JOURNAL OF PROPULSION AND POWER

Vol. 34, No. 4, July–August 2018

Investigation of Cooling Effect on the Aerodynamic Performance

in the Intercooled Compressor

Long-Gang Liu,∗ Xue-Song Li,† Xiao-Dong Ren,‡ and Chun-Wei Gu§

Tsinghua University, 100084 Beijing, People’s Republic of China
DOI: 10.2514/1.B36676
An intercooling technique using convective cooling channels in the compressor stator vanes has been proposed
recently. In this paper, two cooling methods are presented, and numerical simulation is conducted using the conjugate
heat transfer method to study the effect of cooling on the laminar boundary layer and turbulent boundary layer in
compressors. A flat plate and a two-dimensional compressor vane are numerically simulated to study the effect of
cooling on the two-dimensional boundary layer on the flat plate and curved surface, respectively. It is found that
cooling decreases the size of the laminar separation bubble and delays turbulent separation; this reduces the drag
coefficient at both design and off-design angle of attack. Further, a three-dimensional compressor vane with cooling
channels and cooling endwalls is also numerically simulated. It is found that the cooling channels decrease the total
pressure loss at the corner separation region; this proves that the proposed cooling method can be applied as a flow
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

control method in the compressor.

Nomenclature τ = shear stress

Cd = drag coefficient ω = total pressure loss coefficient
Cf = skin friction coefficient
cp = specific heat capacity at constant pressure Subscripts
d = wall distance
a = axial
LC = chord length
e = external boundary of the boundary layer
LG = tangential spacing
in = inlet
Pt = total pressure
out = outlet
p = static pressure
s = static
Re = Reynolds number
t = total
S = pressure coefficient
tur = turbulent
T = temperature
w = wall
Tu = turbulence intensity
U = velocity
uj = components of velocity I. Introduction
x, y = Cartesian coordinates
xj = coordinates of an orthogonal curvilinear coordinate
system T O IMPROVE the cycle efficiency of modern gas turbines,
intercooled recuperated (ICR) gas turbines have been widely
used, especially in marine gas turbines [1–3]. However, the intercooling
y
= nondimensional wall distance
α = angle of attack methods involving the injection of cooling gas [4] or liquid [5,6] into
β = inlet angle the main stream increase the complexity of the system. In the case of the
δ = thickness of boundary layer intercooling method that uses heat exchangers between the different
δ = displacement thickness stages of a compressor, there is an unacceptable increase in weight and
λ = thermal conductivity space for the aeroengines because of the extra heat exchangers and
μ = dynamic viscosity pipes. Hence, it is difficult for the traditional ICR method to be applied
μtur = turbulent eddy viscosity in aeroengines.
ξ = incidence angle Shah and Tan [7] analyzed the performance of an axial compressor
ρ = density with heat extraction via blade surfaces. They reported that cooling
σ = solidity could lead to an increase in the overall pressure ratio, maximum
mass flow capability, and efficiency. Casey and Fesich [8] redefined
polytropic efficiency so that it could be applied without ambiguities in a
Received 6 March 2017; revision received 1 November 2017; accepted for turbocharger compressor with diabatic flows. They found that heating
publication 15 November 2017; published online 21 December 2017. caused an apparent decrease in efficiency. Their results indicated that
Copyright © 2017 by the authors. Published by the American Institute of
cooling via blade surfaces could improve the compressor performance
Aeronautics and Astronautics, Inc., with permission. All requests for copying
and permission to reprint should be submitted to CCC at www.copyright.com;
without much system complexity and increase in weight. Ito and
employ the ISSN 0748-4658 (print) or 1533-3876 (online) to initiate your Nagasaki [9] proposed a new ICR method for aeroengines, as shown in
request. See also AIAA Rights and Permissions www.aiaa.org/randp. Fig. 1. They used the stator vanes in the compressor as the air cooler and
*Ph.D. Candidate, Key Laboratory for Thermal Science and Power the vanes in the bypass duct as the radiator in the intercooling system. In
Engineering of Ministry of Education, Department of Thermal Engineering; the compressor vanes, they used five serial cooling channels for the
[email protected]. convective heat transfer and used supercritical carbon dioxide as the
†
Associate Professor, Key Laboratory for Thermal Science and Power coolant. This coolant has a higher density and a higher specific heat
Engineering of Ministry of Education, Department of Thermal Engineering; capacity than air.
[email protected]. In the analysis of the compressor vanes in the new ICR aeroengine,
‡
Research Assistant, Key Laboratory for Thermal Science and Power
Engineering of Ministry of Education, Department of Thermal Engineering; both aerodynamic performance and heat transfer performance can be
[email protected] (Corresponding Author). affected by cooling. Ito et al. [10–12] experimentally investigated the
§
Professor, Key Laboratory for Thermal Science and Power Engineering of heat transfer in the compressor vanes in the new ICR aeroengine and
Ministry of Education, Department of Thermal Engineering; gcw@mail. developed several correlations for the Nusselt number on the stator
tsinghua.edu.cn. vane surface and the inner contact surface of the refrigerant.
920
 LIU ET AL. 921

II. Numerical Method

A. Governing Equations
The steady CHT simulation is conducted by the ANSYS-CFX
commercial computational fluid dynamics (CFD) software package
using a cell-centered finite volume method.
The use of cooling channels in the vanes of an intercooled
compressor increases the conjugate effects of heat transfer between
the main stream, coolant, and solid. To accurately calculate the heat
Fig. 1 Intercooled and recuperated jet engine by Ito [9]. flux and temperature on the interface, the CHT method [28], which
involves both solid and fluid domains, is normally used because the
heat transfer conditions are unknown in most cases.
The effect of cooling on heat transfer performance has also been In the fluid domain, the time-varying and compressible forms of
studied by applying discontinuous Galerkin method [13–16] with the the Reynolds-averaged Navier–Stokes equations are expressed in the
all-speed Roe scheme [17,18] in the conjugate heat transfer (CHT) compact Cartesian notation.
simulation by Liu et al. [19]. However, in the studies conducted by The continuity equation:
Shah and Tan or Ito et al., the effect of cooling on the boundary layer
was not investigated. Therefore, it is necessary to investigate the ∂ρ ∂

ρuj   0 (1)
cooling effect on the aerodynamic performance of the boundary layer ∂t ∂xj
in the new ICR aeroengine.
Investigations on the effect of cooling on the attached laminar flow The momentum equation:
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

are available in the literature. Since 1947, several studies have been
conducted on laminar flow control by wall cooling. The stability ∂ ∂ ∂p ∂τij
calculation by Lees [20], which included the effect of viscosity, ρui 
ρuj ui   −
(2)
∂t ∂xj ∂xi ∂xj
illustrated that cooling could significantly stabilize the two-
dimensional boundary layer of the air on the flat plate, whereas
heating destabilized it. It has been found that wall cooling has a strong The energy equation:
stabilizing effect on Tollmien–Schlichting (T-S) instability of the

attached air boundary layer and that it can delay the onset of the ∂ ∂ ∂ ∂T
ρh
ρuj h  λ
λtur 
ui τij
SE (3)
natural transition. In 1980, Lekoudis [21] used linear stability theory ∂t ∂xj ∂xj ∂xj
to study the effect of wall cooling in the region around the leading
edge (LE) of a transonic swept wing. He found that cooling had a mild
where SE is the volumetric heat source, and τij is the stress tensor.
stabilizing effect on the crossflow disturbances. In 1979, Reshotko
The ideal gas equation of state, ρ  pabs M∕R0 T, is used.
[22] presented a method for drag reduction in aircraft by using
The specific heat capacity at the constant pressure cp is
hydrogen cooling to delay the transition on the exterior surface. For
specified as 1004.4 J∕kg ⋅ K. The thermal conductivity λ
the supersonic flow, Mack [23] used the inviscid theory and found
is 2.61 × 10−2 W∕m ⋅ K.
that cooling can stabilize the first mode, but it destabilized the second
Using the Boussinesq hypothesis to model the Reynolds stress
mode. In 2009, Ricco et al. [24] studied the cooling effect on
term, the stress tensor is written as follows:
Klebanoff modes and oblique T–S waves. They reported that wall
cooling stabilized the laminar streaks when their spanwise
2 2
wavelength was much larger than the boundary-layer thickness in τij  2μ
μt sij − μ
μt θδij − ρkδij (4)
both subsonic and supersonic conditions. The unstable waves, which 3 3
appeared in the high-Mach-number subsonic and supersonic
conditions, were stabilized by wall cooling and destabilized by wall According to a report presented by Hao et al. [29], the shear-stress
heating. transport (SST) turbulence model [30] for the fluid domain yields a
In this study, it is assumed that cooling can be employed as a flow better result for the CHT problems. For the precise prediction of the
control technique in the compressor. However, the effect of cooling boundary layer, the SST turbulence model combines the k − ω model
on the boundary layer in the intercooled compressor may be different for the region near the wall and the k − ε model for the region outside
from the effects reported in the literature. First, the formation of the boundary layer.
separation bubbles over the compressor blade rows is a common The transport equations for turbulence kinetic energy and
occurrence because of the adverse pressure gradient [25,26], and the turbulence eddy frequency are
compressors operate under the elevated freestream turbulence levels

(Tu > 1%), which leads to a different transition mechanism that is Dρk ∂u ∂ ∂k
 τij i − β ρωk
μ
σ k μtur  (5)
nonlinear [25,27]. Second, a large turbulent region normally exists on Dt ∂xj ∂xj ∂xj
the surface of the compressor vanes. It is necessary to study not only
the laminar boundary layer but also the turbulent boundary layer after
the separation bubble. Therefore, the cooling effect on the laminar

Dρω α ∂u ∂ ∂ω
and turbulent boundary layers in the compressor with separation  τij i − βρω2
μ
σ ω μtur 
bubble under elevated freestream turbulence levels should be Dt νt ∂xj ∂xj ∂xj
investigated. 1 ∂k ∂ω

2ρ1 − F1 σ ω2 (6)
The remainder of this paper is organized as follows. A brief ω ∂xj ∂xj
description of the numerical method is presented. The method is
applied in a flat-plate case to verify the accuracy of the code and to
in which the stress tensor is computed from the eddy-viscosity
conduct a preliminary investigation on the cooling effect on
concept by a limiter:
the boundary layer without curvature. Next, a two-dimensional
compressor vane is used to investigate the effect of cooling on the
a1 k
two-dimensional laminar and turbulent boundary layers. Finally, a νtur  (7)
three-dimensional compressor blade with cooling channels and maxa1 ω; ΩF2 
cooling endwalls is employed to study the cooling effect in the corner
region. where F1 and F2 are the blending functions.
 922 LIU ET AL.

p " ! #
  
  k 500ν 4ρσ ω2 k γ sep  min 2 ⋅ max
Rev
− 1; 0 ⋅ Freattach ; 2 ⋅ Fθt ;
F1  tanh arg41 ; arg1  min max  ; 2 ; ; 3.235Reθc
β ωy ωy CDkω y2
  Freattach  e−RT ∕20 ;
4
γ eff  maxγ; γ sep 
1 ∂k ∂ω
CDkω  max 2ρσ ω2 ; 10−10 ; F2  tanharg22 ;
ω ∂xj ∂xj
! The transition model interacts with the transport equations for
p
2 k 500ν turbulence kinetic energy in SST turbulence model, as follows:
arg2  max  ; 2
β ωy ωy

∂ρk ∂ρuj k ∂ ∂k

− μ
σ k μtur   Pk;eff − Ek;eff (10)
∂t ∂xj ∂xj ∂xj
Coefficients are calculated by Φ  F1 Φ1
1 − F1 Φ2, where
Φ1 and Φ2 are the coefficients of the k − ω and the k − ε model, where
respectively:
Pk;eff  γ eff Pk ; Ek;eff  minmaxγ eff ; 0.1; 1.0 ⋅ Ek ;
σ k1  1.176; σ ω1  2.000; β1  0.0750; α1  0.5532; p
ρd k
σ k2  1.000; σ ω2  1.168; β2  0.0828; α2  0.4403 F1;eff  maxF1 ; F3 ; F3  e−Ry ∕120 ;
8
Ry 
μ
κ  0.41; β  0.09; a1  0.31
The second-order resolution scheme proposed by Barth and
To accurately simulate the laminar boundary layer and the Jespersen [32] is applied for the advection terms, and the central
transition, the γ − Reθ transition model [31] is employed together difference scheme is used for the diffusion terms in the governing
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

with the SST turbulence model for the turbulence closure, which equations in the fluid domain.
yields a good simulation result for separation-induced transition. The In the solid domain, the Fourier equation can be expressed as
transport equations for the intermittency γ and transition momentum follows:
thickness Reynolds number Re ~ θt are formulated as follows:
∂T ∂qj

   ρS cp;S
0 (11)
∂ργ ∂ρuj γ ∂ μ ∂γ ∂t ∂xj

− μ
tur  Pγ − Eγ (8)
∂t ∂xj ∂xj σ γ ∂xj where
∂T
qj  −λS (12)

 ∂xj
~ θt ∂ρuj Re
∂ρRe ~ θt ∂ ∂Re
~ θt

− σ θt μ
μtur   Pθt (9)
∂t ∂xj ∂xj ∂xj The local time stepping method is applied to enhance convergence
in the steady-state mode. In each time step of the CHT method, the
where Pγ ca1 Flength ρSγFonset 0.5 1−ce1 γ, Eγ ca2 ρΩγFturb continuities in temperature and normal heat flux are ensured on
ce2 γ−1, Pθt  cθt ρ∕tReθt − Re ~ θt 1 − Fθt , t  500μ∕ρU2 , the fluid–solid interface. Using the solid solver, the temperature on
Fturb  e−RT ∕4 . S is the strain rate magnitude. Ω is the magnitude of
4 the fluid–solid interface is calculated and used as the boundary
vorticity rate: condition for the fluid domain in the next step. The normal heat flux
on the interface is calculated by the fluid solver and imposed as the
boundary condition for the solid domain in the next time step. Thus,
ca1  2; ce1  1; ca2  0.06; ce2  50; the exchange process can be achieved in each time step of the CHT
σ γ  1; cθt  0.03; σ θt  2 method.

The transition onset is controlled by the following functions: B. Viscosity and Density in the Compressible Boundary Layer
In the boundary layer of the intercooled compressor, a change in
Fonset  maxFonset2 − Fonset3 ; 0; temperature leads to a change in density and viscosity in the
simulation of the compressible boundary layer; this may alter the
Fonset2  minmaxFonset1 ; F4onset1 ; 2; velocity and temperature profiles.
Rev ρSy2 The density in the boundary layer increases with cooling, which
Fonset1  ; Rev  ; leads to contraction of the stream tube near the wall and results in a
2.193Reθc μ thinner boundary layer because of mass conservation. The thinner
  3 
RT ρk boundary layer makes the velocity profile more full (i.e., velocity
Fonset3  max 1 − ;0 ; RT  ω~ becomes higher than before in the boundary layer at the same distance
2.5 μe
from the wall, as shown in Fig. 2).
~ θt, the blending function Fθt is
In the transport equations for Re
defined as

    
γ − 1∕ce2 2
Fθt  min max Fwake e−y∕δ ; 1 −
4
;1 ;
1–1∕ce2
ρeω~ d2 50Ωy
Fwake  e−Reω ∕10  ; Reω  δ δ ;
5 2
;
μ U BL
15 ~ θt μ
Re
δBL  θ ; θBL 
2 BL ρU

For the separation-induced transition, the constant that controls the

relation between Rev and Reθc is changed from 2.193 to 3.235 at the
separation point [31]. The modification is Fig. 2 Variation of the velocity profile in the boundary layer.
 LIU ET AL. 923

In the laminar boundary layer, dynamic viscosity reflects the ability Table 1 Inlet conditions for T3C4 case
to transfer momentum across a stream surface. The relationship
Parameter Value
between temperature and dynamic viscosity of air can be obtained
using Sutherland’s formula [33]: Inlet velocity 1.37 m∕s
Temperature 293 K
 n   Turbulence intensity 3.0%
μ T T0
Ts
 T > 120 K (13) Eddy viscosity ratio 8
μ0 T0 T
Ts Dynamic viscosity 1.8 × 10−5 Pa ⋅ s

where μ0  1.7161 × 10−5 Pa ⋅ s, T 0  273.0 K, T s  124 K, and

n  1.5. The preceding equation indicates that dynamic viscosity The mesh for the analysis is generated using ANSYS-ICEM. It is
decreases with cooling. In the two-dimensional laminar boundary performed with C-mesh, which consists of 600 × 140 cells; the total
layer with cooling, a lower dynamic viscosity leads to a higher velocity. number of meshes is 500,000. The validation of mesh independence is
Therefore, both density variation and viscosity variation can cause performed. The distance between the solid wall and the first grid line
an increase in velocity near the wall, which can delay the laminar is set to y
< 1 for the SST turbulence model and γ − Reθ
separation in the two-dimensional boundary layer when an adverse transition model.
pressure gradient exists. A thinner boundary layer and weaker
separation can increase the pressure rise capability of the compressor. B. Validation
However, in the turbulent boundary layer, the value of the turbulent
eddy viscosity is of a higher order of magnitude than the dynamic To verify the accuracy of numerical method, an adiabatic boundary
viscosity. A change in temperature influences the flowfield through condition is applied on the flat plate. The result is compared with
changes in the density and turbulent eddy viscosity and not through a ERCOFTAC’s experimental data and the numerical result reported
by Langtry [31]. The results are shown in Fig. 4.
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

change in the dynamic viscosity.

The skin friction coefficient and Reynolds number are calculated
using the equations given in Langtry’s paper:
III. Effect of Cooling on Flat-Plate Boundary Layer τ
Cf;Langtry  (14)
A. Overview of Test Case T3C4 1∕2ρU2e
T3 series experiments on a flat plate were performed by European
Research Community on Flow, Turbulence and Combustion ρUin x
(ERCOFTAC) at Rolls Royce in the 1990s. The freestream Rex  (15)
μ
turbulence intensity in the test cases is greater than 1%, which closely
approximates the actual turbulence level in an axial compressor. As a As shown in Fig. 4, the skin friction coefficient decreases with the
consequence of the high turbulence intensity, bypass transition is the development of laminar boundary in the front part of the flat plate. At
dominant transition mode instead of natural transition. the position corresponding to Rex  118;000, where the value of skin
In test case T3C [34], the geometry consists of a flat plate with a friction coefficient is 0, the laminar flow separates from the wall owing
very small rounded leading edge and a curved upper slip wall, as to the adverse pressure gradient. Subsequently, the disturbances in the
shown in Fig. 3. The length of the flat plate is 1.5 m, and the rounded laminar boundary layer are amplified in the shear layer of the separated
leading edge has a radius of 0.75 mm. The main stream accelerates flow, and transition occurs. Because of the enhanced mixing in the
with a favorable pressure gradient at the front part of the plate and turbulent flow, reattachment takes place at Rex  127;100, where the
decelerates at the back part with an adverse pressure gradient. value of skin friction coefficient is 0. This leads to the formation of a
To verify the accuracy of the simulation code in the condition when laminar-separation/turbulent-reattachment bubble, in which the skin
an adverse pressure gradient exists, test case T3C4 is selected; the friction coefficient is negative.
details of this test case are presented in Table 1. In this test case, a The numerical result is in reasonably good agreement with the
small laminar separation bubble exists under a relatively high experimental data and numerical result reported by Langtry [31].
turbulent level with an adverse pressure gradient, which is similar to Although our result has a slight deviation from the experimental data in
the phenomenon commonly observed in a compressor. the middle part of the laminar boundary layer, the separation bubble
could be accurately simulated, showing separation at Rex  118;300
and reattachment at Rex  128;300. Thus, the simulation code shows
a good resolution of the separation bubble.

C. Cooling Effect in Test Case T3C4

To investigate the cooling effect on the laminar boundary layer and
laminar-separation/turbulent-reattachment bubble, different boundary

Fig. 3 Flat-plate geometry and mesh. Fig. 4 Comparison of skin friction coefficients.
 924 LIU ET AL.

channel is assumed to be half of the thickness of the vane; the inlet

angle β is 60 deg; the solidity σ  LC ∕LG is 1.5, where LC is the
chord length and LG is the tangential spacing; the angle of attack α
(angle between the inlet flow direction and blade chord) is set to 5.0,
7.9, 10.8, 12.9, and 15.0 deg.
In the numerical simulation, the material of the intercooled
compressor vane is SUS304, whose thermal conductivity is
16 W∕m ⋅ K. The density of SUS304 is 7930 kg∕m3 , and the
specific heat capacity is 502 J∕kg ⋅ K.
The inlet conditions are shown in Table 2. The inlet velocity is
50.292 m∕s, and the inlet Reynolds number is 444,000, which is the
same as NASA’s experimental data [35]. On the interface between the
vane and the cooling channels, constant static temperature is given as
the boundary condition of the solid domain. The temperature and
normal heat flux on the interface between the main stream and the
Fig. 5 Distribution of skin friction coefficients with different boundary vane can be calculated by the CHT method.
conditions.
B. Validation
conditions of the flat plate are considered by applying wall The mesh is generated with structural elements using ANSYS-
temperatures of 268, 243, 193, and 143 K. ICEM, employing an HOH mesh topology in the fluid domain, with
The skin friction coefficient can be redefined as follows: an O-grid covering the blade surface. In the solid domain, the O-grid
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

is also used on the surface of the cooling channels; a butterfly mesh

τ ∂U∕∂y (HO mesh topology) is used on the leading edge and trailing edge
Cf  μ (16)
1∕2ρin U2in 1∕2ρin U2in (TE). The distance between the solid wall and the first grid line in the
fluid domain is set to y
< 1. The mesh of the two-dimensional (2-D)
where the shear stress is normalized by the inlet kinetic pressure. Both intercooled compressor vane is shown in Fig. 6.
the dynamic viscosity and density can affect it. For the mesh dependence study, four types of mesh are employed
The comparison of skin friction coefficient distributions with for the simulation with five different angles of attack. For example,
different boundary conditions is shown in Fig. 5. As in the previous for validation at the design angle of attack α  12.9 deg, the
analysis, the velocity increases in the laminar boundary layer owing to following conditions are applied: 11,000 elements; mesh 2: 25,000
the decrease in dynamic viscosity and the increase in density; this elements; mesh 3: 55,000 elements; and mesh 4: 93,000 elements.
delays the separation point. The separation bubble is shortened and The separation bubble occurs where pressure maintains at a constant
even eliminated when the wall temperature decreases to 243 K or lower. value and then changes rapidly. The numerical results of the pressure
coefficient S distribution are shown in Fig. 7. By comparing the
pressure coefficient distributions around the separation bubbles on
IV. Effect of Cooling on a Two-Dimensional the SS and pressure surface (PS), it can be concluded that meshes 3
Compressor Vane and 4 can satisfy the mesh independence condition. Therefore, mesh
A. Overview of Two-Dimensional Compressor Stator Vane 3 is employed in this study.
A two-dimensional compressor vane called NACA 65-(12A2 I8b ) The pressure coefficient is calculated using the following equation:
10 is used to investigate the cooling effect on the two-dimensional
boundary layer. The original profile data and experimental results Pt;in − p
S (17)
were provided by NASA in 1956 [35]. Because the profile has a more 1∕2ρin U2in
favorable convex surface streamwise pressure gradient, it has a
greater tendency to extend the laminar boundary-layer region, and For the validation of the numerical method, the values of drag
transition occurs at the back part of the suction surface (SS). Because coefficient Cd obtained from the simulation are compared with those
the separation bubble is often formed on the vanes in flows with a low in NASA’s experimental data at the different angles of attack, as
Reynolds number, low speed test conditions are selected to study the shown in Fig. 8. The numerical results obtained using CFX with
cooling effect on the separation bubble. adiabatic wall on the interface between the blade and cooling
It is assumed that there are three internal convective cooling channels are in good agreement with the experimental data, where the
channels in the compressor vane to act as the heat exchanger, as drag coefficient [36] is defined as follows:
shown in Fig. 6. The chord length is 127 mm; the width of the cooling
Fd
Cd  (18)
LC ⋅ 1∕2ρin U2in

The pressure distributions at the design angle of attack and a

negative incidence angle are also investigated, and the results are
shown in Fig. 9. The forward part of the separation bubble is a region
with constant pressure. At the design angle of attack α  12.9 deg,
separation occurs at the back portion of the suction surface and
middle portion of the pressure surface, as shown in Fig. 9a. With the
decrease in angle of attack, the separation point moves upstream on
the pressure surface. A separation bubble exists near the leading edge
at α  5.0 deg, as shown in Fig. 9b. By comparing the pressure
coefficient distributions, it can be concluded that the numerical
results are in good agreement with the experimental data.

C. Cooling Effect on the Boundary Layer of Two-Dimensional Vane

To investigate the cooling effect on the laminar boundary layer of
Fig. 6 Mesh of 2-D intercooled compressor vane. the two-dimensional compressor vane, the boundary conditions on
 LIU ET AL. 925

Table 2 Inlet conditions for 2-D compressor vane x∕LC  0.64 to x∕LC  0.78 on the suction surface and from
x∕LC  0.27 to x∕LC  0.36 on the pressure surface.
Parameter Value
The position where the absolute value of the negative skin friction
Inlet velocity 50.292 m∕s
coefficient drastically increases, as shown in Figs. 10 and 11,
Inlet total temperature 331.26 K
Inlet total pressure 133,164 Pa corresponds to the onset of transition. Burgmann and Schröder [37]
Turbulence intensity near the leading edge 1.0% reported that a significant increase in the Reynolds shear stress could be
observed in this condition because of the vortices at the downstream
end of the separation bubble. The fluid is transported toward the wall
owing to these vortices. Because of cooling, the thinner boundary layer
the interface between the vane and the cooling channels are set as shifts the position with the maximum Reynolds shear stress toward the
adiabatic wall and wall temperatures of 280, 230, 180, and 130 K, wall at the downstream end of the separation bubble, as shown in
respectively. The temperature and heat flux on the interface between Fig. 12. As a result of the enhanced mixing in the turbulent flow, the
the vane and main stream are calculated using the CHT method. separation bubble is shortened both on the suction surface and the
The comparisons of skin friction coefficient distributions at pressure surface.
α  12.9 deg with different boundary conditions are shown in It can be seen from Fig. 10 that the amplitude of the skin friction
Fig. 10a (suction surface) and Fig. 10b (pressure surface). The coefficient increases in the turbulent boundary layer after the
observations are similar to those in the previous case. In the front part of separation bubble, which demonstrates that the Reynolds shear stress
the vane, the velocity increases because of cooling. The increase in increases at the same position near the wall because of cooling.
velocity delays the separation in the back part of the laminar boundary Consequently, the enhanced mixing in the turbulent flow delays the
layer. The separation point with zero friction moves downstream from turbulent separation near the trailing edge on the suction surface in
the region from x∕LC  0.89 to x∕LC  0.97, as shown in Fig. 10a;
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

this reduces the separation loss. In contrast, there is no turbulent

separation on the pressure surface, as shown in Fig. 10b.
At α  5.0 deg, the comparison of skin friction coefficient
distributions is shown in Fig. 11a (suction surface) and Fig. 11b
(pressure surface).
On the suction surface, similar conclusions can be drawn, i.e.,
cooling delays the separation. The separation point moves downstream
from x∕LC  0.74 to x∕LC  0.80. The separation bubble is
shortened on the suction surface. After the process of transition and
reattachment, the increase in Reynolds shear stress delays the turbulent
separation in the region from x∕LC  0.88 to x∕LC  0.96 near the
trailing edge, as shown in Fig. 11a.
Nevertheless, on the pressure surface, the separation bubble occurs
near the leading edge because of the large negative angle of attack, as
shown in Fig. 11b. Cooling has little effect on the separation bubble
owing to the short laminar region; it is found to affect mainly the
turbulent boundary layer on the pressure surface at a large negative
angle of attack. The higher velocity delays the weak turbulent
separation on the concave surface under a weak adverse pressure
gradient. It even eliminates the small separation bubble when the wall
temperature decreases to 180 K or lower.

D. Cooling Effect on the Overall Performance of Two-Dimensional

Vane
At both the design angle of attack and a negative angle of attack, it
is observed that cooling has a beneficial effect on the aerodynamic
Fig. 7 Mesh dependence study: comparison of pressure coefficient performance. The laminar separation bubble and the turbulent
distributions (α
12.9 deg). separation at the trailing edge can be limited by cooling, which results
in less separation loss.
Simulation is conducted with different wall temperatures and
different angles of attack. The cooling effect on the drag coefficient is
shown in Fig. 13. Owing to the decrease in separation loss, the drag
coefficient decreases significantly. It is reduced to half of the original
value when the wall temperature is 130 K. Additionally, it can be
observed that the two-dimensional compressor vane has a better
aerodynamic performance in the off-design conditions. Cooling leads
to better operation of the two-dimensional compressor vane over a
wider range of the angle of attack.

E. Effect of Density Variation and Dynamic Viscosity Variation on

the Boundary Layer
To study the effect of density variation and dynamic viscosity
variation, the drag coefficient is compared with that in a case assuming
a constant dynamic viscosity μ  1.831 × 10−5 Pa ⋅ s, as shown
in Fig. 14.
The numerical result with a constant dynamic viscosity (dash dot
lines) indicates that the increase in density by cooling makes the
Fig. 8 Drag coefficient at different angles of attack. boundary layer thinner and increases the velocity, which delays the
 926 LIU ET AL.

Fig. 9 Distribution of pressure coefficient.

Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

Fig. 10 Distribution of skin friction coefficient (α
12.9 deg).

Fig. 11 Distribution of skin friction coefficient (α
5.0 deg).

laminar separation without the dynamic viscosity variation. The drag conductivity decreases by cooling, which slightly weakens the
coefficient decreases by the density variation. effect of wall cooling on density and dynamic viscosity, comparing
However, the effect of density variation (dash dot lines) is not as to the case with constant thermal conductivity. However, the
significant as that with both density variation and dynamic viscosity numerical results indicate that the effect of thermal conductivity
variation (solid lines). With the dynamic viscosity variation, the skin variation on the distribution of skin friction coefficient is not
friction in the laminar boundary layer is lower than that with a constant significant. Thus, the thermal conductivity variation can be
dynamic viscosity, and laminar separation can be further delayed. neglected in this study.
Thus, dynamic viscosity variation makes the drag coefficient lower In general, both the density variation and dynamic viscosity
than that with a constant dynamic viscosity, as shown in Fig. 14. variation have effect on the boundary layer with a large laminar region.
Additionally, simulations are conducted with the thermal The variation of density and dynamic viscosity cannot be neglected in
conductivity variation according to Sutherland’s law. The thermal the simulation.
 LIU ET AL. 927
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

Fig. 12 Distribution of the Reynolds shear stress after the separation bubble on SS (α
12.9 deg).

V. Effect of Cooling on a Three-Dimensional

Compressor Vane
A. Overview of the Five-Stage Compressor
The overall configuration of a highly loaded high-speed five-stage
compressor is shown in Fig. 15. Multiple circular arc airfoils are
employed in the transonic rotor of the first stage, and controlled
diffusion airfoil profiles are used in the remaining subsonic blades. A
full-scale rig test was conducted, and the overall performance data and
casing static pressure were obtained under design conditions and
off-design conditions.
Ning et al. [38] conducted aerodynamic analysis of the five-stage
compressor in detail using full-scale experiments and multistage
CFD. The numerical results were in good agreement with the
experimental data. They found that a large corner separation region
existed in the stator of the last stage. Consequently, the target is to
Fig. 13 Effect of cooling on the drag coefficient. improve the aerodynamic performance of the last stator, especially at
the near stall point.

B. Validation
Before the investigation of cooling effect on the last stator, a
validation study is conducted for the overall performance of the

Fig. 15 Overall configuration of the compressor and the stator vane of

Fig. 14 Comparison of cooling effect on the drag coefficient. the fifth stage with cooling channels.
 928 LIU ET AL.

Fig. 16 Measured and predicted overall performance.

five-stage compressor based on the full-scale test data. The

multistage CFD methods are applied in the simulation. The geometry,
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

mesh, and boundary conditions are the same as in the work by Ning
et al. [38]. The mesh is generated using structured hexagonal
elements; the total mesh size for all the 11 rows, including the IGV
and the five stages, is around 6 million, which has been proved to be
adequate in the study by Ning et al. The total pressure profile and inlet
flow angle profile are specified at the inlet based on the test data. A
radial pressure gradient to match the radial equilibrium condition is
specified at the outlet. However, the code used with the SST
turbulence model and γ − Reθ transition model in this study is
different from that used by Ning et al. Further, CHT method is applied
on the interface between the main stream and stator 5 (the last stator)
without cooling channel.
The experimental data from the tests on a full-scale rig are used for
the validation. The measured and predicted overall performances of the
five-stage compressor at the design speed are shown in Fig. 16. All the
data of overall performance presented later are normalized by the rig Fig. 18 Measured and predicted total pressure profiles at the leading
edges of stator 5 (near stall).
conditions at the design point. It is clear that the numerical result agrees
with the experimental result. The CFD slightly overpredicts the peak
adiabatic efficiency by approximately 1.7%, and the predicted mass
flow rate shows good agreement with the experimental data, with Figure 18 shows a comparison of the measured and the predicted total
errors less than 0.6%. It can be considered that the overall performance pressure profiles at the leading edge of stator 5 at near stall operating
of the five-stage compressor is predicted well. point. In the figure, the total pressure is normalized by the averaged
Figure 17 shows a comparison of the predicted static pressure on outlet total pressure in the test rig at the design point. Except for a
the casing with the test data at the near stall operating point. In the deviation in the tip region, the agreement between the results obtained
figure, position 0 of the relative axial distance represents the inlet of from experiments and CFD can be considered to be satisfactory.
the IGV, and position 1 corresponds to the outlet of stator 5. The static To summarize, the simulation code with the SST turbulence model
pressure on the casing is normalized by the static pressure in the outlet and γ − Reθ transition model can predict the overall performance,
casing at the design point. The predicted casing static pressure is in casing static pressure, and total pressure profile of stator 5 fairly well.
good agreement with the experimental data at the near stall point. This proves that it is a relatively reliable tool for the following
analysis of the flowfields at the stator.

C. Numerical Simulation and Computational Mesh

A single stator vane of the fifth stage with three cooling channels
(Fig. 15) at the near stall point is employed in this study to investigate
the cooling effect on the aerodynamic performance. The total
pressure profile, total temperature profile, and inlet flow angle profile
of stator 5 are used as the inlet conditions, and the mass flow rate
based on the test data is used as the outlet condition. The turbulence
intensity at a short distance upstream from the blade is 1.5%. The
passage flow in a single blade is simulated by applying nonslip
conditions on the blade surface, hub, and shroud. Periodic boundary
conditions are applied on the pitchwise boundaries of the flow
domain.
In this study, three test cases are employed for simulation. The first
one is to perform calculations for both the main stream region and the
stator vane without cooling channel by the CHT method. The second
one, called “cooling method 1”, involves simulation in which the
coolant flows through the cooling channels. In this method, the
Fig. 17 Measured and predicted casing static pressure (near stall). coolant that flows through the channels from the shroud to the hub is
 LIU ET AL. 929

Fig. 20 Three-dimensional streamlines.

Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

D. Three-Dimensional Flow in the Compressor

On the surfaces of the three-dimensional compressor vane,
separation does not occur in the same way as in the case of the two-
dimensional surfaces; the separation can occur with no reverse flow
and nonzero friction. Zhang et al. [40] indicated that the separation in
the three-dimensional compressor is more sensitive than that in the
turbine owing to the adverse pressure gradient. It has been reported
that there are horseshoe vortices, passage vortices, tailing shedding
vortices, corner vortices (CVs), and concentrated shedding vortices
(CSV) in the three-dimensional compressor blade [40]. In this study,
the three-dimensional streamlines and limiting streamlines on the
suction surface are shown in Figs. 20 and 21a, respectively. It can be
Fig. 19 Mesh in the middle span of the stator 5. seen that a large separation occurs in the three-dimensional corner
region, which is a feature formed by the endwall and suction surface.
On the suction surface, two limiting streamlines that start from a
saddle point on the suction surface end at the two focuses or spiral
fully developed turbulent water with a pressure of 6 MPa [10] and nodes (F1 and F2 ), where concentrated shedding vortices (CSV) are
temperature of 274 K. The mass flow rate in each channel is 1 kg∕s. lifted off the surface. This is considered as an inherent feature of the
On both the blade surface and the cooling channels’ surfaces, the closed separation [40,41], which drastically increases the loss.
temperature and heat flux are calculated using the CHT method. To The corner vortex (CV) is shown by the separation line on the hub
investigate the effect of endwall cooling, the third case, called near the suction surface. The corner vortex is induced by the passage
“cooling method 2”, is employed, in which a constant temperature of vortex instead of the horseshoe vortex and is formed from the
274 K is specified on the surfaces of the blade, hub, and shroud. In the boundary layer on the hub. It starts from another saddle point
numerical simulation, SUS304 is specified as the material for the S1 (Fig. 21).
stator vane. In Fig. 20, it is observed that streamlines 3, 4, 5, and 6 belong to the
Meshes are generated in the main stream region, solid domain, and secondary flow, which is driven toward the suction surface by the
coolant region of stator 5. Each layer of the three-dimensional mesh is circumferential pressure gradient. Streamline 3 from the hub is lifted
similar to that in the two-dimensional case. In the coolant regions, off into the main stream after encountering the saddle point on the
butterfly mesh is generated using HO mesh topology. On the blade suction surface [42]. The fluid on streamline 4 flows in the radial
surface, hub, and shroud, the distance between the solid wall and the direction and separates at the focus point F1 , and streamline 5 is lifted
first grid line is set to y
< 1. The mesh in the middle span of the off the surface at the focus point F2 . The fluid on streamline 7 flows in
stator 5 is shown in Fig. 19. For the mesh dependence study, three the reverse direction in the corner region because of the streamwise
series of meshes are employed for simulation in the three test cases. adverse pressure gradient. Then, streamline 7 interacts and converges
For example, in the case of cooling method 1, the error in the with streamline 2 from the main stream, and they are lifted off the
discretization of pressure ratio is sufficiently small, where an surface along the separation line.
approximate value for the exact continuum solution is calculated
using Richardson extrapolation [39], as shown in Table 3. By E. Cooling Effect on Three-Dimensional Turbulent Boundary Layer
comparing the limiting streamlines on the surfaces, it can be
concluded that mesh 3 is sufficiently fine to simulate the details of the Because of the adverse pressure gradient, separation occurs on the
flowfield. suction surface at the leading edge. Owing to the high freestream
turbulent level, the disturbance is amplified, and transition takes
place to enhance the mixing in the turbulent flow, which leads to
Table 3 Mesh dependence study: comparison of pressure ratio reattachment, as shown in Fig. 22.
After the separation bubble, the turbulent boundary layer is present on
Main Cooling Error of pressure ratio
Mesh stream Blade channels (cooling method 1), %
the suction surface. According to the analysis of the two-dimensional
vane, cooling does not have a large effect on the separation bubble at the
1 4,700,000 700,000 3 × 350;000 0.038
2 6,400,000 1,000,000 3 × 400;000 0.017
leading edge. Because of the cooling, the thinner boundary layer shifts
3 8,400,000 1,400,000 3 × 500;000 0.014 the position with the maximum Reynolds shear stress toward the wall in
the middle span of the suction surface, which leads to a larger Reynolds
 930 LIU ET AL.

Fig. 21 Suction surface limiting streamline pattern (dashed line: separation line without cooling; long dashed line: separation line with cooling method 1;
solid line: separation line with cooling method 2).

the chord length to a point at 21% (by cooling method 1) or 25%

(by cooling method 2) of the chord length. Comparing Fig. 21a with
Fig. 21b, it can be observed that the cooling on the blade does not
significantly affect the limiting streamline on the hub. However, by
comparing Fig. 21b with Fig. 21c, it can be concluded that the cooling
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

of the hub shifts the separation line on the hub toward the blade
surface; this indicates that the size of the CV is decreased mainly by
the cooling of the hub.
In Fig. 20, streamline 7 interacts and converges with streamline 2
from the main stream, and they are lifted off the surface along the
separation line. As cooling leads to an increase in the momentum
transfer near the surface in the main stream (streamline 1 and
streamline 2), the spanwise extent of the hub separation decreases
from 36% of span from the hub to 32% (by cooling method 1) or 30%
(by cooling method 2) of span, as shown in Fig. 21.
The velocity contours of the shedding vortices at the trailing edge
are shown in Fig. 24. The low-energy fluid zone shifts toward the hub,
and the size of the shedding vortex decreases slightly.
This demonstrates that cooling decreases the size of the separation
Fig. 22 Separation bubble and transition on SS at the leading edge (50%
region and that the low-energy fluid has less effect on the main
span).
stream.
The cooling effect on the aerodynamic performance can be
inferred by the decrease in the total pressure loss coefficient, which is
defined as follows:

Pt;in − Pt;out
ω (19)
1∕2ρin U2in

where Pt;in denotes the mass-averaged total pressure at each S1

stream surface at the inlet; Pt;out denotes the mass-averaged total
pressure at each S1 stream surface at a short distance on the
downstream side of the blade.
As shown in Fig. 25, because of the weakened separation in the
corner region and less effect of low-energy fluid on the main stream,
the total pressure loss coefficient is reduced. It decreases especially in
the region from 30% of span to 40% of span because spanwise extent
of hub separation drastically affected by cooling (shown in Fig. 21).
At a macro scale, overall performance of the intercooled
Fig. 23 Reynolds shear stress in the boundary layer (50% of span, 50% compressor is analyzed. Cooling method 1 reduces 9.1% of total
of chordwise position). temperature rise in the fifth stage, and that of cooling method 2 is
19.4%, which indicates that cooling can be used in the stators of the
multistage compressor as the intercooling method of ICR Brayton
shear stress and enhanced mixing near the wall, as shown in Fig. 23. cycle to increase the thermal efficiency. Besides, the mass-averaged
Consequently, in the middle span of the compressor blade, the tailing total pressure loss coefficient for the entire passage in the case with
separation can be weakened owing to the larger Reynolds shear stress cooling method 1 decreases by 5.9% when compared with that in
near the wall as in the case of the 2-D compressor vane. the case without cooling. In the case with cooling method 2, it
Besides, the corner vortex (CV) is also controlled by the increase in decreases by 12%. This indicates that cooling leads to an increase in
Reynolds shear stress near the wall, which can be seen by the limiting total pressure rise in the three-dimensional compressor stator.
streamline on the hub. The larger Reynolds shear stress leads to Consequently, the cooling methods that are applied on the compressor
enhanced mixing, which delays the occurrence of saddle point S1 and blades not only increase the thermal efficiency of the Brayton cycle by
decreases the size of the corner separation region. As shown in ICR method but also decrease the size of separation region to improve
Fig. 21, saddle point S1 moves downstream from the point at 20% of the performance of the compressor.
 LIU ET AL. 931

Fig. 24 Shedding vortices at the trailing edge.

References
[1] Thames, J. M., Stueber, H. B., and Vincent, C. T., “Design and
Performance Features of the Marine LM1600 Gas Turbine,” ASME 1990
International Gas Turbine and Aeroengine Congress and Exposition,
ASME Paper 90-GT-203, June 1990.
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

doi:10.1115/90-GT-203
[2] Steven, J. M., and Scott, I., “Integration of the WR-21 Intercooled
Recuperated Gas Turbine into the Royal Navy Type 45 Destroyer,”
ASME Turbo Expo 2001: Power for Land, Sea, and Air, ASME Paper
2001-GT-0531, June 2001.
doi:10.1115/2001-GT-0531
[3] Carniere, H., Willocx, A., Dick, E., and Paepe, M. D., “Raising Cycle
Efficiency by Intercooling in Air Cooled Gas Turbine,” Applied Thermal
Engineering, Vol. 26, No. 16, 2006, pp. 1780–1787.
doi:10.1016/j.applthermaleng.2006.02.008
[4] Gröschel, E., Lipfert, C., Erb, W., and Rusch, D., “Numerical
Computation of the Jet Impingement Cooling of High Pressure Ratio
Compressors,” ASME Turbo Expo 2013: Turbine Technical Conference
Fig. 25 Spanwise total pressure loss coefficient distribution. and Exposition, ASME Paper GT2013-94949, June 2013.
doi:10.1115/GT2013-94949
[5] De Lucia, M., Bronconi, R., and Carnevale, E., “Performance and
VI. Conclusions Economic Enhancement of Cogeneration Gas Turbines Through
Compressor Inlet Air Cooling,” ASME 1993 International Gas
The flow control method by cooling, which has been applied to
Turbine and Aeroengine Congress and Exposition, ASME Paper
date for external flows, is employed for the compressor vanes in this 93-GT-071, May 1993.
study. Numerical simulation using the conjugate heat transfer doi:10.1115/93-GT-071
method is conducted to study the cooling effect on the aerodynamic [6] Zheng, Q., Sun, Y., Li, S., and Wang, Y., “Thermodynamic Analyses of
performance of a flat plate, a two-dimensional compressor vane, Wet Compression Process in the Compressor of Gas Turbine,” ASME
and a three-dimensional compressor stator vane under elevated Turbo Expo 2002: Power for Land, Sea, and Air, ASME Paper GT2002-
turbulence level. 30590, Jan. 2002.
For compressor blades with a large laminar region, the flow control doi:10.1115/GT2002-30590
method by cooling can be applied. Cooling in the laminar boundary [7] Shah, P. N., and Tan, C. S., “Effect of Blade Passage Surface Heat Extraction
layer leads to an increase in the density and a decrease in the dynamic on Axial Compressor Performance,” Journal of Turbomachinery, Vol. 129,
No. 3, 2007, pp. 457–467.
viscosity, which delays the laminar separation point. In the turbulent doi:10.1115/1.2372776
boundary layer after the separation bubble, cooling shifts the position [8] Casey, M. V., and Fesich, T. M., “The Efficiency of Turbocharger
with the maximum Reynolds shear stress toward the wall. It shortens Compressors with Diabatic Flows,” Journal of Engineering for Gas
or even eliminates the laminar-separation/turbulent-reattachment Turbines and Power, Vol. 132, No. 7, 2010, Paper 072302.
bubble on the two-dimensional compressor vanes, and the two- doi:10.1115/1.4000300
dimensional trailing edge separation is also weakened. An improved [9] Ito, Y., and Nagasaki, T., “Suggestion of Intercooled and Recuperated
aerodynamic performance can be obtained by cooling, in both the Jet Engine Using Already Equipped Components as Heat Exchangers,”
design and off-design conditions, which also leads to a wider range of 47th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit,
angle of attack. AIAA Paper 2011-6102, July–Aug. 2011.
doi:10.2514/6.2011-6102
For the three-dimensional compressor vane with a turbulent
[10] Ito, Y., Yamamoto, N., and Nagasaki, T., “Estimation of Heat Transfer
boundary layer, the increase in the Reynolds shear stress near the wall Performance for Compressor Stators Heat Exchangers in a New Intercooled
by cooling weakens the three-dimensional corner separation on the and Recuperated Aviation Gas-Turbine Engine,” 48th AIAA/ASME/SAE/
suction surface. The size of the corner separation region is decreased ASEE Joint Propulsion Conference & Exhibit, AIAA Paper 2012-3937,
by the cooling blade and cooling endwalls, which leads to a decrease July–Aug. 2012.
in the total pressure loss by 5.9% when cooling channels are used doi:10.2514/6.2012-3937
and 12% when the cooling channels are used together with [11] Ito, Y., Inokura, N., and Nagasaki, T., “Conjugate Heat Transfer in
cooling endwalls. Thus, the aerodynamic performance of the three- Air-to-Refrigerant Airfoil Heat Exchangers,” Journal of Heat Transfer,
dimensional compressor can be optimized by cooling. Vol. 136, No. 8, 2014, Paper 081703.
doi:10.1115/1.4027554
A conclusion can be obtained that the laminar flow, separated
[12] Ito, Y., Goto, T., and Nagasaki, T., “Effect of Airflow on Heat Transfer of
flow transition, and turbulent flow are all affected by cooling in the Air-to-Refrigerant Airfoil Heat Exchanger,” 53rd AIAA Aerospace
compressor vanes. When the cooling method is applied in the Sciences Meeting, AIAA Paper 2015-1193, Jan. 2015.
compressor, it not only increases the thermal efficiency of the Brayton doi:10.2514/6.2015-1193
cycle but also improves the performance of the compressor, which can [13] Ren, X. D., Xu, K., Shyy, W., and Gu, C. W., “A Multi-Dimensional
enable more aggressive compressor designs. High-Order Discontinuous Galerkin Method Based on Gas Kinetic
 932 LIU ET AL.

Theory for Viscous Flow Computations,” Journal of Computational No. 9, 1997, pp.1472–1478.
Physics, Vol. 292, July 2015, pp. 176–193. doi:10.2514/2.270
doi:10.1016/j.jcp.2015.03.031 [29] Hao, Z. R., Gu, C. W., and Ren, X. D., “The Application of
[14] Ren, X. D., and Gu, C. W., “Application of a Discontinuous Galerkin Discontinuous Galerkin Methods in Conjugate Heat Transfer
Method on the Compressible Flow in the Transonic Axial Compressor,” Simulations of Gas Turbines,” Energies, Vol. 7, No. 12, 2014,
Applied Thermal Engineering, Vol. 93, Jan. 2016, pp. 707–717. pp.7857–7877.
doi:10.1016/j.applthermaleng.2015.10.048 doi:10.3390/en7127857
[15] Ren, X. D., Xu, K., and Shyy, W., “A Multi-Dimensional High-Order [30] Menter, F. R., “Two-Equation Eddy-Viscosity Turbulence Models for
DG-ALE Method Based on Gas-Kinetic Theory with Application to Engineering Applications,” AIAA Journal, Vol. 32, No. 8, 1994,
Oscillating Bodies,” Journal of Computational Physics, Vol. 316, pp. 1598–1605.
July 2016, pp. 700–720. doi:10.2514/3.12149
doi:10.1016/j.jcp.2016.04.028 [31] Langtry, R. B., “A Correlation-Based Transition Model Using Local
[16] Ren, X. D., and Gu, C. W., “A Numerical Study on the Tip Clearance in Variables for Unstructured Parallelized CFD Codes,” Ph.D. Thesis,
an Axial Transonic Compressor Rotor,” Applied Thermal Engineering, Stuttgart Univ., Stuttgart, Germany, May 2006.
Vol. 103, June 2016, pp. 282–290. doi:10.18419/opus-1705
doi:10.1016/j.applthermaleng.2016.04.082 [32] Barth, T. J., and Jespersen, D. C., “The Design and Application of
[17] Li, X. S., “Uniform Algorithm for All-Speed Shock-Capturing Upwind Schemes on Unstructured Meshes,” 27th Aerospace Sciences
Schemes,” International Journal of Computational Fluid Dynamics, Meeting, AIAA Paper 1989-0366, Jan. 1989.
Vol. 28, Nos. 6–10, 2014, pp. 329–338. doi:10.2514/6.1989-366
doi:10.1080/10618562.2014.936315 [33] Sutherland, W., “LII. The Viscosity of Gases and Molecular Force,”
[18] Li, X. S., and Li, X. L., “All-Speed Roe Scheme for the Large Eddy Philosophical Magazine Series 5, Vol. 36, No. 223, 1893, pp. 507–531.
Simulation of Homogeneous Decaying Turbulence,” International Journal doi:10.1080/14786449308620508
of Computational Fluid Dynamics, Vol. 30, No. 1, 2016, pp. 69–78. [34] Savill, A. M., “A Synthesis of T3 Test Case Predictions,” Numerical
doi:10.1080/10618562.2016.1156095 Simulation of Unsteady Flows and Transition to Turbulence,
Downloaded by DRDO-GTRE on August 1, 2018 | http://arc.aiaa.org | DOI: 10.2514/1.B36676

[19] Liu, L. G., Gu, C. W., and Ren, X. D., “An Investigation of the Conjugate Cambridge Univ. Press, Cambridge, England, U.K., 1992, pp. 404–442.
Heat Transfer in an Intercooled Compressor Vane Based on a [35] Erwin, J. R., Savage, M., and Emery, J. C., “Two-Dimensional
Discontinuous Galerkin Method,” Applied Thermal Engineering, Low-Speed Cascade Investigation of NACA Compressor Blade
Vol. 114, March 2017, pp. 85–97. Sections Having a Systematic Variation in Mean-Line Loading,” NACA
doi:10.1016/j.applthermaleng.2016.11.190 TN-3817, Nov. 1956.
[20] Lees, L., “The Stability of the Laminar Boundary Layer in a [36] Herrig, L. J., Emery, J. C., and Erwin, J. R., “Systematic Two-
Compressible Fluid,” NACA TR-876, Jan. 1947. Dimensional Cascade Tests of NACA 65-Series Compressor Blades at
[21] Lekoudis, G., “Stability of the Boundary Layer on a Swept Wing with Low Speeds,” NACA TN-3916, Feb. 1957.
Wall Cooling,” AIAA Journal, Vol. 18, No. 9, 1980, pp. 1029–1035. [37] Burgmann, S., and Schröder, W., “Investigation of the Vortex Induced
doi:10.2514/3.50852 Unsteadiness of a Separation Bubble via Time-Resolved and Scanning
[22] Reshotko, E., “Drag Reduction by Cooling in Hydrogen-Fueled PIV Measurements,” Experiments in Fluids, Vol. 45, No. 4, 2008,
Aircraft,” Journal of Aircraft, Vol. 16, No. 9, 1979, pp. 584–590. pp. 675–691.
doi:10.2514/3.58571 doi:10.1007/s00348-008-0548-7
[23] Mack, L. M., “Linear Stability Theory and the Problem of Supersonic [38] Ning, T., Gu, C. W., Ni, W. D., Li, X. T., and Liu, T. Q., “Aerodynamic
Boundary-Layer Transition,” AIAA Journal, Vol. 13, No. 3, 1975, Analysis and Three-Dimensional Redesign of a Multi-Stage Axial Flow
pp. 278–289. Compressor,” Energies, Vol. 9, No. 4, 2016, p. 296.
doi:10.2514/3.49693 doi:10.3390/en9040296
[24] Ricco, P., Tran, D. L., and Ye, G., “Wall Heat Transfer Effects on [39] Roy, C. J., McWherter-Payne, M. A., and Oberkampf, W. L.,
Klebanoff Modes and Tollmien–Schlichting Waves in a Compressible “Verification and Validation for Laminar Hypersonic Flowfields, Part 1:
Boundary Layer,” Physics of Fluids, Vol. 21, No. 2, 2009, Paper 024106. Verification,” AIAA Journal, Vol. 41, No. 10, 2003, pp.1934–1943.
doi:10.1063/1.3054155 doi:10.2514/2.1909
[25] McAuliffe, B. R., and Yaras, M. I., “Transition Mechanisms in [40] Zhang, H. L., Wang, S. T., and Wang, Z. Q., “Variation of Vortex
Separation Bubbles Under Low-and Elevated- Freestream Turbulence,” Structure in a Compressor Cascade at Different Incidences,” Journal of
Journal of Turbomachinery, Vol. 132, No. 1, 2010, Paper 011004. Propulsion and Power, Vol. 23, No. 1, 2007, pp.221–226.
doi:10.1115/1.2812949 doi:10.2514/1.17245
[26] Windte, J., Scholz, U., and Radespiel, R., “Validation of the RANS- [41] Wang, K. C., “Separation Patterns of Boundary Layer over an Inclined
Simulation of Laminar Separation Bubbles on Airfoils,” Aerospace Body of Revolution,” AIAA Journal, Vol. 10, No. 8, 1972,
Science and Technology, Vol. 10, No. 6, 2006, pp. 484–494. pp. 1044–1050.
doi:10.1016/j.ast.2006.03.008 doi:10.2514/3.50292
[27] Walker, G. J., “The Role of Laminar–Turbulent Transition in Gas [42] Gbadebo, S. A., Cumpsty, N. A., and Hynes, T. P., “Three-Dimensional
Turbine Engines: A Discussion,” ASME 1992 International Gas Separations in Axial Compressors,” Journal of Turbomachinery,
Turbine and Aeroengine Congress and Exposition, ASME Paper Vol. 127, No. 2, 2005, pp. 331–339.
92-GT-301, June 1992. doi:10.1115/1.1811093
doi:10.1115/92-GT-301
[28] Kao, K. H., and Liou, M. S., “Application of Chimera/Unstructured C. S. Tan
Hybrid Grids for Conjugate Heat Transfer,” AIAA Journal, Vol. 35, Associate Editor