See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228984485

A CFD Suite for Design and Performance Prediction of Single and Multistage
Axial Flow Compressors

Article · August 2007

CITATIONS READS

2 242

2 authors, including:

Carlo Cravero
Università degli Studi di Genova
65 PUBLICATIONS   216 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PO FESR 16 "Ottimizzazione energetica e ambientale del processo di fusione del vetro" View project

Prime Glass Project View project

All content following this page was uploaded by Carlo Cravero on 24 October 2014.

The user has requested enhancement of the downloaded file.

Proceedings of the 8th International Symposium
on Experimental and Computational
Aerothermodynamics of Internal Flows
Lyon, July 2007
Paper reference : ISAIF8-00100

A CFD Suite for Design and Performance Prediction of Single and Multistage
Axial Flow Compressors

Dario Bruna, Carlo Cravero

DIMSET – Università di Genova
Via Montallegro, 1 – 16145 Genova
[email protected]; [email protected]

The development of high performance turbomachinery relies heavily upon the use of Computational Fluid Dy-
namics (CFD). At the early design stage, reduced order modeling methodologies are widely used in order to ex-
plore the design space of a new configuration. The matching of new multistage gas turbine solutions needs a large
number of calculations; different off-design conditions have to be checked with fast computational models. Thus
modern 1-D, 2-D, through-flow and axisymmetric codes are the most appropriate tools to be used for this purpose,
due to their low computational demand. They can also give accurate results if properly tuned. A 3-D Na-
vier-Stokes methodology can then be effectively introduced at later stages in order to check the performance of
the turbomachinery configuration under study. In this paper a complete design/analysis system is described. A
design suite of codes (T-AXI) is coupled to a new blade section modeler (PRFL), while a 1-D Time-Marching
code (COMP1D) and a 2-D axisymmetric Navier Stokes code (NAV-AXI) are used for the design and off-design
performance prediction. The presented methodology has been applied to different turbomachinery configurations.
In the present paper the attention is focused on a single stage transonic compressor and an industrial multistage
compressor.

Keywords: Turbomachinery, CFD, compressor design/analysis, NURBS, PVD airfoil.

Introduction can be fully appreciated in details by a 3-D Na-

vier-Stokes methodology. In fact, as mentioned by
The design and the analysis of high performance tur- Denton and Xu [1], 2-D and Q3D approaches can predict
bomachinery is one of the most demanding and chal- many phenomena involved in turbomachinery, but only a
lenging applications for CFD. In the current industrial 3-D calculation can show the real flow complexity. In
practice is common to routinely perform computational this way the analysis process offers useful information
analysis for the development of new design concepts. that can be used to improve the turbomachinery per-
The three-dimensional nature of the flow field inside formance in a (re)design process. In a detailed review
modern axial flow compressor stages (characterized by Adamcyzk [2] presented how different mathematical
blade sweep, lean, end-bending and endwall profiling) models can be used to simulate turbomachinery flows

Dario Bruna: Ph.D. Student, [email protected]

Carlo Cravero: Professor, [email protected]
2 Proceedings of the 8th International Symposium on Experimental and Computational Aerothermodynamics of Internal Flows

and particularly how the 3-D calculation can help the pressors. An interesting numerical approach, with an
multistage turbomachinery development. The physics improved loss modeling implementation (in order to en-
shown in a 3-D analysis can be understood and the un- sure accurate off-design performance prediction), has
derlying concepts implemented (as lower order model or been more recently presented by Boyer and O’Brien
correlation) in simpler design/analysis tools. In order to [6,7].
explore the design space of a new configuration the re- Here, starting from a 2-D axisymmetric software sys-
duced order modeling tools are widely used in the indus- tem developed in collaboration with the University of
trial practice. Another example of process where quick Cincinnati and the MIT [8] for axial flow turbomachinery
computational tools are of great interest is the gas turbine design, the authors present a complete integrated CFD
matching: a large number of solutions for different suite. A new blade section modeler developed completely
off-design conditions for a complete multistage axial at DIMSET - the PRFL_NURBS code - has been coupled
flow compressor have to be computed. For their quick to a throughflow code. In the second part of the paper the
execution and good overall performance prediction the analysis system, based on a 1-D and a 2-D codes, is pre-
1-D, 2-D, throughflow, streamline curvature and axi- sented. Finally the application of the design/analysis sys-
symmetric codes can be very useful tools. Smith [3], tem has been carried out for two examples of axial com-
Novak [4] and Schobeiri [5] reported how, in last dec- pressors: a single transonic stage and a 17-stage indus-
ades, the streamline curvature methods can be used for trial axial flow compressor.
the meridional plane analysis of multistage axial com-
Nomenclature
a nose width Q conservative variables
b L.E. thickness R rational B-Spline basis functions
c blade chord T.E. trailing edge
h weighting factor Greek letters
k curvature function β pressure ratio
n segments of the NURBS convex hull φ flow coefficient
t NURBS curve parameter, time η efficiency
x Cartesian coordinate ψ load coefficient
y Cartesian coordinate Subscripts
B NURBS curve control points, source term k NURBS curve degree
F flux vector TT total-to-total
L.E. leading edge Superscripts
N B-Spline basis functions • 1st derivate with respect to t
N.R. nose ratio •• 2nd derivate with respect to t

Design System
T-C_DES and T-AXI
The turbomachinery design system here used is the
result of a research collaboration with the University of
Cincinnati and the MIT. The software suite has also been
proposed as an educational tool for teaching design con-
cepts for axial flow turbomachinery [8,9]. A set of files
including the fundamental design parameters is the start-
ing point, then the single or multistage turbomachinery
configuration is automatically obtained. Using the soft-
ware suite freely distributed on the web, the 3-D geomet-
rical model definition and visualization are possible. The
main features of the coupled codes, a 1-D meanline de-
sign (T-C_DES) and a 2-D axisymmetric solver (T-AXI)
have been already presented in previous publications [10].
The reader is addressed to the work by Turner et al. [8]
for the complete and detailed procedure description and
to the work by Bruna et al. for the investigation of a
multistage case-study [9].
Fig. 1 Design system block diagram
D. Bruna, C. Cravero A CFD suite for Design and Performance Prediction of Single and Multistage Axial Flow Compressors 3

The authors are currently coupling the design method- suction side and on the L.E. geometry induced the
ology to a CFD based analysis system (2-D axisymmetric authors to implement NURBS (a more complex and
and a 3-D) including lower order models (1-D) and to a flexible mathematical curve model). In fact even weak
geometrical tool for the blade design based on NURBS, changes in the curvature can influence negatively the
in order to have greater geometry definition capabilities. flowfield, as pointed out in a previous analysis [11].
In fact with the original version of the suite the user
has a limited control over the blade section geometry
definition. Thus the implementation of a powerful blade
section modeler, the PRFL code [11] has been performed.
Figure 1 shows the block diagram of the upgraded design
system coupled to the analysis suite. The PRFL_NURBS
code replaces the original blade section generator, here
referred to as “Geo Q3D” for its capabilities to define the
geometries in the axisymmetric surfaces with streamtube
thickness. The geometries can either be analyzed with the
Q3D flow solver MISES [12,13] or with a 3-D Navier
Stokes code. The in-house NAV3D code is directly in-
terfaced to the 3-D blade geometry, output of T-AXI.

Blade section modeler: PRFL_NURBS

The tool implemented for the blade design is an airfoil
modeler developed at DIMSET [11,14]. It has been con-
ceived for the design of modern blade sections for axial
flow compressor, in a new way with respect to the com- Fig. 2 L.E. detail. Nose Ratio definition
mon turbomachinery practice. In fact the airfoil thickness
is no more distributed over the camberline but the de- In a NURBS curve each curve P is defined with the
signer generates the profile with a direct control over the parameter t included in the range (0-1). Introducing the
main geometrical parameters that have a direct impact on control point coordinates Bi, the rational B-Spline func-
the aerodynamic performance. This in a similar way as tions Ri,k(t) and the weighting factors hi, the mathematical
defined in Ref. [15,16]. With the present tool it is possi- form is:
ble to recreate existing blade profiles or to design new
ones. n +1
Acting on the inlet/outlet metal angles, the wedge an- P(t) = " Bi Ri,k (t) (1)
gles and the local thicknesses the leading edge and trail- i=1
ing edge dimensions are defined. The airfoil is set by the
profile maximum thickness and its location (divided by
the chord); moreover the nose ratio and the pressure side hi N i,k (t)
slope at the T.E., an influencing parameter for PVD air- Ri,k (t) = n +1 (2)
!
foils, let the user to have a fine tuning of the geometry in "h N i i,k (t)
critical locations. In Fig. 2 the nose ratio is shown as ra- i=1
tio of two quantities (a, b).
In this way the blade section with the corresponding
In this second version of the PRFL code both the suc-
control points is fixed. It is then necessary to define a
tion and the pressure side have been defined with only a
mathematical curve model for the suction side and the
couple!of curves. Curve of third degree are used in order
pressure side. In a first code release the authors have
to guarantee the C2 continuity. For the leading edge a 3rd
chosen the Hermite cubic polynomial form: the suction
degree B-Spline is used while the trailing edge is defined
side was defined by three curves and the pressure side
with a circular arc.
only by two curves. In order to complete the geometry
A curvature analysis of the airfoil has been done using
two circular arcs at the L.E. and T.E. have been used.
the following expression in the parametric form:
Already with this curve model, controlling both the posi-
tion and the tangency of the curve end points, it was pos-
sible to accurately define turbomachinery profiles for x˙ y˙˙ " y˙x˙˙
k= (3)
both compressors and turbines [14]. But, as shown in Ref. 2 3
[11], some limitations on the curvature control of the ( x˙ 2
+ y˙ )

!
4 Proceedings of the 8th International Symposium on Experimental and Computational Aerothermodynamics of Internal Flows

A detailed analysis of the suction curvature on the Load (ψ) and flow (φ) coefficient values computed by
PVD airfoils, recently presented by Bruna and Cravero. T-C_DES are coupled to standard stage performance
[11], showed the improvement in the local geometry curves, obtained from open literature [18] and made
definition using NURBS with respect to the previous non-dimensional with respect to design values for the
curve model implemented in the PRFL code. In fact it current stage, as described in Ref. [10].
has been shown that minor oscillations in the profile
curvature (induced by the Hermite parametric form that NAV-AXI
does not guarantee a C2 continuity) have a strong influ- An axisymmetric viscous flow solver based on the
ence on the Mach number distribution, especially in the NAV3D code structure [17] has been developed. It
case of PVD profiles. solves the Reynolds-Averaged Navier Stokes equations
with a time marching approach and an explicit
Analysis System Runge-Kutta multistep scheme. In the present version the
For the understanding of complex structures of inter- code obtains the blade geometry from the same dataset of
nal flows inside a turbomachine a 3-D Navier Stokes the NAV3D solver but the integration of a blade modeler
methodology is nowadays the most well suited analysis is under development. The blade channel is currently
tool [17]. But in the industrial practice simplified ap- modeled using the profile meanline and standard correla-
proaches are also required and intensively used thanks to tions (Carters’ rule and Leiblein’s diffusion factor) are
their low computational resources required. In the fol- used to model the blade performance into the related
lowing section both a 1-D and a 2-D axisymmetric code momentum and energy balance equations. The algebraic
for single/multistage compressor performance prediction turbulence closure from Baldwin-Lomax is used to pre-
are introduced. dict the endwall blockage.
The idea of a highly integrated software system with This preliminary version of NAV-AXI, whose results
reduced external input data required is “driving” the CFD have been compared to the NAV3D calculations for a
suite here presented. multistage industrial axial flow compressor, is currently
under development. In fact a new correlative scheme for
COMP1D total pressure loss prediction, developed by the authors
for modern compressor profiles [11, 19], will be included
The COMP1D code performs a 1-D turbomachinery
into NAV-AXI.
analysis and it is coupled to the 1-D design code
(T-C_DES) that supplies the required geometrical data
and boundary conditions, Fig. 3.
Applications
In the present section a transonic single stage and an
industrial multistage compressors have been chosen in
order to apply the presented CFD design/analysis suite.

Single stage transonic compressor

The E/CO-4 redesign case had been presented with the
original stator [10] and now updated with the UKS-31
stator, based on PVD airfoils. The E/CO-4 is a single
transonic compressor stage developed at DFVLR, docu-
Fig. 3 1-D design/analysis system mented by Dunker [20]. The original configuration was
based on NACA 65 airfoils stator (R 030 – S NACA
The model solves the Euler equations, written in 65/60) and was characterized by a rotational speed of
strong conservative form, with a time marching approach. 20260 rpm, a total pressure ratio of 1.51 and a mass flow
The effects of turbomachinery blades (rotors or stators) rate of 17.3 kg/s at a blade tip speed of 425 m/s. In the
are accounted for in the source term and computed using second configuration the number of blades of the stator
reference stage characteristic curves as described in Ref. row changed from 60 to 31. Using PVD airfoils it was
[10]. A finite volume approach with a multistep explicit possible to guarantee the same efficiency at 100% rota-
Runge-Kutta type scheme, are used to solve the system tional speed and a similar value of the stage pressure ra-
of equations in the following vector form, Eq. (4). tio with a lower solidity. The results of the experimental
test on this stage are reported by Dunker et al. [20, 21,
22].
! Q ! F
+ =B (4) The T-C_DES showed good capabilities to accurately
! t ! x recreate the same meridional channel geometry. In fact
D. Bruna, C. Cravero A CFD suite for Design and Performance Prediction of Single and Multistage Axial Flow Compressors 5

for the hub/shroud lines definition a maximum error of

0.6 mm has been measured, while for the section area the
calculated maximum error is 0.64%. The blades geome-
try has exactly the same axial chord; for this redesign
neither lean nor sweep has been introduced.
Once the meridional channel has been obtained with
the 1-D meanline methodology, the axisymmetric solver
computed the flow-field with the free-vortex assumption.
Table 1 summarizes the performance (pressure ratio and
efficiency) of the redesign compressor with respect to the
design intent and the experimental data. Both T-C_DES
and T-AXI results are reported.

Table 1 E/CO-4 Redesign results

Design
Experimental T-C_DES T-AXI
intent
Fig. 4 Mach distribution comparison on redesigned UKG 30.3
β 1.51 1.55 1.52 1.54

ηTT 0.82 0.84 0.84 0.86

Using the meridional channel geometry, dumped by
the T-C_DES code, the COMP1D code has been used to
predict the performance maps with the same reference
design flow and load coefficients used in the design suite.
Figure 3 shows the geometry of the single stage tran-
sonic with the redesigned UKS-31 stator [20], as mod-
eled with the included blade generator.

Fig. 3 E/CO-4 stage geometry with T-AXI

The PRFL code has been used to redesign three stator

blade sections: hub, midspan and tip. Then the airfoils
have been analyzed using MISES: the boundary condi-
tions set for the calculations respect the original working
conditions reported in Ref. [21]. Figure 4 shows the
Mach number comparison with respect to the design in-
tents and the experimental data for the midspan section
(UKG 30.3).
Fig. 5 E/CO-4 compressor map comparison
(dot: experimental, line: COMP1D)
6 Proceedings of the 8th International Symposium on Experimental and Computational Aerothermodynamics of Internal Flows

The compressor map for different rotational speed (55 Table 2 V64.3A Redesign results
- 100% with respect to the design value) has been calcu- T-C_DES T-AXI
lated. Design and off-design working conditions have Δβ +6.64% -1.85%
been simulated and then the values compared to available
ΔηTT -1.25% +0.45%
experimental results [20, 21, 22].
The comparison in Fig. 5 shows two different results.
In the range of 85-100% rotational speed the code pre- Moreover the NAV-AXI has been used for a prelimi-
dicts the performance (as pressure ratio) with acceptable nary analysis of this multistage configuration. A com-
accuracy. But at lower rotational speeds for both accu- parison between the results from the axisymmetric and
racy and predicted operating range there is a loss of fi- the fully Navier Stokes codes is shown in Figs. 7 and 8.
delity with respect to the available data. The stage effi- The predicted mass-averaged distributions for tempera-
ciency is well predicted at peak values but the curve ture and pressure along the turbomachinery at each trail-
slope is not captured giving a wider range of high effi- ing station are in good agreement with those computed
ciency, for every rotational speed, as result. The authors with the NAV3D code. The above distributions have
believe that the difficulties encountered are to be attrib- been made non-dimensional with respect to the values at
uted to the reference stage performance curves [18] im- the outflow for industrial reasons.
plemented in the code. In fact these curves do not corre-
spond to the modern highly loaded transonic stage but
had been defined for moderately loaded compressor
stage.

Industrial multistage compressor

The industrial multistage compressor here presented is
part of the Ansaldo V64.3A gas turbine. As already pre-
sented in Ref. [10] the turbomachine has been redesigned
using T-AXI, and the first stage analyzed with both the
COMP1D code and the NAV3D. The results from the
two analysis codes have been compared in order to un-
derstand the applicability of the standard performance
curves [18] in the modeling of modern axial flow com-
pressor stages. In Fig. 6 the complete multistage geome- Fig. 7 Temperature level comparison
try, as calculated by T-AXI is reported.

Fig. 8 Pressure level comparison

Fig. 6 V64.3A multistage compressor geometry with T-AXI
Conclusions
A summary of the performance calculated with the
The CFD suite for design and analysis of axial flow
different methodologies is included in table 2. The dif-
compressor here presented has shown the capabilities to
ferences between predicted and design values for the
accurately reproduce an existing meridional channel ge-
compressor pressure ratio (Δβ) and efficiency (ΔηTT) are
ometry, to obtain a complete 3D geometry configuration
reported.
of the turbomachine (including a fine tuning of critical
airfoil geometry aspects) and a good overall performance
 D. Bruna, C. Cravero A CFD suite for Design and Performance Prediction of Single and Multistage Axial Flow Compressors 7

evaluation for single and multistage axial flow compres- [11] D. Bruna and C. Cravero: Modeling the Aerodynamic
sors. The application of the software to reference cases Performance of Modern Axial-Flow Compressor Profiles:
has highlighted some weaknesses of the modeling (espe- A Correlative Approach Using Current CFD Technology,
cially stage performance maps) that will be improved. Proc. 45th AIAA Aerospace Sciences Meeting and Ex-
Work is in progress for the implementation of a blade hibit, Reno, NV, USA, January 8-11, (2007).
modeler that will allow a high flexibility in geometry
[12] M. Drela: Two-Dimensional Transonic Aerodynamic
modeling and a direct data exchange between the differ-
Design and Analysis Using the Euler Equations, MIT
ent codes in the suite.
Aero-Astro, Cambridge, MA, (1985).
References [13] M.B. Giles: Newton solution of Steady Two-Dimensional
Transonic Flow, MIT Aero-Astro, Cambridge, MA,
[1] J. D. Denton and L. Xu: The Exploitation of
1985.
Three-Dimensional Flow in Turbomachinery Design,
[14] D. Bruna: A complete CFD Suite for the Design and the
Proc., IMechE, J. Mechanical Eng. Sci, vol. 213 Part C,
Analysis of Three-Dimensional Blades for Axial Flow
pp. 125-137, (1999).
Compressor, M.S. thesis (in Italian), DIMSET, Genoa,
[2] J. J. Adamcyzk: Aerodynamic Analysis of Multistage
Italy, (2004).
Turbomachinery Flows in Support of Aerodynamic De-
[15] U. Köller, R. Mönig, B. Küsters, H.-A. Schreiber: De-
sign, ASME Journal of Turbomachinery, vol. 122, pp.
velopment of Advanced Compressor Airfoils for
189-217, (2000).
Heavy-Duty Gas Turbines - Part I: Design and Optimiza-
[3] L. H. Smith Jr.: The Radial-Equilibrium Equation of
tion”, ASME Journal of Turbomachinery, Vol. 122, pp.
Turbomachinery, ASME Journal of Engineering for
397-405, (2000).
Power, pp. 1-12, (1966).
[16] U. Köller, R. Mönig, B. Küsters, H.-A. Schreiber, “De-
[4] R. A. Novak: Streamline Curvature Computing Proce-
velopment of Advanced Compressor Airfoils for
dures for Fluid-Flow Problems, ASME Journal of Engi-
Heavy-Duty Gas Turbines - Part II: Experimental and
neering for Power, pp. 478-490, (1967).
Theoretical Analysis”, ASME Journal of Turbomachin-
[5] M. Schobeiri: Turbomachinery Flow Physics and Dy-
ery, Vol. 122, pp. 406-415, (2000).
namic Performance, Chap.11, Springer, Germany,
[17] C. Cravero: Viscous analysis of turbine blades with
(2005).
three-dimensional design, 4th European Conference on
[6] K. M. Boyer, W. F. O’Brein: An Improved Streamline
Turbomachinery Fluid Dynamics and Thermodynamics,
Curvature Approach for Off-Design Analysis of Tran-
Florence, March 20-23, (2001).
sonic Axial Compression Systems, 2002-GT-30444, Proc.
[18] A. Stone: Effects of Stage Characteristics and Matching
ASME Turbo Expo, Amsterdam, The Netherlands, June
on Axial Flow Compressor Performance, ASME Trans-
3-6, (2002).
actions, vol. 80, pp. 1273 - 1293, (1958).
[7] K. M. Boyer, W. F. O’Brein: Application of an Improved
[19] D. Bruna, C. Cravero, M. G. Turner: The Development of
Streamline Curvature Approach to a Modern, Two-Stage,
an Aerodynamic Performance Prediction Tool for Mod-
Transonic Fan: Comparison with Data and CFD,
ern Axial Flow Compressor Profiles, 2006-GT-90187,
2002-GT-30383, Proc. ASME Turbo Expo, Amsterdam,
Proc. ASME Turbo Expo, Barcelona, Spain, May 8-11,
The Netherlands, June 3-6, (2002).
(2006).
[8] M. G. Turner, A. Merchant, D. Bruna: A Turbomachin-
[20] R. Dunker: Test Cases for Computation of Internal Flows
ery Design Tool for Teaching Design Concepts for Ax-
in Aero Engine Component – VI.4 Test Case E/CO-4,
ial-Flow Fans, Compressors, and Turbines,
AGARD-AR-275, pp. 245-285, (1990).
2006-GT-90105, Proc. ASME Turbo Expo, Barcelona,
[21] R. Dunker, H. Recther, H. Starken, H. Weyer: Redesign
Spain, May 8-11, (2006).
and Performance of a Transonic Axial Compressor Stator
[9] D. Bruna, C. Cravero, M. G. Turner, A. Merchant: An
and Equivalent Plane Cascades With Subsonic Controlled
Educational Software Suite for Teaching Design Strate-
Diffusion Blades, ASME Journal of Engineering for Gas
gies for Multistage Axial-Flow Compressors,
Turbine and Power, vol. 106, pp. 279-287, (1984).
2007-GT-27160, Proc. ASME Turbo Expo, Montreal,
[22] R. J. Dunker, H. G. Hungenberg: Transonic Axial Com-
Canada, May 14-17, (2007).
pressor Using Laser Anemometry and Unsteady Pressure
[10] D. Bruna and C. Cravero: Design and Performance Pre-
Measurements, AIAA Journal, vol.18, pp. 973-979,
diction of Multistage Axial-Flow Compressor Using In-
(1980).
tegrated Computational Strategies, ATI Perugia, Italy,
September 14-17, (2006).

View publication stats