Simulation Modelling Practice and Theory 14 (2006) 11431154

www.elsevier.com/locate/simpat

Numerical simulation of combined screw compressorexpander

machines for use in high pressure refrigeration systems
Ahmed Kovacevic *, Nikola Stosic, Ian K. Smith
Centre for Positive Displacement Compressor Technology, City University, London EC1V 0HB, UK

Available online 17 October 2006

Abstract

Recent interest in natural refrigerants has created a new impetus for studies of CO2 as a working uid in vapour com-
pression systems for refrigeration and air conditioning. Two major drawbacks to its use are the very high pressure dier-
ences required across the compressor and the large eciency losses associated with the throttling process in the
refrigeration cycle. It is shown how these disadvantages can be minimised by the use of a screw machine both to compress
the gas and use the expansion process to recover power. Both these functions can be performed simultaneously, using only
one pair of rotors, in a conguration that partially balances out the forces induced by the pressure dierence and hence,
reduces the bearing loads to an acceptable level. A further feature is the use of rotors, which seal on both contacting sur-
faces so that the same prole may be used for the expander and the compressor sections. This enables the rotors perform-
ing both these functions to be machined or ground in the same cutting operation and then separated by machining a
parting slot in them. Computational Continuum Mechanics comprising both, uid ow and structural analysis is used
in this paper for the investigation of uid-solid interaction in such machines.
2006 Elsevier B.V. All rights reserved.

Keywords: Screw compressor; Screw expander; Computational uid dynamics; Mathematical modelling

1. Introduction

Nearly 20% of the electricity produced in developed countries is used to drive compressors, the majority of
which are required for air-conditioning and refrigeration systems. There are therefore considerable environ-
mental advantages to be gained by improving the eciency of such systems and the compressors that drive
them. A further requirement for environmental protection is the need to minimise the use of refrigerants,
which may lead to the breakdown of the ozone layer in the upper atmosphere. As a result of the latter require-
ment, in the last few years there has been a growth of interest in the use of natural uids as refrigerants, in
place of the halogenated hydrocarbons now widely used. One of these natural refrigerants is CO2, which is
totally free from ozone breakdown eects. Despite the environmental advantages of CO2 as a working uid

*
Corresponding author. Tel.: +44 20 7040 8780; fax: +44 20 7040 8566.
E-mail address: a.kovacevic@city.ac.uk (A. Kovacevic).
URL: http://www.city-compessors.co.uk (A. Kovacevic).

1569-190X/$ - see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.simpat.2006.09.004
1144 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

Nomenclature

C constant in the turbulence model

F [N] body force
i unit vector
I unit tensor
k [m2/s2] kinetic energy of turbulence
m [kg] mass
m_ [kg/s] source in pressure correction equation
p [Pa] pressure
P production of turbulent kinetic energy
q source term
Q_ [J/s] source in the energy equation
s [m2] control volume surface
t [s] time
u [m] displacement in solid
v [m/s] uid velocity
V [m3] volume
x [m] spatial coordinate
z [m] axial coordinate
C diusion coecient
e [W/kg] dissipation of kinetic energy
/ Variable in transport equation
k Lame coecient
l [m2/s] viscosity
g Lame coecient
q [kg/m3] density
r Prandtl number
Dt [s] time step used in the calculation

Indices
add added or subtracted
e eective
rec receiver
T turbulent

in vapour compression systems for refrigeration and air conditioning, there are two major drawbacks to its
use. These are the very high pressure dierences required across the compressor, which may be as high as
60 bar, and the large eciency loss associated with throttling over such a large pressure drop in the near crit-
ical region.
The magnitude of this loss can be better appreciated by the analysis of a typical CO2 system, as shown in
Fig. 1. In this case, in an idealised reversible cycle, recovery of the throttle losses, by controlled expansion, will
increase the COP by as much as 72%. In a practical system, this gain would be reduced by the compression and
expansion eciencies, which would reduce the expansion work and increase the compression work but even
then, the potential benets are so large that some form of power recovery from this expansion process is essen-
tial if the resultant coecient of performance (COP) is to be acceptable.
To recover power from the throttle loss in an economic manner, a number of proposals have therefore been
made to use various types of positive displacement machines, mainly of the vane type [4], in such a manner
that compression in one part of them is combined with recovery of work from simultaneous expansion in
 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154 1145

Fig. 1. Typical CO2 cycle with throttle valve.

another part. The principle of this is shown in Fig. 2. However, how well they operate with high-pressure dif-
ferences across the vanes has not yet been conrmed.
For many years, the authors have been investigating the use of twin screw machines as both expanders and
compressors of conventional halocarbon refrigerants [9]. Their mode of operation is shown in Fig. 3. In this
case the compression process is shown in Fig. 3a, where the suction port is located at the low-pressure end and
where the trapped volume is largest. Compressed gas is then discharged through the high-pressure port at a
position where the trapped volume is reduced in size. The expulsion of the gas is complete when the trapped
volume is zero since the clearance volume in these machines is negligible. Reversing the direction of rotation,
as shown in Fig. 3b, causes the direction of uid ow through the machine to reverse. Gas, then enters through
the high-pressure port and is discharged through the low-pressure port. The machine then acts as an expander.

Fig. 2. CO2 cycle with combined machine.

1146 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

Fig. 3. Screw compressor (a) and Screw expander (b) working principles.

The machine will also work as an expander when rotating in the same direction as a compressor, if the suction
and discharge ports are positioned in the same axial position but on the opposite sides of the casing to
those shown in Fig. 3b, since this is eectively the same as reversing the direction of rotation relative to the
ports.
Although these machines have many potential advantages over other types of positive displacement
machine, when applied to CO2, the huge bearing forces associated with the pressure distribution within them
have hitherto made them appear to have limited range in which they can work reliably and economically. Cur-
rent practice is for a maximum discharge pressure of 85 bar and a maximum dierence between suction and
discharge of 35 bar. Rinder, [8], presented a comprehensive analysis of the eects of such pressures and Arbon,
[1], gave a good review of current trends in the design, manufacture and use of high pressure screw compres-
sors. The use of CO2 (R744) in refrigeration cycles requires both maximum pressures and pressure dierences
beyond these limits. Accordingly, till recently, screw compressors have not been considered for this purpose.
The arrangement of a twin screw machine to both compress and expand the refrigerant, as shown in Fig. 4,
can overcome most of these disadvantages. Firstly, the axial forces on the rotors in the two sections virtually
balance, while the inlet and discharge ports are located to partially balance out the radial rotor forces. The net
eect is to reduce the total bearing load and thereby, reduce mechanical friction losses. This implies that the
machine can operate under higher dierential pressures than would be possible if the two functions were per-
formed in separate machines.
Secondly, if the rotor proles are generated to form a seal on both contacting surfaces, then both the expan-
sion and compression processes can be equally ecient, regardless of which pair of surfaces are in contact. The
rotors can then be formed in a single manufacturing operation and, by means of slot subsequently cut in them,
the compression and expansion chambers can be formed within a single casing, divided by a pair of split
plates, inserted during assembly of the machine.
Also, the large mass of oil injected into the working chamber to seal, lubricate and cool the rotors and cas-
ing, minimises thermal distortion of the rotors.
The net result of all these features is to make it possible to produce a twin screw machine to full both func-
tions at low cost with a range of operating pressure dierences large enough to consider it for use in CO2
refrigeration applications.
Computational Continuum Mechanics (CCM) has been used here to model the pressure and temperature
changes within the machine. This has been developed further to estimate how the solid components deform as
 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154 1147

Fig. 4. Combined compressorexpander with split chambers.

a result of these changes and hence how the uidsolid interaction resulting from the deformation alters the
performance.

2. Three-dimensional CCM calculation of the combined screw machine

Improvements in both, computational speed and numerical methods over the last two decades have
greatly increased the scope and power of Computational Fluid Dynamics (CFD), and Computational Con-
tinuum Mechanics (CCM), in engineering design. Consequently, vendors of such software packages have
developed facilities for their use in a wide range of engineering applications. However, because of the com-
plex geometry and processes within screw machines, none of the standard packages were capable of anal-
ysing them, so their designers have hitherto made little or no use of three-dimensional Computational
Continuum Mechanics.
The usual practice for the analysis of solid body deformation or uid-structure interaction is by coupling
nite volume (FV) code with nite element (FE) solvers using a specially designed interface. Most of CFD and
FE vendors do that procedure to take advantages of both FV and FE procedures. Although well established,
this procedure in many situations may not be entirely suitable for calculation. One important example is that
of conjugate heat transfer, where heat transfer in both solid and uid has to be calculated simultaneously. This
is the case of the interaction of uid ow and solid deformation in a screw machine. In this case, the small
deformations of solid parts are caused by the large pressure and temperature gradients, which are generated
as a consequence of the uid ow. Although these deformations are relatively small, they are of the similar
order of magnitude as the compressor clearance, and may thus signicantly change the ow within the
machine.
CCM enables simultaneous calculation of both the uid ow and the structure behaviour to determine the
eects of changes in the compressor geometry on internal heat and uid ow and vice versa. However, such an
approach can produce reliable predictions only if calculated on a regular numerical mesh. Todays grid gen-
erators are still not powerful enough to produce a regular mesh that can be conveniently deformed and moved
1148 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

to simulate the processes within a screw machine. Even if a grid generator could be found to produce regular
cells in so complex geometry, where the ratio of the largest cell size to that of the smallest may be up to 1000:1,
it can be achieved only with substantial number of computational cells. Increasing the number of grid points
usually increases the accuracy of calculations but it demands substantial computer resources and time, often
unacceptable for practical calculations. The authors aim was therefore to develop a procedure to reduce the
number of grid points, without sacricing accuracy, in order to make three-dimensional ow analysis possible
in screw machines.

3. Grid generation

The authors have developed an automatic numerical mapping method for arbitrary screw compressor
geometry, as explained in [5], which was later used for the analysis of the processes in screw compressors.
This method took advantage of the work done by Demirdzic, [2] and Peric, [3] who showed that by the use
of moving frames on structured and unstructured grids, a common numerical method could be used for the
simultaneous solution of uid ow and structural analysis. On that basis, the authors have developed an
interface program called SCORG (Screw COmpressor Rotor Geometry Grid generator), which also enables
a grid, generated by the program, to be directly transferred to a commercial Computational Fluid Dynamics
or Computational Continuum Mechanics code through its own pre-processor. A number of commercial
numerical solvers are currently available, of which the authors decided to employ Star CDs COMET for
calculation of screw machine processes. That code oers the possibility to calculate both the uid ow
and its eects on the solid structure simultaneously by the application of Computational Continuum
Mechanics.
The interface employs a novel procedure to discretise rotor proles and to adapt boundary points for each
particular application. An analytical transnite interpolation method with adaptive meshing is used to obtain
a fully structured three-dimensional numerical mesh, which is directly transferable to a Computational Fluid
Dynamics or Computational Continuum Mechanics code. This is required to overcome problems associated
with moving, stretching and sliding rotor domains and to allow robust calculations in domains with signi-
cantly dierent ranges of geometry features [6].

4. Governing equations

Both, the compressor ow and the relationship between stresses and strains in the structure of the com-
pressor parts are fully described by the mass averaged conservation equations of continuity, momentum,
energy and space, which are accompanied by an equation of state and turbulence model equations, as
shown in [2]. The continuity and mass equations are then balanced through the articial pressure correction
equation, which nally closes the relation between the pressure, velocity and temperature. In the case of
multiphase ow, where oil is injected in the working chamber, a concentration equation is added to the sys-
tem. The numerical solution of such a system of partial dierential equations is then made possible by the
inclusion of constitutive relations in the form of Stokes, Ficks and Fouriers law for the uid momentum,
concentration and energy equations respectively and Hookes law for the momentum equations of a thermo-
elastic solid body.
All these equations are conveniently written in the form of the following generic transport equation:
Z Z Z Z Z
d
q/ dV q/v  vs  ds C/ grad/  ds q/S  ds q/V  dV 1
dt V S S S V

Here / stands for the transported variable, e.g. Cartesian components of the velocity vector in uids vi,
enthalpy h, etc. C/ is the diusion coecient. The meaning of surface and volume source terms, q/S and
q/V in all transport equations is given in Table 1.
The resulting system of partial dierential equations is discretised by means of a nite volume method in the
general Cartesian coordinate frame. This method enhances the conservation of the governing equations while
at the same time enables a coupled system of equations for both, solid and uid regions to be solved simul-
taneously [3].
 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154 1149

Table 1
Terms in the generic transport Eq. (1)
Equation / C/ q q/V
h/S  i
Fluid momentum vi le leff gradvT  23 leff div v p I  ii fb, i
T
Solid momentum ui g [g(grad u) + (kdiv u  3KaDT)I] ii fb, i
k lt k oe
Energy e oe=oT rT  oe=oT op  gradp T: grad v + h
Concentration ci qDi, e 0 sci
1
Space q 0 0 0
lt
Turbulent kinetic energy K l rk 0 P  qe
lt 2
Dissipation of turbulent kinetic energy e l re 0 C 1 P ke  C 2 q ek  C 3 qediv v

5. Boundary conditions

This mathematical scheme is accompanied by boundary conditions for both the solid and uid parts. A
novel treatment of the compressor and expander uid boundaries was introduced in the numerical calculation,
as suggested in [5]. Both, the compressor and the expander were positioned each between two or more rela-
tively small receivers at their suction, discharge and the oil injection points. Together, the compressor, expan-
der and receivers form a system separated from the surroundings by adiabatic walls only. The system
communicates with its surroundings through the mass and energy sources or sinks placed in these receivers
in order to maintain constant pressures at suction, discharge and the oil injection pipes. The mass source,
which exists in the pressure correction equation, is calculated as the dierence between the pressure in the cell,
obtained by the solution of the system of equations in the current iteration, or time step, denoted by p, and the
pressure prescribed for the entire receiver, prec as
 
dm p  p qV
m_ add  rec 2
dt pconst prec Dt
The energy of the uid added to or subtracted from a cell of either receiver is calculated as
 
dm
Q_ add hadd m_ add hadd 3
dt pconst
That energy is then added to the energy equation through the volume source denoted by h in Table 1. The
enthalpy hadd in Eq. (3) is a function of the cell pressure and temperature calculated from the system of
equations.
The solid part of the system is constrained by zero displacements in restraints and zero tractions elsewhere.
Because of the nature of space discretisation the connections between the solid and uid parts are explicitly
determined. Temperatures calculated from the uid ow are extrapolated on solid boundaries as their own
boundary temperatures for the energy equation. On the other hand, the pressure acting on boundary forms
a boundary force as the boundary condition for the solid momentum equation. At the same time, displace-
ments in the solid body make the numerical grid of the ow passages to change for the next iteration or time
step.

6. Presentation and discussion of the results

A balanced rotor two-chamber combined compressor expander, shown in Fig. 4, is considered here for a
high pressure CO2 application. Gears are not used to synchronize the rotors since a fair amount of oil is
injected in the system to seal and lubricate them and cool the high pressure gas. The oil is also used to maintain
the temperature of the rotors at a similar level to that of the main uid.
The rotors for the compressorexpander considered in this paper, for use in a CO2 refrigeration system, are
shown in Fig. 5. The machine presented in this gure was designed with the aid of the three-dimensional
1150 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

Fig. 5. CAD model of the expandercompressor.

Mechanical Desktop 6 model and used as the basis for numerical three-dimensional analysis. The numerical
grid generated by SCORG, shown in Fig. 6, contains 1,378,960 numerical cells in both the uid and solid
parts. The rotors of the combined machine are mapped with 393,752 hexahedral numerical cells while the
remaining cells belong to the uid part of the machine of which 538.050 cells belong to the compressor part.
Even with well over one million cells, a solution could be obtained from a PC powered by an ATHLON 2000+
processor with 1.5 GB RAM.

Fig. 6. Numerical mesh of the compressorexpander rotors.

 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154 1151

Fig. 7. Radial and axial forces for compressor, expander and combined compressorexpander.
1152 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

The operation of the machine was simulated through 24 time steps for one interlobe rotation giving an
overall number of 120 time steps for one full rotation of the male rotor. The time step length was synchronised
with a compressorexpander machine speed of 5000 rpm. An error reduction of 4 orders of magnitude was
obtained after approximately 50 outer iterations at each time step. This took approximately 10 min of com-
puter running time per the time step. The overall compressor parameters such as the torque, volume ow,
forces, eciencies and specic power of both, the compressor and the expander, as well as the combined
machine, were then calculated. Additionally, pressuretime diagrams of the compression and expansion pro-
cess, the ow and pressure patterns in the compressor chambers and the rotor deformation are provided.
Calculations obtained for the two chamber porting arrangement of the expander-compressor result in the
balancing of the axial load on the rotors. As shown in Fig. 7 axial forces that are in excess of 300 kN act on
both, the compressor and the expander, but since they are in opposing directions, the overall axial force is
reduced to as little as an average of 40 kN on the male rotor. Also, since the pressure loads in the compressor
and the expander are located almost opposite each other near the centre of the rotors, the radial loads on the
bearings are thereby reduced as shown in Fig. 8 by as much as 20%, if all the bearings are taken into account.
Consequently, the arrangement with only two pairs of bearings, for the combined compressorexpander
machine, in comparison with 4 pairs of bearings, for a separate compressor and expander, gives lower bearing
mechanical losses.
The reduction of the bearing load is of particular importance when considering the design of CO2 systems
since it is the value of the sustainable bearing load, which limits the maximum permissible operating pressure
dierence.
Additionally, in the combined machine, the deformation of the rotors is substantially smaller than in the
case of the compressor or expander alone, as can be seen from Fig. 9. The deformations of compressor rotors
are presented in the top gure while the deformation of the combined compressorexpander machine for the
same conditions are shown in the bottom gure. More details about the analysis of the solid uid interaction
in dierent types of screw compressors is given in [7]. Some experimental results have been presented there
which agree well with the predicted values. It has shown there that CO2 refrigeration screw compressor rotors
deform by about 20 lm, thereby increasing the critical clearance between the rotors in the discharge area of

Fig. 8. Position of radial and axial forces for compressor, expander and combined compressorexpander.
 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154 1153

Fig. 9. Deformations of the compressor rotors (top) and combined compressorexpander rotors (bottom).

the machine. Deformations presented in Fig. 9 are magnied 5000 times to show the dierence between the
combined compressorexpander machine and a screw compressor used for the same application. The defor-
mation of the compressorexpander rotors is much smaller as a consequence of the balancing forces on rotors.
Together, the reduced rotor deformations and the rotor prole, which results in a stronger female rotor,
increase the range of pressure for which a screw machine can be used for CO2 refrigeration applications.

7. Conclusions

Simultaneous compression and expansion by means of a single pair of screw rotors, maintained in separate
chambers of the same casing, has been examined as a means of improving the performance of vapour com-
pression refrigeration and air conditioning systems, with CO2 as the working uid.
A full three-dimensional Computational Continuum Mechanics simulation of uid ow and structure
deformations has been carried out to determine what forces are induced by the pressure and temperature
changes, during the processes of compression and expansion, and how these inuence internal distortion
within such machines. Preliminary results show that such an arrangement not only increases the Coecient
1154 A. Kovacevic et al. / Simulation Modelling Practice and Theory 14 (2006) 11431154

of Performance of the system by recovery of power otherwise lost through throttling, but also reduces both the
radial and the axial forces acting on the rotors. This enables the maximum permissible operating pressure dif-
ferences for twin screw compressors to be increased and the mechanical bearing friction losses to be reduced.

References

[1] I. Arbon, High pressure screw compressors, in: IMechE International Conference on Compressors and Their Systems, London, 2001.
[2] I. Demirdzic, S. Muzaferija, Numerical method for coupled uid ow, heat transfer and stress analysis using unstructured moving
mesh with cells of arbitrary shape, Comput. Method. Appl. M. 125 (1995) 235255.
[3] J.H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, Springer, Berlin, Germany, 1996.
[4] M. Fukuta, T. Yanagisawa, Y. Ogi, Cycle performance of CO2 cycle with vane compressorexpander combination, in: IMechE
International Conference on Compressors and Their Systems, London, 2001, pp. 315324.
[5] A. Kovacevic, N. Stosic, I.K. Smith, Three-dimensional numerical analysis of screw compressor performance, J. Comput. Method. Sci.
Eng. 3 (2) (2003) 259284.
[6] A. Kovacevic, Three-dimensional numerical analysis of ow prediction in positive displacement screw machines, Ph.D. Thesis, City
University London, 2002.
[7] A. Kovacevic, N. Stosic, I.K. Smith, The Inuence of Rotor Deections upon Screw Compressor PerformanceConference on Screw
Type Machines, VDI-Schraubenma-chinen, Dortmund, Germany, 2002.
[8] L. Rinder, Oil-injected screw compressor of high-pressure application, in: IMechE International Conference on Compressors and Their
Systems, London, 1999, pp. 3748.
[9] N. Stosic, I.K. Smith, A. Kovacevic, Twin screw machines to replace throttle valve in refrigeration systems, J. Mech. Eng. 8 (2001) 484.