NASA Technical Memorandum 106636
/1\/ °7
J~a iq~
S
f
esign
o
a
Low_Aspect
Ratio Transonic
ompressor
Stage
Using
FD
Techniques
Nelson
L
Sanger
Lewis Research Center Cleveland Ohio
~ ;
--
-----=--~-.- ...~
~~~
Prepared for the
----
. -----39th International Gas Turbine Aeroengine Congress Exposition sponsored
by
Oe
Ariiencan SocietY-of MechamcafEngirieers The Hague, NetherIands,Jime-13-16,
1994---~
•
ational
Aerooautics
and
Space
Administration
NASA-TM-I06636)
DESIGN
OF
A
LOW
ASPECT
RATIO
TRANSONIC COMPRESSOR
STAGE
USING CFD TECHNIQUES NASA.
LewiS
Research
Center
18
p
~ ~
N94-33971
Unclas
G3/01 0013491
Design
of
a
Low
spect
Ratio
Transonic Compressor
Stage
Using CFD
Techniques
Nelson Sanger
NASA
lewis
Research
Center Cleveland
OH
44135
USA
ABSTRACT
A transonic compressor
stage has been
designed
for
the
Naval
Postgraduate
School
Turbopropulsion Laboratory. The
design
relied heavily
on
CFD
techniques
while
minimizing conventional empirical
design
methods.
The
low aspect
ratio
(1.2)
rotor
has been
designed
for
a
specific
head
ratio
of
.25
and
a
tip
relative inlet
Mach
number of
1.3.
Overall stage
pressure ratio is
1.56.
The rotor
was
designed
using
an
Euler code augmented
by
a distributed
body force model
to
account
for
viscous
effects.
This
provided a relatively quick-running
design
tool,
and
was
used
for
both
rotor
and
stator calculations. The initial stator sections were
sized
using a compressible, cascade
panel
code.
In
addition
to
being
used
as
a
case study
for teaching purposes,
the
compressor
stage will
be
used as
a research stage. Detailed measurements, including non-intrusive LDV,
will
be compared
with
the design
computations,
and with
the
results
of
other
CFD
codes,
as
a
means
of
assessing
and
improving
the
computational codes
as design
tools.
NOMENCL TURE
C
p
specific
heat
at constant pressure D diffusion factor P pressure
PR
pressure ratio, total-ta-total R gas constant s entropy T temperature
U
rotor
wheel
speed
V velocity specific
weight
flow
p
V
{3
flow
angle
Y
ratio of specific
heats
o
deviation angle
{32
-
/(2
i incidence angle
{31
-
/ 1
blade
metal
angle
p
gas
density
(1
blade solidity , chord/blade
spacing
¥
-specific
head
rise
G1
total pressure loss coefficient
Superscripts:
, relative
Subscripts:
o
total inlet
2
outlet
is
isentropic process value
t
tip
INTRODUCTION
The
Naval
Postgraduate
School
axial transonic compressor stage
and
test
facility
were designed
in
1968 by
Prof.
Michael
H.
Vavra. The purpose of
the
stage
was to
provide
an
advanced test vehicle for operational
and
research experience for
naval
officers. The completed
design,
aerodynamic
and
mechanical,
was
performed
manually
by
Prof.
Vavra,
a notable accomplishment. However;
in
almost
the last step
in
the design
calculations,
an
error
in sign
occurred
which
resulted
in an
erroneous radial distribution
of
blade
setting
angle. This distribution
was
built
into
the
rotor.
A
second inaccuracy arose
from
an
assumption
made
for the
through
flow
into
the
rotor.
These,
and
other
aspects
of
the
Vavra
design
are described
by
Erwin
1983).
The
major
effect of
the
design
errors occurred over
the outer
25
of
span
where the
rotor blade
setting angles were more open
than
the
design
intent,
and
over
the
inner
~5
of
.s~
where the
incidence
was
again too
high
because
axial
~el~It1c:s
were
lower
than
had
been
assumed.
The
higher resulung
mCldence
ang.les
toward
the tip increased supersonic expansion around
the
.
eading
edge,
producing
larger
shock
los~
~d
lower pressure nse
than
design
intent.
This,
and the higher lflcldence angles toward
~e
hub,
led
to
a
radial
distribution of
flow
angle from the
rotor
which was
not
matched
to the
distribution
used
in
the design
of
the
stator. Experimental results for
the stage,
and
for
the
rotor only,
were
reported
by
Neuhoff
1985
&
1986).
While
the
Vavra stage
design was not
a correct one, it
was
qui~
successful
in
that it
facilitated the development
of a
transoruc
compressor
test
capability
at
the
av~
postgrad~te
Schoo~
and
of the
development
of
high
response mstrumentation techniques
to
determine
flow
behavior
in
transonic
stages
Shreeve
&
~euhoff
1984
and
Neuhoff et al,
1986).
However, a
new
stage
deslg~
was
sought, one for
which
detailed
meas~rements
woul~
proVide
a
meaningful evaluation
of current
design
and
analysIs
method~.
NASA
Lewis Research Center agreed
to
perform
the aerodynarruc and mechanical design
of
such
a
stage. This provided the
opportunity
to use the newly-emerging
CFD
tools,
supplemen~
by
traditional
methods
and, later,
to
have the
procedure
tested agamst
experimental results. The
purpose
of
~e
present
pa~r.
is to
present the
design
that
was
obtained, With. a descnption of
the
design/analysis
process that
was
used
to
amve at
the
final geometry.
T ST
FACn.ITY The
open loop
test
facility
and
present compressor stage
is
shown in
Figure
1.
Air
enters
through
filters
in
a
housing
which surrounds
the inlet piping.
Within
the
housing an
intake throttle
valve is
attached
to
a settling
chamber.
The throttle
valve
consists of
two
plates
having
identical
hole patterns;
the
plates rotate
with
respect
to
each
other
and
are
driven
by
an
hydraulic actuator. The
flow then
passes
through
perforated
plates
in
the O.813-meter 32-inch)
diameter
settling chamber; following the
settling chamber,
the
flow
passes
into
a 0.457-meter (l8-inch) diameter
pipe
in
which
there is a calibrated
nozzle.
The
flow
enters
the
compressor
through
a
O.279-meter
(ll-inch)
diameter
inlet pipe
and
exits
radially. A
honeycomb section following the
stator
removes
any
circumferential velocity
component
from
the
flow. Thus the
torque
supplied
to
the
rotor
is
equal
to
the torque
experienced
by
the
stator
and
honeycomb
section. The stator
section
is
mounted on
ball
races
and
is free
to
rotate against
flexures instrumented
with
strain
gauges.
The strain gauge
reading
is
then
a
measure
of
the
torque
supplied
to
the
rotor. The
compressor
rotor
is
driven by
an
opposed-rotor single stage
air
turbine
supplied by
air
from
the
laboratory
compressed
air
supply.
The
compressed
air
is
supplied
to
the
laboratory
by
a
12-stage Allis
Chalmers
axial
compressor.
DESIGN
PROCEDURE
Preliminary Design.
Design
goal
as
to
achieve
as high
a loading
and specific weight
flow
as
was
practical,
while keeping
rotor
tip Mach number
at a
moderate
level. Chief constraint
was
power
available
from
the
drive
turbine. The
results of a parametric
study
are
shown
in
Figure 2.
Specific
head-rise
was
calculated
for
a series
2
of
tip speeds
and
pressure ratios
( 'If
=CpTOl
PR
y-1fT)
-1
/
rf
.
After consideration of
various
configurations, a target
design was
specified
which has
higher
loading
and
lower
tip speed
than
the
original Vavra
design.
The
loading
is
also
slightly higher
than
the
NASA
Rotor
67,
which is
also
plotted
on
Figure
2.
The power required
by the
target
design
341
kw)
was set
very near the
available (limiting) power available
from
the
turbine drive
354
Jew).
To minimize fabrication expenses, the
external
flow
path and
stator
mechanical design
was
retained.
This
restricted
aerodytiamic design
choices, but
was
considered necessary
for
economic
reasons. A
new
conical
spinner
for
the
rotor
was
designed having
a constant 28-degree
ramp angle. This
angle
is
effectively
set by
the
choice
of pressure ratio
and
aspect
ratio. A
review
of
in-house and
contractor
designs
led to
the choice
of
an
approximately
constant radial distribution of exit
total
pressure; exit
total
temperature
energy
addition)
was
chosen
to
be nearly
radially constant,
but
higher
in
the tip region
to
compensate for shock losses
(see
figure
16
later).
These
distributions of
total
pressure
and
temperature,
the
overall pressure
and
temperature ratio,
flow
rate
and
the
new flow
path were entered into
a streamline curvature
design
program. During
the
preliminary
design
process, before
final design pcuameters
were
chosen,
various combinations
of
parameters were assessed
until
detailed distributions of parameters
such
as
diffusion
factor
(
D=1-V
'/V;+ :..v;/2av;
)
and
loss appeared
reasonable
in
light of contemporary experience. Specific empirical loss sets or
deviation angle
distributions
were not used. Deviation angle estimates
for
the
controlled-diffusion stator were
based
on
. experimental results
from the
controlled-diffusion stator
designed by
Sanger (Sanger,
1982) and
tested
at
the Naval
Postgraduate
School
cascade
facility. A
summary
of
the design
parameters
is
given
in
Table
I.
Computer Codes.
Several computer
codes
were
used
in
the
design
process.
Streamline curvature:
The
NASA
in-house
streamline-curvature
code
was used
for
preliminary
calculations
and
production
of
fabrication
coordinates. The
main-frame version
of
the code was documented
in
NASA
TP-1946
Crouse
&
Gorrell,
1981).
An
improved
PC-version, available
from and
supported
by
James
E.
Crouse,
was used in
this design. This version has
shown
less
sensitivity
to
blade section stacking
in
high hub
ramp angle
cases
as
encountered
in
this design.
Steady,
axisymmetric flow
is
assumed, thus reducing the problem
to
solving the two-dimensional
flow
in
the meridional plane.
The
full
radial
equilibrium equation
forms
the basis
of
the code.
The
aerodynamic solution gives
velocity
diagrams
on
selected
streamlines of revolution at
the
blade edges.
There
are
no
calculation stations inside
the blade row.
There.
are
numerous
options for controlling
the
form
of
input
and
for
specifying
the'
amount
of output.
Blade element code:
The
geometry
portion of
the
streamline-' curvature
code
has
been extracted
and
combined with
in-house graphics
to
run on
the Lewis main-frame
computer. This
short
code affords a convenient
means
for
designing
individual blade
sections
according
to
the
criteria
established
in
the
streamline-curvature code.
The meanline
of
the blade
is
described
by
two polynomial segments, each
of
which
can
be
specified
by up
to
quartic polynomial. The polynomial is a fit
of
local blade angle against mean-line distance. The distribution
of
blade thickness about the mean-line is also specified
by
two polynomials, each
of
which
may
be quartics. The thickness is added symmetrically on either side of
the
mean-line. A more complete description
of
this process is given
by
Sanger (1982).
Delllon TlP D:
The 3D analysis code employed
was
a hybrid code developed
by
Denton and
named
TIP3D.
t
was
originally reported in
1986
(Denton, 1986), but
has
since
been
expanded
and
upgraded.
t
uses
the
original Denton Euler code
as
a base (Denton, 1982), but couples it with a relatively simple viscous approximation
in
order
to
account
for
blockage and secondary
flow
effects.
In
the approximation, viscous effects are taken into account
by
including a body force
term
in the momentum equation
and
a source term
in
the
energy equation.
Wall
shear stress
is
approximated
by an
empirical equation. The distribution of shear stress
from
the wall
is
obtained
from
the Boussinesq eddy viscosity
model.
The viscous effects
can be
cut off at some fraction
of
passage width
by
providing that fraction
as
input. With sufficient grid points within the boundary layer one
can
obtain a reasonably
good
prediction
of
the
flow
in
a
real
turbulent boundary layer,
and
at a considerable saving in computing time
and
cost.
An
additional improvement
to
the original code
is
the inclusion
of
tip clearance flow. The code
is
not capable of calculating the details
of
the
flow
in a tip clearance region, but is capable
of
including the effect
on
the primary
flow and
on the overall performance. The leakage
is
predicted simply
by
reducing the blade thickness to zero in the clearance gap, and
by
applying a periodicity condition within the gap
in
exactly the same
way
periodicity
is
applied upstream
and
downstream
of
the blade
row.
This enables
flow
to
pass over the blade tip,
and
ensures that
no
load
is
carried within the gap. A typical
run
time for a transonic rotor blade
row was
20 minutes
on
a Cray
YMP.
anel
Code:
Because the stator
flow
field
is
subsonic, a subsonic 2D panel code
was
used to
screen blade sections. The entire stacked 3D stator
was
then analyzed
by
the TIP3D code. The panel code (McFarland, 1984), employs a surface singularity method which solves the
invisCid,
irrotatfonal,
cOmpressible
blade-to-blade
flow
equations on a surface
of
revolution. Streamsheet thickness
can
be incorporated as a function of meridional distance. The governing equations are linearized
by
approximating compressibility effects, and solved using an integral technique (panel method). Blade sections were created with the blade geometry code
and
analyzed with the panel code and
an
integral boundary layer code (described below). This provided a quick screening method
and
allowed development
of
a series
of
blade sections which showed no boundary layer separation at the design point. A typical run time on a 386
PC
(25
MHZ
CPU
is
30 seconds.
Boundary
ayer
Code:
An
integral boundary layer code (McNally, 1970)
was
used
in conjunction with the panel code
to
screen the
2D
blade sections. The
design
point surface velocity distributions calculated
by
the panel code were entered into
the
boundary layer code
and
the condition of the turbulent boundary layer
was
monitored. No blade section
was
accepted
which
indicated that turbulent separation
had
occurred (incompressible
form
factor must be less
than
2.0)
3
The McNally. code
uses
integral methods
to
solve the
two-
dimensional,
Compressible
laminar
and
turbulent boundary layer equations
in
an
arbitrary pressure gradient. Cohen
and
Reshotko's (1956)
method is
used
for the laminar boundary layer, transition
is
predicted
by
the
Schlichting-Ulrich-Granville (Schlichting,
1979
method,
and
Sasman's and Cresci's (1966)
method
is used for the turbulent boundary layer. A
typical
run time on the Lewis Amdahl 5870 main-frame running under
the
VM operating system
is
10
seconds. Rotor Design. Velocity triangles were set
with
the streamline curvature code. Blade shapes were then fitted
to the
triangles using both quasi-3D
and
full-3D calculations. Rotor
tip
solidity
was
set at 1.3. At that solidity
and
a
tip
inlet Mach relative number
of
1.3, it
was
verified that
the
leading edge oblique shock would
be
contained
in
the blade passage, striking the suction surface at about
80
to 90 percent
of
chord. The radial distributions
of
inlet relative Mach number
and
of D-Factor
for
the rotor are
shown
in
Figure 3. Initially, rotor blade sections were screened
using
a quasi-3D version
of
the Denton Euler code which
had
a transpiration model
to
simulate the boundary layer blockage. Surface boundary layer condition
was
checked using
the
McNally integral boundary layer code. Criteria
used
for
design
were
to
minimize supersonic acceleration
on
the uncovered portion
of
the blade
to
reduce shock strength,
and then to
control diffusion in
the
covered portion
to
prevent or delay boundary layer separation. This procedure provided a quick, initial screening
of
blade shapes, but with the advent
of
the full-3D code with body forces, it probably would be more effective to
use
the
3D
code and develop a
3D
geometry package
to
fit
and
stack tentative blade designs. Although a quasi-3D code
is
quick, the result
may
be inaccurate because transonic sections are quite sensitive
to
streamtube height, a userspecified quantity. When
the 3D
body force code became available, the blade sections originally defined using the quasi-3D code were
used as
a base design; subsequent alterations
to the
geometry were made based
on
the 3D calculations. Three blade sections
from
the final design are shown
in
Figure 4. Leading edge radii non-dimensionalized
by
chord varied
from
.002 at
the
tip
to
.0058 at
the
hub (Fig.
5 .
Trailing edge radius
to
chord
had
a similar distribution. Maximum thickness followed a linear variation
from
3-112
%
of
chord
at
the tip to 8
%
of
chord at the hub. The rotor leading edge wedge angles in the supersonic region
of
the blade were minimized
by
positioning the
maximum
thickness location well back
on
the blade. The maximum thickness location
was
varied
from
76
%
of
chord at the tip to 50 at the
hub
(Fig. 6). A meridional
view
of
the grid
used
in
the full-3D calculations
is
shown in Fig. 7. Slightly more
th n
100,000 points were used, with 97 points
in
the meridional direction
and
33
each
in
the radial
and
circumferential planes. The grid points were packed close
to
the blade surfaces,
hub and
casing,
and
meridionally from the leading and trailing edges
of
the blades
using
a geometric packing scheme. Expansion factors were 1.25
in
the
spanwise direction, 1.3
in
the blade-to-blade direction, 1.25
in
the meridiorJal
direction upstream
and
downstream
of
the blades,
and
1.1 from the
blade leading and trailing edges
in
the meridional direction inside the blade passage. A complete calculated rotor
map
is presented
in
Figure
8.
The calculated performance projects a pressure ratio
of
1.61
at design