16th Australasian Fluid Mechanics Conference

Crown Plaza, Gold Coast, Australia

2-7 December 2007

Numerical Investigation of a Small Gas Turbine Compressor

J. Ling, K.C. Wong & S. Armfield

School of Aerospace, Mechanical & Mechatronic Engineering
University of Sydney, NSW, 2006 AUSTRALIA

Abstract
This paper presents the initial development and study of a Introduction
centrifugal compressor undertaken at the Micro Propulsion The application of small gas turbine engines ranges from
Group (MPG), University of Sydney. The compressor is used in hobbyist model jet aircrafts to target drones, and Unmanned
a small gas turbine engine (108mm diameter). Gerendás et al [1] Aerial Vehicles (UAV). Due to the low efficiency and thrust-to-
refer to gas turbines of this size as small aero-engines. The gas weight ratio, they are limited to short range applications.
turbine chosen for the initial study is the KJ66, which is one of However, the inability for spark ignition engines to work at high
the most robust and primitive small gas turbine designs available. altitude warrants further development to improve the efficiency
This project is aimed at improving the performance and of these miniature gas turbines. Furthermore, gas turbines in the
efficiency of a KJ66 small gas turbine, which currently utilises a form of turbojet or turbofan are capable of providing propulsion
66mm diameter centrifugal impeller and radial wedge type at higher speed, where spark ignition engines are not applicable
diffuser. The proposed replacement impeller is 71mm diameter, with air screws.
coupled to a mixed flow vaned diffuser to produce a higher Most small gas turbines for aero-propulsion use centrifugal
pressure ratio at design conditions. The changes were made in compressors similar to those from automotive turbochargers.
such a way that the original engine’s diameter of 108mm was The dissimilarity is that turbochargers usually have their radial
maintained. The geometry of the diffuser vanes was redesigned diffuser coupled to a volute with delivery outlet tangential to the
to ensure effective diffusion, while directing the flow from the rotor, while aero engines do not benefit from this design due to
radial to axial direction in a smooth turn. In addition to space and design restrictions. The inlet and outlet of small aero
increasing the impeller diameter, having smooth turns in the engines such as the KJ66 are both axial even though the
stage components is expected to increase compression efficiency, compressor wheel is of a centrifugal type. The passage turns
at least in a certain operational range. immediately after the diffuser into the axial direction, to feed
The work presented here uses Computational Fluid Dynamics pressurised air into the combustion chamber right behind the
(CFD) simulations. The performance and efficiency charts of the compressor (Figure 1).
original compressor stage, with a KKK2038 compressor wheel The ideal thermal efficiency of a gas turbine is related to the
coupled to a wedge-type diffuser, were mapped. The pressure ratio and can simply be estimated by a well-known
performance chart of the new design was also mapped and equation [2]:
comparisons are made. The performance of the new stage design
is notably better than the original design, within a certain 1
η th = 1 −
operational range. The power requirement to drive the (P2 P1 )k −1 k (1)
compressor was also checked to ensure its compatibility with the
turbine drive. In such small engines the ideal thermal efficiency ranges between
10 and 20%. However, losses due to the nature of these engines
running at high rotational speed and high flow velocity relative to
Nomenclature the size of air passages plague these engines, further lowering the
η Efficiency efficiency. Due to the small diameter, the compressor rotors
P Pressure have to run at nominal speeds of around 100,000 rpm. At this
k Specific air ratio speed, a 10cm diameter engine (with 62~68mm impeller wheel
Pr Pressure ratio and a diffuser section which makes up the rest of the diameter)
μ Work Coefficient can only produce a pressure ratio of around 1.8 ~ 2. This is
W Work
h Enthalpy normally achieved by incorporating a centrifugal compressor due
M Mach Number to the fact that they are more compact compared to axial
C Absolute flow velocity machines and capable of achieving a higher pressure ratio per
U Impeller local tangential speed stage. The high rotational speed also gives high frictional losses
Cp Constant pressure specific heat in mechanical parts.
It is therefore proposed to improve the engine by redesigning the
Subscripts compressor so that it will produce a higher pressure ratio and run
t Total condition at lower speeds, if possible. However, the pressure rise is
th Thermal properties proportional to the square of impeller tip Mach number [3]
tt Total to total conditions
00 Stagnation to stagnation conditions
1 Impeller inlet conditions
(
Prtt = 1 + (k − 1)ημM U2 2 )(
k k −1)
(2)
2 Impeller tip conditions In other words, to maximise performance-to-rotational speed
θ Tangential component ratio, the diameter of the compressor wheel must be increased to
maintain the tip speed, while the rotational speed reduced.

961
However, within the space constraints of the small engine, this The original design of the compressor stage in the KJ66 has a
can only be achieved by compromising the diffuser section 66mm compressor wheel coupled to a radial wedged diffuser
design. section (Figure 1). The flow is turned from radial to axial
The diffuser is necessary to recover the kinetic energy of the flow immediately after the diffuser passages. The wedge section is
and transform it into pressure. The diverging rate must be positioned at an angle of 3° to the radial. The inner diameter of
correctly designed within the working range to ensure high the casing is 107.4mm.
efficiency conversion of kinetic energy. If a larger compressor The aerodynamic design of the compressor wheel normally starts
wheel is employed, there is not much space left for the diffuser with 1D equations and is correlated with empirical data [4]. The
section. geometry data from initial 1D analysis is then used to produce a
Fortunately, compressors in aero-engines do not need wide mass physical geometry for the wheel, defined by a series of curves.
flow range at a specific rpm when compared to industrial With the help of some simple measuring devices and the
compressor systems or automotive turbochargers. Therefore only compressor design software ANSYS BladeModeller, the
the performance in a certain range is necessary to be improved. geometry of the KKK2038 compressor wheel was digitally
It should be noted that this compressor is driven by an axial reproduced. The performance of the compressor stage from 1D
turbine through a direct shaft. This enables control of the flow by analysis agrees well with the manufacture’s specification for
selection of an appropriately sized turbine, so that the compressor maximum rpm. The impeller tip diameter is 66mm, with inlet
will only work within the matched range. This allows us to have hub and inlet shroud diameters 12.5mm and 45mm respectively.
a diffuser section designed in such a way that the performance is The blade geometry definition is shown in Figure 2. A clearance
altered to give a head-to-flow curve more similar to an axial of 0.3mm between the shroud tip and the casing was modelled.
stage.
Centrifugal stages have flat head-to-flow curve, while axial 0
stages have a steep curve. Axial stages usually have much higher
flow rates compared to a centrifugal stage of similar diameter [5]. -10
In a compact engine such as the KJ66, a centrifugal stage is
chosen due to the compactness and higher stage pressure ratio.
The proposed centrifugal stage design achieves a higher pressure -20
Angle (beta definition)

ratio at lower flow rate, but with a rapidly reduced pressure ratio
with higher flow rates. Under such circumstances it may likely -30
necessary to redesign the turbine stage (which is not a subject of
discussion in this publication) to ensure the mass flow rate
always falling within the optimised range. -40
Span = 0.0

To accommodate a larger compressor wheel, the original radial Span = 0.5

wedge type diffuser no longer performs well; there is no longer -50

enough radial distance to form diffuser passages with the Span = 1.0
required divergence rate. Therefore a vaned diffuser section -60
which gradually guides the air into the axial direction while the 0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00
passage expands was designed, requiring a careful selection of Meridional Length
the radial-to-axial curvature to minimise separation and
recirculation. After several attempts, a decision was made to Figure 2: Impeller blade angle definition
increase the size of the impeller wheel to 71mm so that a diffuser
of correct turning rate could be accommodated.
CFD Model of Original Stage
Original Compressor Stage The CFD results were obtained using CFX commercial CFD
package. Only one passage of each of the impeller and diffuser
sections was meshed for the simulations. A rotational periodic
boundary condition was applied.
Meshes for the impeller domain and the diffuser domain were
generated independently. The mesh of the impeller domain was
generated using ANSYS Turbogrid and contained only hexahedra
elements; while the diffuser section was meshed using CFX-
mesh with tetrahedral/prism elements.

Table 1: Mesh information

Domain Nodes Elements
Impeller 110170 99705
Diffuser 18292 46071
All Domains 128462 145776

The two domains were connected by a General Grid Interface

(GGI) with non matching grid, which iterates and checks
balances in fluxes. Steady state solutions were obtained using the
k-ε model. Although it does not show the flow details such as
Figure 1: 3D model of original stage design of KJ66.
flow separations, grid independence studies showed this would
be sufficient to obtain accurate performance curves.

962
Total pressure inlet and mass flow outlet boundary conditions It is expected the actual efficiency will be lower than the
were chosen so that the solutions for a range of mass flows could simulation results shown in Figure 4 due to minor losses, such as
be obtained. The inlet turbulence intensity was set to be 3% with heat loss through the casing and mechanical friction, (which were
the length scale auto-computed by CFX. The inflow vector was not taken into account in the modelling). But the efficiency
set to be axial to the rotational axis with a stationary frame total curves still show the correct trend, which is sufficient for
temperature of 300K. The entire computational volume of the comparative studies.
impeller passage was placed in a rotational frame except the inlet
face and the shroud surface. 0.85
Both stationary and moving walls were modelled as smooth non-
slip walls. The mesh generation method was based on a 0.8
boundary layer with 0.1mm layer thickness, expansion rate of
1.33 and a 0.002mm wall offset. Mesh details are given in Table
1. All residuals were required to be converged to at least 1.0e-4, 0.75

with high efficiency points converged to 1.0e-6.

Total Isentropic Efficiency

0.7
Results of Original Stage
Pressure ratio and efficiency curves for the original stage running
0.65
at 80,000, 100,000, 117,000 and 120,000 rpm are shown in
Figure 3. The specification of the KJ66 engine states the
maximum speed of the engine is 117,000rpm, at which the 0.6 80,000rpm

compressor produces a pressure ratio of 2.2 at a flow rate of 100,000rpm

0.22kg/s. This is well matched by the results from CFD 0.55 117,000rpm
simulations as shown. 120,000rpm
Both Figure 3 and Figure 4 show the pressure and efficiency
0.5
curve features of typical centrifugal compressors, where pressure
0.10 0.15 0.20 0.25 0.30 0.35
and efficiency peak at a certain flow rate. Increasing or
Mass Flow Rate [kg s^-1]
decreasing the flow rate from this optimum point results in a
reduction in both pressure and efficiency. Pressure drops rapidly Figure 4: Stage efficiency curves.
with the increase in mass flow rate and this will lead to chocking
at a certain point. On the other hand, pressure also drops Figure 4 shows the efficiency of the stage drops as the speed
accordingly as the mass flow rate is reduced from that optimum increases from 80,000rpm to 120,000rpm. Even though the stage
point and the stage will quickly start to surge. The exact points is capable of producing a higher pressure ratio, it has limitations
for surging and choking would not be predicted. in flow rate and the efficiency will drop accordingly.

2.4 New Compressor Design

80,000rpm The improved design of the compressor stage incorporates a
2.2 100,000rpm
71mm impeller wheel with a larger inlet flow area and a mixed
flow diffuser section with rounded turn. The 71mm impeller
117,000rpm
wheel alone was designed to give higher mass flow. However,
2 120,000rpm
this diffuser section which turns the air 90° does not work very
well with higher flow velocity. Earlier trial and error
investigations showed that if the flow velocity is too high, which
Pressure Ratio

1.8
is often the case for flow leaving the impeller outlet at a high
flow rate; the flow in the diffuser section separated quickly and
1.6
formed an aerodynamic blockage.

1.4

1.2

1
0.10 0.15 0.20 0.25 0.30 0.35
Mass Flow Rate [kg s^-1]

Figure 3: Pressure ratio – mass flow curve of the compressor stage.

Total isentropic properties are used for the efficiency definition

in this case, as both pressure and velocity at the stage outlet are a
useful form of energy in a gas turbine. Kinetic energy in the flow
which is not yet transformed into pressure, can be used to
enhance mixing in the combustion process or used to build up
more pressure later by further diffusion. Total isentropic
efficiency is then defined by
Δht (3) Figure 5: 3D model of the new compressor stage incorporate 71mm
η isen = impeller wheel
Winput

963
Unlike the radial wedge diffuser, where the flow velocity was CFD Model of New Design
already slowed down before it entered the turn, the mixed flow The CFD model of the stage with a 71mm impeller wheel was
diffuser slows the flow while making the turn simultaneously. meshed similarly to that of original design, as discussed earlier.
Therefore either the volumetric flow rate or the impeller tip The only difference was that the geometrical nature of the vaned
velocity needed to be reduced before the flow turns into the axial diffuser enabled the section to be meshed with hexahedra
direction. The higher tip backswept angle on the 71mm impeller elements providing greater computational efficiency. Mesh
blade is designed to lower the absolute flow velocity leaving the details are given in Table 2. Simulations were made with the
impeller outlet. stage running at 100,000rpm and 120,000rpm.

-10

-20
Angle (beta definition)

-30

-40

Span = 0.0
-50

Span = 0.5
-60
Span = 1.0

-70
0 10 20 30 40 50
Meridional Length
Figure 8: Hexahedra mesh on the 71mm impeller wheel (showing two
Figure 6: Impeller blade angle definition (71mm wheel) passages).

Figure 6 shows the new compressor has blades with a higher

backswept angle. The inlet hub and shroud diameters are Table 2: Mesh information of new compressor stage.
increased to 18mm and 56mm respectively. The impeller blade Domain Nodes Elements
tip height in this case was increased from 6mm in the original Impeller 141156 128926
impeller to 6.2mm. Diffuser 59997 63760
In the diffuser section, the passages are separated by vanes
instead of wedges, to control the rate of diffusion while making All Domains 201153 182686
the turn into axial direction. Without increasing the overall
engine diameter, the diffuser makes a smooth turn within a
shorter distance while maintaining the engine’s 107.4mm inner
diameter. The angle of the diffuser vanes is approximately linear
(Figure 7), with an inlet angle of 70°, which was determined by
earlier CFD simulations for flows in the impeller passages alone.

75

70

65
Angle (beta definition)

60

55

50 Span = 0.0

Span = 0.5 Figure 9: Hexahedra mesh on the diffuser passage (showing two
45 passages).
Span = 1.0

40 Results of New Design

0.00 5.00 10.00 15.00 20.00 25.00 30.00 Simulation results showed the new design did not produce higher
Meridional Length flow rates when coupled to the vaned diffuser, despite the
Figure 7: Diffuser blade angle definition. increased inlet eye diameter. At 100,000rpm, the pressure ratio
drops rapidly as the flow rate is increased and the new design
gives a lower pressure ratio than the original compressor when
the flow rate exceeds 0.228kg/s (Figure 10). Even with the

964
impeller operating at 120,000rpm it is not expected to give higher The main improvement over the old design is the pressure ratio at
mass flow rate compared to the original stage. a lower mass flow rate. At this lower end, both pressure and
A similar trend is observed in the efficiency plot (Figure 12). At efficiency are significantly higher than those of the original stage.
100,000rpm, efficiency drops below the efficiency of the old Unlike compressors in industrial applications, where the flow rate
stage at 0.225kg/s flow rate. of a stage is controlled by a separate system such as an electric
motor, the mass flow rate in a gas turbine depends on a well-
matched turbine, connected by a common shaft, with the mass
2.8 flow increasing proportionally with the rpm. Therefore, with a
Old 100,000 rpm correctly matched turbine, the flow rate could be controlled
2.6 within the optimised range.
Old 120,000 rpm

2.4 New 100,000 rpm

New 120,000 rpm 0.85

2.2
Pressure Ratio

2 0.8

1.8
0.75

Total Isentropic Efficiency

1.6
0.7
1.4

0.65
1.2
New 100,000rpm
1 0.6
New 120,000rpm
0.10 0.15 0.20 0.25 0.30 0.35
Mass Flow Rate [kg s^-1] Old 100,000rpm

Figure 10: Pressure ratio curves for new design compared to the old 0.55
Old 120,000rpm
design.
0.5
The rapid performance degradation at higher flow rates may be 0.15 0.17 0.19 0.21 0.23 0.25 0.27 0.29 0.31
related to the increase in recirculation in the turning section of the Mass Flow Rate [kg s^-1]
diffuser. Figure 11 shows the velocity distribution is more even Figure 12: Efficiency curves of new design compared to ole ones
at the outlet with the lower mass flow rate (b), compared to the
velocity contour for higher mass flow rate (a). Discussion and Conclusion
CFD simulations show the pressure ratio and efficiency of the
compressor stage were significantly enhanced at the operational
flow rate of the engine at corresponding rpm, by utilising a larger
impeller wheel and a rounded turn in the diffuser section.
With larger inlet flow area on the 71mm impeller wheel, the
impeller allows more mass flow. This will allow flat pressure
ratio curves to extend further into higher mass flow regime with a
radial diffuser configuration. However, the larger inlet mean
radius does not help give a higher pressure ratio. Equation (2)
can also be written as
k ( k −1)
⎛ η ⎞
Prtt = ⎜1 +
⎜ C T
(U 2 Cθ 2 − U 1Cθ 1 )⎟⎟ (4)
(a) ⎝ p 00 ⎠
With a larger inlet mean radius, the velocity term U1Cθ1 is
relatively large and therefore reduces the pressure ratio according
to Equation (4). However, as the engine does not require higher
flow rate, as well as the fact the new vane diffuser was not
designed to do so, the inlet hub and shroud diameters can be
reduced. This will allow a higher pressure build up at the
impeller exit.
Figure 13 shows the power requirement of the new stage
compared to the original one running at 100,000rpm and
120,000rpm. At lower flow rates there is no significant increase
in power requirement. Therefore, it is likely the original turbine
of the KJ66 (maximum flow rate of 0.22kg/s at 117,000rpm) will
match acceptably well with this new compressor stage.
Proposed future work includes detailed CFD analysis using k-ω
turbulence model with SST wall model. This will enable studies
(b) of flow separation and recirculation in the diffuser section and
Figure 11: Meridional velocity contours in the diffuser with flow rate (a) their correlation to the overall performance. Also, studies of the
0.228kg/s, and (b) 0.186kg/s, with the impeller operating at 100,000rpm. effect of having diffuser sections of different turning radii will be

965
undertaken to determine whether it is possible to utilise an even [2] R. Sonntag, C. Borgnakke & G. Van Wylen, Fundamentals
larger impeller wheel or reduce the engine’s diameter. A of Thermodynamics, 5th Ed John Wiley & Sons 1998
compressor wheel of 71mm diameter but with lower flow [3] D. Japikse, Centrifugal Compressor Design and
coefficient will also be tested with the rounded diffuser section. Performance, Concepts ETI Inc 1996
[4] D. Japikse & N. Baines, Introduction to Turbomachinery,
Concept ETI 1994
40000 [5] R. Aungier, Centrifugal Compressors: A Strategy for
Old 100k rpm Aerodynamic Design and Analysis, ASME 2000
Old 120k rpm
[6] D. Flaxington & E. Swain. Turbocharger Aerodynamic
35000
Design. Proc Instn Mech Engrs 1999 Vol 213 Part C.
New 100k rpm
pp.43-57
New 120k rpm [7] P. Dalbert, B. Ribi, T. Kmeci & M.V. Casey. Radial
30000
Compressor Design for Industrial Compressor. Proc Instn
Mech Engrs 1999 Vol 213 Part C. pp.71-83
Power (W)

25000 [8] P.M. Came & C.J. Robinson. Centrifugal Compressor

Design. Proc Instn Mech Engrs 1999 Vol 213 Part C.
pp.139-155
20000 [9] A. Lohmberg, M. Casey & S. Ammann. Transonic Radial
Compressor Inlet Design. Proc Instn Mech Engrs 2003 Vol
217 Part A. pp.367-374
15000 [10] J.D Denton & W.N. Dawes. Computational Fluid Dynamics
for Turbomachinery Design. Proc Instn Mech Engrs 1999
Vol 213 Part C. pp.107-124
10000 [11] R. Aghaei Tog, A. Mesgharpoor Tousi & M. Soltani.
0.10 0.15 0.20 0.25 0.30 0.35
Design and CFD Analysis of Centrifugal Compressor for a
Mass Flow Rate [kg s^-1]
Microgasturbine. Aircraft Engineering & Aerospace
Figure 13: Power requirement of the compressor. Technology: An International Journal 79/2 (2007) 137-143
[12] S. Ibaraki, T. Matsuo, H. Kuma, K. Sumida & T. Suita.
Aerodynamics of a Transonic Centrifugal Compressor
References Impeller. Transactions of the ASME J. Turbomachinery
[1] M. Gerendás & R. Pfister, Development of a very small 2003 Vol 125 pp. 346-351
aero-engine. 45th ASME International Gas Turbine and [13] ANSYS CFX, Release 11.0, ANSYS Inc. Southpointe
Aeroengine Technical Congress and Exposition. Procedings December 2006.
of the ASME Turbo Expo 2000.

966