof
of the separator but not connected to the steam outlet tube
nor the separator wall except for a tube connection. The
position of the bell mouth inlet “close off” a space between
it and the roof of the vessel, refer here as the trap space. It
also forms two traps to remove water creeping up walls.

Figure 16 – Equivalent Inlet Loss of Webre Separator

Similarly, the piping system in the conceptual design
(Figure 17) can be said to consists of

• 1 x bell mouth entry

• 2 x LR bends

Figure 15 – Bell Mouth Inlet

As water creeps up the vertical section of the separator
wall, it enters the trap space by its upward momentum.
Inside the space, it follows the curve wall of the semi-
ellipsoidal head. It soon loose its momentum and drop from
the roof into a tray formed by the bell mouth.

Water that escapes this first trap may continue to creep up

the face of the bell mouth against an increasing centrifugal
force exerted by the spinning steam. If it does get into the
mouth, centrifugal force will hold it to the side of the wall
as it climb into a second trap formed by the smaller steam
outlet and the bell mouth. This type of trap is commonly
used in centrifuge to decant one product from another

Lowering the pressure slightly in the trap space may assist

in trapping the water. It is quite possible, (experimentation Figure 17 – Equivalent Inlet Loss of New Separator
is required) to create a slightly lower pressure in the trap One can immediately see that there is an improvement in
space by ducting the trap space into the “vena contracta” of pressure loss even without doing any calculation.
the expanding fluid in the separator inlet nozzle. Cooling
the steam in this space has similar effect. Both need to be For an 18” to 20” steam outlet pipe, the total resistant
investigated. coefficient (K factor in Crane) for the two configurations
are 1.22 and 0.4 approximately or a ratio of 3:1.
The trapped brine is simply piped away to the main brine
system. If the pressure drop in the steam system can be lowered, the
life of a geothermal well can be extended as separation can
take place at a lower pressure. A typical well characteristic

6
 Foong

shows increase mass flow at lower pressure and the steam 8. CONCLUSIONS
fraction for a given fluid enthalpy increases as pressure The problem with water creep was re-cognized by Pollak
reduces. The result is more steam. and Ter Linden but not addressed in current separator
design. The efficiency equation proposed by Lazalde-
7.3. Settling Efficiency Crabtree certainly did not recognize it
With the center tube removed the cross sectional area for a
given diameter separator available for upward flow is The conceptual separator recognized this problem and a
increased. This leads to lower upward velocity and would third term is required in the proposed efficiency equation to
improve the settling efficiency. account for this. In addressing the problem improvements
were made in other areas.
One can of course reduce separator size to maintain the
same upward velocity with theoretically no change in In summary, the anticipated improvements with this new
efficiency. design are as follows:

• Trap and reduce re-entrainment of creep water

7.4. Accessibility
from the vessel wall
Inside the Webre separator, the internal steam pipe is
located in the middle of the cylindrical shell of the separator • Improve settling efficiency marginally by
leaving a very small annular space between them. Part of decreasing up draft velocity
the maintenance requirement is to inspect the internals of
the separator for weld cracks, wear and debris accumulation • Reduce pressure loss in the separator.

An operator using the manhole to access the Webre • Significantly improve access to the inside of the
separator is immediately confronted with the steam tube separator
directly in front. Generally, peering into the separator is all
that can be done to inspect the separator. In a larger REFERENCES
diameter separator, the operator might be able to squeeze
P. Bangma 1961: The development and performance of a
into the annular space between the vessel wall and steam
tube. steam-water separator for use on geothermal bores.
Hugo Lazalde-Crabtree 1984: Design approach of steam-
With the steam pipe removed from the center of the water separators and steam dryers for geothermal
separator, it opens up the space inside the separator and applications. Geothermal Resource Council Bulletin
inspection is made much easier. It also reduces the available September 1984
wall area for water creep
R. McKibbin 1998: Fluid flow in flashing cyclone separator
Design approach of steam-water separators and steam
dryers for geothermal applications.
Pollak A. and Work L.T: The Separation of Liquid From
Vapor, Using Cyclones., ASME 1942
Ter Linden, A.J.: Cyclones as Drop Separators Chemie-
Ingenieur-Technik Vol 25 (1953) p. 328
Crane Co: Flow of Fluids Through Valves, Fittings and
Pipe

Figure 18 – The Concept for a New Separator