Mechanics & Industry 20, 107 (2019)

© AFM, EDP Sciences 2019 Mechanics

https://doi.org/10.1051/meca/2019004 &Industry
Available online at:
www.mechanics-industry.org

REGULAR ARTICLE

A one-dimensional stage un-stacking approach to reveal flow

angles and speeds in a multistage axial compressor at the design
operating point
Fu Hai Alan Koh1,2 and Yin Kwee Eddie Ng1,*
1
School of Mechanical and Aerospace Engineering, Nanyang Technological University (NTU), 50 Nanyang Ave,
639798 Singapore, Singapore
2
Lloyd’s Register Global Technology Centre, 1 Fusionopolis Place, #09-11 Galaxis, 138522 Singapore, Singapore

Received: 30 April 2018 / Accepted: 23 January 2019

Abstract. Stage stacking methods commonly use a one-dimensional (1D) through flow analysis at the mean line
to design individual axial compressor stages and stack these to form a multistage axial compressor. This phase of
design exerts a great influence on each stage’s pressure and temperature ratio. The design process for an
individual stage is usually guided by design values and rules developed in previous designs. This study develops a
1D stage un-stacking method (SUSM), which uses a minimal set of data from an actual axial compressor, while
reducing the needed number of assumptions. Proceeding from the premise that an actual axial compressor design
fulfills all thermodynamic requirements, velocity triangle requirements and design guidelines simultaneously,
this proposed SUSM calculates the pressure, temperature, velocities and flow angles as a set of dependent data at
each stage of the axial compressor. In approximating a possible axial compressor design for the LM2500 gas
turbine that achieves the known pressure ratio distribution, the suggested stage loading coefficient (SLC)
distribution is more appropriately considered an initial well-informed estimate and further improvements to this
SUSM are needed to infer the actual SLC distributions used.
Keywords: Multistage axial compressor / stage un-stacking approach / velocity triangle / flow design /
aero-derivative gas turbine

1 Introduction the blade rows are unwrapped from the rotational axis), 2D
streamline curvature, 1D analysis of radial variation across
The need for greater efficiency drives each new gas turbine the blade span resulting in span-wise blade twist angles and
model towards higher overall pressure ratios and power three-dimensional (3D) analysis to simulate the challeng-
outputs, inevitably keeping the gas turbine relevant. The ing actual turbulent flows at the blade root and tip (which
design ideas implemented in actual gas turbines are contribute heavily to losses). Despite its relative simplicity,
complex and delicately optimized across various techno- the 1D analysis at the mean line exerts great influence on
logical aspects to achieve performance, reliability and cost- the design of an axial compressor because, at this design
effective maintenance. The compressor is often the axial phase, each stage’s pressure and temperature are defined
flow design, to pass higher mass flow rates through a before detailed design work begins and assembled together.
relatively smaller frontal area and achieve generally higher In 1D analysis through the turbomachine, while the
stage pressure ratios with lower losses compared to the corrected mass flow rate is examined for the effects of
centrifugal design. With each successive design delivering pressure, Cumpsty [1] demonstrated that obtaining correct
higher overall pressure ratios, the relevance of the stage stacking or matching will be challenging, because the
multistage axial flow compressor is not diminishing. mass flow rate, effective flow area and pressure are in an
The design process for an axial compressor from simple intricate relationship. This stage matching challenge is
to challenging are: one-dimensional (1D) analysis at the further compounded after incorporating inter-disciplinary
mean line, two-dimensional (2D) cascade analysis (where aspects into the design space. The gas turbine developed
traditionally along distinct components lines and has been
tremendously refined to achieve component efficiencies
* e-mail: mykng@ntu.edu.sg over 90%. The design space is now arguably more complex
2 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

owing to more disciplines, including inter-disciplinary This paper is divided into the following Sections. The
work. Ghisu et al. [2,3] have developed an integrated design literature is reviewed in Section 1.1. In Section 2, the SUSM
approach using a 1D mean line solver embedded in an is described. The results of stage un-stacking and testing on
optimisation routine, to defer fixing the components an approximated 3D model are presented in Section 3.
interface design parameters till later phases in the design Section 4 discusses the results and is followed by
process, resulting in better explorations of the design space conclusions.
and hence harnessing the gains from trade-offs between
different components and disciplines. From the observa- 1.1 Literature
tions of Jarrett and Ghisu [4], the search for an optimised
balance between time spent on configuration selection and The literature contains a number of 1D stage stacking
refining a selected configuration during the design process methods for designing an axial compressor from individual
reveals that in the best designs, configuration selection stages and stage stacking these together to form a
with 1D mean line solvers consumes half to three-quarters multistage axial compressor.
of the design time.
The stage stacking process is the core of 1D analysis at 1.1.1 RSRR approach
the mean line. The stage stacking approaches recorded in
the literature focus on building up a compressor stage by The repeating-stage repeating-row (RSRR) model from
stage, with stage temperature and pressure (or equivalent Mattingly [8], also introduced in an aircraft engine design
information) available. For the gas turbine manufacturers, book by Mattingly et al. [9], is one of the simplest design
the design process may be guided by design values and rules approach and therefore provides a suitable initial design.
developed in previous compressor designs. Sehra et al. [5] However, the constant mean line in this model often does
apply existing knowledge and design techniques from an not match that in actual compressors and actual
aviation gas turbine to design the compressor of a utility compressor stages are often not repeating as seen in the
gas turbine. Smed et al. [6] report evolving the design of clearly varying stage axial velocity.
compressor within a family of gas turbine models. Smith [7]
begins unifying compressor models into families based on 1.1.2 STGSTK code
similar design rules. Mattingly [8] shows that the design
process is often iterative as a multitude of performance The STGSTK code by Steinke [10] is an early code used to
requirements must be fulfilled simultaneously. In the light predict the off-design performance of an axial compressor-
of this, the stage flow angles, flow speeds, stage character- based performance at the design operating point. The
istics, among other variables are then the inputs to the stage analysis is performed at the mean-line with velocity
stacking method. The ability of the blades to maintain un- triangles at the rotor inlets and outlets. The compressor
separated flow at reduced or excessive mass flow rate within a is built up cumulatively through stage stacking to obtain
range of off-design flow angles is often summarized as a the multistage design and overall performance. However,
model, which is usually empirical, derived from experimental the few critical parameters that build the compressor stage
data or from performance data of a previous related design. by stage are required inputs to STGSTK; therefore, the
When tasked to determine the stage details of a multi- STGSTK code is unable to provide guidance on the stage-
stage axial compressor designed by others, the current to-stage variation of these critical parameters.
methods require a large number of inputs, which unfortu-
nately are not known with certainty. While estimates and 1.1.3 LUAX-C code
design guides may be helpful sources of input design
variables, there are few systematic ways to un-stack a A more recent 1D steady state operation stage stacking
multi-stage axial compressor, other than iteratively testing model is the LUAX-C by Falck [11]. This model is under
with a range of input values for each design variable. The active development again in 2013 by Perrotti [12]. This
available information on an axial compressor is usually the model is much more flexible than the RSRR model and has
overall pressure ratios and the overall efficiency, but do not the potential to obtain most of the geometric, thermody-
mention how flow angles and flow speeds relate within the namic and flow conditions in each blade row of each stage.
machine. The aim of this paper is thus to present a stage un- A number of experimentally based enhancements are
stacking method (SUSM) that uses minimum information, incorporated, such as empirical relations for incidence
and applies a feasible relationship between adjacent stages angles and deviation angles at each blade row, blade profile
temperature ratio and pressure ratio to infer the flow angles losses and endwall losses. However, this model is not used
and flow speeds at each stage for the axial compressor as it required stage solidity and stage reaction as inputs.
operating at the design point. Another required input which discouraged use is the
Presently, this method is restricted to operating inflexible distribution of stage loading.
conditions at a compressor’s design point only. While 2D
streamlines curvature methods are the common approach 1.2 Deliberately working with limited data
to define the blade geometry and flow angles at each axial
stage in relation to the next axial stage, this paper details a Stage stacking methods that use more inputs generate
1D approach that trades off calculating for realistic flow more feasible designs and are capable of more realistic
physics at blade surfaces for quicker calculation of the performance predictions. This is not a challenge for the
mean flow variables. knowledgeable original equipment designer. However, the
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 3

reminder of the non-designers looking to analyse the axial

compressors and have no intimate access to one, the
available information tends to be sparse. A few pieces of
necessary information for the studied axial compressor are
taken from Pedersen [13]:
– PWSD load on power turbine (kW);
– PS2 static pressure at the compressor outlet (Pa);
– T2 stagnation temperature at the compressor outlet (K);
– W2 mass flow rate (kg s
1);
– NGG shaft rotational speed (RPM).

1.3 Supporting information

To supplement the insufficient information gathered in

Section 1.2, further inputs for stage un-stacking an axial
compressor are sought from only publicly available and
conveniently accessible information. For example, approxi-
mate radius and axial station coordinates are derived using
the following coarse estimation approach to reduce reliance
on detailed data. The radius of the casing and hub and axial
stations are estimated from a compressor schematic on the
gas turbine manufacturer’s marketing datasheet [14]. No
further inputs are sought from either the original equipment
manufacturer or equipment owners, so that the proposed
method is also applicable for preliminary analysis of other
designs and by students without access to extensive libraries.
For the axial compressor examined here, there is much more
information available from Klapproth et al. [15] and Wadia
et al. [16]. To develop a robust method capable of estimating
a compressor’s mean line performance close to the actual,
information in Klapproth et al. [15] and Wadia et al. [16] are
not used to develop the SUSM but only in validating its
effectiveness.
Fig. 1. Overview of the stage un-stacking method. Part 1.
Thermodynamic model: Sections 2.10.1 and 2.10.2. Part 2. Axial
2 Methods variation model: Sections 2.2 and 2.10.3; Sections 2.3 and 2.10.4;
Section 2.4; Sections 2.5 and 2.10.5. Part 3. Velocity triangle
Stage stacking assembles individual stages at the design model: Sections 2.6 and 2.10.6.
operating point, where the incidence angle at the blade
leading edge is small, the flow of the working fluid follows
the curvature of the blade and the velocity triangles at the 2.2 Stage loading coefficient (SLC) design rule
stage outlets are very similar to the velocity triangles at the
Between the compressor inlet and outlet, the SLC indicates
following inlets. the amount of energy imparted to the flow through the
specific stagnation enthalpy rise at each stage. Since specific
2.1 Overview of the stage un-stacking method stagnation enthalpy is known only at the compressor inlet
and outlet, an SLC model is needed to suggest a feasible axial
The SUSM is developed in three main parts and works distribution of SLC at the compressor’s design operating
according to the flowchart of Figure 1. Since an actual gas point. The maximum feasible specific static enthalpy rise in a
turbine fulfills simultaneous requirements, iterative calcu- stage Dh may be approximated by equation (29) in Bullock
lations are used to match the flow quantities throughout and Prasse [17], with subscripts included for clarity.
the gas turbine. For parts 1–3 of the model, after the inlet

guide vanes (IGV) outlet and outlet guide vanes (OGV) 2sV REL:1 U WHEEL V REL:2
Dh ¼ Dþ
1 : ð1Þ
inlet angles are selected as iteration variables, the J V REl:1
remaining uncertain design variables are allowed to vary
within acceptable bounds. The two largest unknowns are The diffusion factor D is defined by equation (13) in
found in the axial variation model (part 2): the axial stage Lieblein et al. [18].


loading coefficient (SLC) and the static pressure ratio V REL:2 DV u
distribution. The axial variation of these and other design D¼ 1
þ ð2Þ
V REL:1 2sV REL:1
variables defined within the axial variation model are
presented next. where DV u ¼ jvREL:2
vREL:1 j and v ¼ jvj .
4 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

In early compressors, Bullock and Prasse [17] reported distribution, dividing the overall pressure ratio into near-
that the compressor was also designed for reduced outlet equal stage pressure ratios or near-equal diffusion processes
axial velocity so that excessive or abrupt deceleration for all stages.
before the combustor was avoided. This meant that the Based on the compressor’s pressure rise and tempera-
rear stages must use reduced axial velocities and would see ture rise, a corresponding overall specific entropy rise is
smaller specific static enthalpy rises than the front stages. already incurred. After apportioning the compressor
In recent combustor designs, the diffuser design incorpo- pressure rise and temperature rise nearly equally across
rated after the combustor inlet has improved greatly, all stages, each stage is able to see a small but unavoidable
incurring acceptable stagnation pressure losses while specific entropy rise.
slowing down the flow. This has removed the need for
the compressor to produce greatly reduced axial speed for 2.4 Axial velocity design rule
the combustor which Mattingly et al. [19] demonstrated in
the design approach for the combustor. For minimal buildup of wakes and boundary layers to
The RSRR compressor design approach in Mattingly maximise effective flow area for greater mass flow rates,
[8] treats the diffusion factor D as a design variable. This Bullock and Prasse [17] point out the need to minimise
gives the designer greater flexibility to distribute SLC more abrupt changes at the mating surfaces of the stage inlets
evenly throughout the compressor and one feasible axial and outlets. Implementing this guide, the casing and hub
distribution of SLC is defining the specific stagnation walls are constructed to vary smoothly from compressor
enthalpy rise as a fixed proportion of U 2WHEEL , resulting in inlet to outlet so that changes to the boundary layers and
constant SLC. then wakes are gradual. The axial velocity distribution
The overall design of the engine is optimised for cost model takes in the resulting smoothly varying cross-section
and weight saving. The compressor is of no exception and areas and also requires that the axial temperature and axial
therefore is likely close to the optimum least weight when pressure distributions return a density distribution that is
finalized, as Smith [7] points out. From his wealth of design varying smoothly. This results in a smooth variation of
experience, Smith [7] emphasizes the importance of loading axial velocity for use in further analysis with velocity
each stage appropriately through advising designers and triangles.
designs to work with proven loading criteria achievable
through practical mechanical clearances. 2.5 Axial blockage design rule
When operating at a compressor’s design point, this
study uses an equal or near equal distribution of SLC, with Due to boundary layers and wakes accumulating from
the following arguments. stage to stage, the effective flow area at each stage
– An axial compressor may improve its overall pressure gradually decreases. While the blockage in the compressor
ratio by adding more stages. Each added stage increases increases, the ideal distribution must be smoothly changing
weight and machine complexity such that the design so that the available flow area is able to give a smooth axial
must extract maximum useful output from any stage. velocity profile. This works in tandem with Smith’s [7]
Consequently, each stage is then designed applying the advice of removing all forward facing steps and obtaining
same utmost improvements in aerodynamic insight. surface finishes to appropriate smoothness. This study
– An axial compressor may also increase its pressure ratio includes the effect of viscosity as an additional increase in
through increasing the individual stage pressure ratio via blockage over mechanical blockage from the rotor and
improved aerodynamic insight and design of the blade stator blades.
rows. Since constraints on weight and machine complex-
ity demand the least number of stages, each stage 2.6 Velocity triangles design rule
receives the same improved aerodynamic insight.
– When no information is available, weight and machine The velocity triangle rule is implemented in part 3 of the
complexity constraints do indicate that each stage shares SUSM. At the design operating point, a compressor is
the compression burden. IGV and OGV stages add and performing at its ideal state aerodynamically, based on the
remove swirl respectively and must be presented so the advice of Smith [7] that the blades are uniquely designed for
inlet stage and the outlet stage of the compressor are able the design operating point. The working fluid follows the
to impart the same amount of work on the working fluid curvature of the blades with minimum deviations from the
as the other stages as explained in Mattingly [8]. design intention. For minimum variation in flow angles
between the exit plane of a blade row to the inlet plane of
the next blade row, it would be a reasonable argument that
2.3 Pressure ratio design rule the outlet velocity triangle of a blade row is the same or
While a compressor is designed to be highly efficient, there very similar to the inlet velocity triangle of the next blade
will be losses and specific entropy rise across each stage, row for the following reasons. This minimizes the onset of
stemming from irreversibility in compression. Considering flow separation in the adverse pressure gradient on the
the need to minimize weight and complexity again, the suction side of each compressor blade and that in turn
relationship between the stage pressure ratios would be minimizes the onset of stall and maximizes the diffusion
similar to that for SLC; each stage bears a similar burden. taking place within adjacent blades to give maximum
A feasible axial distribution of pressure ratios when compressor efficiency. To begin solving for the stage inlet
operating at a compressor’s design point is a near-equal and outlet velocity triangles, the straight forward RSRR
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 5

design guideline is used to initialise the design process. This

initial design assumption results in similar flow angles at
the stage inlet and outlet. As the compressor design
matures, the stage inlet and outlet flow angles variations
are however allowed.
In the studied compressor, the axial velocity and the
wheel speed are not constant throughout the compressor,
implying varying velocity triangles. Wheel speed is a
product of high pressure spool rotational speed and the
mean radius or Eulerian radius at each stage. Each stage
has been designed with a different Eulerian radius; hence,
wheel speed is not constant. Attempting to obtain similar
absolute velocity and relative velocity flow angles at the
stage inlet and outlet is too stringent. Among the many
possible departures from the RSRR design guide, this study Fig. 2. Velocity diagram at the rotor inlet and rotor outlet.
retained only the design rule of similar relative velocity flow
angles at the stage inlet and outlet, until minor changes are sufficient for the SUSM. The diffusion factor is a detailed
necessary to achieve a converged set of flow angles for all design criterion requiring further compressor details
blades in all stages. which may not be available.
5. The inlet velocity triangle of a stage is very similar to the
outlet velocity triangle of the previous stage.
2.7 Iteration variables
Ready with all the inputs in the categories listed above,
there are still two unknown flow angles that influence the 2.9 Stages without variable guide vanes
flow speeds within the compressor. As the IGV flow
turning angle is unknown, the absolute inlet flow angle of As greater compression takes place in the high pressure
the first stage is also unknown. The other unknown angle stages where there is typically no VGV, the SLC and flow
is the absolute outlet flow angle at the last stage of angle relationship is examined next. At non-VGV stages,
compressor before the OGV, which is also referred to as the angles of the velocity triangles at both the rotor inlet
the absolute OGV inlet angle. These flow angles will be the and the rotor outlet need to be fairly consistent across a
iteration variables to explore the design space of flow range of rotational speeds, v to achieve the intended
angles for each stage. When a combination of IGV flow pressure rise. This implies at the inlet and outlet stations
turning angle and absolute OGV inlet flow angle returns a of a blade row, the ratio of absolute speed VABS:relative
set of flow angles, from all the stages, where the differences speed VREL:wheel speed UWHEEL in Figure 2 must remain
between inlet velocity triangles and outlet velocity similar to maintain velocity triangles of similar propor-
triangles are acceptably small, this set of flow angles tions when varying rotational speed.
and their corresponding flow speeds is considered a The Euler whirl equation, eWHIRL is
converged solution of the SUSM. eWHIRL ¼ DhTOT:STG ¼ vr2 j~
v 2 j
vr1 j~
v 1 j: ð3Þ

2.8 Design guidelines for uncertain information Combining the Euler whirl equation and SLC, c for an
axial stage gives
To reduce assuming a fixed value for uncertain informa-
tion, the following minimum design guidelines are DhTOT:STG vr2 j~v 2 j
vr1 j~
v1j
c¼ ¼ : ð4Þ
implemented. ~ WHEEL j
jU 2 ~ WHEEL j
jU 2
1. The axial flow speed, outer and inner radii are smooth
varying in the axial direction to reduce boundary layer ~ WHEEL j at station 1 as
Defining wheel speed jU
build up. ~
jU WHEEL:1 j,
2. The stator and rotor blades are likely to have similar
camber angles at the initial design stage. Since flow is vr2 j~
v 2 j
vr1 j~
v1j vr2 j~
v2j vr1 j~
v1j
more energetic across the rotor, the rotor blade is c¼ ¼
: ð5Þ
~
jU WHEEL:1 j 2 ~
jU WHEEL:1 j 2 ~
jU WHEEL:1 j2
allowed more camber than the stator blade as the
calculation progresses.
3. Stage reaction is initially assumed at 0.5 and is allowed Using the wheel speeds at stations 1 and 2, and noting
to vary between 0.0 and 1.0. that v is common since the same shaft is used,
4. The de Haller number is calculated at each stage and
checked against the historical achievement of >0.72, as ~ WHEEL:1 j jU
jU ~ WHEEL:2 j
used in Falck’s [11] design approach. Saravanamuttoo v¼ ¼ ð6Þ
r1 r2
et al. [20] discusses the de Haller number as only an
initial design criterion. Due to the lack of more intimate a relationship between wheel speeds at different stations of
machinery details in this study, the de Haller number is varying radius is found
6 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

~ WHEEL:1 j ¼ r1 ~
jU jU WHEEL:2 j: ð7Þ 2.10.1 Inlet guide vane model
r2
Before setting up the thermodynamic model of the
SLC becomes compressor, the IGV is treated separately as it only turns
r22 vr2 j~v2j vr1 j~
v1j the flow. As the IGV is a flow device with no moving parts
c¼
and only smooth, gently curving walls, the blockage
~ WHEEL:2 j2 jU
r21 jU ~ WHEEL:1 j2 estimated to accumulate at the end of the IGV is assumed
ð8Þ at 1.00%, due to boundary layer growth. From the same gas
r2 j~
v2j j~
v1j
¼ 22
: turbine manufacturer, Holloway et al. [22] report a
~ WHEEL:2 j jU
r1 jU ~ WHEEL:1 j blockage of 3% at the first rotor inlet when designing a
highly loaded 10-stage axial compressor with an overall
The components of absolute speed gives a relation pressure ratio of 23 for aviation. With no further
between j~
u 1 j and j~
v 1 j. information, a small 0.01% of the entire specific entropy
rise is assumed to take place in the IGV, with the remainder
j~ ~ ABS:1 jcos a1 ¼ j~
u 1 j ¼ jV
v1j
: ð9Þ
through the compressor. The IGV is also assumed to incur
tan a1 no flow losses given its short length and smooth gradual
flow turning angle. The compressible Bernoulli equation in
~ WHEEL:1 j
v 1 j and jU
The subsequent analysis is to relate j~ equation (14) is one of the governing equations for the IGV,
with both flow angles a and b. Using the two triangles in selected for its conservation of specific stagnation enthalpy.
Figure 2 found above and below the axial velocity vector,

the wheel speed, jU ~ WHEEL:1 j, is related to the axial speed, g1 P1 1 2
þ ðu þ v21 Þ
u 1 j by equation (10).
j~ g 1
1 r1 2 1


g2 P2 1 2 1
¼ þ ðu þ v22 Þ þ kLOSS u21 : ð14Þ
j~
u1j 1 g 2
1 r2 2 2 2
¼ : ð10Þ
~WHEEL:1 j
jU tan b1 þ tan a1
Station 1 is the IGV inlet while station 2 is the IGV
outlet. The loss coefficient kLOSS is assumed zero but should
Removing j~ u 1 j and introducing j~
v 1 j with equation (9) at it be needed, the reference velocity is u1 instead since this is
station 1 gives the only known velocity and v1 is zero at the inlet. The
other governing equation is the second law of thermody-
j~
v1j tana1
¼ : ð11Þ namics, where specific entropy is determined as a function
~WHEEL:1 j
jU tanb1 þ tana1 of temperature and pressure.

DsIGV ¼ s2
s1 ¼ s2 ðT 2 ; P 2 Þ
s1 ðT 1 ; P 1 Þ ≥ 0: ð15Þ
Similarly at station 2,
Due to flow turning, increasing rotational speed and
j~
v2j tan a2 maintaining mass flow rate, the kinetic energy increases
¼ : ð12Þ
~
jU WHEEL:2 j tan b2 þ tan a2 at the expense of thermal energy and pressure through
the IGV.
Returning to SLC in equation (8), SLC becomes
2.10.2 Thermodynamic model
r22 tan a2 tan a1 The unknown inlet conditions of the compressor are
c¼
: ð13Þ estimated as 100 000 Pa and 278.15 K. At the compressor
r21 tan b2 þ tan a2 tan b1 þ tan a1
outlet, there is only information on stagnation tempera-
ture and static pressure. This information is transferred
Within a non-VGV stage, equation (13) indicates that onto the OGV inlet, equivalent to one stator blade row
SLC depends on the flow angles and the variation between ahead of the compressor outlet. The outlet static
g 1 and g 2. Since in rear high pressure stages with no VGV temperature is determined at the OGV inlet instead,
capability, flow angles need to remain constant at varying through iteration with the help of the estimated velocity
rotational speed, v, SLC is necessarily constant when components ~ u and~v (u and v). The orientation of ~
u and~
v is
varying v at each non-VGV stage. shown in Figure 2.
Viscosity and the shed wakes build up throughout the
2.10 Building a basic axial compressor compressor, increasing blockage for the axial flow. The
minimum blockage accumulated at the end of the
The compressor performance data from Pedersen [21] at compressor is 10.0%, based on the available flow area in
the highest power output is considered the design point for Pedersen [21] and the blockage model then defines the
this axial compressor and used for inferring a possible blockage value B as 90%.
compressor design. In the following sections, the inferred At the OGV inlet, v depends on the stator outlet flow
design is studied by examining the effects of implementing angle, which is also an unavailable piece of information.
various design rules. However, v may be efficiently defined as a fraction, ku of
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 7

wheel speed from equation (16). The form of the SLC model based on constant
specific static enthalpy rise shares the same inspiration
n ¼ ku vrE ð16Þ as the constant temperature rise SLC model and
supported by a rule of thumb for constant stage energy
E is the Eulerian radius
whereprffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi at each blade row and rise in [24]. Specific static enthalpy is the variable as this
rE ¼ 0:5ðrCASE 2 þ rCORE 2 Þ. The outlet static tempera- is also common in turbomachinery analysis.
ture T is found through the outlet stagnation temperature. 4. Varying (decreasing) SLC across the stages: the SLC
With OGV inlet temperature and pressure determined, the at the last stage is a percentage lower than the SLC at
specific entropy rise in the whole compressor is calculated. the first stage, with SLC varying linearly across all
the middle stages. Using the operating point with the
highest power output, equation (21) gives the specific
2.10.3 Stage loading coefficient model
stage stagnation enthalpy rise in terms of the design SLC
In this SUSM, four possible SLC distribution models are of each stage, cDS,STG. Using the specific stagnation
considered. There is no preferred model as each gas turbine enthalpy rise for the full compressor, equation (22)
has its unique heritage and possibly additional stages were determines the design SLC of each stage.
designed differently as evident in the account by Smith [7]
for a gas turbine manufacturer. DhTOT:STG ¼ cDS:STG ðvrE:STG Þ2 ð21Þ
1. Constant SLC for non-VGV stages and constant SLC for
VGV stages: at each level of power output, the specific STG
XMAX
stagnation enthalpy rise through the compressor, DhTOT:COMP ¼ DhTOT:STG : ð22Þ
DhTOT.COMP with corresponding rotational speed v is STG¼1
used in equation (17) to determine the SLC, c,
distribution at that power level. This decreasing SLC model was inspired by [25] where
the front stages are deliberately highly loaded. An
STG
XMAX increasing SLC distribution may be possible too.
2
DhTOT:COMP ¼ cv rE:STG 2 : ð17Þ To determine the SLC distribution, the axial and
STG¼1 rotational components of the absolute velocity must be
known. However, these are found only after the velocity
Using the set of data from Pedersen [13] with 11 triangle analysis. Therefore, a more comprehensive
power output levels and thermodynamic data from solution requires iteration. To test the robustness of the
Cengel [23], the third highest power output gives the solution procedure, the SLC in equation (4) is approxi-
largest constant c at all stages. This is considered the mated by arguing that the specific stage stagnation
design SLC, cDS. enthalpy rise is similar to the specific stage static enthalpy
2. Constant temperature rise for all stages: based on the rise, DhTOT.STG ≈ DhSTG and this removes the need to
highest power output, equation (18) gives the tempera- iterate as DhSTG may be determined without velocity
ture at each stage outlet. The velocity at the mean inputs.
line fixes the specific stage stagnation enthalpy in
equation (19) and in turn, determines the design SLC for 2.10.4 Pressure ratio model
each stage, cDS,STG.
In this SUSM, three possible pressure ratio models are
△T COMP available, each built with an efficiency model and a specific
T STG:OUT ¼ T COMP:IN þ ðSTGÞ ð18Þ
STGMAX entropy model.
1. Pressure ratio guided by small stage polytropic efficiency
1 using reference specific entropy (function of temperature)
hTOT:STG:OUT ¼ hðT STG:OUT Þ þ ðuSTG2 þ vSTG2 Þ: ð19Þ 2. Pressure ratio guided by small stage polytropic
2
efficiency using specific entropy (function of temper-
This SLC model, which is based on constant ature, pressure)
temperature rise through all stages, is inspired by a 3. Pressure ratio guided by fully isentropic compression
worked example in Mattingly [8], where a preliminary based specific entropy (function of temperature, pres-
compressor is designed without detailed stage tempera- sure)
ture information. Each pressure ratio model suggests a feasible relative
3. Constant specific static enthalpy rise for all stages: using distribution of maximum stage pressure ratios for all the
the operating point with the highest power output, stages. There is no best model as pressure ratio is
equation (20) calculates the specific stagnation enthalpy determined stage by stage to meet the overall pressure
at each stage outlet, which in turn determines the design ratio, which in turn fulfills several possible objectives. The
SLC for each stage, cDS,STG. gas turbine could have been designed for maximum overall
pressure ratio or the most economical maintenance
DhCOMP package. An inference using the overall pressure ratio is
hTOT:STG:OUT ¼ hCOMP:IN þ ðSTGÞ
ðSTGMAX Þ at best an estimate and cannot compensate for unavailable
1 information, considering that an actual design has as many
þ ðuSTG2 þ vSTG2 Þ: ð20Þ
2 as a million details as pointed out by Ghisu et al. [2].
8 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

In pressure ratio model 1, the actual specific stage

entropy rise is not known beforehand. However, the stage
efficiency may be approximated by the compressor’s
polytropic efficiency (23), with a mean g,g determined from
the conditions at the inlet and outlet of the compressor.
h2S h1 g 1
nSTG ˆ ˆ nPOLY ˆ
h2 h1 g
ln p2 =p1 †
: 23†
ln{ 1=h†‰ P 2 =P 1 † g 1†=g 1Š ‡ 1}
The specific stage outlet static enthalpy, h2 and the
specific stage isentropic outlet static enthalpy, h2s lie on the
same isobar on the h–s diagram, experiencing the same
pressure rise. To avoid the still unknown actual specific Fig. 3. Blockage models in this study.
stage entropy rise, isentropic compression calculation
is carried out with h2s. The specific entropy rise in
equation (24) is set to zero and using T2s = T(h2s), a feasible STGMAX
stage pressure ratio is estimated from P2/P1. OPR ¼ P RACT:1 ∏ P RACT:STG ð27Þ
STG¼2
P2
DsSTG ¼ S0 ðT 2S Þ
S 0 ðT 1 Þ
R ln ¼ 0: ð24Þ where P RACT:STG ¼ K PR P RMOD:STG and P RACT:1 ¼ kPR:1
P1
P RMOD:1 .
The specific reference entropy s0 is from the thermody- Next, the stage temperature, gas law and mass flow rate
namic table for air from Cengel [23]. conservation determine the no-blockage axial velocity at
Pressure ratio model 2 is similar to pressure ratio each stage outlet. Iteration continues till the no-blockage
model 1 except specific entropy is a function of velocity distribution from stage 1 to stage 3 varies linearly
temperature and pressure based on the s–T–P chart by with compressor axial coordinate x, in a straight line fit of
Aartun [26]. Using the stage inlet temperature, T1 and the 3 points. The no-blockage velocity was selected as it
stage isentropic outlet temperature, T2S the specific depends only on geometry since the boundary layer build
entropy rise in equation (25) determines the stage pressure up is assumed small at the beginning of the compressor.
ratio, P2/P1.
2.10.5 Blockage model
DS STG ¼ sðT 2S ; P 2 Þ
sðT 1 ; P 1 Þ ¼ 0: ð25Þ
Blockage in a gas turbine compressor is challenging to
In pressure ratio model 3, knowing that for a given define accurately due to the accumulation of wakes from
temperature rise, T1–T2, the maximum pressure rise occurs multiple bodies interrupting the flow. Falck’s [11] LUAX-C
when the specific entropy rise is zero. With available inlet design code suggests that the blockage increases at 0.5%
and outlet temperatures from the selected SLC model, per stage and stabilises after stage 8. This idea was adapted
specific entropy is found as a function of temperature and to give blockage models 1 and 2 in Figure 3. The final
pressure based on the s–T–P chart by Aartun [26]. blockage was set to 10% for both a smooth and abrupt
Equation (26) determines the maximum stage pressure transition at about stage 8. The data from Pedersen [21]
ratio, P1/P2. indicates that the maximum available flow area through
the compressor follows a steep “s” shaped curve and reaches
DS STG ¼ sðT 2 ; P 2 Þ
sðT 1 ; P 1 Þ ¼ 0: ð26Þ a steady level of 90% availability in cross-section area in the
last few stages of the compressor and is designated as
The actual stage pressure ratio is lower than maximum blockage model 3 in Figure 3.
stage pressure ratio, and in this way, accounts for the specific
stage entropy rise. The specific stage entropy rise is 2.10.6 Velocity triangle model
proportional to the maximum pressure rise in a stage for
the same stage efficiency and assuming nearly linear isobars The velocity triangles of the last stage of this compressor
on the T–s diagram. In this way, the stage losses are are determined first since at the outlet of the last stage, all
distributed according to the stage compression capability. the flow quantities are known. Based on the analysis of non-
The lower actual pressure ratio is also determined in such a VGV stages in Section 2.9, non-VGV stages require similar
way that a smooth axial velocity distribution results at the velocity triangles which in turn need similar SLC. Solving
front of the compressor. All stages are subjected to the same from the last stage to the first stage results in more similar
fraction except the first stage. In the studied compressor, the flow angles in the last stages as departures from the RSRR
same fraction kPR is applied on stages 2 to 16 and kPR,1 for design guide are admitted only at the front stages. This
stage 1. The iteration variable is kPR. Each guessed kPR is approach is found to ease solution convergence difficulties.
automatically constrained by equation (27) such that all For each stage, the solution is a set of blade angles and
stage pressure ratios achieve the overall pressure ratio of the flow angles where the Euler whirl matches the specific
compressor, OPR as kPR,1 takes the slack. stagnation enthalpy rise, where the de Haller number is
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 9

met for both rotor and stator and the stage reaction is
reasonable. The stage reaction is initially 0.5 and allowed to uROT
ðuSTA þ DuÞ
eSOFT ¼ : ð33Þ
vary as iteration proceeds. The rotor blade is given more uSTA þ Du
curvature than the stator blade as the solution iterates.
At each stage, the solution process begins at the stage When the stage degree of reaction and acceptable angle
outlet, where specific stagnation enthalpy, the flow difference Du between the rotor and stator blade angles do
velocities, blade angles and thermodynamic properties not result in feasible set of blade and flow angles, the inlet
are found. Moving to the stage inlet, this SUSM applies relative flow angle b1 is adjusted with the eHARD error
only initially the design rule of similar relative velocity flow variable. The design variables VDS are updated with their
angles at the stage inlet and outlet, b1 = b3. The b1 angle error variables e and a relaxation factor, fRELAX to prevent
enables determining the stage inlet flow angles, blade over-correction, according to equation (34).
angles and thermodynamic properties. The rotor-stator
interface quantities in the middle of the stage are V DS:NEW ¼ V DS:OLD ð1 ± f RELAX eÞ ð34Þ
determined by degree of reaction, RREAT in equation (28),
where the subscripts “NEW” and “OLD” refer to the
h2 ¼ h1 þ RREACT ðh3
h1 Þ ð28Þ updated and previous values, respectively.

and conservation of specific stagnation enthalpy across 3 Results

stator in equation (29). This sets up the specific static
enthalpy and specific stagnation enthalpy to determine the
A basic axial compressor from Section 2.10 is “designed”
absolute velocity.
with the following inputs to study the influence of the IGV
hTOT:2 ¼ hTOT:3 : ð29Þ flow turning angle, the OGV inlet angle, the pressure ratio
models and the SLC models. The SLC is constant for the
Using the absolute velocity, the rotor–stator interface non-VGV stages and also constant SLC for the VGV
blade angles and thermodynamic properties are found. stages (SLC model 1). The distribution of stage pressure
With all flow angles and thermodynamic properties ratios is based on fully isentropic compression using
available, the stage Euler whirl in equation (3) is the specific entropy as a function of temperature and pressure
“hard” criteria used to test the suitability of stage reaction applied to each stage (pressure ratio model 3). The
and inlet relative velocity angle. The first “soft” criterion is blockage from boundary layer shedding off the blades as
the stage degree of reaction, which ranges about 0.5 for wakes and stationary wall boundary layers is set to grow
the axial compressor in this study. The next “soft” criteria to a maximum of 10% of the cross-sectional area, using
are the difference in blade angles, Du from blade angles, blockage model 1. The “design” goal is to reduce the mean
uROT ¼ b2
b1 and uSTA ¼ a3
a2 . amount of mismatch in flow angles at the stage-outlet–
stage-inlet interfaces. The simplified solution procedure,
Du ¼ uROT
uSTA : ð30Þ where, the specific stage stagnation enthalpy rise is
assumed similar to the specific stage static enthalpy
The de Haller number, dH is the last “soft” criteria, rise, DhTOT:STG ≈ DhSTG is used to remove the need to
which is a ratio of relative outlet speed to relative inlet iterate, sacrificing improved results for speed.
speed for the rotor blades and absolute outlet speed to Using the stage un-stacking approach in Section 2, a
absolute inlet speed for the stator blades. large collection of data is generated for a range of IGV
flowing turn angles and OGV inlet flow angles (before the
dℋROT ¼ V REL:2 =V REL:1 OGV straightens the flow), to locate the design space where
ð31Þ
dℋSTA ¼ V ABS:3 =V ABS:2: the mismatches in flow angles at each stage interface is
minimal. This would indicate a possible axial compressor
Since there is insufficient design information to allow a design at the operating point.
direct calculation of the flow and blade angles, a suitable
range of values are accepted for the design variables, 3.1 Effect of increasing IGV flowing turn angle
RREACT, Du are dℋ, according to the design guidelines for
uncertain information in Section 2. Being a “hard” criteria, To study the effect of increasing the IGV flow turning
Euler whirl, eWHIRL is used in an error variable eHARD in angle, Figure 4 shows the data generated for IGV flowing
equation (32) to iteratively suggest a Du that results in an turn angles from 20° to 34° and for selected rotational
Euler whirl which matches the specific stage stagnation components of the absolute velocity at the OGV inlet. The
enthalpy rise. rotational component v is a fraction, varying from
0.42 to 0.52, of the OGV inlet wheel speed, defined in
eWHIRL
ðhTOT:3
hTOT:1 Þ equation (16). Each data point represents a set of design
eHARD ¼ : ð32Þ
hTOT3
hTOT:1 choices that results in a feasible set of flow angles for the
whole compressor.
Consequently, the stage degree of reaction is also As both the IGV flow turning angle and v, the fraction
affected and is updated with another error variable eSOFT in of wheel speed, increase towards the optimum combination
equation (33). of IGV flow turning angle of about 31° and v close to 0.46 of
10 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

Fig. 4. The effect of flow turning angle in the IGV on the mean mismatch in flow angles per stage at each stator-outlet–rotor-inlet
interface.

the OGV inlet wheel speed, it becomes possible to obtain a enthalpy at the OGV inlet requires the rotational
set of feasible flow turning angles for the axial compressor component v before allocating sufficient specific static
as a whole that achieves minimal mismatch in flow angles and stagnation enthalpy to each axial compressor stage.
at each stator-outlet–rotor-inlet interface between stages. Using the same generated data for IGV flow turning
However, as the fraction of wheel speed increases further, it angles, the effect of OGV inlet flow angle is examined by
again becomes increasingly difficult to obtain a set of varying v and presented in Figure 5. Each data point in
feasible flow turning angles for the whole compressor. Since Figure 5 represents a set of design choices that results in
the axial compressor’s design is highly optimized, the many feasible set of flow angles across the whole compressor.
design parameters simultaneously match narrowly only For clarity, only the even IGV flow turning angles are
about the design point. shown.
Increasing v as a fraction of wheel speed at the outlet
3.2 Effect of OGV inlet flow angles has only a small effect on reducing the amount of mismatch
in flow angles at the stator-outlet–rotor-inlet interface.
The velocity components u and v at the OGV inlet Adjustments to v hardly minimise the flow angle
determine the outlet flow angle at the last stage’s stator. mismatches unless the IGV flow turning angle is close to
An improved estimate of the specific stagnation 31° for this axial compressor.
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 11

stage pressure ratio is closer to the actual from only stage 5

onwards. The “PR GE LM2500” data in Figure 7B is taken
from Klapproth et al. [15].
The resulting axial velocity profile in Figure 8 is smooth
at the front stages even after factoring the effect of
blockage. However, when moving from the last VGV stage
to first non-VGV stage, there is a rougher transition. This
would happen as this SUSM has no provision for transiting
the flow from a VGV stage to a non-VGV stage. For the
same mass flow rate, the axial velocity distribution from
pressure ratio model 3 in Figure 8 does indicate lower axial
velocities, implying higher density from higher pressure
ratios in the front stages.
Considering that the actual axial velocity distribu-
Fig. 5. The effect of increasing v as a fraction of wheel speed at tion (“Act. Ax. Vel.” in Fig. 8) from Klapproth et al. [15]
the OGV inlet on the mean mismatch in flow angles per stage at is not always smooth, the rough transition in axial
each stator-outlet–rotor-inlet interface. velocity is acceptable. According to the approximate
compressor schematic on the gas turbine manufacturer’s
marketing datasheet [14], there is a proportionally longer
stator at stage 8. The 8th stage stator houses the bleed
air valve, according to Klapproth et al. [15]. In the non-
VGV stages, the axial speeds show the greatest
difference. This difference is traced to several possible
sources, the OGV, bleed air, annulus cross-sectional area
and blockage.
The OGV is mentioned by Klapproth et al. [15] but
there is too little information on the cross-sectional area
and the axial length for a proper analysis of the flow speed.
Should the OGV be part of the analysis, it would contribute
only a small pressure rise. The OGV would only very mildly
Fig. 6. Comparison of mean flow angle mismatch between pressure reduce the stage loading at all stages as a small pressure
ratio models. and temperature rise is obtained through straightening the
flow. The corresponding mild reduction in stage compres-
sion would give lower stage density at the last stages and
3.3 Effect of pressure ratio model the axial speeds would increase to maintain the same mass
flow rate. Therefore, the OGV excluded from the analysis is
The effects of different pressure ratio models on an axial not a cause of higher axial flow speed at the last stages. In
compressor’s flow angles at the design point are examined addition, inferring the radius from the same approximate
with a basic compressor from Section 2.10, where the IGV compressor schematic on the gas turbine manufacturer’s
flow turning angle is 28° and at the OGV inlet, v = 0.45 marketing datasheet [14], there is little change in cross-
of the wheel speed. The SLC model 1 sets the axial sectional area after the last compressor stage, implying
temperature distribution, assuming DhTOT:STG ≈ DhSTG . axial mass flux might keep constant after flow straighten-
The combined effect of the choice of pressure ratio model ing. Since pressure increases to give higher density through
and the method that smooths the velocity distribution the OGV, lower axial speed results after the OGV to
has minimal effect on the mean flow angle mismatch at maintain mass flow rate. This would not affect the axial
the stator-outlet–rotor-inlet interfaces as seen in Figure 6. speeds of the last compressor stages. Bleed air for film
The pressure ratio models only distribute available cooling in the high pressure turbine is interpreted as
pressure among the stages of the axial compressor. Within obtained from the compressor discharge according to the
the overall compression ratio of this axial compressor, any technical section of this 3rd party market analysis [27], and
additional rise in pressure ratio in a stage is compensated does not affect the axial speed at the last stages. Similar
with a lower pressure ratio in another stage. For example, no-blockage annulus cross-sectional area is found by
the lower pressure ratios at the rear stages suggested estimating the radius from the same approximate com-
by pressure ratio model 3 are complemented by higher pressor schematic on the manufacturer’s marketing data-
pressure ratios in the front stages and the reverse occurs sheet [14] and using data in Pedersen [13]. The estimated
pressure ratio models 1 in Figure 7B. radius from the same approximate compressor schematic
The choice of pressure ratio model, however, does affect on the manufacturer’s marketing datasheet [14] results in a
the maximum pressure ratio available at each stage as seen slightly larger annulus cross-sectional area. The blockage
in Figure 7A, where pressure ratio model 3 delivers more model supported by data in Pedersen [21] shows a clear
compression in the front stages compared to pressure ratio 10% reduction in available cross-sectional area after stage 9
model 1 and 2. This effect is present in a small way in the as indicated in Figure 3. This level of blockage is used at the
actual stage pressure ratios in Figure 7B. The estimated OGV inlet. To resolve this obvious difference in axial speed
12 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

Fig. 7. Maximum pressure ratios and used Pressure ratios with different pressure ratio model (stage outlets indicated by numbers in
italics).

In the actual compressor, by the same argument, the

much lower axial speed will result in larger blade cambers,
resulting in differences between the blade cambers
obtained in this stage unstacking method and the actual
(“Tip Camber” and “Hub Camber” in Fig. 9).
By deliberate design in this SUSM, the b1 angle is
consistent or allowed to vary only slightly across the stages.
Since the b1 angle could keep constant in Figure 10A with
only slight variations at the first 3 stages to accommodate
determining the other flow angles, this assumed b1 design
guide is feasible. Assuming that the effects of the RSRR
initial design persist after many design iterations, the a1
angle would also likely be as consistent as b1 at each stage.
Without constraining the a1 angle, this SUSM has
determined a distribution of a1 angles that is relatively
Fig. 8. Stage axial velocity with different pressure ratio models consistent, if not smoothly varying, especially for the non-
(stage outlets indicated by numbers in italics). VGV stages. It is possible that the studied axial compressor
shows signs of the RSRR initial design, especially for the
in the 2nd half of the compressor especially at the OGV non-VGV stages. Similar to the b1 angle, the a1 angle varies
inlet, certain modelling improvements are possible. The smoothly in Figure 10B. The a2 and b1 angles from stage
extraction of bleed air must be considered and a better unstacking are much closer to the actual. Should lower
estimated annulus cross-sectional area is needed. As the axial speed be used at the last stage, the last stage b1 angle
analysis is at the design operating point and approximate, would be larger. Since this stage unstacking method has the
further information on the VGV operation from Klapproth capability to slightly reduce the b1 angle as the analysis
et al. [15] for this gas turbine is not used. proceeds towards the front of the compressor, it would be
For the same SLC model, the rotor blade row needs to possible that beginning with a larger b1 angle at the last
deliver a similar specific stage stagnation enthalpy rise in stage, this SUSM is able to estimate a b1 angle distribution
all three pressure ratio models. When pressure ratio model closer to the actual.
3 delivers lower stage axial velocity (Fig. 8) compared to
the other pressure ratio models, the rotor needs to turn the 3.4 Effect of SLC distribution
flow by a greater angle to deliver a larger rotational
component of the absolute velocity that compensates for The effects of different SLC models on an axial compres-
the reduced axial component of the absolute velocity, sor’s flow angles at the design point are examined with the
according the Euler whirl equation. This leads to pressure same basic compressor from Section 2.10, except that
ratio model 3 generating slightly larger rotor blade angles pressure ratio model 3 is used and the SLC model varies.
compared to the other pressure ratio models in Figure 9A. The IGV flow turning angle is still 28° and at the OGV
Consequently, since the rotor blade row is delivering a inlet, v = 0.45 of the wheel speed.
larger rotational component of absolute velocity with Compared to the different pressure ratio models, the
pressure ratio model 3, and the stator blade row need only choice of SLC model has a larger effect on mean flow angle
conserve specific stagnation enthalpy, the inlet and outlet mismatch at the stator-outlet–rotor-inlet interfaces than
velocity triangles of the stator blade row are more similar. for the choice of pressure model as seen in Figure 11. Using
Consequently, the stator blade angles are slightly smaller these design options, the SLC model 3, where there is a
compared to that obtained in the other pressure ratio constant specific static enthalpy rise for all stages could not
models in Figure 9B. support generating a complete feasible set of flow angles
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 13

Fig. 9. Blade angles with different pressure ratio models.

Fig. 10. Flow angles with different pressure ratio models.

stages. At the same time, the front stages must impart

higher specific stagnation enthalpy to the working fluid
when lesser rotational kinetic is available as the Eulerian
radius, rE is smaller.
The specific stage static enthalpy rise in Figure 13
shows the same trend as stage SLC since the simplifying
assumption that DhTOT:STG ≈ DhSTG is used in the SLC
definition. Because SLC model 3 could not support in
finding a complete set of flow angles for the whole axial
compressor, the specific stage static enthalpy distributions
in SLC models 1, 2 and 4 indicate that specific stage static
enthalpy rise must increase stage after stage for this
axial compressor. The specific stage static enthalpy rise in
model 2 is not a smooth curve as the temperature-specific
Fig. 11. Comparison of mean flow angle mismatch between SLC static enthalpy function is only sufficiently piecewise smooth.
models. In this SUSM, the stage pressure ratios cannot be fixed
while varying the stage load coefficient because only the
variation in pressure ratios is available for smoothing the
that fit all stages. While generating the velocity triangles axial velocity especially from the IGV to the 2nd stage of
from the OGV inlet to the inlet with SLC model 3, the the axial compressor. However, the difference is stage
available energy and requirements could not set up a pressure ratios in Figure 14 are small from stage 5 onwards.
proper velocity triangle at a middle stage within this axial The major difference is found in stages 1–4, where higher
compressor. specific static enthalpy rises (Fig. 13) correspond to higher
The stage SLC distribution in Figure 12 shows that pressure ratios (Fig. 14).
model 3 with constant specific stage static enthalpy rise has Since the compressor inlet axial speed is known, each
the lowest SLCs for the non-VGV stages. The compressor SLC model works with the same pressure ratio model to
designed with SLC model 3 imparts lesser specific generate the same inlet density, resulting in pressure and
stagnation enthalpy to the working fluid when higher temperature moving in tandem at the inlet. In the
rotational kinetic energy (vrE) is available at the rear subsequent stages, for the same mass flow rate and known
14 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

Fig. 12. Stage SLC distribution with different SLC models. Fig. 14. Stage pressure ratios with different SLC models.

Fig. 13. Specific stage static enthalpy rise with different SLC Fig. 15. Stage axial velocity with different SLC models.
models.

4.1 IGV flow turning angle

smooth cross-sectional area variation, a smooth density
variation is needed to generate a smooth axial velocity While the implemented design options have identified a
profile. This smooth density variation further maintains design space where a range of inputs are able to generate
the differences in pressure ratio for the front few stages. zero flow angle mismatches between the stages, the most
By the 2nd stage in the axial compressor, Figure 15 influential design option that reduces angle mismatches to
shows that the axial velocity is no longer influenced by the zero is the IGV flow turning angle.
conditions at the compressor inlet. In SLC model 1, due to In terms of reducing the mean mismatches in flow
the initial lower pressure ratios, the resulting lower density angles per stage, varying the OGV inlet flow angle by
flow is compensated by increasing the axial velocity varying v as a fraction of wheel speed at the OGV inlet
through the rest of the axial compressor. The opposite exerts little influence. In this LM2500 compressor,
occurs for SLC model 3, which has the highest initial increasing v causes no increase in OGV inlet specific
compression of the 4 SLC models. stagnation enthalpy since the compressor’s outlet stagna-
tion temperature is fixed. However, increasing v reduces the
specific static enthalpy only slightly as the velocity
4 Discussion contribution to specific stagnation enthalpy is small. A
10% increase in v from 42% to 52% of wheel speed at the
This SUSM is applied to the design operating point only. OGV inlet results in only about 0.05 kJ/kg change in
The actively adjustable design options are the choice of specific stagnation enthalpy at the stage before the OGV
SLC model, pressure ratio model, blockage model and inlet. Furthermore, beyond this small 10% range of v, there
amount of blockage. These factors work through the are either no feasible sets of flow angles for the whole
method to determine the overall specific stage static compressor or the flow angle mismatch cannot reach zero.
enthalpy rise, stage pressure ratios and axial velocity In contrast, adjusting the IGV flow turning angle exerts
profile through the axial compressor. These in turn a greater influence on reducing the mean mismatches in
determine velocities and flow angles within each stage of flow angles per stage. For solutions surrounding the
the axial compressor, which in turn determine the amount optimum IGV flow turning angle that results in zero flow
of flow angle mismatch. angle mismatches across all stages, a 1° increase in IGV flow
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 15

The GE LM2500 and Hitachi 17-stage are land based

turbines. In the design of the Hitachi 17-stage compressor,
the front 3 stages are highly loaded [25]. The GE LM2500
compressor does not deliberately load the front stages.
While the design perspectives differ, the stage pressure
ratio distributions are similar with a peak in pressure ratio
at one of the front stages. The “GE NASA EEE” refers to
the Energy Efficient Engine (EEE) and is meant to
support aviation engine development. The detailed design
phase of compressor of the “GE NASA EEE” began with
specifying the specific stagnation temperature rise at each
stage for the design operating point seen in Figure 14 from
Holloway et al. [22]. This distribution was overall
decreasing with stage number except the middle stages
Fig. 16. Pressure ratio distribution of selected gas turbines. 6 and 7, where the decrease was more obvious. The initial
distribution of the stage static temperature rise in
turning angle causes a 0.2–0.5 kJ/kg specific static Figure 27 from Holloway et al. [22] is similar. In the final
enthalpy drop at the IGV outlet. For a more quantitative configuration, the static temperature distribution through
comparison, a 3° change about the optimum IGV flow the stages did not change too much as seen also in
turning angle of about 31° (≈10% change) causes about Figure 27 from Holloway et al. [22]. While the “GE NASA
0.6–1.5 kJ/kg change in specific stagnation enthalpy, at the EEE” pressure ratio was not specified directly, it
IGV outlet (stage 1 inlet), which is easily 10 times greater synchronizes closely with the temperature variation.
than the specific static enthalpy changes implemented at Therefore, the distribution of stage pressure ratio is a
the OGV inlet. design choice, driven by temperature variation and
Determining matching flow angles between stages and therefore SLC.
for the Euler whirl to match the specific stagnation These 3 designs show that aggressive compression in the
enthalpy rise at each stage for all 16 stages of this early stages is more feasible than in the later stages. The
compressor is akin to simultaneously satisfying 32 sets of important detail is which stage then delivers the highest
requirements. These requirements and the compressor’s pressure ratio. The absolute pressure rise of the first stage
overall pressure and temperature ratios are therefore in the EEE would be smaller than that of the other two
actually coupled together to characterize the design of the compressors, since the EEE design operates primarily in a
studied axial compressor. This large number of constraints low ambient pressure environment of 22 631.264 Pa at
naturally greatly narrows the solution range of both the 11 km altitude according to Frei’s [29] online atmospheric
IGV flow turning angle to near 31° and v to about 0.46 of pressure calculator. However, more pressure ratio distri-
the OGV inlet wheel speed for this axial compressor. butions from actual compressors are needed before a firm
However, as v at the OGV inlet has much lesser influence on design guideline may be constructed.
flow angle mismatches than the IGV flow turning angle, the The pressure ratio models, and their supporting SLC
angle mismatches are efficiently first reduced by adjusting models, suggested so far are suited only for preliminary
the IGV flow turning angle and if needed, followed by fine analysis. The difference between the shapes of the pressure
tuning with v at the OGV inlet. ratio distribution curves estimated in Figure 7B and the
actual in Figure 16 can only be resolved with a more
4.2 Pressure ratio distribution comprehensive SLC distribution. For this LM2500 com-
pressor, the axial temperature distribution is not available.
Smith [7] documents the rich design heritage of this gas This impedes inferring the actual SLC distribution. A clear
turbine which has experienced 2 rounds of stage additions improvement to this SUSM is in the SLC models. A more
at the rear. While there is little doubt of an improved realistic SLC distribution inferred for the LM2500
understanding of the underlying physics of flow and of compressor is shown in Figure 17 and this is based on
materials, the account by Smith [7], shows instead that the the stage pressure ratios in Figure 16. This distribution
proven aspects of the current design are retained in the contains details that presented simpler SLC models do not
interest of reliability when adding more stages. Wadia et al. have, hence the differences between the estimated and
[16] report the most recent addition of the 0th stage to the actual stage pressure ratios.
front of the compressor where little of the original gas
turbine is altered. This implies that the design approach is 4.3 Acceptable range for camber angles
possibly not consistent in all aspects across the compressor.
In Figure 16, the shape of the pressure ratio distribution In a worked example in Mattingly [8], the RSRR approach
curve of the studied compressor is compared to another is applied to stage stack an axial compressor, generating a
compressor by the same manufacturer “PR GE NASA large database of possible values for the design variables.
EEE” referred to by Bruna et al. [28] and a similar 17 stage The difference between the inlet and outlet flow angles of a
machine (PR Hitachi 17 stg) by Kashiwabara et al. [25]. blade row ranged from 7° to 30°. For an NACA 65, a = 1.0
The inlet is 0 and the compressor outlet is 1 on the series airfoil, this translates to a camber of 1.53% to 6.58%
horizontal axis. of the chord where the camber lines are circular arcs.
16 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

stage appears similar. Currently, SLC Model 1 represents

this best. For the same SLC distribution, the pressure ratio
model 1 gives a slightly more balanced distribution across
all stages. The blockage model is 1. However, this suggested
SLC distribution is more appropriately considered an
initial well-informed estimate.

5 Conclusion
This paper presents an SUSM that determines the
velocities and flow angles within a multistage axial
compressor. The design options in this method are built
upon inferred possible design guidelines implemented in
Fig 17. Estimated SLC distribution for the compressor in the actual gas turbines. The design options are implemented in
LM 2500 gas turbine. four SLC models, three pressure ratio models and three
blockage models which, working with other required
information, determines the stage velocities and stage flow
Table 1. Span wise radial mean blade curvature angles.
angles for an axial compressor operating at design point.
The method implements a calculation procedure that
Experiment Blade row Spanwise
simultaneously fulfills thermodynamic requirements, ve-
(radial direction)
locity triangle requirements and user-selected design
mean blade
options for the whole axial compressor at the mean line.
camber angle
The outputs are specific static enthalpy, pressure, axial
Test Case Rotor 51.81° velocity and flow angles at each stage interface of the axial
E/CO-1 compressor. The SUSM is tested with a set of operation
Test Case Rotor 17.27° data from an actual aero-derivative gas turbine. Through
E/CO-3 adjusting of the IGV flow turning angle and OGV inlet flow
Stator 47.30° angle, the method identifies a small range of design options
that suggests an axial compressor with minimal mismatch
Test Case Rotor for stage 1 and 2 38.36°
of flow angle at each stage interface. Among the adjustable
E/CO-5
design options, the method shows that the IGV flow
Stator for stage 1 and 2 43.78° turning angle has the most influence on minimising flow
angle mismatches compared to the other design options.
Noting that incidence angles and deviation angle are
usually small at the design operating point, the estimated
blade angles from this SUSM are well within this range. Nomenclatures
While solving for the stage flow angles, a stage with a
relatively higher stage axial velocity makes for less, Uppercase
therefore easier, rotational velocity contributions to
specific stagnation enthalpy, which in turns requires less A Station cross sectional area (m2)
flow turning in the rotor blade row. This eventually lowers B Blockage in terms of fraction of available station
the blade curvature and consequently also lowers the cross sectional area for flow (dimensionless)
tendency for boundary layer growth and separation on the D Diffusion factor, a measure of the blade loading
suction surface. (dimensionless)
The flow features surrounding the blades are described J Mechanical equivalent of heat, 778.2 ft-lb/Btu
in relatively greater detail in Test Case E/CO-1 by Serovy or 1 J/J where 1 ft-lb = 1.35582 J and 1 Btu =
and Dring [30], Test Case E/CO-3 by Ginder and Harris 1055.06 J (dimensionless)
[31] and Test Case E/CO-5 by Serovy and Dring [32] for the OPR Overall pressure ratio of the compressor (di-
purpose of bench marking computational fluid dynamics mensionless)
codes. The blade curvature angles used in the blade rows of P Absolute pressure in Pascal (Pa)
the three relevant test cases are calculated across the span PRMOD Stage pressure ratio determined from the
in the radial direction and shown in Table 1. The estimated pressure ratio models (dimensionless)
blade angles from this SUSM are also well within this range. P RACT Actual stage pressure ratio used (dimensionless)
R Gas constant for air (J/kg.K)
4.4 Selected stage loading coefficient and pressure RREACT Stage degree of reaction (dimensionless)
ratio models c The stage load coefficient (dimensionless)
cDS The design SLC, which could be a single value for
The pressure ratio distribution of the LM2500 spans a all stages or a distribution of values by stage
relatively small range indicating that the loading on each (dimensionless)
 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019) 17

T Absolute temperature in Kelvins (K) r Density of the working gas which is air in this study
U~WHEEL The rotor tangential speed, also referred to as (kg/m3)
wheel speed UWHEEL (m s
1) s Blade solidity which is the ratio of chord to pitch
V~ABS The absolute velocity in the velocity triangles for (dimensionless)
turbo-machinery analysis and also referred to as v Shaft rotational speed (rad/s)
VABS (m s
1)
~
VREL The free stream velocity outside the boundary Script
layer, which is the velocity relative to the blade
and is the relative velocity in the velocity dℋ The de Haller number (dimensionless)
triangle analysis for turbo-machinery and also
referred to as VREL (m s
1) Subscript
~
VDS The general symbol for a design variable
(depends on variable) 1 The blade leading edge which is also considered to
be equivalent to the upstream measuring station
Lowercase of the blade row or blade row inlet station
2 The blade trailing edge which is considered to be
CP Specific heat of air at constant pressure (J/kg.K) equivalent to the downstream measuring station
CV Specific heat of air at constant volume (J/kg.K) of the blade row or blade row outlet station
eWHIRL Euler whirl (J/kg) ABS Absolute
fRELAX The general symbol for a relaxation factor CASE The compressor casing
(dimensionless) CFD Computational fluid dynamics
h Specific enthalpy (J/kg) COMP Full compressor
kLOSS The coefficient of loss for flow in the Bernoulli CORE The compressor hub
equation (dimensionless) DS Design
KPR Actual fraction of the stage pressure ratio found HARD Strict or fixed conditions or criteria
in the pressure ratio models (dimensionless) IGV Inlet guide vanes
KPR.1 The KPR for stage 1 of the compressor (dimen- IN The stage inlet station
sionless) MAX Maximum value of a variable
Ku The fraction of wheel speed at the OGV inlet Mean Mean value
(dimensionless) MIN Minimum value of a variable
m The mass flow rate (kg/s) NEW Updated values of a design variable
g CASE Casing radius (m) OLD Previous values of a design variable
rCORE Hub radius (m) OUT The stage outlet station
gE Eulerian radius (m) POLY Polytropic
S Specific entropy (J/kg.K) REL Relative to the blade
SO Specific reference entropy from thermodynamic ROT Rotor
tables (J/kg.K) S Isentropic condition
~
u or u The axial component of V ABS also referred to as u SOFT Less strict or flexible conditions or criteria
(m s
1) STA Stator
~ or V The tangential component of V ABS also referred
V STG Stage
to as V~ (m s
1) SUSM Stage un-stacking method
x Axial coordinate of the compressor (m) TOT Total or stagnation properties

Greek Acknowledgements. The authors would like to acknowledge the

support of Lloyd’s Register Singapore, Energy Research Institute
a The flow angle of the V ABS makes with the axial @ NTU (ERI@N) and Singapore Economic Development Board
direction in velocity triangle analysis for turbo- (EDB), under the EDP-IPP programme in undertaking this work.
machinery (°)
b The flow angle of the V REL makes with the axial
direction in velocity triangle analysis for turbo-
machinery (°)
References
g The ratio of specific heats, CP/CV (dimensionless) [1] N.A. Cumpsty, Some lessons learned, J. Turbomach. 132
g Mean ratio of specific heats between two or more (2010) 041018-1
stations in the compressor (dimensionless) [2] T. Ghisu, G.T. Parks, J.P. Jarrett, P.J. Clarkson, An
∊ The general symbol for an error variable (dimen- integrated system for the aerodynamic design of compression
sionless) systems-part I: development, J. Turbomach. 133 (2011)
n Thermodynamic efficiency (dimensionless) 011011-1
uROT The angle of the curvature of the rotor blade (°) [3] T. Ghisu, G.T. Parks, J.P. Jarrett, P.J. Clarkson, An
uSTA The angle of the curvature of the stator blade (°) integrated system for the aerodynamic design of compression
Du The difference in angle of curvature of the rotor and systems-Part II: application, J. Turbomach. 133 (2011)
stator blade (°) 011012-1
18 F.H.A. Koh and Y.K.E. Ng: Mechanics & Industry 20, 107 (2019)

[4] J.P. Jarrett, T. Ghisu, Balancing configuration and refine- [19] J.D. Mattingly, W.H. Heiser, D.T. Pratt, Engine component
ment in the design of two-Spool multistage compression design: combustion systems, in: J.S. Przemieniecki (Ed.),
systems, J. Turbomach. 137 (2015) 091008-1 Aircraft Engine Design, 2nd edn., AIAA Education Series,
[5] A. Sehra, J. Bettner, A. Cohn, Design of a high-performance AIAA, Reston, Virginia, USA, 2002, Chap. 9, pp. 325–418
axial compressor for utility gas turbine, J. Turbomach. 114 [20] H.I.H. Saravanamuttoo, G.F.C. Rogers, H. Cohen, Axial flow
(1992) 277–286 compressors, in: Gas Turbine Theory, 5th edn., Dorling
[6] J.P. Smed, F.A. Pisz, J.A. Kain, N. Yamaguchi, S. Kindersley (India) Pte. Ltd., Licensees of Pearson Education
Umemura, 501F compressor development program, J. Ltd. in South Asia, New Delhi, India, 2001, Chap. 5, pp. 181–262
Turbomach. 114 (1992) 271–276 [21] A. Pedersen, Ignition probability of a flammable mixture
[7] L.H. Smith, Axial compressor aerodesign evolution at exposed to a gas turbine, MSc. thesis, Department of Energy and
general electric, J. Turbomach. 124 (2002) 321–330 Process Engineering, NTNU, Trondheim, Norway, 2006
[8] J.D. Mattingly, Turbomachinery, in: J.J. Corrigan, J.W. [22] P.R. Holloway, G.L. Knight, C.C. Koch, S.J. Shaffer, Energy
Bradley (Eds.), Elements of Gas Turbine Propulsion, Efficient Engine High Pressure Compressor Detailed Design
International Edition, McGraw Hill Book Co., Singapore, Report, NASA, Lewis Research Center, OH, 1982
1996, Chap. 9, pp. 615–756 [23] Y.A. Cengel, Appendix 1 Property tables and charts (SI units),
[9] J.D. Mattingly, W.H. Heiser, D.T. Pratt, Engine in: Introduction to Thermodynamics and Heat Transfer, 2nd
component design: rotating turbomachinery, in: J.S. Prze- edn., McGraw Hill, New York, USA, 2008, pp. 765–808
mieniecki (Ed.), Aircraft Engine Design, 2nd edn., AIAA [24] M.P. Boyce, Chapter 1 An overview of gas turbines, in: Gas
Education Series, AIA, Reston, Virginia, USA, 2002, Chap. Turbine Engineering Handbook, Butterworth-Heinemann,
8, pp. 253–324 an imprint of Elsevier, Oxford, UK, 2012, pp. 3–87
[10] R.J. Steinke, STGSTK: A Computer Code for Predicting [25] Y. Kashiwabara, Y. Matsuura, Y. Katoh, N. Hagiwara, T.
Multistage Axial Flow Compressor Preformance by a Hattori, K. Tokunaga, Development of a high-pressure ratio
Meanline Stage Stacking Method, NASA Technical Paper axial flow compressor for a medium-size gas turbine, J.
2020, 1982 Turbomach. 108 (1986) 233–239
[11] N. Falck, Axial flow compressor mean line design, M.Sc. [26] I. Aartun, Using the program Allprops at the Center for
thesis, Department of Energy Sciences, Lund University, Applied Thermodynamic Studies, University of Idaho, 2002.
Lund, Sweden, 2008 https://www.studentlitteratur.se/fileaccess/private/fid8263/
[12] D. Perrotti, Two dimensional design of axial compressor
produkt/37354EnBe/torr_luft.pdf
an enhanced version of LUAX-C, M.Sc. thesis, Department [27] Forecast International, “www.forecastinternational.com,”
of Energy Sciences, Lund University, Lund, Sweden, November, 2010. www.forecastinternational.com/samples/
2013 F649_CompleteSample.pdf
[13] A. Pedersen, Ignition probability of a flammable mixture [28] D. Bruna, C. Cravero, M.G. Turner, A. Merchant, An
exposed to a gas turbine, Project Report for M.Sc. educational software suite for teaching design strategies for
Programme, Department of Energy and Process Engineer- multistage axial flow compressors, J. Turbomach. 134 (2012)
ing, NTNU, Trondheim, Norway, 2005 051010-1
[14] GE Marine, LM2500+ marine gas turbine data sheet, GE, [29] S. Frei, ICAO Standard Atmosphere, Swiss Aviation Resources,
Cincinnati, Ohio, USA, 2006 1997–2005. http://www.aviation.ch/tools-atmosphere.asp
[15] J.F. Klapproth, M.L. Miller, D.E. Parker, Aerodynamic [30] G.K. Serovy, R.P. Dring, Section 6.1 Test case E/CO-1
development and performance of the CF6-6/LM2500 Single low speed compressor rotor, in: L. Fottner (Ed.),
compressor, in: American Institute of Aeronautics and AGARD advisory report no. 275: Test cases for computation
Astronautics, 4th International Symposium on Air Breath- of internal flows in aero engine components, NATO AGARD
ing Engines, Orlando, FL, USA, 1979-7030 Propulsion and Energetics Panel, Working Group 18, 1990,
[16] A.R. Wadia, D.P. Wolf, F.G. Haaser, Aerodynamic design pp. 152–164
and testing of an axial flow compressor with pressure ratio of [31] R.B. Ginder, D. Harris, Section 6.3 Test case E/CO-3 Single
23.3:1 for the LM2500+ gas turbine, J. Turbomach. subsonic compressor stage, in: L. Fottner (Ed.), AGARD
124 (2002) 331–340 advisory report no. 275: Test cases for computation of
[17] R.O. Bullock, E.I. Prasse, Chapter 2 Compressor design internal flows in aero engine components, NATO AGARD
requirements, in: I.A. Johnsen, R.O. Bullock (Eds.), Propulsion and Energetics Panel, Working Group 18, 1990,
Aerodynamic Design of Axial Flow Compressors, NASA pp. 214–244
SP-36, Washington D. C., USA 1965, pp. 9–51 [32] G.K. Serovy, R.P. Dring, Section 6.5 Test case E/CO-05 Low
[18] S. Lieblein, F.C. Schwenk, R.L. Broderick, Diffusion Factor speed two stage compressor, in: L. Fottner (Ed.), AGARD
for estimating losses and limiting blade loadings in axial- advisory report no. 275: Test cases for computation of internal
flow-compressor blade elements, NACA RM E53D01, flows in aero engine components, NATO AGARD Propulsion
Washington D.C., 1953 and Energetics Panel, Working Group 18, 1990, pp. 286–298

Cite this article as: F.H.A. Koh, Y.K.E. Ng, A one-dimensional stage un-stacking approach to reveal flow angles and speeds in a
multistage axial compressor at the design operating point, Mechanics & Industry 20, 107 (2019)