Professional Documents
Culture Documents
net/publication/290057720
CITATIONS READS
16 2,726
3 authors, including:
Some of the authors of this publication are also working on these related projects:
The 13th Conference on Sustainable Development of Energy, Water and Environment Systems – SDEWES2018 View project
All content following this page was uploaded by Elisio FILHO Caetano Filho on 30 May 2016.
Elisio F. Caetano
ABSTRACT
1. INTRODUCTION
The twin-screw multiphase pump has been studied as an
alternative system to substitute the conventional one (fluid The twin-screw multiphase pump is one of several
separation, liquid pumping and gas compression) in petroleum alternatives being considered to reduce production costs in
boosting. By “pumping” simultaneously gas and liquid, the offshore petroleum exploitation. By simultaneously “pumping”
multiphase pump could reduce production costs in deepwater gas and liquid, this pump could reduce capital and operational
activities. This paper presents a thermodynamic model of a twin- expenditures, particularly in deep-water petroleum production,
screw multiphase pump to determine performance parameters among other advantages (see CAETANO et al. (1997)).
such as: absorbed power, discharge conditions and efficiency. Although many prototypes have being successfully tested,
To overcome problems with the complex flow field inside of this there are just some few works concerning the twin-screw
novel equipment, the multiphase flow was divided into a multiphase pump modeling. This lack of information became
sequence of simpler processes. Such approach helps determine clear when PETROBRAS decided to make a comparative
energy and mass balances and enables the use of a process analysis between a novel (SBMS-500) subsea multiphase
simulator (Hysys.Process v2.1) to construct the model. The pumping system (see KUJAWSKI and CAETANO (1999)),
model prediction when compared to the literature show that the equipped with a twin-screw multiphase pump, and a
assumption of a smooth turbulent flow, considering the pressure conventional one (fluid separation, liquid pumping and gas
loss in the entrance and discharge of the gap, fits better the compression) . A model allowing predictions of the absorbed
phenomena than the turbulent flow when calculating the flow power, backflow rate and discharge conditions could support not
through the gaps. In addition, the comparison for absorbed only comparative analysis but also the pump design and studies
power indicates that the assumption of gaps filled only with about thermodynamic behavior of a system operating with such
liquid is not valid under all operation conditions. pump.
1 Copyright © 2002 by ASME
The main objective of that project was then to develop a pump, the rotation and meshing of the screws create one or more
model that would be able to predict the thermodynamic behavior chambers which moves continuously from the suction to the
of a twin-screw multiphase pump. On the basis of suction discharge region (Fig.1). Therefore, any type of product (liquid,
conditions, the model would determine, among other aspects, the gas or solid) that can be introduced into the cavities will be
absorbed power, the pump discharge conditions and the pressure carried along to the discharge. Such characteristic allows the
profile along the screws. pump to transport multiphase fluids with GVF varying from 0 to
95% (100% if 5% of liquid can be recirculated).
2. TWIN-SCREW MULTIPHASE PUMP
The twin-screw pump has a series of clearances (peripheral, One working model was proposed by CHRISTIANSEN (in
radial and flank clearance) between the casing and the screws. WINCEK (1992)). In this model, the liquid is considered to
These clearances together with the screw rotation and pressure being impelled by centrifugal forces so it is kept in the periphery
difference allow some fluid to flow back. Therefore, there must of the screw. With this characteristic, just liquid would fill pump
be some liquid to fill these clearances , otherwise the pump clearances and, thus, only liquid would flow backwards through
would lose pumping capacity. However, as a result of the high these clearances. Such recirculation will compress the gas
centrifugal forces in the pump, a small quantity of liquid is contained in a particular chamber. Once there is a reduction in
enough to keep the pump capacity. Due to the backflow, the real gas volume, its original space must be filled with liquid. So, as
volumetric flow is always lower than the pump theoretical one chamber displaces towards the discharge, its pressure
capacity. increases and there is liquid accumulation. This accumulation is
possible only if the backflow rate, entering the chamber, is
chambers higher than its outflow. So, the pressure difference must be
higher near the discharge region.
The pressure profile along the screws depends on the GVF
of the pumped fluid. For GVF=0, the fluid is incompressible and
suction discharge the pressure profile is linear. However, with compressible fluid,
as the multiphase one, the profile is widely flat and with a
greater pressure gradient near the discharge region.
Further details about pumping principles and pump
casing characteristics can be found in the literature (COOPER et al.
peripheral
(1996), e.g.).
clearance
radial
clearence flank
clearence
60 95% de gás
líquido the screw peripheral region are admitted to be concentric.
40
Effects of shaft deflection are not taken into account.
20
Where the indicated value 1.5 at the right hand part of the And thus describes the radial clearance geometry between
equation corresponds to the local losses in entrance and internal and external diameters of the twin screws.
discharge regions of a clearance. The coefficient depends on In the radial clearance the coefficient for local losses is not
the occurring flow regime. necessary due to the absence of corners.
WINCEK (1992) developed equations for the back-flow
through each clearance – including their particular geometry – c) Flank clearance:
on a twin-screw multiphase pump. Such clearances were
considered completely filled with liquid and under rough Velocity v under laminar and smooth turbulent flow, with
turbulent flow. Thus, the volumetric flow through the clearances local losses, is given respectively by:
would be driven by two factors: (1) pressure difference between 2
24l F l
two chambers; and, (2) screw rotation (referred, respectively, as 12 l p 24l F l
differential and rotational components of back-flow). Local s 2 2
F sF
losses were not contemplated in this model. Thus, in the present v (3.6)
model, the referred equations for backflow were modified in 3 l
order to take into account the local losses. In addition, the 0.25
occurrence of smooth turbulent flow was assumed, as had been l 2p
0.1397 l F v 1.75 1,5v 2 0 (3.7)
proposed by EGASHIRA et al. (1996). A comparison between .s 5 l
these two approaches will be presented. Equations for laminar l F
and rough turbulent flow, without local losses, and for the The total backflow of the pump is given by adding all of its
rotational component are the same as the ones proposed by components occurring on the previously referred clearance
WINCEK (1992) and will not be presented here. On the areas.
contrary, modified equations will be shortly described in the With the expressions for back-flow rate, which is a
following items. For further details, one should refer to function of shaft speed and pressure difference between two
WINCEK (1992) or to NAKASHIMA (2000). cavities, it is possible to determine the multiphase pump pressure
profile. Initially, the model starts with any pressure profile. After
a) Peripheral clearance: a compression stage takes place, there is a reduction in the
volumetric flow due to gas compression. If the back-flow rate
Equations for average velocity under laminar and smooth that comes from the downstream stage, added to the volumetric
turbulent flow, with local losses, are given respectively by: flow that comes from the upstream stage, equals the theoretical
volumetric flow rate, then the pressure of the stage is kept
constant. However, if the resulting volumetric flow rate is higher
2
24 B l
12 l .p 24 B l
than the theoretical one, the stage pressure must be raised in
s 2 2 order to reduce the gas volumetric flow and the back-flow rates.
P sP
w (3.2) On the contrary, if the flow rate is lower than the theoretical one,
3 l the pressure must be reduced. This correction is performed
successively until the volumetric flow rate along the model is
constant and equals the pump theoretical one.
2 2
u w w
2
b)
( x , y ) (3.10)
y x y
GFV (%) W (kW) W / W th
0 283.1 1.0000
The velocity profile (u and w in equation 3.10) in the
30 282.1 0.9964
screw, which is necessary to calculate friction power, is given by
60 275.9 0.9746
CAMPBELL et al. (1996). Once both parts are calculated, the
efficiency of pumps in each model’s stage will be stated by: 95 264.2 0.9332
V p i When a chamber is displaced from the suction to the
b
(3.11)
discharge, the gas volume is continuously reduced due to the
V p i W P W c
increase in pressure. The space previously occupied by gas is
4. RESULTS filled with the liquidflowing back and, therefore, for a given
reduction in the gas volume there is an increase of the same
In a first stage (items 4.1 and 4.2), an attempt was made to magnitude in the volume of liquid. The sum of the volumes of
check how the model, working with efficiencies (volumetric gas and liquid is constant since the volume of the chamber is
efficiency of pumps and compressors) of 100%, was capable of fixed. With the separation of the process into a sequence of
reproducing the behavior of an ideal pump. Some parameters simpler processes, there is a difference in relation to the actual
were also evaluated to see how they would affect the model. process and how it is represented in the model - as shown in Fig.
Initially, a fixed pressure profile was used for the model and the 4. The dashed line indicates the process involving a continuous
backflow rates were adjusted so that the volumetric flow rate increase in pressure and volumetric flow rate that actually
along the model would be kept constant (and equal to the one at occurs. In the model, however, the correction in the liquid
the suction). Then, modifications that would allow the model to volume and the heat exchange between the phases occur only at
determine the behavior of a real pump, including the backflow discrete points, and not continuously as in the actual process.
and the pressure profile, were included. Finally, the simulation This difference is directly reflected in the calculation of the
results were compared with experimental data (item 4.3). power as shown in Fig. 4. The process (1-3) involving the
simultaneous increase in pressure and liquid volumetric flow
4.1 GVF and of the number of stages rate is represented in the model by a simple pumping process (1-
2) followed by a correction in the volume of the liquid (2-3)
Tables 1a and 1b show the theoretical power consumption
through a backflow stream. Area 1-2-3 represents the power that
predicted by the model (W ) compared to the expected the model fails to take into account, which leads to a predicted
theoretical power smaller than the expected one. In the case of
theoretical power ( W th ), as a function of the GVF, for discharge compression (process 4-6), it is represented by a compression
pressures (pD) of 2,600 and 4,528 kPa, respectively. The (4-5) and a cooling (5-6) process due to the mixture with the
efficiency adopted values for the pumps and compressors were liquid. Thus, the compression power calculated by the model is
100%. The suction pressure (pS) was equal to 672 kPa. larger than the actual power, and the difference is represented by
area 4-5-6. By further observing Fig. 4, it is possible to note that
the area of region 1-2-3 is larger than that of region 4-5-6. In
5 Copyright © 2002 by ASME
other words, the higher compression power which is only partly Table 3. Effect of pressure profile in the simulations of the
calculated by the model compensates the reduction in the model (pS=672 and pD=2600 kPa): a) =100%; b) =50%
calculation of the pumping power. Thus, the resulting theoretical a)
pumping power calculated by the model is smaller than the
Pressure profile W (kW) TD (oC)
expected one. If the GVF is low, the slope of the curve is smaller
and as a result the cited areas (and so the deviation from the Increasing gradient 132.7 52.6
expected value) will also be smaller. Linear gradient 135.9 52.7
Decreasing gradient 133.4 52.4
V
b)
3
1 2
4 5 6 Pressure profile W (kW) TD (oC)
Increasing gradient 264.2 71.2
Linear gradient 271.3 71.9
2 Decreasing gradient 265.9 71.1
4 1 3
5
Backflow The pressure profile has little effect on the behavior of the
6
model (for power calculations) since, regardless of which
p
pressure distribution is used in the model, it will be representing
Figure 4- Differences between the actual process and the the same process, but different points thereof. If the model has a
model representation. linear pressure distribution, for instance, this does not mean that
it is representing a multiphase pump with a linear pressure
The results analyzed up to now indicate that the use of a profile, but that it is simulating the points of the pump in which
larger number of stages could reduce the difference between the the process has reached that pressure, as shown in Fig. 5.
predicted and the expected theoretical power. In fact, the results
obtained with the use of 16 stages are closer to the theoretical p p
power as shown in Table 2. However, as the deviations are lesser Linear pressure distribution Correpondent geometric points
and not expected to affect significantly the simulations, it is in the model in multiphase pump
more convenient to keep the model’s number of stages equal to pD
the pump’s number of stages. Then, it is possible to establish an
exact correspondence between one stage of the model and one
stage of the pump. The deviation in theoretical power can be
avoided by considering the theoretical power simply as a pS
product of total pressure by the theoretical volumetric flow.
Table 2. Predicted ( W ) and theoretical power ( W th ) -
comparison for calculation with 8 and 16 stages (pS=672 and
pD=2600 kPa; GVF=95%).
Figure 5- Significance of linear pressure distribution in the
No. of stages W (kW) W / W th model.
8 132.7 0.9371 Initially, it was found that an exact correspondence
16 138.4 0.9775 between the pressure profile of the model and that of the actual
pump was important for power and temperature calculations.
4.2 Pressure profile influence However, with the results presented herein, one can conclude
that deviations in relation to the actual pump are allowable for
Due to the lack of information regarding the pressure the calculation of the absorbed power. This does not mean that
profile, it was decided to check the impact that occasional flaws the calculation of pressure profile is not important. Just that
in the profile calculation of the model could cause in its when the model has calculated the pressure profile there is an
behavior. Studies were conducted for three different pressure exact correspondence between one stage of the model and one
profiles: with an increasing pressure gradient, with a linear stage of the pump, which is important for evaluation of losses
pressure gradient and with a decreasing pressure gradient. The (backflow and friction power).
results obtained, as shown in Tables 3a and 3b, indicate that 4.3 Results analysis
even large variations in the pressure profile have little effect on
the absorbed power and discharge temperature (TD) for different Figure 6 shows a comparison between calculated data and
efficiencies () in pumps and compressors. experimental data from EGASHIRA et al. (1996) for a pump
working with water only. Experimental points, whose tendency
is given by the continuous line, are shown in Fig. 6. The
6 Copyright © 2002 by ASME
1.0
calculated values obtained with the simulation were plotted over
Dimensionless pressure
Test 1
EGASHIRA’s et al (1996) figure. Points under the experimental 0.8 Mesured
curve are the values obtained using equations for smooth
difference [ - ]
Egashira
turbulent flow with local losses. Above the experimental curve 0.6 Model
are values calculated with equations for rough turbulent flow, 0.4
with a roughness 4,0x10-6 (m) and without local losses. Inside
the investigated region, the values obtained with equations of 0.2
Dimensionless pressure
1200 rpm Test 5
Backflow [m3/h]
difference [ - ]
Re 1780 rpm Model
flu 30 0.6
xo
0.4
de 20
Va 0.2
10
zã
0.0
o 0.0 0.2 0.4 0.6 0.8 1.0
0
0 490.3 980.7 1471.0 1961.3 2451.7 2942.0 3432.3
Dimensionless length [ - ]
Pressure
Diferença difference[kPa]
dePressão [kPa] (b)
Figure 6: Comparison between calculated and experimental Figure 7: Pressure profile – comparison against
backflow results. experimental data: a) 1200 rpm; b) 1800 rpm.
Figure 7 shows the comparison between calculated and Figure 8 shows the influence of GVF and shaft speed
measured pressure profiles for water-air mixtures (EGASHIRA on the pressure profile calculated by the model. As shown in the
et al. (1996)). There is a good agreement, as shown in Fig. 7a picture, lower GVF’s and higher shaft speeds contribute to
and 7b, although it is not possible to make a quantitative generate a flat parabolic pressure profile instead of the linear one
comparison since experimental results were presented in graphic obtained when the pump operates with liquid only. Some other
format only. Dimensionless length and pressure difference used simulations show that low suction pressure and pressure
in pictures are given respectively by: difference and high viscosity flatten the pressure profile. This
behavior is in agreement with experiments and theory found in
the literature.
i p pS
l dim ensionless ; p dim ensionless i
Table 4 shows values of backflow rate calculated by the
N BM 1 pD pS model, with the assumption of clearances completely filled with
liquid, and those obtained experimentally by EGASHIRA et al.
(4.1) and (4.2) (1996). Experimental values were calculated by subtracting real
volumetric flow from pump theoretical flow. As can be observed
The comparison using several profiles for the pressure in the table, despite the fact that the simulated values are
shows that the flatter the profile the higher the deviation following the same tendency dictated by the experimental data,
between experimental and calculated data. Flat profiles indicate there is still some difference between these values.
higher amounts of gas and, also, less liquid accumulation in the
cavities since gas volume reduction occurs suddenly near 1.0
discharge region. These conditions reduce the possibility of
Dimensionless pressure
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Dimensionless length [ - ]
(a)
0.6
shaft speed increase. However, there is no simple cause-effect
relation for this deviation as in the case of power. Probably, the
0.4 isentropic efficiency of the compression has an important
influence on temperature besides the friction power.
0.2
Dimensionless length [ - ]
300
(b)
Power (kW)
Figure 8: Influence of (a) GVF (n=1200 rpm) and (b) shaft 200 Modelo
Model
Potência
speed (GVF=60%) on pressure profile. Texaco
HATTON
100
Texaco
200
HATTON
Figure 9 shows that with lower shaft speeds and GVF’s
(Figure 9a), the calculated power for the multiphase pump 100
agrees with HATTON (1993) measurements. Higher GVF’s and 0
higher shaft speeds (Fig. 9b), however, make the model 0 2000 4000 6000
overestimate the absorbed power. Again, this behavior is
explained by the little liquid accumulated in pump cavities when p (kPa)
it operates under these conditions. In this case the gas, now
present in pump clearances, reduces average viscosity causing (b)
the real friction power to be less than the calculated one, which
was based on liquid viscosity. In Fig. 9b, when the differential Figure 9: Absorbed Power - comparison against
pressure values increase, the calculated and experimental experimental data: a) 2000 rpm, GVF=50%; b) 3000 rpm,
power becomes closer again. Table 5, with calculated errors, GVF=90%.
shows this behavior in details. The higher differential
pressures induce higher backflow rates and so there is an
increase in liquid accumulation inside chambers. With more
Table 5: Absorbed Power and Discharge Temperature - comparison against experimental data:
a) 2000 rpm, GVF=50%; b) 3000 rpm, GVF=90%.
ACKNOWLEDGEMENTS
REFERENCES
CAETANO, E.F. ; SILVA, R.M. ; da SILVA, R.G. ; CAMARGO,
R.M.T. and ROHLFING, G. (1997). “Cooperation on Multiphase Flow
Pumping”. Proceedings of the 1997 29th Annual Offshore Technology
Conference, OTC'97, v.3, Houston, TX, USA, p. 109-118.