39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit AIAA 2003-4833

20-23 July 2003, Huntsville, Alabama

PREDICTION OF MAIN/SECONDARY-AIR SYSTEM FLOW INTERACTION IN A

HIGH-PRESSURE TURBINE

Roger L. Davis1 Juan J. Alonso2

Stanford University and the Stanford University
University of California, Davis Stanford, California 94305
Davis, California 95616

Jixian Yao3 Roger Paolillo4 and Om P. Sharma5

General Electric Global Research Center Pratt & Whitney
Niskayuna, New York 12309 E. Hartford, Connecticut 06108

ABSTRACT INTRODUCTION
The results from a steady, viscous flow simulation of the The design of modern gas-turbine engines has increasingly
main flow path of a modern jet engine transonic high-pressure relied upon the use of three-dimensional computational fluid dy-
turbine, transition duct, and first vane of the low-pressure turbine namics procedures to optimize aerodynamic performance, re-
coupled with the secondary-air system under the high-pressure duce weight and cost, and maximize durability. Several nu-
turbine is presented. The secondary-air system configuration in- merical procedures have been developed to predict the flows in
cludes the lower cavity near the centerline, the mid-cavity with the compressor and turbine turbomachinery components of the
the tangential on-board injection nozzle and pump, and the upper gas-turbine engine. These procedures can solve for either the
rim-cavity. In addition, the flow through the brush and labyrinth “steady” or unsteady-flow through multi-stages of those compo-
seals that separate the cavities is included. The purpose of this in- nents.
vestigation has been to determine the feasibility of using compu-
Steady-flow predictions are often used as a three-
tational procedures to predict the steady-flow interaction effects
dimensional detailed extension of through-flow analysis and are
between the main and secondary-air system flow, the ability to
especially useful as a design tool. The steady-flow prediction
predict flow splits in the secondary-air system, and the numeri-
methodology only requires the solution of a single passage of
cal issues associated with such a computation. A description of
each blade-row in the machine. A mixing-plane boundary con-
the numerical procedure along with technical details of the ini-
dition, in which the primary variables or fluxes are circumferen-
tial and boundary conditions, and convergence of the numerical
tially averaged and exchanged at the interfaces between blade-
procedure are presented. Details of the predicted flow physics in
rows, is used to communicate the axisymmetric flow through the
the secondary-air system flow-path including the rim-cavity and
machine. As such, steady-flow predictions can be performed
brush/labyrinth seals, the main flow-path, and the interaction re-
with relatively fast turn-around solution times and reasonable
gion at the junction between the two flow-paths are presented.
computer resources. These solutions provide information re-
The effect of the secondary-air system flow on the performance
garding work-load distributions, aerodynamic performance (ef-
of the main flow-path blade rows is also discussed.
ficiency) and airfoil pressure loading. In addition, these steady-
state solution procedures can also be used to predict the radial
distribution of heat loads through the machine which is especially
1 UC-Davis Professor, Stanford Consulting Professor, Senior Member of
important for the turbine.
AIAA
2 Assistant Professor, Member of AIAA Combustor burners often produce temperature peaked pro-
3 Research Engineer
files that can enter the turbine where they interact with the tur-
4 Air Systems Design-Integration Discipline Chief
5 Chief Technologist
1
American Institute of Aeronautics and Astronautics

Copyright © 2003 by Roger L. Davis. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
bine blade and accumulate in the time-averaged flow toward the CONFIGURATION
blade pressure side and near the tip. The temperature profile at Figure 1 shows the transonic high-pressure turbine (HPT)
the exit of the combustor may be biased by design toward the stage, transition duct (TD), and low-pressure turbine (LPT) first
hub endwall of the turbine to avoid blade tip and outer air-seal vane in the main flow-path along with the high-pressure turbine
burning due to hot-streak migration. In this case, however, the secondary-air system that was used in the present numerical in-
hot gases from the combustor can be ingested into the rim-cavity vestigation. This geometry is representative of a modern high-
that separates the high-pressure turbine vane and blade. If large pressure turbine design consisting of 36 and 60 high-pressure
amounts of gases are ingested, the temperature near the hub and turbine vanes and blades, respectively, and 48 low-pressure tur-
in the rim-cavity can become supercritical leading to burning of bine first vanes. The secondary-air system is modeled as an ef-
the hub platform and blade disk. fective axisymmetric representation of the real geometry but with
three-dimensional flow in order to reduce the computational grid
To avoid hot-gas ingestion into the rim-cavity, compressor
requirements. The circumferential extent of the secondary-air
bypass air is used to purge the disk and rim-cavities and main-
system is the same as the high-pressure turbine blade.
tain a positive leakage flow back into the turbine main flow path.
As shown in Figure 2, flow enters the secondary air-system
During the design, a one-dimensional flow network analysis is
near the centerline and mid-way radially in the lower secondary
typically used to predict the pressure and flow-rate necessary to
air-system cavity. This air comes from the high-pressure com-
maintain this positive leakage flow and prevent hot-gas ingestion
pressor. In addition, air enters the tangential on-board injection
into the rim-cavity. In predictions of the turbine main flow path,
(TOBI) nozzle from the combustor burner bypass. Flow exists
leakage flows through gaps in the hub and casing endwalls be-
the high-pressure secondary air-system near the centerline where
tween the rotating and stationary blade rows are typically mod-
it passes to the low-pressure turbine secondary air-system. In ad-
eled using a prescribed transpiration or injection boundary con-
dition, flow exists between the TOBI pump and the blade disk
dition treatment in the computational fluid dynamic procedure.
near the top of the TOBI pump where it proceeds into the high-
The mass flow and stagnation temperature of the leakage flow
pressure turbine blade as film-cooling air (not modeled in the
are typically based upon the one-dimensional network analysis.
current simulation). Finally, some of the air from the secondary
Because of the limitations in the one-dimensional network air-system exits into the main flow-path at the top of the rim-
analysis and the complex flow physics at the hub where the rim- cavity. This air is critical to maintain a positive purge from the
cavity flow interacts with the main flow-path, designs can often secondary air-system into the main flow-path in order to keep
have either too little purge flow resulting in hot-gas ingestion and combustor hot gases from being ingested into the rim-cavity and
disk and/or hub burning, or too much purge flow resulting in lost burning the blade disk. The boundary conditions used for the in-
engine thrust and degraded engine efficiency. Any inaccuracy lets and exits of the secondary air-system were determined from
in the network analysis prediction in purge flow also effects the a one-dimensional network analysis used during the preliminary
main flow-path predictions. Thus, there has been a desire to use design of the secondary air-system.
more sophisticated analyses, including computational fluid dy- Swirl vanes that are present at the end of the TOBI nozzle
namics, to predict the flow in the secondary air-system and its are absent from the present simulation. A body-force model has
interaction with the main flow path. It is believed that with addi- been implemented in the current numerical procedure to swirl the
tional fidelity in the analysis, the reliability and accuracy of the flow in place of the TOBI vanes. Also, holes in the TOBI pump at
predictions can increase thereby leading to further optimization the end of the TOBI nozzle have been replaced with a slot with
in the amount of purge flow and the secondary-air system geom- the same effective area as the holes. These modifications were
etry. made in the numerical model in order to simplify grid generation
and reduce the computational grid requirements.
This paper presents the predicted results in which the
secondary-air system (SAS) is coupled directly with the main
flow-path of a modern high-pressure turbine (HPT) in a single NUMERICAL SIMULATION
computational fluid dynamic simulation. Several issues, includ- The three-dimensional, multi-block, parallel flow solver,
ing placement of the main flow-path inter-blade row mixing- TFLO (8) has been developed under the DoE ASCI program in
plane treatment, boundary conditions, flow initialization, and an effort to step up to large-scale parallel steady- and unsteady-
computational efficiency are discussed. The flow physics in the flow multi-stage turbomachinery simulations. The Reynolds-
various regions of the secondary-air system, including the lower averaged Navier-Stokes (RANS) equations are solved in the
centerline cavity, the tangential on-board injection pump, the TFLO procedure using a cell-centered discretization on arbitrary
mid-cavity, the brush and lab seals, and the rim-cavity are pre- multi-block meshes. The solution algorithm is based on an ef-
sented. The effect of the secondary air-system flow on the per- ficient explicit 5-stage Runge-Kutta integration scheme (4) cou-
formance of the main flow-path blade-rows is also discussed. pled with multi-grid, implicit residual smoothing, and local time-

2
American Institute of Aeronautics and Astronautics
stepping convergence acceleration techniques. Wilcox’s (7) k-ω path. By holding the exit static to inlet total pressure ratio in the
two-equation turbulence model is used to predict the turbulence main flow path, the mainstream flow rate was predicted as part
viscosity in the field. The solver is parallelized using domain de- of the overall solution.
composition, an SPMD (Single Program Multiple Data) strategy, At the three different inlets to the secondary-air system, the
and the Message Passing Interface (MPI) standard. Validation stagnation temperature and absolute flow angle was held fixed
of the TFLO procedure for steady and unsteady turbine flows is to specified values determined from a one-dimensional network
presented in references (8; 9; 3). design analysis. In addition, either the stagnation pressure or
mass-flow rate may be prescribed. In the current investigation,
Computational Grid the mass-flow through each of the secondary-air system inlets
Figure 2 shows the multi-block structured-grid used in the was prescribed to be the same as that determined from the one-
main/secondary air-system flow paths (note that every-other grid dimensional network analysis. The flow angle was held to be
point is shown). The secondary air-system computational grid normal to each inlet boundary. The mass-flux computed from the
is point-matched to a grid block constructed to extend across the specified mass flow, ṁin and inlet area, Ain is used to determine
main flow path between the end of the HPT stator hub and the be- the inlet density from the enthalpy relationship and equation of
ginning of the rotating HPT blade hub. Comunication between state according to:
the main and secondary air-system flows in TFLO do not have
to be performed directly using point-matched grids. A special  
ṁ
interface that allows point mis-matched grids to be used across P  P 2 2 Ttin  A in 2
in 
any block boundary (6) may be used as an alternative to point- R R cp
ρnew
(1)
matched grids in order to eliminate grid stretching or skew prob- 2 Ttin
lems that might occur. This special interface allows different
computational grid systems or non-point-matched grids to com- where P is the static pressure extrapolated in the direction
municate with each other through use of a three-dimensional in- normal to the boundary from inside of the domain, R is the gas
terpolation process. In addition, this interface allows for multi- constant, and Ttin is the specified stagnation temperature. Once
ple TFLO executions (each possibly corresponding to a different the inlet density is determined, the velocity components are com-
flow region) to run and communicate with each other in parallel. puted from the prescribed mass-flow rate and inlet area assum-
An H-grid was used to discretize the main flow-path. The ing that the flow angle is normal to the inlet. It can be shown
mesh sizes used for the main flow-path, in the axial circum- that this boundary condition treatment is the same as extrapolat-
ferential radial directions, can be summarized as follows: 1st ing the static temperature and iteratively changing the inlet total
vane: 129 57 81, strip block to merge with secondary-air pressure until the desired flow-rate is obtained.
system: 33 57 81, rotor: 129 57 81 for the main pas- One of two possible boundary condition treatments may be
sage and 97 33 17 for the tip gap, transition duct and 2nd used at the exits of the secondary-air system. At these exits, the
vane: 377 57 81. A total of 97 points where used in the static pressure or the mass-flow rate may be held at a specified
axial direction along the airfoil chord between the leading and value. The flow rate through each of the secondary-air system
trailing edges for each airfoil row. The secondary-air system exits, and thus, the flow splits are predicted if the exit static pres-
grid consisted of 183 structured-grid blocks with approximately sure is prescribed. If the mass-flow rate through each exit is
6 million grid points. This resulted in a total of approximately 9 prescribed, then the exit pressure of each secondary-air system
million points for the coupled simulation. The steady-flow sim- exit is automatically adjusted by first calculating the exit Mach
ulation was executed on 9 nodes (144 processors) of the DoE number, Mexit , from the total mass-flow parameter through the
Frost Pacific IBM SP3 at Lawrence Livermore Laboratory. The iterative solution of:
total turn-around time for this initial baseline simulation was ap-
proximately 105 days. As will be mentioned below, convergence 
 Ttexit R
acceleration techniques are currently being developed to speed
 γ  1 2 γ 1
2 γ  1
ṁ pres  γ
up this type of simulation by a factor of 3 to 5. Mexit 1 Mexit 
(2)
2 Ptexit Aexit
Boundary and Initial Conditions
The flow conditions used for this investigation correspond to where ṁ pres is the prescribed mass-flow rate, Ptexit is the pre-
the design cruise conditions. At the inlet to the main flow-path, a dicted stagnation pressure, Ttexit is the predicted stagnation tem-
total pressure and total temperature profile representative of the perature, and Aexit is the area of the exit. Two to three iterations
combustor exit flow was used as the boundary condition. A uni- are typically required to determine the exit Mach number from
form exit static pressure was held as the exit boundary condition this relationship. Once the exit Mach number is determined, the
downstream of the low-pressure turbine vane in the main flow- exit static pressure required to deliver the prescribed mass-flow

3
American Institute of Aeronautics and Astronautics
rate can be found from the isentropic total-to-static pressure re- the secondary-air system decreases with radius. Near the center-
lationship and the local stagnation pressure. line, in the lower secondary-air system cavity, the local speed of
The focus of the current investigation has been on determin- sound can become very large and thereby cause the local time-
ing the ability of computational fluid dynamics to predict the flow step size to become very small. Many time-steps are required
physics in the cavities and the interaction between the main and to convect the flow from the lower cavity to the brush seal just
secondary-air system flow using the same secondary-air system below the tangential onboard injection (TOBI) nozzle where the
flow splits determined from the one-dimensional network analy- flow rapidly accelerates and chokes (in the blade relative frame).
sis. As such, the mass-flow rate of each of the secondary-air sys- This type of convergence limitation could benefit from a psuedo-
tem exits was prescribed to the 1-D network design intent values. compressibility preconditioning scheme such as that described
The mixing plane between the high-pressure turbine vane by Merkle and Choi (5). This type of preconditioning scheme is
and blade required for the steady-flow simulation was located at currently being implemented into the TFLO solution procedure.
the end of the vane hub. By placing the mixing plane at this loca- Another strategy might be to use an implicit time-integration pro-
tion, the secondary-air system flow was allowed to mix with the cedure to solve the flow in the secondary-air system or the cou-
blade mainstream flow. An alternative to this placement might pled domain.
be to locate the mixing plane at the blade hub leading edge. If Global convergence of the flow is limited by the choked flow
mixing planes were used at both the end of the vane hub and at in the brush seal just below the TOBI nozzle and labyrinth seals
the blade hub leading edge, then the possibility arises of using an at the top of the mid-cavity in the secondary-air system. The
axisymmetric analysis for the secondary-air system rather than incoming flow from either the lower cavity or from the TOBI
allowing for three-dimensional flow. This strategy may have su- nozzle inlet must pass through one or both of these seals prior
perior convergence and reduced solution time but it was felt that to making its way to the exit of the rim-cavity and into the
the solution accuracy may not be sufficient. As mentioned above, main flow-path. Thus, even with the use of a low-speed pre-
the secondary-air system had the same circumferential pitch as conditioning convergence acceleration scheme, the global con-
the blade. The flow in the secondary-air system was computed vergence of the coupled solution is somewhat limited by the
along with the blade main-stream flow in the blade-relative coor- choked flow in these seals. The flow that makes its way through
dinate system. the holes (modeled as a slot with effective area) of the TOBI
For the present simulation, the initial conditions were set in pump into the region adjacent to the blade disk can either flow
the secondary-air system to be axial flow at a pressure and tem- downward to the centerline or upward to the blade hub and into
perature equal to that at the junction between the secondary-air the blade where it becomes blade internal cooling and ultimately
system and the main flow path between the high-pressure turbine film-cooling air (not modeled in the current investigation). The
vane and blade. The initial conditions in the main flow path were lower portion of the flow adjacent to the blade disk flows between
set such that the velocity approximately followed the initial ve- the convergent/divergent nozzle created by the lower portion of
locity triangles of the airfoils and the pressure and temperature the TOBI pump. The flow in this region exits near the centerline
varied linearly between the inlet and the exit. As will be de- and migrates into the low-pressure turbine secondary-air system
scribed below, this initialization strategy could be improved by cavity.
accounting for the swirl in the secondary-air system flow based Finally, the recirculating flow behind the seals and in the
upon the local radius and the blade rotational speed. lower-, mid-, and rim-cavity can break-up into multiple re-
circulating zones during convergence. The natural unsteadiness
Convergence and flow instability in and around these cavity recirculations also
There are three factors that limit the convergence-rate of the limits convergence rate.
coupled simulation to a steady-state solution, the low-speed flow
in the secondary-air system, the choking of the flow through the Predicted Flow Physics
brush and labyrinth seals, and the regions of recirculating flow The absolute temperature contours throughout the main and
downstream of the seals and in the secondary-air system cavities. secondary-air system flow-paths of the converged solution are
In the current investigation, the complex flow in the secondary- shown in Figure 3. The locations where solution convergence
air system took 1-2 orders in magnitude longer to converge than was monitored, as shown in Figure 4, are delineated as SW
the flow in the main flow-path and was the bottleneck to the over- (stationary-wall), DP (disk-pump), and BHLE (blade hub lead-
all convergence of the coupled domain. ing edge). The temperature is non-dimensionalized by the inlet
The local convergence of the flow field in the lower por- velocity-squared to the main flow-path divided by the real-gas
tion of the secondary-air system is limited by the low velocity constant in all of the visualization figures of this paper. The
of the fluid. Since the TFLO procedure uses an explicit time effects of the hot combustor gas is shown to greatly increase
integration scheme, the local time-step size is inversely propor- the surface temperature in the high-pressure turbine. However,
tional to the local speed of sound. The velocity of the flow in the cooler temperature from the high-compressor permeates the

4
American Institute of Aeronautics and Astronautics
secondary-air system cavities and keeps the temperature much pattern of Figure 6 when constrained to the X, R plane. The flow
lower. Figure 3 show that the cooler secondary-air system flow in the secondary-air system is difficult to visualize since one wall
lowers the hub endwall temperature through the blade and into is stationary and the opposite wall (blade disk and TOBI pump)
the transition duct. is rotating at the blade wheel speed. At the exit of the rim-cavity
Figure 4 shows the convergence history of the temperature where the flow merges with the main flow path, it is useful to
field in the high-pressure turbine stage secondary air-system at visualize the flow in both the stationary and rotating reference
various points along the stationary wall, the rotating TOBI disk- frames.
pump, and the blade hub leading-edge plane. This figure shows A close-up of the Mach number contours along with the
that the temperature field in the secondary air-system takes a sig- streamlines around the brush and labyrinth seals and the rim-
nificant number of iterations to converge due to the low speed and cavity are shown in Figure 7. This figure shows the complexity
high swirl in the secondary air-system. The temperature history of the flow with large gradients and multiple regions of swirling,
across the pitch of the high-pressure turbine blade at the lead- recirculating flow. In addition to the issues related to solution
ing edge is also shown in Figure 4. Five pitch-wise stations at convergence mentioned above, these large gradients in the flow
0%, 25%, 50%, 75%, and 100% pitch are shown. Since the flow can lead to numerical instabilities during convergence that also
along the hub at these locations is effected by the secondary-air restrict the convergence-rate of the procedure.
system flow exiting the rim-cavity, it is one of the last regions Figure 8 shows a close-up of the velocity vectors and rela-
in the domain to converge. The temperature at mid-pitch at the tive stagnation temperature contours at the junction between the
blade leading-edge plane decreases significantly during conver- rim-cavity exit and the main flow-path upstream of the blade hub.
gence and creates a temperature gradient across the pitch of the This figure and Figure 9 which shows the streamlines, gives some
blade at the hub. This temperature distribution is described fur- indication of the complexity of the flow at this junction. The
ther below. flow exiting the high-pressure vane, flows over the rim-cavity
Figure 5 shows the average non-dimensional velocity com- exit where it stagnates on the blade hub endwall near the lead-
ponents in the secondary-air system as well as the non- ing edge. At this location, the flow splits with some of the flow
dimensionalized total and static pressure and temperature as a reversing into the rim-cavity where it meets with the rim-cavity
function of the radius from the engine centerline. The velocity exit flow leading to a separation and a flow recirculation. This
components were non-dimensionalized by the main flow-path in- flow recirculation region is largest near the blade and smaller
let total velocity. The pressure was non-dimensionalized by the near mid-pitch. The complex streamline pattern at this junction
main flow-path inlet total pressure and the temperature was non- is shown in Figure 9. This figure shows that the flow emerging
dimensionalized by the main flow-path inlet maximum total tem- from the rim-cavity is not uniform across the pitch of the blade
perature. This figure shows that the technique used to initialize but is affected by the blade endwall secondary flow and pressure
the flow field for the current investigation was reasonable. The field. Although it is believed that ingestion of main flow-path hot
converged average axial and radial velocity in the secondary-air gases into the rim-cavity is probably induced by an unsteady in-
system is small. The absolute swirl velocity, Cu, is small at the teraction, Figures 8 and 9 show that another contributing factor
lower radii and grows in magnitude from the labyrinth seal lo- may be the interaction between the rim-cavity exit flow and the
cation outward to the main flow-path. The relative circumferen- secondary flow of the blade (and possibly the vane).
tial velocity exiting the secondary-air system into the main flow
path is positive corresponding to a small negative incidence to Secondary-Air System Inlets/Exits
the blade at the hub. Improvements in the initialization for the Table 1 gives a comparison of the flow, total temperature,
secondary-air system could be made by accounting for the swirl. and total pressure between the computational analysis and the
Contours of the absolute and relative Mach number along one-dimensional network analysis at the various inlets and ex-
with streamlines constrained to the X, R plane are shown in Fig- its of the secondary-air system. This table shows that the total
ure 6. The Mach number in the secondary-air system is lowest pressure at the inlet to the TOBI nozzle predicted by the cou-
near the engine centerline and is high at the exit of the TOBI pled simulation was lower but that the total pressure at the lower
nozzle as the flow enters the TOBI pump slot as well as in the cavity entrances was higher than that determined by the network
vicinity of the seals. The streamlines shown in this figure illus- analysis. The lower total pressure at the TOBI pump entrance
trate the various recirculating regions in the secondary-air system may be due to the differences in how the vanes at the exit of
flow. However, the streamlines as shown constrained to the X, R the TOBI pump are modelled between the two analyses. Table 1
plane are somewhat misleading. By far, the largest component of also shows that the predicted total pressure and total tempera-
the velocity in most regions of the secondary-air system except ture leaving all of the secondary-air system exits in the coupled
in the TOBI nozzle and near each inlet, is the circumferential simulation is greater than that predicted by the network analysis.
(swirl) velocity. The unconstrained streamlines project primarily The higher total pressure and total temperature of the flow exit-
in the circumferential direction and only show this recirculating ing the rim-cavity into the main flow-path has a strong effect on

5
American Institute of Aeronautics and Astronautics
 Lower Lower TOBI Pump TOBI Pump TOBI Pump Rim
Parameter Cavity Cavity Combustor Lower Upper Cavity
Centerline HPC Bypass Bypass Centerline Blade Exit
Entrance Entrance Entrance Exit Exit
ṁ
Computational and Network Flow Ratio, ṁmain 0.001 0.018 0.061 0.004 0.064 .012
PT
Computational Pressure Ratio, PT main 0.896 0.900 0.772 0.680 0.848 .715
PT
Network Pressure Ratio, PT main 0.550 0.556 1.047 0.347 0.601 .483
TT
Computational Temperature Ratio, TT mainmax 0.304 0.477 0.431 0.565 0.531 .895
TT
Network Temperature Ratio, TT mainmax 0.304 0.477 0.431 0.438 0.440 .508
Table 1. Flow, Pressure, and Temperature Ratios of Computational and Network Analyses

HPT HPT LPT

Vane Blade Vane
PT PTexit
Relative Total Pressure Loss, inlet
PTinlet , Coupled Simulation 0.049 0.102 0.044
PTinlet CPTexit
Relative Total Pressure Loss, PT , with Transpiration 0.034 0.185 0.073
inlet
Sexit Sinlet
Entropy Change, Cp , Coupled Simulation 0.0188 0.0286 0.0107
Sexit Sinlet
Entropy Change, Cp , with Transpiration 0.0136 0.0532 0.0197
Table 2. Relative Total Pressure and Entropy Change For Each Blade-Row with and without Secondary-Air Leakage Flow

the aerodynamic performance of all three blade-rows as will be The absolute temperature distribution on the hub of the main
discussed next. flow-path is shown in Figure 10. Of particular interest is the tem-
perature on the hub of the blade. As mentioned above, the cooler
Blade-Row Performance temperature flow from the secondary-air system tends to be split
A comparison of the aerodynamics losses in terms of the by the secondary flow in the turbine blade passage. Figure 10
relative total pressure loss and entropy change across each blade shows that the secondary-air flow is split by the blade and mi-
row from the coupled simulation and an isolated main flow-path grates toward the suction-side/hub corner of the blade passage.
simulation where the leakage flow was modeled with a transpira- This leads to a temperature gradient across the pitch of the blade
tion boundary condition is given in Table 2. In the isolated main hub surface. For comparison, the temperature contours along the
flow-path simulation, the mass-flow rate and stagnation temper- hub resulting from the more traditional simulation of the isolated
ature of the transpiration leakage flow was prescribed at the lev- main flow-path with the secondary-air leakage flow modeled as
els determined from the one-dimensional secondary-air system a transpiration boundary condition is shown in Figure 11. The
network analysis. This table shows that the losses in the high- high-pressure turbine vane and blade endwall temperature dis-
pressure turbine vane of the coupled simulation are greater than tributions resulting from the coupled secondary-air system/main
those of the isolated main flow-path simulation. In addition, the flow-path simulation is quite different than that predicted from
table shows that the relative losses in the high-pressure turbine the main flow-path simulation. Even though the flow-rates enter-
blade and downstream low-pressure turbine vane are lower for ing the main flow-path in the two simulations are the same, the
the coupled simulation compared to that of the isolated main leakage stagnation pressure, stagnation temperature and velocity
flow-path simulation. These differences can be attributed to the of the coupled simulation are predicted to be greater than those
higher total pressure and velocity exiting the rim-cavity in the prescribed for the transpiration leakage flow. As a result, the mo-
coupled simulation. For the coupled simulation, the rim-cavity mentum of the secondary-air system flow entering the main flow-
exit flow increases the momentum of the blade hub flow leading path is greater in the coupled simulation. As shown in Figures 8
to a reduction in losses and a redistribution of the flow across the and 9, the rim-cavity exit flow and its interaction with the blade
span of the blade.

6
American Institute of Aeronautics and Astronautics
secondary flow is quite complex and not likely to be modeled by
transpiration leakage boundary conditions.
Entropy contours along spanwise slices at the blade near-
hub, midspan, and tip positions are shown in Figure 12. This
figure shows that the predicted entropy at the endwalls from the
coupled simulation is lower but higher near midspan compared
to the isolated main flow-path results. The high momentum flow
exiting the rim-cavity in the coupled simulation has increased the
flow incidence to the blade near the hub and midspan compared
to the isolated main flow-path simulation. However, the inci-
dence at the blade tip has decreased in the coupled simulation
leading to a smaller tip vortex and lower tip losses. So the lower
blade losses shown in Table 2 from the coupled simulation are
not only due to the high momentum leakage flow, but also from
a redistribution of the blade flow and a reduction of tip losses.

CONCLUSIONS Figure 1. High-Pressure Turbine Stage with Secondary-Air System,

A steady, viscous flow simulation of the coupled main flow- Transition Duct, 1st Vane Low-Pressure Turbine Configuration
path and secondary-air system for a modern jet engine high-
pressure turbine is presented and compared with results from
a more “traditional” main flow-path simulation in which the Experimental Results,” ASME Paper 2000-GT-0446, ASME
secondary-air leakage flow is modeled with transpiration. De- Turbo Expo 2000, Munich, Germany.
tails of and issues related to the initial and boundary conditions Davis, R. L., Yao, J., Clark, J. P., Stetson, G., Alonso, J. J.,
in the coupled simulation are described along with a discussion Jameson, A., Haldeman, G. W., and Dunn, M. G., “Unsteady In-
of the solution convergence bottlenecks. Predicted temperature, teraction Between a Transonic Turbine Stage and Downstream
Mach number, and streamlines are presented to illustrate the flow Components,” ASME Paper GT2002-39364, June, 2002.
physics in the various regions of the flow. The complex interac- Jameson, A., “Time Dependent Calculations Using Multi-
tion between the rim-cavity exit flow and the main flow-path is grid, with Applications to Unsteady Flows Past Airfoils and
described and its effect on the main flow-path blade-row perfor- Wings,” AIAA Paper 91-1596, AIAA 10th Computational Fluid
mance is discussed. Dynamics Conference, Honolulu, HI, June 1991.
Merkle, C. L. and Choi, Y. H., “Computation of Low
Speed Flows with Time-Marching Procedures,” International
ACKNOWLEDGMENT
Journal for Numerical Methods in Engineering, Vol. 25, pp.
The authors would like to thank the U.S. Department of En-
293-311, 1985.
ergy (DoE) for its generous support under the ASCI program.
Shankaran, S., Alonso, J. J., Liou, M., Liu, N., and Davis, R.
The authors would especially like to recognize the support of the
L., “A Multi-Code-Coupling Interface for Combustor Turboma-
staff at the Lawrence Livermore National Laboratory. We would
chinery Simulations,” AIAA Paper 01-0974, January, 2001.
also like to thank the managers at Pratt & Whitney for their sup-
Wilcox, D. C., Turbulence Modeling for CFD, DCW In-
port. Finally, the authors would like to thank Edward Clutter for
dustries, Inc., La Cañada, CA, 1998.
generating the secondary-air system computational grid for this
Yao, J., Jameson, A., Alonso, J. J., and Liu, F., “De-
investigation.
velopment and Validation of a Massively Parallel Flow Solver
REFERENCES for Turbomachinery Flows,” AIAA Journal of Propulsion and
Busby, J. A., Davis, R. L., Dorney, D. J., Dunn, M. G., Power, Vol. 17, No. 3, May-June, 2001.
Haldeman, C. W., Abhari, R. S., Venable, B. L., and Delaney, Yao, J., Davis, R. L., Alonso, J. J., and Jameson, A., “Un-
R. A., 1998, “Influence of Vane-Blade Spacing on Transonic steady Flow Investigations in an Axial Turbine Using the Mas-
Turbine Stage Aerodynamics, Part II: Time-Resolved Data and sively Parallel Flow Solver TFLO,” AIAA Journal of Propul-
Analysis,” ASME 98-GT-482. sion and Power, 2001.
Clark, J. P., Stetson, G. M., Magge, S. S., Ni, R. H.,
Haldeman, C. W., and Dunn, M. G., 2000, “The Effect of Air-
foil Scaling on the Predicted Unsteady Loading on the Blade of
a 1 and 1/2 Stage Transonic Turbine and a Comparison with

7
American Institute of Aeronautics and Astronautics
 Stationary Wall Temperature History
130
SW1
120 SW2
SW3
SW4
110
SW5
SW6
100 SW7
SW8

Twall / Tref
90

80

70

60

50

40
0 100000 200000 300000 400000 500000 600000
Iteration

Disk Pump Temperature History

130 DP1
DP2
120 DP3
DP4
DP5
110
DP6
DP7
Figure 2. Main Flow-Path and Secondary-Air System Computational 100 DP8
Twall / Tref
Grid (every other grid point shown) 90

80

70

60

50

40
0 100000 200000 300000 400000 500000 600000
Iteration

Blade Hub/LE Temperature History

120
BHLE1
BHLE2
115 BHLE3
BHLE4

110
Twall / Tref

105

100

95

90
0 100000 200000 300000 400000 500000 600000
Iteration

Figure 3. Main Flow-Path and Secondary-Air System Absolute Temper- Figure 4. Computed Temperature History at Various Points in High-
ature Contours Pressure Turbine Secondary Air-System and Main Flow Paths

8
American Institute of Aeronautics and Astronautics
 8

4
Cx, Cr, Cu, U, Wu

Brush Seal
-2
Labyrinth Seal

-4
Cx, Non-Dim. Axial Velocity
Cr, Non-Dim. Radial Velocity
-6 Cu, Non-Dim. Abs. Circ. Velocity
U, Non-Dim. Wheel Speed
Wu, Non-Dim. Rel. Circ. Velocity
-8
0 5 10 15 20 25 30
Radius (cm)
Secondary-Air System Absolute Mach Number
Average Non-Dimensional Velocity Components
and Streamlines (Confined to X, R plane)

1.2

1
Pt, Ps, Tt, Ts

0.8

0.6

Brush Seal
0.4 Labyrinth Seal
Pt, Non-Dim. Abs. Stagnation Pressure
Ps, Non-Dim. Static Pressure
Tt, Non-Dim. Abs. Stagnation Temperature
Ts, Non-Dim. Static Temperature

0.2
0 5 10 15 20 25 30
Radius (cm)

Average Non-Dimensional Pressure and Temperature Secondary-Air System Relative Mach Number
Figure 5. Average Non-Dimensional Velocity, Pressure and Temperature Figure 6. Secondary-Air System Mach Number Contours and Stream-
in Secondary-Air System Flow Path lines (Limited to Axisymmetric Plane)

9
American Institute of Aeronautics and Astronautics
 Labyrinth Seal Absolute Mach Number and
Streamlines (Confined to X, R plane)

Figure 8. Velocity Vectors at Rim-Cavity Exit along Plane Aligned with

Blade Leading Edge and Relative Stagnation Temperature Contours (Vec-
tors Colored by Relative Stagnation Temperature Level

Brush Seal Absolute Mach Number and

Streamlines (Confined to X, R plane)

Figure 9. Streamlines at Rim-Cavity Exit and Relative Stagnation Tem-

Rim-Cavity Absolute Mach Number and
perature Contours
Streamlines (Confined to X, R plane)
Figure 7. Main Flow-Path and Secondary-Air System Temperature Con-
tours

10
American Institute of Aeronautics and Astronautics
Figure 10. Main Flow-Path Hub Absolute Stagnation Temperature Con-
tours from Coupled Main/Secondary Simulation

Coupled Simulation

Figure 11. Main Flow-Path Hub Absolute Stagnation Temperature Con- Isolated Main Flow-Path Simulation
tours from Isolated Main Flow-Path Simulation with Secondary-Air Mod- With Transpiration Leakage Model
eled as Transpiration Leakage Figure 12. Entropy Contours Near-Hub, Midspan, and Tip of Blade

11
American Institute of Aeronautics and Astronautics