Design and CFD analysis of centrifugal

compressor for a microgasturbine

Reza Aghaei tog and A. Mesgharpoor Tousi
Department of Aerospace Engineering, Amir Kabir University of Technology, Tehran, Iran, and
M. Soltani
Niroo Research Institute of Iran, Tehran, Iran

Abstract
Purpose – This paper aims to show how a good compressor can be designed and modeled with CFD steady models and to explain reasons for
discrepancies between experiment (1D design) and 3D CFD analysis.
Design/methodology/approach – A model with only one impeller channel was used to compare 1D design data, which were obtained from
centrifugal compressor design code, written and developed by the authors. The often used model for CFD analysis of turbo machinery, known as
“frozen-rotor” model, only yields satisfying results for efficiency and pressure ratio, at and near the point of best efficiency. For this case, the static
pressure shows a nearly uniform circumferential distribution at the inlet of the diffuser, which numerically leads to more homogeneous flow rates
through the single vane channels, and thus to a more realistic time averaged flow distribution.
Findings – The numerical results with respect to performance data showed quite good agreement with experimental data at and near the operating
point of best efficiency.
Originality/value – This paper offers a combined 1D and 3D numerical approach in turbo machinery design, especially in radial compressible turbo
machines design.

Keywords Turbines, Centrifugal force, Diffusion

Paper type Technical paper

1. Introduction especially in the rotating impeller, makes the CFD modeling

very difficult. There has been done much research in
The centrifugal compressor has very large field of applications analyzing the flow and the different viscous phenomena, but
in the process industry. The pressure levels vary from 50 kpa there are still no programs that can calculate the time
up to 50 Mpa and the pressure ratio may go up to ten. dependent flow explicit (Pitkanen and Esa, 2000).
Microturbines are small gas turbines. In a micro turbine, a
centrifugal compressor compresses the inlet air. The static
pressure rises due to the decrease of the relative velocity, but
2. One-dimensional flow design
also due to the centrifugal forces. After the impeller, the flow
enters a radially diffuser. The design of centrifugal compressor impeller usually
The first part of the paper deals how the design of a assumes two distinct stages. A preliminary design, making
centrifugal compressor starts with the definition of the initial use of 1D analysis based on previous experience, is sketched
parameters, which are used to calculate the 1D flow in the out to specify the inlet and outlet blade angles and the
corresponding program. This program, developed by authors, “skeletal” dimensions. This is followed by a detailed design in
and test results from previous design to give preliminary which the complete blade and channel geometry is specified
results about the compressor performance. and then subsequently refined by means of successive
In second part of paper, three-dimensional (3D) design and aerodynamic and stress analyses (Ibaraki et al., 2003).
CFD analysis of centrifugal compressor will be described. At first, the design team gathered data from literature and
The initial parameters of compressor impeller geometry and visited to other universities and research institute and was
flow are used in the 3D design process, which is used to define convinced that the design of a centrifugal compressor starts
the whole compressor wheel geometry. with the definition of the initial parameters, which are used to
The compressor flow is calculated with 3D CFD program calculate the 1D flow in the corresponding program.
(FLUENT 6). Results of CFD calculation and 1D Gradually, we were able to have design parameters from our
calculations are compared, and the design procedure own developed code for centrifugal compressor design
repeated, if necessary. The high complexity of the flow, (CCD). This code uses empirical data from literature and
test results to give preliminary results about the compressor
The current issue and full text archive of this journal is available at performance as well as some impeller and diffuser geometrical
www.emeraldinsight.com/1748-8842.htm parameters.
The input data includes compressor pressure ratio, inlet
pressure, rotating speed, and mass flow rate. The program
Aircraft Engineering and Aerospace Technology: An International Journal includes variables that can later be altered: specific speed,
79/2 (2007) 137– 143
q Emerald Group Publishing Limited [ISSN 1748-8842] different geometrical ratios, slip factor, inlet flow angles, etc.
[DOI 10.1108/00022660710732680] The main calculation results are:

137
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

.
Aerothermodynamics properties at different position of ends of the control volume. The depth of the control volume
impeller wheel. is b2 and gradients in the X-direction are not considered.
.
Main geometry parameters, specific speed and power Hence, integration is very, very small but finite angle. Hence,
consumption. there are no variations in the f direction to be considered.
.
Isentropic efficiency and pressure ratio in each point, diff. However, the velocities are time-dependent entering the
pressure recovery. control volume (Figure 1b and c). Integration will be over a
.
Velocity triangles, deviation angle and relative Mach period t and hence, for example, Cm is Cm2p during the time
number (inlet outlet). interval 0 to (1 2 1)t and it is Cm2s during the time interval
(1 2 1)t to t Parameters on the r þ dr side are space and time
This program able to calculate the off-design performance
averaged by definition, hence yielding the so-called mixed-out
and produces the compressor performance parameters for
state. The historical assumption of an unloaded tip is used,
further studies on the performance map.
hence P2s ¼ P2p ¼ P2. Thus, we get:
2.1 Basic equations for one-dimensional design Z t
Z D f Z b2
A further important step is necessary before design 2 rðtÞCðtÞCm ðtÞr 2 dfdx
0 0 0
optimization calculation can be conducted, that is a Z D f Z b2  Z tX ð4Þ
complete description of the flow process, in term of the þ r2m C 2m C m2m r 2 dfdx dt ¼ FðtÞdt
basic governing equations. The basic governing equations are 0 0 0
presented here, but only for mixed out flow (Japikse, 1996).
Using the proper control volume shown in Figure 1 (Japikse, This reduces to radial momentum:
1996). mp C m2p þ ms Cm2s ¼ mCm2m þ ð p2m 2 p2 Þ2pr 2 b2 ð5Þ
Mixed-out state equations (m):
Z t
 Z tX By the same approach, the tangential component of the
trðtÞCðtÞ½CðtÞdA
dt ¼ FðtÞdt ðmomentumÞ ð1Þ momentum equation becomes:
0 c:v: 0
Z t
 r2p C u2p ðC m2p A2p Þ þ r2s Cu2s ðC m2s A2s Þ
trðtÞCðtÞdA dt ¼ m ðmassÞ ð2Þ ð6Þ
c:v:
0 2 r2m C u2m ðC m2m A2 Þ ¼ 0
Z t
 X
tDh0 ½rðtÞCðtÞdA
dt ¼ Pi ðenergyÞ ð3Þ This reduces to tangential momentum:
0 c:v:
i
mp C u2p þ ms C u2s ¼ mC u2m ð7Þ
where CðtÞ ¼ C m ðtÞi^ m þ C u ðtÞ^i u and dA ¼ dxrdu^i m and Pi
includes all parasitic power terms (2 ) and the overall shaft Similarly the energy equation (3) above becomes to energy
power (þ ), where ^i m and ^i u are unit vectors in the meridional equation:
and tangential directions. Hence, CðtÞdA ¼ ^C m rdAdx. mp C P T 02p þ ms C P T 02s þ ðP disc_friction þ P recirc Þ ¼ mC P T 02m
The control volume (Figure 1a) has an infinitesimal radial
extent (dr), hence there are no shear stresses on the sides or ð8Þ

Figure 1 These figures are equivalent for classical tow-zone modelling

138
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

And the continuity equation (2) above becomes: 3. Three-dimensional design and CFD calculations
m ¼ 2pr 2 b2 C m2m r2m ðconservation of massÞ ð9Þ In radial turbomachines, the first blade passage measurements
were published by Eckardt (1976) for a centrifugal
C 22m compressor. The computation methods which have been
T 2m ¼ T 02m 2 ðstagnation temperatureÞ ð10Þ devised and published, large in number and diverse in type to
2C P
combination with Eckardt useful results to have detail
p2m information about flow field in centrifugal impeller passages.
r2m ¼ ðequation of stateÞ ð11Þ
RT 2m Many of researchers such Mc. Nally and Sockol (1985),
Japikse (1976-1996), were used CFD methods to
C 22m ¼ C 2u2m þ C 2m2m ðvector identityÞ ð12Þ turbomachines (Whitfield, 1990). Moore et al. (1984) used
C 2m 3D viscous CFD method to examine the flow in a medium
M 2m ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðMach numberÞ ð13Þ pressure ratio impeller. The CFD results, although not
kRT 2m
directly compared with measurements, showed several aspects

 of loss production in the impeller. Hirsch et al. (1996) using
T 02m k=ðk21Þ
P 02m ¼ p2m ðequivalanceeÞ ð14Þ Krain’s data (1981) and performed numerical simulations
T 2m quoted by theoretical notions concerning secondary flow
(Larosiliere and Skoch, 1997). In the most nearly years, the
Plus assorted dependent variables:
C u2m use of CFD for the computation of turbo machinery flows has
l2m ¼ ðswirl parameterÞ ð15Þ significantly increased, and by combination with
C m2m
measurements it provides a completely tool for simulation,
m2m ðl2m 2 tan b2b Þ design, optimization and most important of all, analysis of the
s2m ¼ ðslip factorÞ ð16Þ flow field inside the turbomachines (Tamma et al., 1999). The
l2m
main purpose of the presented work is to compare 1D and 3D
C u2m results to complete design process of a centrifugal compressor
m2m ¼ ðwork input factorÞ ð17Þ
U2 for a micro turbine.

3.1 Numerical approaches for turbo machinery

2.2 Diffuser equations The solution of flows in moving reference frames can be best
The vane less diffuser is frequently employed in process handled using “moving” cell zones. The computation domain
compressors. For a vane less diffuser, the conservation can consist only of one moving cell zone or a combination of
equations are solved by integration through the vane less moving and stationary cell zones. The single rotating frame
diffuser. In a 1D compressor analysis, this is done as a option “RRF” can be used to model flows in turbo machinery
succession of states through the vane less diffuser. The full set where the flow is unsteady in an inertia frame because the
of equations account for wall friction effects in both the linear rotor/impeller blades sweep the domain periodically.
momentum equations (Japikse, 1996): However, in the absence of stators or volute, it is possible
to perform calculations in a domain that moves with
dC m C 2u C 2 cos b 1 dp the rotating part. In this case, the flow is steady relative to
Cm 2 þ Cf þ ¼0 ð18Þ the rotating frame, which simplifies the analysis. If stators are
dr r b sin f r dr
present in addition to a rotor, that it is not possible to render
dC u C m C u C 2 sin b the computational problem steady by choosing a calculation
Cm þ þ Cf ¼0 ð19Þ
dr r h sin f domain that rotates with the rotor or impeller. The CFD
programs offers different approaches for treating this
1 dp 1 dC m 1 dh 1 problem, all of them are using multiple frames of Tamma
þ þ þ ¼0 ð20Þ
r dr C m dr h dr r et al. (1999):
.
the frozen rotor model;
p
r¼ ð21Þ .
the mixing plane model; and
RT .
the sliding mesh model.
T þ C2 Figure 2 shows a resume from the different numerical
T0 ¼ ð22Þ
2C P approach used in turbo machinery. In this work, a micro
turbine compressor, which designed one-dimensionally with
where the f is the diffuser inclination. This set of equations CCD code, have been simulated three-dimensionally using
can be solved directly by Runge-Kutta integration. The the commercial CFD code FLUENT 6.
equation: The often used model for turbo machinery, known as

0:2
1:8 £ 105 “frozen-rotor”-model only yields satisfying results for
Cf ¼ k
Re efficiency and pressure ratio, at and near the best efficiency
(Tamma et al., 1999).
where k is a constant has been used (Japikse, 1996). A value
of k ¼ 0.01 is frequently used based on a review of different 3.2 Grid generation
vane less diffusers in industrial machines. One of the most important and time-consuming tasks in the
Based on these equations, the 1D design of compressor has process of a CFD simulation is the generation of
been done and results of preliminary design will be used in the computational grid. First of all is the decision whether
3D design and CFD calculations. the domain shall be discretiz with hexahedral, tetrahedral or

139
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

Figure 2 CFD approaches used in turbo machinery

hybrid cells. Using tetrahedral cells can save a lot of time walls in the inertia frame of reference, which are part of the
because the meshing of the volume works almost rotating reference frame, are treated as moving walls with a
automatically. The grids for compressor channel were rotational speed of absolute zero. The walls are adiabatic.
completely generated with the pre-processor GAMBIT,
which offers some CAD tools as well as meshing tools. 3D
3.4 Equation for a rotating reference frame
models of the machine described above were generated. In the
When the equations of motion are solved in a rotating frame
centrifugal compressor, due to rotational periodicity of the
of reference, the acceleration of the fluid is augmented by
blades, only one channel consisting of pressure and suction
additional terms that appear in the momentum equations.
side of the main blade had to be generated (FLUENT Inc.,
FLUENT allows you to solve rotating frame problems using
1998). The whole impeller is made up of 13 blades. The
either the absolute velocity, V, or the relative velocity Vr, as
whole impeller geometry had to be divided into several
the dependent variable. The tow velocities are related the
hexahedral volumes, each of which could be meshed with
following equation (FLUENT Inc., 1998):
hexahedral cells (Figure 3).
Vr ¼ V 2 V £ r ð23Þ
3.3 Numerical implementation
3.3.1 Model parameter Here, V is the angular velocity vector and r is the position
The impeller region was modeled as a reference rotating with vector in the rotating frame.
respective to rotary speed. The most appropriate The left-hand side of the momentum equation appears as
discretization scheme was the segregated implicit solver. It follows for an inertia reference frame:
immediately turns the initial backflow, and solution parameter
come to convergence. Turbulence is modeled using the ›
ðrV Þ þ 7 · ðrVV Þ ð24Þ
standard k 2 1 model without realizable modifications. Also ›t
the energy equation has to be solved additionally. The For a rotating reference frame, the left-hand side written in
medium is treated as ideal gas with viscous heating. terms of absolute velocity becomes:
The momentum equation is solved using the QUICK
scheme, the pressure correction with the SIMPLE scheme. ›
ðrV Þ þ 7 · ðrV r V Þ þ V £ V ð25Þ
The under relaxation factor pressure must be very ›t
conservative, about 0.3 or 0.2; otherwise the pressure In terms of relative velocities the left-hand side is given by:
correction might diverse.
› ›V
3.3.2 Boundary conditions ðrV r Þ þ 7 · ðrV r V r Þ þ 2V £ V r þ V £ V £ r þ r £ r ð26Þ
For the compressible flow, mass flow rate normal to boundary ›t ›t
was used as the inlet boundary condition. At the inlet total where 2V £ V r þ V £ V £ r is the Coriolis force. Note that
temperature, turbulent intensity and length scale must be FLUENT neglects the rð›V=›tÞ £ r term, so it cannot
defined too (Tu ¼ 1.5 per cent, L ¼ 10 mm, FLUENT Inc., accurately model a time-varying angular velocity (FLUENT
1998). A velocity at the outlet, the static pressure and backflow Inc., 1998).
properties must be defined. All in inertia reference frame Using the relative velocity formulation for flows in rotating
moving walls are part of the rotating reference frame. These domains, the equation for conservation of mass, or continuity
walls are treated as moving walls with rotational speed of 0 equation, can be written as follows for both the absolute and
relative to the adjacent cell zone, which is rotating. Non-rotating the relative velocity formulations:

140
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

Figure 3 Modelling and Grid generation by GAMBIT

the design process of centrifugal compressor will be complete.

›r ›
þ ðrV r Þ ¼ S m ð27Þ Not that by using of CFD results we able to improve
›t ›xi performance parameters, and describe the flow field in
Energy equation: FLUENT solves the energy equation in the compressor channel that measurements unable to give this
following form: ability. The compressor data in the point of best efficiency
(BEP) are shown in Table I.
› › The flow field inside the computational domain is shown in
ðrEÞ þ ðut ðrE þ pÞÞ
›t ›xi Figure 4. Contours of static pressure (Figure 5) and
0 1
ð28Þ tangential velocity (Figure 4) are presented in passage of
› @ ›T X
¼ keff 2 hj 0 J j 0 þ uj ðT ij Þeff A þ S h channel. Static pressure contours, for impeller also are shown
›xi ›xi j0 in Figure 6. It must be emphasized that the flow field inside
the impeller shows a good qualitative agreement with the 1D
where keff is the effective conductivity, and Jj is the diffusion data. In areas of high static pressure, the radial velocity is
flux of species j0 . The first three terms on the right-hand side rather small; the effect has an influence on the flow field inside
of energy equation represent energy transfer due to the impeller. High radial velocities are located in areas of low
conduction, species diffusion, and viscous dissipation, static pressure. These results are in different mass flow rates
respectively, Sh includes heat of chemical reaction, and any through each single impeller channel.
other volumetric heat sources you have defined. In energy The comparison of 1D (base on experimental data) and
equation: numerical analysis of the centrifugal compressor is shown in
Table II. As seen, static pressure and velocity magnitude for
p u2
E ¼h2 þ i : impeller, in the BEP show good agreement with 1D data, and
r 2
maximum difference between CFD calculations and 1D
Where sensible enthalpy h is defined for ideal gases design is about 15 per cent.
P A good match between 1D and CFD calculations, offer a
as P h ¼ j 0 mj 0 hj 0 and for incompressible flows as
h ¼ j 0 mj 0 hj 0 þ P=r.R In this equations, mj 0 is mass fraction of little change in design process. It means that design of
T compressor was successfully done, and designers can obtain
species j0 and hj 0 ¼ T ref C P; j 0 dT (where Tref is 298.15 K).
these data, and use them to improve performance of

4. Results and comparison to 1D data

Table I Compressor data in the BEP
As said above, the main purpose of this work was to design a
centrifugal compressor for a micro turbine. Both of the Efficiency h 82 per cent
preliminary design and CFD methods were described, and Rotating speed N 5,000 (rad/s)
said that the 1D design has been done with CCD code and Mass flow rate m_ a 1.45 (kg/s)
essential data for 3D design and CFD calculation were Pressure ratio Pr 4
obtained from this code. Now with comparison these results,

141
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

Figure 4 Tangential velocity

Figure 5 Static pressure contours

compressor or for a new design. By using CFD results, also it interaction of rotating and stationary parts is the frozen
is possible to describe impeller passage flow phenomena. rotor method, which is useful when the circumferential flow
variation of each blade passage experiences is large during a
full revolution. With this option, computations are again
5. Conclusions performed in a steady state mode, based on the assumption of
The centrifugal compressor for a micro turbine has been quasi steady flow around the rotating component at every
designed completely. 1D design was complete with CCD rotating angle. The detailed CFD model of compressor and it
code, which written and developed by authors, then by using comparison with the experimental data are explained and
FLUENT code, CFD analyses base on 1D data have been shown in this work. A good match between CFD and 1D data
done. The common CFD package offers different approaches was seen for a centrifugal impeller of approximately 4:1
to simulate systems with both stationary and rotating pressure ratio. Overall, the CFD gave a good prediction of the
components. With the general single or multi frames, using performance and resolved enough of the local flow details. By
local reference frames, which are stationary or rotating as using this information of flow phenomena, it is possible to
appropriate, can steady state analyses in turbo machinery be improve compressor performance, or make a new design of
performed. One of these different ways to model the centrifugal compressor for different machines.

142
Design and CFD analysis of centrifugal compressor for a microgasturbine Aircraft Engineering and Aerospace Technology: An International Journal
Reza Aghaei tog, A. Mesgharpoor Tousi and M. Soltani Volume 79 · Number 2 · 2007 · 137 –143

Figure 6 Static pressure contours

Table II Comparison of 1D and numerical analysis data Pitkanen, H. and Esa, H. (2000), Centrifugal Compressor
Design and Testing in the Finnish High Speed Technology,
1D CFD (average) Department of Energy Tech., Lappeenranta.
Impeller Inlet Outlet Inlet Outlet Tamma, A., Gurge, M. and Stoffel, B. (1999), Experimental
and 3-D Numerical Analysis of the Flow Field in
Static pressure (kpa) 84.099 241.643 92.05 242.185
Turbomachines Part One and Two, Darmstadt.
Radial velocity (m/s) 0 298.24 0 332.203
Whitfield, A. (1990), Design of Radial Turbomachines,
Tangential velocity (m/s) 28 361.309 44.53 335.021
Longman, Harlow.
Relative velocity (m/sec) 243 363.36 275.75 416.35

Further reading
References Niazi, S., Stein, A. and Sankar, L.N. (1998), Development and
Application of a CFD Solver to the Simulation of Centrifugal
FLUENT Inc. (1998), Fluent5 User’s Guide and GAMBIT
Compressors, School of Aerospace Engineering, Georgia
Modelling Guide, FLUENT Inc., Lebanaon, NH.
Institute of Technology, Atlanta, GA.
Ibaraki, S., Matsuo, T. and Kumo, H. (2003), “Aerodynamics of
Pitkanen, H., Esa, H., Sallinen, P. and Larjola, J. (1999),
a transonic centrifugal compressor impeller”, ASME, Vol. 125. Time-Accurate CFD Analysis of a Centrifugal Compressor,
Japikse, D. (1996), Centrifugal Compressor Design and Department of Energy Tech., Lappeenranta.
Performance, ETI, Wilder, VT.
Larosiliere, L.M. and Skoch, G.J. (1997), “Aerodynamic
synthesis of a centrifugal impeller using CFD and
Corresponding author
measurements”, 107515, AIAA-97-2878. Reza Aghaei tog can be contacted at: Reza_tog@yahoo.com

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com

Or visit our web site for further details: www.emeraldinsight.com/reprints

143