This action might not be possible to undo. Are you sure you want to continue?
2L
2W
nonwetting
material
weting
material
2002 ME Graduate Student Conference
Apri l 13, 2002
HEAT TRANSFER AND FLUID FLOW IN AN IDEALIZED MICRO HEAT PIPE
J in Zhang
M.S. Candidate
Thesis advisor: Prof. Harris Wong
ABSTRACT
Micro heat pipes have been used as a heatdissipating
device in many systems, such as micro electronic
components and the leading edge of hypersonic aircraft.
[1, 2]
Micro heat pipes transfer heat by evaporation, convection,
and condensation, same as the conventional heat pipes.
However, the effective thermal conductivity of micro heat
pipes is only 1/40 that of conventional heat pipes. Due to
the complexity of the coupled heat and mass transport, and
to the complicated threedimensional bubble geometry
inside micro heat pipes, there is a lack of rigorous analysis.
As a result, the relative low effective thermal conductivity
remains unexplained. This work conceptualized an idealized
micro heat pipe that eliminates the complicated geometry,
but retains the essential physics. The simplified bubble
geometry allows a direct comparison between theoretical
predictions and experimental data.
The idealized micro heat pipe is a rectangular heat pipe,
the top portion of which is made of a nonwetting material,
and the bottom portion a wetting material. The lower
portion is filled to the rim by a wetting liquid, and the
upper portion is filled by its vapor. This configuration
ensures that the contact line of the liquidvapor interface is
pinned at the interception between the wetting and non
wetting materials. Pinning of the interface allows a capillary
pressure gradient to drive the liquid flow. When this micro
heat pipe is driven at small temperature differences, the
interface should be roughly flat, allowing the analysis to be
greatly simplified.
The evaporation and condensation in the idealized
micro heat pipe is analyzed. It is found that the interface can
be separated into two regions: an inner region near the wall
where evaporation occurs and an outer region away from the
wall. The evaporation rate is solved by the method of
matched asymptotic expansions, and the leading order
evaporation rate is obtained as εlnε, where ε measures the
ratio of conductive heat flux in the liquid to evaporative
heat flux at the interface. The small parameter ε is
where kf is the liquid thermal conductivity, T
∞
is the wall
temperature, c=(2πRT
∞
)
1/2
with R being the universal gas
constant per unit mass of the vapor as determined by the
kinetic theory, ρ
∞
is vapor density, hfg is liquid latent heat
and W is the half width of the pipe. The Stokes flow
induced by the surface tension gradient (Marangoni stress)
along the interface and by the evaporation at the interface is
solved using a finitedifference method.
Fluid flow
and heat transfer along the micro heat pipe
are also studied. Liquid temperature distribution along the
micro heat pipe is given in Figure 2. It is found that the
temperature profile is relatively flat except the region near
the evaporator, which is the evaporation region. The length
scale of the region is calculated as
where Aw and Af are crosssectional area of the wall and
liquid, respectively, and kw is wall thermal conductivity.
For a micro heat pipe with larger L/W the length of the
evaporation region is shorter. Vapor pressure distribution
along the micro heat pipe is also given in Figure 3. It is
clear that the pressure goes approximately linearly and not
affected strongly by L/W. Effective thermal conductivity keff
is evaluated. A longer or wider micro heat pipe will have a
larger keff. And it is found that increasing the evaporation
area at the evaporator will increase keff. It is also affected by
the latent heat of the working fluid. A fluid with larger
latent heat will produce larger keff.
FIGURES AND TABLES
Figure 1 an Idealized Micro Heat Pipe
δ =
π Awkw +Afkf
2 kf  ln ε
,
ε =
k
f T∞
c ρ∞hfg
2
W
,
2
ACKNOWLEDGMENTS
This work was supported by NASA and LaSPACE.
REFERENCES
1. G. P. Peterson, Appl. Mech. Rev. 45 (1992) 17589.
2. P. Dunn & D. A. Reay, in "Heat Pipes" (Pergamon
Press, 1982) 1618.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
L/W = 20
L/W = 50
L/W = 100
L/W = 200
p
*
z
*
p
z
0
50
100
150
200
250
0.0002 0.0004 0.0006 0.0008 0.001
L/W = 20
L/W = 50
L/W = 100
L/W = 200
k
eff
W
W
keff
Figure 4 Effective Thermal Conductivity along an
Idealized Micro Heat Pipe
Figure 3 Pressure Distribution along an
Idealized Micro Heat Pipe
Figure 2 Temperature Distribution along an
Idealized Micro Heat Pipe
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
L/W = 20
L/W = 50
L/W = 100
L/W =200
T
*
z
*
T
z
3
2002 ME Graduate Student Conference
April 13, 2002
PHYSICAL PARAMETER ESTIMATION OF VIBRATING STRUCTURE FROM ITS
SPECTRAL DATA: A NEW MATHEMATICAL MODEL
Kumar Vikram Singh
Ph.D. Candidate
Faculty Advisor: Y. M. Ram
ABSTRACT
The problem of reconstructing a model with prescribed
spectral data is known as inverse eigenvalue problem.
Reconstruction of the distribution of physical parameters of
a continuous vibratory system by using its eigenvalues is
addressed here. Considering a unit length piecewise
continuous rod as shown in figure 1. The eigenvalue
problem associated with this rod is given by the following
set of differential equations
¹
¹
¹
¹
¹
'
¹
′ · ′ ·
· ′ ·
· · + ′ ′
· · + ′ ′
) ( ) ( ) ( ) (
0 ) ( 0 ) 0 (
where, , 0
where, , 0
2
2
a v a u a v a u
L v u
sv v r
qu u p
γ
ω λ λ
ω λ λ
. (1)
Applying the boundary and matching conditions of
displacement and force leads to problem of finding the non
trivial solution of
o ·
,
_
¸
¸
1
1
1
1
1
1
1
]
1
¸
−
−
− −
4
3
2
sin cos 0
sin cos cos
cos sin sin
z
z
z
L L
a a a
a a a
β
ω
β
ω
β
ω
β
ω
β
ω
β
γω
β
ω
β
γω
α
ω
α
ω
β
ω
β
ω
α
ω
. (2)
We named this problem the Transcendental Eigenvalue
Problem (TEP). The general form of this problem is
( ) o z A · ω . (3)
Frequently the classical finite element and finite difference
formulation are used in approximating such a continuous
system. The characteristic equation of the obtained
eigenvalue problem is a polynomial. In contrast, the
continuous systems are characterized by TEP [1]. By using
finite element or finite difference method, the TEP is
transformed into an algebraic eigenvalue problem. It has
been concluded by [2,3] that the solution to the discrete
problem is not a good approximation to the continuous one.
Past research associated with inverse problems of the
continuous vibratory system can be found in [4,5,6,7,8].
Since the behavior of a finite dimensional polynomial is
fundamentally different from the transcendental function,
such an approach may involve inaccurate approximation of
the physical parameters, as illustrated in figure 2.
For the given continuous system in figure 1, the inverse
problem can be defined as follows:
Given the resonant frequencies
2 1
,ω ω , antiresonant
frequency
1
µ and the total mass of the rod.
Determinethe physical parameters
2 1
, p p and
2 1
, q q .
The problem now is of determining the roots of the system
of transcendental frequency equations,
( ) ( )
( ) ( )
( ) ( )
¹
¹
¹
¹
¹
'
¹
· ·
· ·
· ·
·
·
·
0 , , det ) , , (
0 , , det ) , , (
0 , , det ) , , (
1
2
1
3
2
1
µ ω
ω ω
ω ω
γ β α γ β α
γ β α γ β α
γ β α γ β α
A
A
A
F
F
F
, (4)
for the given values of
2 1
,ω ω and
1
µ . The research aims at
developing low dimensional analytical models allowing
estimation of the physical parameters of the structures from
measured vibration test data. The main idea presented here
is to replace the continuous system with variable physical
parameters by a continuous system with piecewise uniform
properties as shown in figure 3. The boundary and matching
conditions between the various parts of the continuous
model can be expressed in the TEP form. A rapidly
converged algorithm is used for evaluation of the physical
parameters of the system. The algorithm implements the
Newton’s iterative method for determining the physical
4
parameters of the system. Formulation of such mathematical
models for nonuniform axially vibrating rods and
reconstruction of their area distribution by using this
algorithm, as illustrated in figure 4, is presented. This
proposed solution of TEP can also be used to solve classical
direct problems in structural dynamics such as buckling [9]
and vibration control [10].
FIGURES AND TABLES
Fig.1. Piecewise continuous axially vibrating rod
Fig.2.Physical parameter Identification of piecewise rod
from its associated discrete model
Fig.3. New mathematical model used for the approximation
of a nonuniform rod
Fig.4. Reconstruction of the shape of the exponential rod
with model order n=4 and n=8
ACKNOWLEDGMENTS
The research work presented here is supported by a
National Science Foundation research grant CMS9978786.
REFERENCES
1. Singh K. V. and Ram Y. M., “A mathematical model to
overcome the discrepancies between continuous
systems and their discrete approximation”, ASME ETCE
2002.
2. Boley D. and Golub G.H., “A Survey of matrix eigenvalue
problems”, Inverse Problems, Vol. 3, pp. 595622, 1987.
3. Paine J., “A numerical method for the inverse Sturm
Liouville problem”, SIAM Journal on Scientific and
Statistical Computing, Vol. 5(1), pp. 149156, 1984.
4. Ram Y.M. and Caldwell J., “Physical parameters
reconstruction of a freefree massspring system from
its spectra”, SIAM Journal of Applied Mathematics, Vol.
52(1), pp. 140152, 1992.
5. Frieland S., Nocedal J. and Overton M.L., “The
formulation and analysis of numerical methods for
inverse eigenvalue problems”, SIAM Journal of
Numerical Analysis, Vol. 244, pp. 634667, 1987.
6. Ram Y.M. and Elhay S., “Constructing the shape of a
rod from eigenvalues”, Communications in Numerical
methods in Engineering, Vol. 14, pp. 597608, 1998.
7. Gladwell G.M.L., “Inverse problems in vibration”,
Applied Mechanics Review, Vol. 39, pp. 10131018, 1986.
8. Gladwell G.M.L., “Inverse Problem in vibration”, Martin
Nijhoff publishers, First Edition, 1986.
9. Singh K.V. and Ram Y.M., “The Transcendental
Eigenvalue Problem and Its Applications”, Accepted for
publication in AIAA Journal, 2002.
10. Singh K.V. and Ram Y.M., “Dynamic Absorption by
Passive and Active Control”, ASME Journal of vibration
and acoustics, Vol. 122(4), pp. 429433, 2000.
1
1
−
−
n
n
q
p
1
1
q
p
2
2
q
p
3
3
q
p
L
L
n
n
q
p
L
L
L
L
1
x
2
x
3
x
1 − n
x
n
x
n
u
1 − n
u
2 − n
u
3
u
1
u
2
u
h h
L
1
1
−
−
n
n
q
p
1
1
q
p
2
2
q
p
3
3
q
p
L
L
n
n
q
p
L
L
L
L
1
x
2
x
3
x
1 − n
x
n
x
n
u
1 − n
u
2 − n
u
3
u
1
u
2
u
h h
L
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
n=4 n=8
Original shape Estimated
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
n=4 n=8
Original shape Estimated
L
2 2 2
, , A E ρ
1 1 1
, , A E ρ
a
¹
'
¹
< < ·
< < ·
L x a A E r
a x A E p
2 2
1 1
0
p r s r q p · · · γ β α , ,
¹
'
¹
< < ·
< < ·
L x a A r
a x A q
2 2
1 1
0
ρ
ρ
Axial Rigidity Density
0 2 4 6 8 10 12 14 16 18 20
0
1
2
3
4
5
6
Identified area
Theoretical area
0 2 4 6 8 10 12 14 16 18 20
0
1
2
3
4
5
6
Identified area
Theoretical area
Identified area
Theoretical area
5
2002 ME Graduate Student Conference
April 13, 2002
EFFECT OF WALL THICKNESS OF CENOSPHERES ON THE COMPRESSIVE
PROPERTIES OF SYNTACTIC FOAMS
Nikhil Gupta
Ph.D. Candidate
Faculty Advisor: Dr. Eyassu Woldesenbet
ABSTRACT
Cenospheres are incorporated in polymeric materials to
obtain composites of low density and high compressive
strength, known as syntactic foams. Some studies on the
modeling and experimental behavior of such composites are
available in the literature [114]. However, no comprehensive
studies could be found which characterize the behavior of
syntactic foams with respect to various parameters like
cenosphere wall thickness (density) and size distribution.
This experimental work investigates the effect of wall
thickness of cenospheres on the compressive properties of
syntactic foams. As the matrix material D.E.R. 331, a diepoxy
resin, manufactured by DOW Chemical Company was
selected. This resin is called diglycidyl ether of bisphenol A
(DGEBA). To lower the viscosity of the resin a diluent is
added to it. It is difficult to mix large volume of cenospheres
in the epoxy resin if the viscosity is very high. Adding 5% of
diluent, C12C14 aliphatic glycidyl ether, brings down the
viscosity of the resin from about 9000 cps at 20°C to about
2000 cps at the same temperature. Average equivalent
epoxide weight (EEW) of the diluent is 285. For a 95 wt%
resin and 5 wt% diluent mixture the EEW is 177.5.
Triethylene tetramine (TETA), C6H18N4, is used as curing
agent. This chemical is commercially known as D.E.H. 24 and
manufactured by DOW Chemical Company. Molecular
weight of this hardener is 146.4 and weight per active
hydrogen is 24.4. Phr (parts per hundred parts of resin) of
amine hardener for 955 resindiluent mix was calculated to
be 13.74. Stainless steel molds having inner dimensions of
9×9×0.5 in
3
are used to cast the syntactic foams. Four
different types of cenospheres were used for the fabrication
of syntactic foam specimens. These microballoons were
manufactured and supplied by 3M. Specimen size used for
the test was 12.7×12.7×25.4 mm
3
. Compression test was
conducted at constant crosshead movement rate of 0.5
mm/min. Minimums of five specimens of each type of
syntactic foam were tested. Stressstrain cures for each type
of specimens are presented here [Figs. 25]. Trends of peak
stress [Fig. 6] and modulus [Fig. 7] of the specimens with
respect to cenosphere density are also presented.
FIGURES AND TABLES
Table 1. Particle size distribution of microballoons.
Ceno
sphere
type
Average
Particle Size
(µm)
Top Size
(µm)
Cenosphere
density
(g/cc)
Syntactic
foam density
(kg/m
3
)
S22 35 75 205 493
S32 40 80 320 545
K37 40 85 380 575
K46 40 80 460 650
Fig.1. Compressive fracture behavior of syntactic foam.
S22, ASTM D695
0
5
10
15
20
25
30
35
0 0.05 0.1 0.15
Strain (mm/mm)
S
t
r
e
s
s
(
N
)
Fig. 2. Compression test results of S22 syntactic foam.
6
S 32, ASTM D695
0
10
20
30
40
50
0 0.05 0.1 0.15
Strain (mm/mm)
S
t
r
e
s
s
(
N
)
Fig. 3. Compression test results of S32 syntactic foam.
K 37, ASTM D695
0
10
20
30
40
50
60
0 0.02 0.04 0.06 0.08 0.1
Strain (mm/mm)
S
t
r
e
s
s
(
M
P
a
)
Fig. 4. Compression test results of K37 syntactic foam.
K 46, ASTM D695
0
20
40
60
80
0 0.02 0.04 0.06 0.08
Strain (mm/mm)
S
t
r
e
s
s
(
N
)
Fig. 5. Compression test results of K46 syntactic foam.
Change in Peak Compressive Strength
6000
7000
8000
9000
10000
11000
12000
200 300 400 500
Cenosphere Density (kg/m
3
)
P
e
a
k
S
t
r
e
s
s
(
M
P
a
)
Fig. 6. Dependence of Peak strength of syntactic foam on
cenosphere density.
Cenosphere Density Dependence of
Modulus
1000
1200
1400
1600
1800
2000
2200
200 250 300 350 400 450 500
Density of Cenospheres (Kg/mm
3
)
M
o
d
u
l
u
s
(
M
P
a
)
Fig. 7. Dependence of Modulus of syntactic foam on
cenosphere density.
REFERENCES
1. Gupta, N., Kishore, Woldesenbet, E., Sankaran, S.; J.
Mater. Sci., 36, 18 (2001) 44854491.
2. Gupta, N., Brar, B. S., Woldesenbet, E.; Bull. Mater Sci.,
24, 2 (2001) 219223.
3. Gupta, N., Kishore, Sankaran, S.; J. Reinf. Plast. Compo.
18, 14 (1999) 13471357.
4. Gupta, N., Karthikeyan, C. S., Sankaran, S., Kishore;
Mater. Charact., 43, 4 (1999) 271277.
5. Gupta, N., Woldesenbet, E., Kishore, Sankaran, S.; J.
Sand. Str. and Mater., accepted, in press.
6. Gupta, N., Woldesenbet, E., Kishore; J. Mater. Sci.,
accepted, in press.
7. Gupta, N., Woldesenbet, E.; in proceedings of ETCE
2002, Feb 2002, Houston, TX.
8. Gupta, N., Woldesenbet, E.; (accepted) in proceedings of
10
th
USJapan Conference on Composite Materials,
September 1618, 2002, Stanford University, CA.
9. Gupta, N., Woldesenbet, E.; in proceedings of ASC 16
th
Annual Conference, Blacksburg, VA, Sept. 912, 2001.
10. Gupta, N., Woldesenbet, E.; in proceedings of ETCE
2001, Feb 2001, Houston, TX.
11. Rizzi, E., E. Papa and A. Corigliano, Intl. J. Solids and
Struct. 37 (2000) 5773.
12. Corigliano, A., E. Rizzi and E. Papa, Compo. Sci. Tech. 60
(2000) 2169.
13. Ishai, O., C. Hiel and M. Luft, Composites 26, 1 (1995) 47.
14. Malloy, R. A., J. A. Hudson, in Intl. Encylop. of Compos.,
Ed. S. M. Lee (VCH, 1990).
7
2002 ME Graduate Student Conference
April 13, 2002
EFFECT OF TIP GEOMETRY ON BLADE TIP FLOW AND HEAT TRANSFER
David Kontrovitz
M.S. Candidate
Faculty Advisor: Dr. Srinath Ekkad
ABSTRACT
In an attempt to increase thrust to weight ratio and
efficiency of modern gas turbines, engine designers are
always interested in increasing turbine operating
temperatures. The benefits are attributed to the fact that
higher temperature gases yield a higher energy potential.
However, the detrimental effects on the components along
the hot gas path can offset the benefits of increasing the
operating temperature. The HPT first stage blade is one
component that is extremely vulnerable to the hot gas.
The cause for tip failures are fairly well understood and
can be explained as follows. A clearance gap between the
rotating blade tip and stationary shroud is necessary to
allow for the blade’s mechanical and thermal growth during
operation. Unfortunately, the gap allows for leakage flow
from the pressure side to the suction side. The gas is
accelerated as it passes through the small gap. This leads to
enhanced heat load to the blade tip region. Leakage flow, or
clearance flow, also leads to undesirable aerodynamic losses
not unlike the losses associated with airplane wing tips. In
fact, one third of the losses through the turbine section can
be attributed to leakage flow. Other relevant studies by Azad
et al. [12], Bunker et al. [3], Bunker and Bailey [45]
presented detailed heat transfer results on highpressure
turbine blade tips with different pressure ratios. The effect of
tip geometry was also considered in some of these studies.
The present study explores the effects of gap height
and squealer depth on heat transfer and flow distribution.
This investigation differs from those in the other studies
because the tip profile is for an inservice High Pressure
Turbine of an aircraft engine. Other experiments have used
t he E
3
test blade or a power generation blade that have
different characteristics. The pressure ratio used was 1.2,
which is lower than the actual pressure ratio this blade sees
in service (PR = 1.7). A transient liquid crystal technique
was used to obtain the tip heat transfer distributions.
Pressure measurements were made on the blade surface and
on the shroud for different tip geometries and tip gaps.
FIGURES AND TABLES
Figure 1. Leakage Flow
Figure 1 shows the typical leakage flow behavior for a
turbine blade [1]. The plain tip is a flat tip and flow leaks
through a constant area across the blade. The squealer tip
has a groove cut on top of the blade which increases the
area and stalls the flow thus creating back pressure and
restricts leakage flow and reduces heat transfer. In this
study, we focus on the plain tip and two different squealer
depths (shallow and deep).
8
Figure 2: Blade Pressure Distribution
Figure 2 shows the pressure distribution on the blade
surface at different span of the blade. The 100% span is on
the tip of the blade with clearance gap. The pressure
distribution changes as the blade span moves towards the
gap as expected. Figure 3 shows the pressure distributions
on the shroud for a blade with squealer tip. The reduced
static pressures are the cause of reduced leakage flows.
Figure 3: Shroud Measurements
Figure 4 presents a typical heat transfer distribution on
the blade tip with a shallow squealer. The heat transfer
distributions show the local hot spots near the leading edge
on the floor of the cavity and the reduced heat transfer
towards the trailing edge of the blade. The leakage flow is
stronger at the leading edge and weaker along the trailing
edge of the blade.
Figure 4: Heat Transfer Distribution
ACKNOWLEDGMENTS
This study was sponsored by the NSF through a
GOALI grant. The author would like to acknowledge Drs.
Srinath Ekkad and Sumanta Acharya for their instruction and
advice. Also, thanks to Dr. Ron Bunker, at G.E. Corporate
Research and Development, for his input to this project.
Acknowledgments are also due to my colleagues in the
Turbine Blade Research Lab.
REFERENCES
1. G.S. Azad, J.C. Han, R.S. Bunker, C.P. Lee, Effect of
Squealer Geometry Arrangement on Gas Turbine Tip
Heat Transfer, in Proceedings of the ASME International
Mechanical Engineering Congress and Exposition, New
York, November 2001, HTD24314
2. G.S. Azad, J.C. Han, R.J. Boyle, Heat Transfer and Flow on
the Squealer Tip of a Gas Turbine Blade, in Proceedings
of the 2000 ASME Turboexpo, Munich, Germany, May
2000, 2000GT1955
3. R.S. Bunker, J.C. Bailey, A.A. Ameri, Heat Transfer and
Flow on the First Stage Blade Tip of a Power
Generation Gas Turbine Part1: Experimental Results, in
Proceedings of the 1999 ASME International Gas Turbine
Conference, Indianapolis, 1999, ASME 99GT169.
4. R.S. Bunker, J.C. Bailey, Effect of Squealer Cavity Depth
and Oxidation on Turbine Blade Tip Heat Transfer, in
Proceedings of the 2001 ASME International Gas Turbine
Conference, New Orleans, June 2001.
5. R.S. Bunker, J.C. Bailey, An Experimental Study of
Heat Transfer and Flow on a Gas Turbine Blade Tip
With Various Tip Leakage Sealing Methods, in
Proceedings of the 4th ISHMT / ASME Heat and Mass
Transfer Conference, Pune, India, 2000, HNT2000055 p.
411416.
9
2002 ME Graduate Student Conference
April 13, 2002
LARGE EDDY SIMULATIONS OF ROTATING SQUARE DUCT WITH NORMAL RIB
TURBULATORS
Mayank Tyagi
Ph.D. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
The goal of the present work is to develop a generalized
Large Eddy Simulations (LES) methodology for complex
geometries, and to apply this methodology to gas turbine
blade cooling applications. Results from one such
application, that of internal cooling in a turbine blade, are
reported in this abstract.
The internal cooling configuration selected in this study
corresponds to the experimental study of Wagner et al.
(1992). The computations were performed at a Reynolds
number (Re) of 12,500, rotation number (Ro) of 0.12 and the
inlet coolanttowall density ratio (∆ρ/ρ) of 0.13. The rib
heighttohydraulic diameter ratio (e/D) is 0.1 and the rib
pitchtoheight ratio (P/e) is 10. The ribs are square in cross
section and are placed transverse to the flow in the duct. Of
specific interest in this study are the dynamics of the
coherent structures, and how they influence the heat
transfer.
R
i b
A
R
ib
B
T
ra
ilin
g
W
a
ll
L
e
ad
i n
g
W
a
ll
BackWall
Front Wall
Inflow
Rotation
L
z
L
y
L
x
Rib Height : e
Rib Xsection : Square
L
x
= L
y
= L
z
= L
e/L = 0.1, R
mean
= 49.5
P/L = 1.0
Re
m
= 12500
Ro = 0.12
P/2
P/4
∆ρ/ρ· 0.13
Figure 1: Schematic of the computational domain
A novel direct method for computing the source term in
the energy equation (which arises due to unsteadiness and
uniform wall temperature in periodic geometries) is
presented. This is in contrast to the iterative approach of
Wang and Vanka, 1995, which has been used to date.
The details of flow field and the temperature field
obtained from the LES are presented and analyzed in this
study. The LES procedure is based on a dynamic mixed
model for subgrid stresses and scalar flux.
RESULTS
Rib A
Rib B
Rib A
Rib B
Timeaveraged Velocity Vectors at Y/D = 0.5
Figure 2: Details of timeaveraged vectors near the ribs
Figure 2 shows the timeaveraged flow field at the spanwise
centerplane, and reveals the recirculating eddies generated
in the vicinity of the ribs. The coherent structures (Figure 3)
are educed from the instantaneous flow field using the
positive levels of Laplacian of pressure field. Near the
trailing wall, the Coriolis induced secondary flow lead to a
breakup of the flow structures along the spanwise center
plane. The roller vortices formed on the rib are connected
with the streamwise or braid vortices originating either at the
side walls or around the centerplane, thus forming a hairpin
coherent structure. The unsteady dynamics show several
10
packets of such hairpin vortices between the ribs at trailing
wall. The vortices near the leading wall are weaker than the
vortices near trailing wall. Moreover, the roller vortices
generated at the ribs are stretched and tilted into the core of
duct by the secondary flow. Hairpin packets on these ribbed
walls evolve in different fashions.
X
Z
Y
Leading wall
Traili ng wall
X
Y
Z
Leading wall
Traili ng wall
Figure 3: Coherent structures in a rotating ribbed duct
The proper orthogonal decomposition (POD) of the flow
field using the method of snapshots is performed (Sirovich,
1987). About 99% of the turbulent energy is captured in the
first 75 POD modes (Figure 4). Clearly, the chaotic dynamics
of these coherent structures can be modeled by a low
dimensional system.
POD mode number
E
n
e
r
g
y
c
o
n
t
a
i
n
e
d
i
n
t
h
e
P
O
D
m
o
d
e
50 100 150 200
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
Figure 4: Distribution of energy in POD modes based on
200 snapshots.
Heat transfer calculations show the unsteady hot streaks on
the duct walls. The distribution of the timeaveraged Nusselt
number is in agreement with the experimental observations.
It is illustrated that scalar mixing is related to the scalar
dissipation of the temperature.
ACKNOWLEDGMENTS
Discussion with Dr. A.K. Saha on the preliminary results
is gratefully acknowledged.
REFERENCES
1. Sirovich, L. (1987) Turbulence and the dynamics of
coherent structures, Part IIII, Quarterly of Appl. Math.,
XLV(3), pp. 56182.
2. Tyagi, M., Saha, A.K. and Acharya, S. (2001) Large eddy
simulations of rotating square duct with normal rib
turbulators, DNS/LES: Progress and Challenges, 3
rd
AFOSR Intl. Conference, Eds. C. Liu, L. Sakell and T.
Beutner, pp. 807814.
3. Wagner, J.H., Johnson, B.V., Graziani, R.A. and Yeh, F.C.
(1992) Heat transfer in rotating serpentine passages
with trips normal to the flow, J. Turbomachinery, Vol.
114, pp. 847857.
4. Wang, G. and Vanka, S.P. (1995) Convective heat transfer
in periodic wavy passages, Int. J. Heat Mass Transfer,
Vol. 38, pp. 32193230.
1
2002 ME Graduate Student Conference
April 13, 2002
NODAL CONTROL OF A VIBRATING BEAM
Akshay N Singh
Ph.D. Candidate
Faculty Advisors: Dr. Y. M. Ram and Dr. SuSeng Pang
ABSTRACT
Vibration control is an important engineering problem
and many methods for both active and passive vibration
absorption have been developed. This paper deals with
developing a method to achieve nodal control at the point of
excitation in a BernoulliEuler beam. Singh and Ram in [3]
have shown that under certain conditions that have been
characterized in [3] the steady state motion of a certain
degree of freedom in a harmonically excited conservative
system may be absorbed by both passive and active means.
Ram in [1] has developed a method to eliminate the steady
state motion of a prescribed location in a continuous system
like rod under the influence of a harmonic excitation. He has
presented a closed form solution for the control gain in
terms of infinite product of eigenvalues. This thesis extends
the approach in [1] to achieve nodal control for suppressing
vibration at prescribed location in beams and provides a
simpler formula for the control gain in terms of
eigenfunctions. It is established that, for a uniform
BernoulliEuler beam, the steady state motion at the point of
excitation can be absorbed by means of a control force
determined from displacement information at the point of
application. A closed form solution for the control gain is
presented and a criterion for implementing the control by
active and passive means is developed. The result for the
control gain is generalized for the case of a nonuniform
beam. It is also shown through some examples that the
theory can be also applied to eliminate the steady state
motion at any desired location other than the point of
excitation. Analysis is also performed to determine the
optimal control force and investigate the stability of the
overall system. Several controllability graphs are shown and
meaningful conclusions are drawn from these graphs. An
experiment is designed to validate the proposed theory and
display its practicality.
The developed theory will provide a strong foundation
for realizing realistic and convenient methodologies in
control applications in cases like surgical procedures,
drilling and turning operations etc. However, one of the
many direct applications of this method is structural
vibration control in an aircraft wing. Several measurements
such as vibrational response, air temperature, wind velocity
etc are required in order to monitor flight conditions in an
aircraft. These data also assist the pilot in flying the aircraft.
Sensors and data collection circuitry form an entire network
of the electrical wiring all in and around the airplane body.
Data acquisition devices are also located on the wing of the
airplane. Shielding of these devices from undesirable
vibration of the wing is critical in order to avoid noise in the
gathered data and prevent physical damage. Exclusion of
steady state vibration at the locations of these devices
provides the motivation for this investigation.
Consider a uniform BernoulliEuler beam of length L .
Suppose that the beam is excited by a harmonic force
( ) t t f ω cos · , as shown in Figure 1(a). The steady state
motion of a prescribed point of the beam may be vanished
by applying a concentrated control force ) , ( t a u at a x· as
shown in Figure 1(b).
( ) ( ) t a w t a u , , α · ,
(1)
where α is the control gain. The partial differential equation
for the controlled BernoulliEuler beam shown in Figure 1(b),
for L x < < 0 , 0 > t , is given as
( ) w a x
t
w
A
x
w
EI
x
α δ ρ − ·
∂
∂
+
,
_
¸
¸
∂
∂
∂
∂
2
2
2
2
2
2
,
(2)
and the boundary conditions are
0 ) , 0 ( · t w no deflection,
(3)
( )
0
, 0
·
∂
∂
x
t w
, no slope,
(4)
2
( )
0
,
2
2
·
∂
∂
x
t L w
, no moment,
(5)
( )
t
x
t L w
EI
x
ω cos
,
2
2
·
,
_
¸
¸
∂
∂
∂
∂
shear force.
(6)
The work here focuses on determining a closed form
solution for the control gain α that absorbs the motion of
the beam at L x · .
A closed form expression for the control gain α
obtained from mathematical manipulation is expressed as
( )
( ) ( )
( )
,
_
¸
¸ ′ ′ ′ − ′ ′ ′
·
a v
a v a v
EI a
1
1 2
,ω α
,
(7)
where
1
v and
2
v are the deflections of the beams with span
a x < < 0 and L x a < < respectively. The values of the
control gain
) , ( η αa
for 2 . 0 · a , as function of the non
dimensional parameter
L β η·
are shown in Figure 2.
FIGURES
Figure 1: Vibration control of a harmonically excited beam
Figure 2: Plot of control gain α against nondimensional
parameter η
ACKNOWLEDGMENTS
The research was supported in part by a National
Science Foundation research grant CMS9978786.
REFERENCES
1. Ram Y. M., Nodal Control of a Vibrating Rod,
Mechanical Systems and Signal Processing, 2002
(to appear).
2. Den Hartog J. P., Mechanical Vibrations, 4th
edition, McGraw Hill, 1956.
3. Singh K. V., Ram Y. M., Dynamic Absorption by
Passive and Active Control, Journal of Vibration
and Acoustics, Vol. 122 (4), pp. 429433, October
2000.
4. Inman D. J., Engineering Vibration, Prentice Hall,
1994.
5. Ram Y. M., Elhay S., The Theory of a Multi Degree
of Freedom Dynamic Absorber, Journal of Sound
and Vibration, pp. 607615, Vol. 195, 1996.
6. Singh K. V., Ram Y. M., Numerical Algorithm For
The Transcendental Eigenvalue Problem, AIAA
Journal (accepted).
7. Elhay S., Gladwell G.M.L. Golub, G.H. and Ram
Y.M., On some EigenvalueEigenvector Relations,
SIAM Journal on Matrix Analysis and
Applications, 20(3), pp. 563574, 1999.
8. Mottershead J.E., On the zeros of structural
frequency response functions and their
sensitivities, Mechanical Systems and Signal
Processing, Vol. 12, pp. 591597, 1998.
9. Mottershead J.E. and Lallement G., Vibration nodes,
and the cancellations of poles and zeros by unit 
rank modifications to structures, Journal of Sound
and Vibration, Vol. 222, pp.833851, 1999.
x
( ) t t f ω cos ·
( ) t x w ,
L
(a) Uncontrolled beam
x
( ) t t f ω cos ·
( ) t x w ,
L
(b) Controlled beam
( ) ( ) t a w t u , α ·
a
x
( ) t t f ω cos ·
( ) t x w ,
L
(a) Uncontrolled beam
x
( ) t t f ω cos ·
( ) t x w ,
L
(b) Controlled beam
( ) ( ) t a w t u , α ·
a
η
0 2 4 6 8 10 12
α
3000
1000
0
1000
3000
η
0 2 4 6 8 10 12
α
3000
1000
0
1000
3000
α
3000
1000
0
1000
3000
13
2002 ME Graduate Student Conference
April 13, 2002
ON THERMALLY INDUCED SEIZURES (TIS) IN JOURNAL BEARINGS
Rajesh Krithivasan
M.S. Candidate
Faculty Advisor : Dr. Michael M Khonsari
ABSTRACT
Thermally induced seizure (TIS) in journal bearings is a
mode of failure that can occur quite suddenly and end up
with a catastrophic damage to the system. A failure, as such,
can occur quite suddenly and often the damage to the
system is catastrophic. Although it can take place in
lubricated bearings, thermally induced seizure is
predominant when a hydrodynamic bearing happens to
operate in the boundary or mixed lubrication regimes. These
conditions occur during start up or in an event of lubricant
supply blockage. The objective of this work is to perform a
comprehensive study of seizure in bearings during start up
and arrive at a seizure time evaluation formula that is a
function of the various operating parameters. Dufrane and
Kannel
1
analyzed the catastrophic seizure of bearings due to
dry friction by a simple 1D equation relating the seizure time
to the bearing operating parameters and material properties.
Hazlett and Khonsari
2
performed a detailed finite element
analysis to gain insight into the nature of the contact forces
and encroachment of the mating pair leading to TIS of a dry
bearing during start up.
The finite element modeling is done using ANSYS 5.7
3
.
First, the TIS analysis of Hazlett and Khonsari
2
was
recreated. The finite element model of the present work
employs a finer mesh than the mesh used by Hazlett and
Khonsari to evaluate the contact forces with more accuracy.
The analysis of a bearing undergoing TIS during start up
was done by the following steps: 1. A 2D static contact
analysis was performed to determine the contact forces and
the contact angle. 2. A transient heat transfer analysis was
done to model thermal effects of dry frictional heating on the
journal and the bearing. 3. A transient thermoelastic analysis
was performed to study the interactions of the journal
bearing pair during bearing start up. The variation of radial
clearance, contact forces and ovalization of the bearing were
studied in this analysis.
The loading applied in the thermal analysis consists of
the heat generated by the frictional contact at the shaft 
bushing interface, which is a function of the load, speed and
coefficient of friction. The heat generated is applied to the
journal and the bushing according to the areas of contact on
the shaft and the bushing
2
and cooled convectively at the
areas not in contact. The external surface is also cooled by
free convection as shown in Fig 2. The loading for the non
linear thermoelastic analysis consists of the thermal loads
applied as nodal temperatures and the radial force acting on
the journal. The time dependent thermal load is obtained
from the results of the transient thermal analysis. The static
load, W is applied to act in the negative ydirection on the
shaft. As the model utilizes halfsymmetry, a load of W/2 is
applied. Symmetry boundary conditions are used to model
the onehalf symmetry as shown in Figure 3. The constraint
of the bearing on its outer surface is modeled by fixing the
bearing at the node under the shaft on the outer edge of the
bearing on the symmetry plane. This constraint
approximates the boundary condition on the bottom surface
of a pillow block type of bearing as shown in Figures 1 and
3.
Due to the rise in temperature, the encroachment of the
shaft on to the bushing with concomitant reduction in the
clearance continues until TIS occurs due to the increase in
frictional torque. This process is a complex, nonlinear
phenomenon. Analysis shows that TIS is initiated by the
ovalization of the bearing combined with the uniform
outward expansion of the shaft yielding contact between the
top of the shaft and the inner bushing surface. This leads to
an increase in the contact forces and the formation of
multiple contact areas. Increase of contact forces raises the
frictional heat flux and sets up a positive feedback that
accelerates the loss of clearance. Analyses show that the
increase in the frictional torque is abrupt once the
ovalization of the bearing causes the shaft to encroach the
bushing, as there is further loss in the operating clearance.
Seizure Criterion  Frictional torque is the torque
resisting the driving torque exerted by the motor. When the
frictional torque increases beyond the extent of the driving
torque capability, it can be concluded TIS is imminent. The
contact forces acting on the gap elements at any instant of
time determine the frictional torque at any time. The frictional
torque increased to exceedingly large values within typically
14
3 seconds after the first instance of establishment of new
areas of contact immediately after ovalization. See Figure 4.
The seizure time can be written as a function of the
speed, load, shaft radius, clearance, friction coefficient and
the bearing length. i.e.
t
s
= g (N, W, R
s
, C, f, L).
The variation of the seizure time during the system start up
is studied when the operating parameters (variables N, W, R
s
,
C, f, L) are varied. Then a generalized equation is derived
depending on the individual relationships of the operating
parameters with the seizure time. Using the range of data
from 72 sets of simulations and applying a statistical
analysis
4
, we obtain the following expression for the seizure
time.
( ) ( )
74 . 8 17 2 . 1 26 . 1 5 . 14 . 62860 7
10 7 . 5 1 10 5 . 1
− − − −
× + × ·
s
L C
s
R W fN e e t
The empirical relationship is verified for its validity using
some of the results published by Hazlett and Khonsari
2
and
Wang et.al.
5
. See comparitive results shown in Table 1.
FIGURES AND TABLES
REFERENCES
1. Dufrane K. and Kannel J. Thermally induced seizures of
journal bearings. ASME Journal of Tribology, 1989, 111,
28892
2. Hazlett T.L. and Khonsari M.M. Finite element model of
journal bearing undergoing rapid thermally induced
seizure. Tribology International, 1992a, 25, No.3, 17782
3. ANSYS 5.7 Online Users Manual, 2001, ANSYS Inc.
4. Hamrock B.J. Elastohydrodynamic lubrication of point
contacts. Ph.D Thesis, Institute of Tribology,
Department of Mechanical Engineering, The
University of Leeds, 1976, 4466, pp. 93102
5. Wang H., Conry C. and T.F., Cusano, Effects of
Cone/Axle Rubbing Due to Roller Bearing Seizure on
the Thermomechanical Behavior of a Railroad Axle.
ASME Journal of Tribology, 1996, Vol.118, pp.311319
15
2002 ME Graduate Student Conference
April 13, 2002
AN EIGENVALUE CONFORMING MODEL FOR A VIBRATIING ROD
Jaeho Shim
Ph.D. Candidate
Faculty Advisor: Dr. Y. M. Ram
ABSTRACT
The natural frequencies determined by using finite
element or finite difference models of order n are fairly
accurate only about 3 n of lower eigenvalues of the
underlying continuous system. The natural frequencies of a
uniform rod with 60 · n and exact solution are demonstrated
in Figure 1. The new model to improve this existing problem
has been developed. The model named as a spectral
conforming discrete model can estimate the n lowest
eigenvalues of the continuous system with uniform
accuracy. The essential ingredient in building of such a
model is the inverse eigenvalue problem of reconstructing a
chain of massspring system with prescribed spectral data
[1].
Consider a nonuniform axially vibrating rod of length
L , axial rigidity ( ) x p , and mass per unit length ( ) x ρ ,
which is fixed at 0 · x and free to oscillate at L x · , as
shown in Figure 2. The axial motion ( ) t x u , of the rod at the
time t is governed by the differential equation (1) and two
boundary conditions (2).
( ) ( )
2
2
t
u
x
x
u
x p
x ∂
∂
·
,
_
¸
¸
∂
∂
∂
∂
ρ
(1)
( ) 0 , 0 · t u ,
( )
0
,
·
∂
∂
x
t L u
, (2)
This nonuniform rod shown in Figure 3 may be
approximated as a piecewise continuous rod with r uniform
parameters
i
p and
i
ρ within the i th element. In order to
determine a higher order spectral conforming element model,
a uniform rod element of length L, axial rigidity p, and mass
per unit length
ρ
in Figure 4 is considered. A matrix A
which is reconstructed from spectral data is defined as
2
1
2
1
− −
· KM M A . (3)
The matrix Ais symmetric tridiagonal with positive diagonal
element
l
α and negative off diagonal
j
β .
1
1
1
1
1
1
]
1
¸
·
−
− − −
r r
r r r
α β
β α β
β α β
β α
1
1 1 2
2 1 1
1 1
O O O A
(4)
From the matrixA, the stiffness matrix K and the mass
matrix M can be determined using the Lanczos method [2].
The essential novelty in the spectral conforming model
introduced here is that the dynamic response of the
continuous system is fitted to the spectrum of the discrete
estimating model.
As an example, an exponential rod of length L ,
constant Young’s modulus of elasticity E , constant density
γ
, and variable cross sectional area
x
e A · , is presented.
The rod is fixed at 0 · x and free to oscillate at L x · , as
shown in the figure 5. The eigenvalues of this system have
been approximated by using finite differences and spectral
conforming model. The finite difference model implemented
60 · n elements of equal length. The spectral conforming
model used 4 · n elements of equal length, each of order
15 · r . Hence, the global matrices in the three approximating
methods used are of the same dimension 60 60× . The
various results obtained are shown in Figure 6. As expected
the spectral conforming model yields superior overall
estimation with uniform accuracy for all eigenvalues
predicted.
Future research in this topic involves extending the
method to include tapered elements that can better capture
the geometry of a nonuniform rod. Broadening the method
over two and threedimensional elements appears to be a
challenging problem.
16
FIGURES AND TABLES
Figure 1. Predicted natural frequencies using Finite
Difference Model, Finite Element Model and Exact Solution
for 60digreeoffreedom model.
Figure 2. Nonuniform axially vibrating rod.
Figure 3. An rdegreeoffreedom element
Figure 4. An rth order spectral conforming element model
Figure 5. Exponential rod
Figure 6. Predicted natural frequencies using Finite
Difference Model, Spectral Conforming Model and Exact
Solution.
ACKNOWLEDGMENTS
The research work presented here is supported by a
National Science Foundation research grant CMS9978786.
REFERENCES
1. D. Boley and G.H. Golub, A survey of matrix inverse
eigenvalue problem, Inverse Problems, Vol. 3, pp. 595
622, 1987.
2. C. de Boor and G.H. Golub, The numerically stable
reconstruction of a Jacobi matrix from spectral data,
Linear Algebra and Its Application, Vol. 21, pp. 245
260, 1978.
3. F.P. Gantmakher and M.G. Krein, Oscillation Matrices
and Kernels and Small Vibration of Mechanical Systems,
State Publishing House for TechnicalTheoretical
Literature, MoscowLeningrad, 1961 (Translation: US
Atomic Energy Commission, Washington DC).
4. J. Paine, A numerical method for the inverse Sturm
Liouville problem, SIAM J. Sci., Stat. Comput., Vol. 5, pp.
149156, 1984.
5. J.W. Paine, F. de Hoog and R.S. Anderssen, On the
correction of finite difference eigenvalue approximations
for SturmLiouville problems, Computing, Vol. 26, pp.
123139, 1981
0 10 20 30 40 50 60
0
50
100
150
200
250
k
k
η
Finite Difference Model
Finite Elements Model
Exact Solution
0 10 20 30 40 50 60
0
50
100
150
200
250
k
k
η
Finite Difference Model
Finite Elements Model
Exact Solution
Finite Difference Model
Finite Elements Model
Exact Solution
) ( ), ( x p x ρ
dx
) , ( t x u
L
x
) ( ), ( x p x ρ
dx
) , ( t x u
L
x
0 10 20 30 40 50 60
0
20
40
60
80
100
120
140
160
180
200
Exact Solution
Finite Difference Model
Spectral Conforming Model
λ
i
0 10 20 30 40 50 60
0
20
40
60
80
100
120
140
160
180
200
Exact Solution
Finite Difference Model
Spectral Conforming Model
Exact Solution
Finite Difference Model
Spectral Conforming Model
λ
i
) ( ), ( ), ( x x A x E ρ
L
) ( ), ( ), ( x x A x E ρ
L
1
k
2
k
1
m
2
m
3
k
r
k
r
m
( )
( )
( ) e
e
e
A
E
ρ
( ) e
L
D
( ) e
L
1
k
2
k
1
m
2
m
3
k
r
k
r
m
( )
( )
( ) e
e
e
A
E
ρ
( ) e
L
DD
( ) e
L
1
k
2
k
1
m
2
m
3
k
r
k
r
m
p , ρ
L
1
k
2
k
1
m
2
m
3
k
r
k
r
m
p , ρ
L
1
2002 ME Graduate Student Conference
April 13, 2002
BUCKLING ANALYSIS OF GRID STIFFENED COMPOSITE CYLINDERS
Samuel Kidane
M.S. Candidate
Faculty Advisor: Dr. Eyassu Woldesenbet
ABSTRACT
Due to their high stiffness to mass ratio, stiffened cylindrical
composite shells are major components of Aerospace and
Aircraft industries. These structures are employed in
fuselage and fuel tank applications, and are usually
subjected to combinations of compressive, shear or
transverse loads. Usually the failure mode associated with
these structures is buckling. This failure mode is further
subdivided into ‘local skin and/or stiffener buckling’, and
‘universal buckling’.
In this paper buckling investigation of a grid stiffened
composite cylinder is presented using analytical model,
Finite elements model and experimentation. The cylinder
under discussion has orthotropic stiffeners integrally made
with an orthotropic shell. All the buckling analysis is based
on a uniaxial compressive load condition.
An analytical model is first developed that reduces the
grid/shell cylinder panel to an equivalent orthotropic
laminate (Fig. 1). This model makes use of the force and the
moment interactions and derives the A, B and D matrix of the
equivalent laminate model. Consequently buckling loads are
calculated making use of the energy method. Using the
analytical model developed, parametric analysis is preformed
to determine optimum configuration of stiffeners and shell.
A Finite elements model is also built using ANSYS for
the same cylinder. Buckling analyses are performed on
different models built with different configurations of
stiffener and shell parameters. The effect of shell thickness
variation on buckling load and buckling mode is studied in
detail (Fig 2). Based on this the optimum skin thickness is
determined for a given stiffener configuration. In this section
correlation is made between failure mode and optimum skin
thickness. In addition to skin thickness the effect of
stiffeners angle variation is also analyzed using ANSYS. The
result is plotted and conclusions are drawn on optimum
stiffener orientation.
Buckling experiment was also performed on a stiffened
composite cylinder specimen (Fig 3). The test setup and the
results obtained are discussed briefly.
Finally this paper tries to compare the different results
obtained using the three analysis methods. An attempt is
made to account for certain differences observed between
the three analysis methods. Conclusions are drawn on the
reliability of the analytical model developed and remarks
made on limitation of model.
FIGURES AND TABLES
ε l
θ
x
φ
ε t
Fig. 1 Unit cell.
Fig. 2 FEM analysis (local skin buckling)
18
35
30
25
20
15
10
5
0
Ti me ( s)
ε
Load
Strain Gauge1
Strain Gauge2
h/2
1
2
Fig. 3 Experimental results.
ACKNOWLEDGMENTS
Support for this research was provided by a grant from
the NASA/Louisiana Space Consortium and the Louisiana
Board of Regents under LaSPACE (BOR12662) and LEQSF
(200104)RDB03.
REFERENCES
1. Helms JE, Li G, Smith BH. Analysis of Grid Stiffened
Cylinders. ASME/ETCE 2001.
2. Brush DO, Almroth BO. Buckling of Bars, Plates, and
Shells. McGrawHill Book Company, New York, NY
1975.
3. Bruhn EF. Analysis and Design of Flight Vehicle
Structures. Jacobs Publishing, Inc., Carmel, IN June
1973
4. Navin J, Norman FK, Damodar R. Formulation of An
Improved Smeared Stiffener Theory of Buckling
Analysis of GridStiffened Composite Panels. NASA
technical Memorandum 110162, June 1995.
5. Ramm E. Buckling of Shells. SpringerVerlag, Berlin
1982.
6. Gerdon G, Gurdal Z. Optimal Desing of Geodesically
Stiffened Composite Cylindrical Shells. AIAA Journal,
November 1985; 23(11):17531761.
7. Troisky MS. Stiffened Plates, Bending, Stability and
Vibrations. Elsevier, 1976.
8. Phillips JL, Gurdal Z. Structural Analysis and Optimum
Design of Geodesically Stiffened Composite Panels.
Report NASA CCMS9050, (VPIE9008), Grant NAG
1643, July 1990.
9. Whitney JM. Structural Analysis of Laminated
Anisotropic Plates. Technomic, 1987.
10. Agarwal BD, Broutman LJ. Analysis and Performance of
Fiber Composites. John Wiley and Sons, 1990.
11. Dow NF, Libove C, Hubka RE. Formulas for Elastic
Constants of Plates with Integral Wafflelike Stiffening.
NACA RM L53L1 3a, August 1953.
19
2002 ME Graduate Student Conference
April 13, 2002
SYNTHESIS, PROPERTIES AND CHARACTERIZATION OF CRDLC
NANOCOMPOSITE FILMS
Varshni Singh
Ph.D. Candidate
Faculty Advior: Dr E.I. Meletis
ABSTRACT
Diamondlike carbon (DLC) films have been extensively
studied over the past decade, due to their unique
combination of properties. One of the drawbacks with DLC
films is that they are thermally unstable beyond 350
o
C [1].
Above 400
o
C the changes are more profound and
graphitization of the film occurs by conversion of C bonds
from sp
3
to sp
2
, a phenomenon that is also observed during
wear at hot spots [2]. Thus for more than a decade
researchers have focused on metalcontaining DLC (Me
DLC) films in an effort to improve wear resistance, adhesion,
thermal stability and toughness. A number of studies on
synthesis and characterization of MeDLC films have been
conducted on Si, Ti, Ta, W and NbDLC [310]. Even
though Cr is a carbide former and possesses an attractive
combination of other properties (corrosion resistance, wear
resistance, etc.) little work has been reported in this area
[6,11]. The purpose of the present work was to initiate a
systematic study of the processingstructureproperty
relationship in CrDLC films as a function of Cr content. The
objective is to develop a better understanding of this system
and identify possible compositional ranges where
tribological performance and thermal stability are
significantly improved.
CrDLC nanocomposite films were deposited on Si
(100) substrate, by reactive magnetron sputtering utilizing an
intensified plasmaassisted processing system. The
processing parameters (chamber pressure, bias voltage,
magnetron current, etc.) were varied to synthesize CrDLC
films, with Cr content ranging from ~0.1 at. % to 28 at. %.
Carbon and chromium content was determined by
wavelength dispersive spectroscopy (WDS) utilizing a JEOL
JXA 733 super electron probe microscope. Xray diffraction
(XRD) experiments were performed, using a Rigaku Miniflex
2θ diffractometer with a Cu  K
α
source and transmission
electron microscopy (TEM) was conducted in a JEOL JEM
2010 electron microscope. Pinondisc experiments were
conducted by utilizing an ISC200 tribometer and the wear
rate was calculated by a Veeco 3D optical profilometer.
Mechanical properties, of the films were studied by
nanoindentation measurements, using a Hysitron
Triboscope© instrumented nanoindentation/ nanoscratch
device incorporated on a Digital Instrument Dimension 3100
atomic force microscope. The short range order structure of
the films is being studied by the xray absorption fine
structure (XAFS) spectra collected by using bending
magnet radiation at the double crystal monochromator 1
beamline. The thermal stability experiments were conducted
by utilizing a DSC7 differential scanning calorimeter.
XRD patterns of DLC and CrDLC films, show that all
the films exhibited nearly the same XRD pattern, indicating
an amorphous structure. Electron diffraction and high
resolution TEM studies show that the films, with ~ 9 at. %
Cr, deposited using low (200 V) and high (1000 V) specimen
bias during processing are composed of nanocrystalline
metallic Cr and nanocrystalline cubic chromium carbide,
respectively surrounded by an amorphous matrix. Fig. 1
show the dark contrast clusters, diameter 1 – 5 nm,
surrounded by an amorphous matrix corresponding to
nanocrystalline Cr carbide. The XANES spectra of the Cr
DLC films show that the Cr content (5 to 28 at. %) in the Cr
DLC films, deposited at –1000 V bias, has little effect on its
structure and Cr atoms are incorporated in the carbon
network. This initial result is in agreement with that of the
TEM results. Furthermore, such preference for metal atoms
to be incorporated into the local carbon structure has also
been shown by our recent XAFS experiments on SiDLC
films with ~10 at. % Si [12].
Fig. 2 presents the variation of the coefficient of
friction (µ) and wear rate of CrDLC films (deposited at –1000
V) with Cr content. The results in Figs. 2 show that in
general all CrDLC films exhibit a low µ (less than 0.2) for
both alumina and 440 stainless steel pins. It is interesting to
note that µ remains at low levels (less than 0.15) for up to a
Cr/C ratio of 0.24. At a higher Cr level, the results indicate
that the coefficient of friction increases. Very similar
behavior has been observed previously for WDLC films [5].
The wear rate was also found to be relatively independent of
20
pin material. The wear rate was low and remained at almost
the same levels in films with a Cr/C ratio less than 0.24. At
higher Cr content (Cr/C equal to 0.35), the wear rate
increased significantly (by at least an order of magnitude),
which is consistent with the relatively higher µ observed for
that film. The present results are in general agreement with
the previous reports however, doesn’t show high wear rates
at very low Cr content.
The DSC result qualitatively suggests that the thermal
stability of the CrDLC films increases with increasing Cr
content, due to stabilizing effect of Cr on the DLC matrix
network, up to the point where the DLC network is
completely stabilized. Nanoindentation results suggest that
the hardness of the films reduces with increasing Cr content
to ~3 atomic % and then it stabilizes around 13 GPa. With the
exception of an initial deep, reduced modulus (E/(1 ?
2
))
increases with increasing Cr content and stabilizes around
118 GPa at ~11 atomic % Cr. The H/E/(1 ?
2
) exhibits a peak
value of ~0.17 at 0.05 atomic % Cr and then gradually
decreases and stabilizes to values around 0.11. Compared to
other MeDLC films, the present profile of H/E/(1 ?
2
) for Cr
DLC exhibits this interesting region yet to be explored
between 0.05 at. % Cr and ~2.75 at. % Cr. Presently, the
reason for this peculiar behavior is unknown.
At present, the indepth analysis of the spectra
obtained from CrDLC films is underway. Study of CrDLC
films deposited at lower substrate bias is planned for the
next period. In addition, a couple more compositions in the
aforementioned range between 0.05 at. % Cr and 2.75 at. %
Cr are planned to explore the lower range of the Cr content.
So as to completely understand the effect of Cr content and
the substrate bias on the short range order around Cr in the
CrDLC films further analysis and experiments are underway.
ACKNOWLEDGMENTS
This work was supported by the Army Research Office
grant DAAG559810279 and Louisiana Board of Regents.
TEM was performed at MCC facility of LSU with the help of
Dr J. Jiang. Nanoindentations were performed with the help
of Ms Tracy Morris and XAFS spectroscopy was done with
the help of Dr V. Palshin and Dr R. Tittsworth in CAMD,
LSU. WDS was performed using electron microscopy
facility of the Geology Department at LSU with the help of
Dr. Xie. Assistance of Mr. Pankaj Gupta in the deposition
experiments is also acknowledged.
REFERENCES
1. Z.L. Akkerman, H. Efstathiadis, and F.W. Smith, J. Appl.
Phys., 80(5), 30683075 (1996).
2. Y.Liu and E.I. Meletis, J. Mater. Sci., 32, 3491 (1997).
3. A. Varma, V. Palshin, K. Fountzoulas and E.I. Meletis,
Surface Engineering 15(4), 301306 (1999).
4. K. Oguri and T. Arai, J. Mater. Res., 7(6), 1313 (1992).
5. C.P. Klages and R. Memming, Materials Science Forum,
5253, 609644 (1989).
6. Y.L. Su and W.H. Kao, J. Mater. Eng. Perf., 9(1) (2000)
3850.
7. M. Fryda, K. Taube and CP Klages, Vacuum, 41(46)
(1990) 12911293.
8. H. Dimigen and CP Klages, Surf. Coat. Technol., 49
(1991) 543547.
9. W.J. Meng and B.A. Gillispe, J. Appl. Phys., 84(8) (1998)
43144321.
10. K. Bewilogua, C.V. Cooper, Surf. Coat. Technol., 132
(2000) 275283.
11. X. Fan, E.C. Dickey, S.J. Pennycook and M.K. Sunkara,
Appl. Phys. Let., 75(18) (1999) 27402742.
12. V.A. Palshin, R.C. Tittsworth, K. Fountzoulas and E.I.
Meletis, Journal of Materials Science 37, (2002).
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
1E9
1E8
1E7
1E6
wear rate
W
e
a
r
R
a
t
e
(
m
m
3
/
N

m
)
Pin Material
440 Steel
Alumina
Cr/C (atomic ratio)
0.1
0.2
0.3
0.4
0.5
Load 2.5 N
Pin dia 9.5 mm
Sliding speed 0.1m/s
µ
C
o
e
f
f
i
c
e
i
n
t
o
f
f
r
i
c
t
i
o
n
Figure 1 HRTEM image of CrDLC films with Cr 9
at. % deposited at a bias of –1000V.
Figure 2 Coefficient of friction and wear rate of CrDLC
films of Cr/C ratio.
21
2002 ME Graduate Student Conference
April 13, 2002
NUMERICAL SIMULATION OF FLOW AND HEAT TRANSFER IN MICRO HEAT
EXCHANGERS
Readul Islam
M.S. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
Pin fin arrays have been used for turbine blade cooling
applications in the trailing edge where they can fulfill a
structural as well as a heat transfer function. A recently
proposed concept involves covering the outer surface of
turbine blades with an array of very short pin fins and
adding a coating on top that is exposed to external gases [1].
The coating completely enshrouds the turbine blade, leaving
a micrometersized gap bridged by the pin fin array, and
preventing contamination of the coolant by the high
temperature gases outside. Cooling air bled from the
compressor stage flows through the gap, protecting the
blade surface.
A preliminary computational study using a commercial
CFD program (Fluent™) was undertaken to provide support
and future guidance to the experimental evaluation of the
concept under way at LSU. The problem was modeled as
periodic flow with conjugate heat transfer through an array
of cylindrical pin fins bounded by parallel flat plates. The
energy input from the external gases was modeled by a
uniform heat flux perpendicular to the fin axis applied to the
top endwall, while the bottom endwall was assumed to be
adiabatic.
To validate the computational model, simulations were
run to compare previous experimental work with various
types of pin fin arrays performed by Chyu [2] and Metzger
[3]. Good agreement with published friction factor and
Nusselt number results therein inspires a level of confidence
in the current computational model.
The computational study investigated the effects of
varying Reynolds number and Prandtl number on the friction
factor and heat transfer characteristics of the basic flow
configuration. Table 1 displays that, for the range of
Reynolds numbers in the present study, the heat transfer
enhancement increases with increasing Reynolds number.
However, this increased enhancement comes with rises in
the friction factor as flow becomes more turbulent. Friction
factors and Nusselt numbers can be incorporated into a
single quantity that contains information about the pressure
penalty paid for increased heat transfer, known as the
thermal performance factor (TPF):
TPF = Nu · ƒ
–?
Table 1 indicates a maximum TPF that falls within the range
of Reynolds numbers used.
Simulations were also performed to evaluate the effects
of modification of the fin array geometric parameters—
changing the fin height (H/D), and the streamwise (L/D) and
spanwise fin (W/D) array density. Results presented in Table
Table 1: Results of increasing Reynolds number on friction
factor and Nusselt number (normalized by corresponding flat
channel values), and associated TPF
Re ƒ/ƒ
flat
Nu/Nu
flat
TPF
675 0.3974 1.79 2.43
1350 0.2950 13.49 20.26
2025 4.0011 17.66 11.12
3038 7.5495 22.72 11.58
Table 2: Results of changing geometry parameters (base
case: L/D = W/D = 5; H/D = 2.5) on friction factor and
Nusselt number (normalized by corresponding flat channel
ƒ/ƒ
flat
Nu/Nu
flat
TPF
1.25 15.7 7.9 3.16
2.5 4.6 10.3 6.20
H/D
5 3.0 16.6 11.47
2 5.3 12.3 7.06
5 4.6 10.3 6.20 L/D
8 4.8 8.8 5.19
2 29.8 23.0 7.42
5 4.6 10.3 6.20 W/D
8 1.9 6.2 4.95
22
2 indicate that from a heat transfer viewpoint, larger H/D
values are desired—this seems to reflect the rising
effectiveness of the fluid to remove energy as the fin height,
and thus, the channel height is increased. Heat transfer can
also be increased by decreasing L/D and W/D, which means
increasing the fin array density. However, as Table 2
demonstrates, increasing fin density comes with severe
pressure penalties. Nusselt numbers are more responsive to
W/D changes than L/D, possibly because the wake of the
flow (L/D) isn't as important as the shear layers that form
around the fin region. Taken overall, the best performance
will be obtained from closely spaced arrays with large height
to diameter ratios—essentially, the current results are
predicating a move toward transverse flow across closely
spaced tube bundles. This observed trend matches well with
[4], where it was noted that the heat transfer enhancement
using pin fin arrays of intermediate H/D is lower than that for
classic tube bundles operating at the same Reynolds number
(until very high Reynolds numbers are reached).
While the pin fin arrays mentioned so far consist of
cylindrical pin fins, it is by no means certain that a circular
cross section produces the most effective heat transfer
enhancement. The computational study was extended to
investigate the effects of a variety of fin shapes at a
Reynolds number of 3000. The table associated with Figure 1
illustrates performance factors for square, diamond and
elliptical fins oriented parallel and perpendicular to the flow.
In terms of TPF, the last case offers a viable alternative to
cylindrical pin fin arrays.
The results presented allow optimal choices of
promising geometric parameters and fin shapes to be made
for further experimental study.
ACKNOWLEDGMENTS
This study is supported by DARPA. The author wishes
to thank Dr. Acharya and Dr. Kelly for valuable insight and
guidance, and Christophe Marques for fruitful discussions.
Criticism from Dr. A. K. Saha, constructive much more often
than not, is gratefully acknowledged.
REFERENCES
1. J. C. COYNEL, MS thesis, Louisiana State University,
Baton Rouge, Louisiana, 1999.
2. M. K. CHYU, Y. C. HSING and V. NATARAJAN, J.
Turbomachinery. 120 (1998) 362367.
3. D. E. METZGER, C. S. FAN and S. W. HALEY, J. Eng.
For Gas Turbines and Power. 106 (1984) 252257.
4. D. E. METZGER, R. A. BERRY and J. P. BRONSON, J.
Heat Transfer. 104 (1982) 700706.
A B
C
D
A B
C D
Figure 1. Nusselt number contours on top surface of channel for different pin shapes for L/D = W/D = 5, H/D = 2.5
arrays and Reynolds number = 3x10
4
, and associated performance factors
23
2002 ME Graduate Student Conference
April 13, 2002
ADVANCED TURBULENCE MODELING FOR INDUSTRIAL APPLICATIONS
Raymond M. Jones
Ph.D. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
Computational fluid dynamics ( CFD ) in industry is
typically performed using the Reynoldsaveraged Navier
Stokes equations ( RANS ). With RANS complex turbulent
flows at high Reynolds numbers can be solved with
reasonable accuracy and within an acceptable amount of
time. Other methods such as large eddy simulation ( LES )
and direct numerical simulations ( DNS ) can be used to
obtain more accurate results but require extensive
computational resources, even for complex turbulent flows
at moderately low Reynolds numbers.
With RANS, a turbulence model is needed to close the
momentum equations. Twoequation turbulence models
such as the kε model have been primarily used in industry
because of their robustness. There are several different
types of twoequations turbulence models but they are
similar in that they all use damping functions to accurately
represent near wall turbulence. These damping functions
are usually derived for simple flows, such as flow in a
channel, and are ill suited for most complex flows.
In the kε model, the eddy viscosity is computed as
ε ν
µ
2
k C
t
· where
µ
C is typically 0.09. Figure 1 shows
computed values of
t
ν for 09 . 0 ·
µ
C and 075 . 0 ·
µ
C for a
turbulent channel flow. It can be seen that
t
ν is
overpredicted for both values of
µ
C . This shows that
t
ν
computed by the kε model can not accurately predict the
near wall eddy viscosity as long as
µ
C is a constant. The
damping functions mentioned above are used to replace
µ
C
with a function.
Based on this observation Durbin (1991) introduced an
alternative eddy viscosity formulation defined as
ε ν
µ
k v C
t
2
· where 2 . 0 ·
µ
C and
2
v is the velocity
fluctuation normal to the wall. This formulation can also be
seen in Fig. 1. Durbin (1991) developed the ε k f v −
2
model, which solves a scalar transport equation for k, ε, and
2
v which are used to computed the eddy viscosity as
ε ν
µ
k v C
t
2
· . The ε k f v −
2
model has shown to
produce better results compared to conventional two
equation models (Durbin (1992), Parneix et al. (1998)) .
Despite the improved accuracy, the ε k f v −
2
model has
shown to be numerically stiff even for simple flows. In the
present work a new model has been developed which uses
the general framework of the original ε k f v −
2
model but
has shown to be more robust. In the new model, the ε
equation in the ε k f v −
2
model has been replaced by the ω
equation of Wilcox and is called the ω k f v −
2
model. A new
relationship between ω and ε is introduced based on
dimensional arguments, and several of the closure constants
calibrated with respect to DNS data.
The ω k f v −
2
model has been tested for predicting the
channel flow of Moser (1999) for Re
τ
.=395. The ω k f v −
2
model was compared with the kω model and the ε k f v −
2
model and the results for the streamwise velocity predictions
can be seen in Fig. 2. All of the models accurately predict
the correct velocity profile. Figure 3 shows kinetic energy
predictions. It can be seen that the ω k f v −
2
model shows
improved predictions compared to both the kω model and
the ε k f v −
2
model. The ω k f v −
2
model was tested for
predicting the backward facing step of Kasagi (1993). Figure
5 shows that the ε k f v −
2
model and the ω k f v −
2
model
produce better near wall velocity predictions compared to
the kε model.
24
FIGURES AND TABLES
ν
+
y
+
0 0.02 0.04 0.06 0.08 0.1
0
50
100
150
200
DNS
0.09k
2
/ε
0.075k
2
/ε
0.2kv
2
/ε
Fig. 1 Exact eddy viscosity compared with the kε model
(blue and orange) and the ε k f v −
2
model (red line)
y
+
u
+
10
0
10
1
10
2
0
4
8
12
16
20
24 Re
τ
= 395
v
2
fkω
kω
v
2
fkε
Fig. 2 Mean Velocity
k
+
y
+
0 1 2 3 4 5 6
0
80
160
240
320
400
Re
τ
= 395
v
2
fkω
v
2
fkε
kω
Fig. 3 Kinetic Energy
y
+
d
i
s
s
p
r
o
d
0 25 50 75 100
0.3
0.2
0.1
0
0.1
0.2
0.3
Re
τ
= 395
v
2
fkω
kω
v
2
fkε
Fig. 4 Production and Dissipation
x/H
y
/
H
0 1 2 3 4 5 6 7 8 9 10 11
0
0.5
1
1.5
2
2.5
3
Fig. 5 Streamwise velocity for the kε model (solid line),
ε k f v −
2
model (dashed), ω k f v −
2
model (dotted)
x/H
y
/
H
0 1 2 3 4 5 6 7 8 9 10 11
0
0.5
1
1.5
2
2.5
3
Fig. 6 Kinetic Energy for the kε model (solid line),
ε k f v −
2
model (dashed), ω k f v −
2
model (dotted)
ACKNOWLEDGMENTS
I would like to acknowledge The Dow Chemical for their
support on this project.
REFERENCES
1. P. A. DURBIN, AIAA Journal. 33 (1991) 659664.
2. P.A. DURBIN, Annual Research Briefs, (1992) 316.
3. N.KASAGI, A. MATSUNAGA, and S. KAMARA, J.
Wind Eng. Ind. Aero. 46 (1993) 821829.
4. R.D.MOSER, J.KIM, N.MANSOUR, Phys. Fluids. 11
(1999) 943945.
5. S.PARNEIX, M.BEHNIA, P.DURBIN, Annual Research
Briefs, (1998) 149164.
25
2002 ME Graduate Student Conference
April 13, 2002
TRIBOLOGICAL BEHAVIOR OF NANOSTRUCTURED NICKEL
Dean J. Guidry
M.S. Candidate
Faculty Advisor: Dr. Efstathios I. Meletis
ABSTRACT
The present study reports the effects of electroplating
parameters on the microstructure, and thus the mechanical
and tribological properties, of nanostructured nickel.
Electroplating was conducted in a Watt’s type bath at
current densities of 30 mA/cm
2
and 15 mA/cm
2
in
electroplating bath temperatures of 30°C and 50°C. The PH
of the bath was maintained at 3.0 using sulfuric acid. The
electroplating was carried out using a direct current in
galvanostatic mode with a nickel anode contained in a
titanium wire basket. Average grain sizes and uniformity of
grains were determined from TEM and SEM micrographs.
Tribological tests were carried out on a pinondisc type
tribometer. The same tests were conducted on Ni200 for the
purpose of comparison. Wear rates were calculated for the
nickel surfaces using optical profilometry and for the
alumina pins using optical microscopy. Nanoindention
techniques provided the nanohardness, stiffness, and
reduced modulus values for all samples. Microhardness
readings were also recorded to further study the surface
properties. Result s show how grain size, mechanical
properties, and wear properties change with the variations in
plating parameters. Grain size reduction shows surface
hardness increases and improved tribological properties.
Plating bath temperature increases showed a decrease in
grain uniformity.
ACKNOWLEDGMENTS
The Author acknowledges Dr. Efstathios I. Meletis, Dr.
Kun Lian, Dr. Jie Chao Jiang and Varshni Singh for their help
and advice throughout this research project.
REFERENCES
1. G. Robinson, Electronic Engineering Times, 958 (1997)
33.
2. T. E. Buchheit, T. R. Christenson, D. T. Schmale, D. A.
Lavan, Mater. Res. Soc. Sym., 546 (1998) 126.
3. A. M. ElSherik, U. Erb, J. Mater. Sci., 30 (1995) 5743.
4. V. Provenzano, R. Valiev, D. G. Rickerby, G. Valdre,
NanoStructured Mater., 12 (1999) 1103.
5. F. H. Froes and C. Suryanarayana, J. Mater. Sci., June
(1989) 12.
6. H. Gleiter, Acta Materialia, 48 (2000) 1.
7. H. Gleiter, NanoStructured Materials, 6 (1995) 3.
8. C. Meneau et. al., Surface and Coatings Technology,
100 (1998) 12.
9. F. Ebrahimi, G. R. Bourne, M. S. Kelly, T. E. Matthews,
NanoStructured Mater., 11 (1999) 343.
10. Z. N. Farhat, Y. Ding, D. O. Northwood, A. T. Alpas,
Mater. Sci. Eng. A, 206 (1996) 302.
11. S. W. Banovic, K. Barmak, A. R. Marder, J. Mater. Sci.,
33 (1998) 639.
12. E. O. Hall, Proc. Phys. Soc., B64 (1951) 747.
13. N. J. Petch, J. Iron Steel Inst., 174 (1953) 25.
14. L. S. Stephens, K. W. Kelly, E. I. Meletis, S. Simhadri,
Mat. Res. Symp. Proc., 518 (1998) 1.
15. A. H. Chokshi, A. Rosen, J. Karch and H. Gleiter, Scripta
Metall. Mater., 23 (1989) 1679.
16. K. Lu, W. D. Wei and J. T. Wang, Scripta Metall.
Mater., 24 (1990) 2319.
17. G. E. Fougere, J. R. Weertman and R. W. Siegel, Scripta
Metall. Mater., 26 (1992) 1879.
18. G. Palumbo, U. Erb and K. T. Aust, Scripta Metall.
Mater., 24 (1990) 2347.
19. N. Wang, Z. Wang, K. T. Aust, U. Erb, Acta Metall.
Mater., 43 (1995) 519.
20. R. Z. Valiev, N. A. Krasilnikov, N. K. Tsenev, Mater. Sce.
Eng. A, 137 (1991) 35.
21. S. M. Myers, J. A. Knapp, D. M. Follstaedt, M. T.
Dugger, J. Appl. Phys., 83 (1998) 1256.
22. Y. Ichida, K. Kishi, Trans. of the ASME, 119 (1997) 110.
23. D. A. Rigney, Mat. Res. Innovat., 1 (1998) 231.
24. V. Singh, X. Nie, P. Gupta, E.I. Meletis, 6
th
Nat. Cong.
Mech. Proc., 2 (2001) 1.
25. K. Laul, M. Dorfman, Adv. Mater. and Proc., 158 (2000)
46.
26. M. Legros et. al., Phil. Mag. A, 80 (1999) 1017.
27. A. B. Witney, P. G. Sanders, J. R. Weertman, Scripta
Metall. Mater., 33 (1995) 2025.
26
28. X. J. Wu et. al., NanoStructured Materials, 12 (1999)
221.
29. T. R. Christenson, T. E. Buchheit, D. T. Schmale, R. J.
Bourchier, Mat. Res. Soc. Symp. Proc., 518 (1998) 185.
30. E. Bonetti et. al., NanoStructured Mater., 11 (1999) 709.
31. S. Greek, F Ericson, Mat. Res. Symp. Proc., 518 (1998) 51.
32. A. Cziraki et. al., Thin Solid Films, 318 (1998) 239.
33. G. Auner, et. al., Thin Solid Films, 107 (1983) 191.
34. R. V. Nandedkar et. al., Phys. Status Solidi, 72 (1982) 89.
35. J. C. Pivin et. al., J. Mater. Sci., 22 (1987) 1087.
36. P. S. Barlow, R. A. Collins, G. Dearnaley, J. Phys. D: Appl
Phys., 22 (1989) 1510.
37. D. M. Mattox, J. Electrochem. Soc, 115 (1968) 1255.
38. K. Bouslykhane, J. P. Villain, P. Moine, Tribology
International, 29 (1996) 169.
39. L. Jun, W. Yiyong, W. Dianlong, H. Xinguo, J. Mater.
Sci., 35 (2000) 1751.
40. F. Ebrahimi, A. J. Liscano, Mater. Sci. Eng. A, 301 (2001)
23.
41. R. I. Pratt, G. C. Johnson, Mat. Res. Soc. Symp. Proc.,
518 (1998) 15.
42. T. Bieger, U. Wallrabe, Microsystem Tech., 2 (1996) 63.
43. I. M. Hutchings, Mater. Sci. Eng. A, 184 (1994) 185.
44. J. Ferrante, Phys. World, July (1991) 46.
45. S. Jahanmir, in N. P. Suh and N. Saka (eds),
Fundamentals of Tribology, MIT Press, Cambridge,
(1980) 455.
46. J. F. Archard, J. Appl. Phys., 24 (1953) 981.
47. D. J. Tillack, E.B. Fernsler, ASM Handbook Ninth
Edition, 10 (1996) 754.
48. LECO Corporation, Metallography Principles and
Procedures, (1994) 38.
27
2002 ME Graduate Student Conference
April 13, 2002
INTRINSIC STRESS DEVELOPMENT IN TIC:H CERAMIC NANOCOMPOSITE
COATINGS
B. Shi
Ph. D. Candidate
Faculty Advisor: Dr. W. J. Meng
ABSTRACT
The development of intrinsic stresses within titanium
containing hydrocarbon (TiC:H) coatings was monitored by
insitu substrate curvature measurements using a multibeam
optical sensing (MBOS) technique. Stress as a function of
the TiC:H layer thickness was monitored in a wide range of
specimens, from nearly pure amorphous hydrocarbon (a
C:H) to nearly pure titanium carbide (TiC). The intrinsic
stress within TiC:H was found to vary significantly in
magnitude and depend systematically on the Ti
composition.
Ticontaining hydrocarbon (TiC:H) coatings,
consisting of a nmscale mixture of crystalline titanium
carbide (TiC) and amorphous hydrocarbon (aC:H)
1
, form a
prototype of pseudobinary ceramic nanocomposites. TiC:H
coatings possess mechanical properties and tribological
characteristics which depend systematically on coating
composition
2
, demonstrating the potential of engineering
ceramic nanocomposite coatings for specific applications.
The dependence of tribological characteristics of TiC:H
coatings on the Ti composition has been related to a
percolation type transition
3
. In addition to the influence of
plasma characteristics, the coating composition may
therefore exert a significant influence on intrinsic stress
development within ceramic nanocomposite coatings.
This paper addresses the dependence of intrinsic
stresses within TiC:H coatings on the Ti composition. Ti
C:H deposition was accomplished using a highdensity
plasma assisted hybrid chemical vapor deposition
(CVD)/physical vapor deposition (PVD) process. Under
nominally identical plasma conditions, a series of TiC:H
coatings, ranging from nearly pure aC:H to nearly pure TiC,
was deposited onto Si(100) substrates. The development of
intrinsic stresses was monitored by insitu measurements of
substrate curvature change. Our results show that the
intrinsic stress within TiC:H coatings depends
systematically on the Ti composition. A significant increase
in stress was observed as the Ti composition increases
beyond 30 at. % and suggested to be related to the
percolation type transition in the coating mi crostructure.
Details of experimental setup and procedures will be
described in the presentation.
Figure 1 shows a typical substrate temperature – time
history during a complete TiC:H deposition run. The Ti
cathode current was 1.0A during TiC:H deposition. During
the etch and cool stages, the substrate temperature rose
from ~ 25
o
C to ~ 225
o
C and fell to ~ 150
o
C. It stayed ~ 150
o
C during Ti interlayer deposition, and rose to ~ 225
o
C
during TiC:H deposition. During the entire deposition run,
the temperature difference between front and back beam
surfaces was ≤ 5 K. Such a temperature difference across a
300 µm thick Si wafer would induce a substrate curvature
change ∆K of ~ 1/21 m
1
. Figure 1 shows that, during Ti
interlayer and TiC:H deposition, change in relative reflected
laser spot spacing ∆D/D
0
, which is measured experimentally
and linearly related to the curvature change, induced by
temperature gradient across the Si substrate is substantially
smaller than 10 %.
A multitude of TiC:H/Ti/Si(100) specimens were
deposited. During each deposition, the curvature change of
the Si(100) beam substrate was monitored by MBOS. Figure
2 shows the average composition of the TiC:H layers as a
function of the Ti cathode current obtained by combining
RBS and ERD measurements. The Ti and hydrogen
compositions respectively increase and decrease in a
monotonic fashion with increasing Ti cathode current. The
observed trend is consistent with our previous results and
supports the fact that TiC:H coatings are pseudobinary
TiC/aC:H nanocomposites, in which hydrogen inclusion
occurs only through incorporation into the aC:H phase
4
.
Figure 3 shows measured ∆D/D
0
during TiC:H deposition as
a function of time. The time origin coincides with the
beginning of TiC:H deposition. Only the curvature change
due to TiC:H deposition is taken into account, as ∆D/D
0
was set to zero at time zero. In the early stage of growth, 0 –
400 sec, ∆D/D
0
increases approximately linearly to ~ 30%,
independent of the Ti composition. In the late stage, 400 –
2500 sec, ∆D/D
0
continues to increase linearly with time, but
in most cases with a distinctly different slope as compared
to the early stage.
28
Temperature measurements, such as the one shown in
Figure 1, showed that during the early stage of TiC:H
growth, the substrate temperature rose ~ 70 K. In the late
stage, the substrate temperature change was ≤ 25 K at all Ti
cathode currents. This temperature change ∆T would induce
a substrate curvature change ∆K
T
due to the difference in
thermal expansion between Si and the TiC:H coating ∆α
sc
,
where Y
s
, t
s
, Y
c
, and t
c
are respectively the biaxial modulus
and thickness of the substrate and the coating. ∆α
sc
was
taken to be ~ −4×10
6
K
1
, according to measurements on W
C:H coatings
5
. For the present measurements, a
conservative estimate yielded ∆D/D
0
≤ +5% for ∆T = +100 K
6
. It is thus concluded that thermal contribution to the
present measurements can be neglected, and that in all cases
the measured ∆D/D
0
reflects intrinsic stress development.
The linear dependence of ∆D/D
0
on time during late stage
growth indicates a constant level of incremental intrinsic
stress as the TiC:H layer thickens.
In summary, a detailed experimental study of the
dependence of intrinsic stress within TiC:H coatings on the
Ti composition was performed by measuring insitu
substrate curvature change. The intrinsic stress wit hin Ti
C:H layers was found to vary significantly in magnitude and
depend systematically on the Ti composition. The observed
stress dependence on coating composition is suggested to
correlate with a percolation type transition in the coating
microstructure, and may be one generic characteristic of
ceramic nanocomposite coating systems.
ACKNOWLEDGMENTS
We gratefully acknowledge partial project support from
NIST through ATP program 70NANBH0H3048, NSF through
grant DMI0124441, and Louisiana Board of Regents
through contracts LEQSF(200003)RDB03 and
LEQSF(200104)RDA07. The ion beam analysis work at the
Argonne National Laboratory was performed by L. E. Rehn
and P. M. Baldo and supported by the DOE Office of
Science, Basic Energy Sciences, under contract #W31109
ENG38.
FIGURES AND TABLES
REFERENCES
1. W. J. Meng, R. C. Tittsworth, J. C. Jiang, B. Feng, D. M.
Cao, K. Winkler, V. Palshin, J. Appl. Phys., 88, 2415
(2000).
2. W. J. Meng, R. C. Tittsworth, L. E. Rehn, Thin Solid
Films 377/378, 222 (2000).
3. B. FENG, D. M. CAO, W. J. MENG, L. E. REHN, P. M. BA
4. LDO, G. L. DOLL, THIN SOLID FILMS 398/399, 210
(2001).
5. D. M. Cao, B. Feng, W. J. Meng, L. E. Rehn, P. M.
Baldo, M. M. Khonsari, Appl. Phys. Lett. 79, 329 (2001).
6. J. S. Wang, Y. Sugimura, A. G. Evans, W. K. Tredway,
Thin Solid Films 325, 163 (1998).
7. For the present Si(100) beam substrates, Y
s
= 180 GPa
and t
s
= 300 µm. t
c
is taken to be 800 nm, the entire
thickness of the TiC:H layer. The maximum biaxial
modulus for TiC:H coatings is ~ 180 GPa (ref. 8), thus
Y
c
is taken to be 180 GPa.
Time (sec)
0 500 1000 1500 2000 2500
T
e
m
p
e
r
a
t
u
r
e
(
C
)
0
50
100
150
200
250
Ti cathode 1.0A, front surface
Ti cathode 1.0A, back surface
Figure 1. Temperature – time history during an entire
TiC:H deposition run: 0 – 500 sec, substrate etching;
500 – 700 sec, cool down; 700 – 1300 sec, Ti interlayer
deposition; 1300 – 2500 sec, Ti C:H deposition.
Ti cathode current (A)
0.0 0.5 1.0 1.5 2.0
T
i

C
:
H
c
o
m
p
o
s
i
t
i
o
n
(
a
t
.
%
)
0
10
20
30
40
50
60
70
80
Ti composition
C composition
H composition
Figure 2. Composition of TiC:H coatings as a
function of the Ti cathode current during deposition.
0 5 0 0 1 0 0 0 1500 2000 2500
0
5 0
1 0 0
1 5 0
2 0 0
2 5 0
3 0 0
1 a t . %T i
1 4 a t . %T i
2 0 a t . %T i
2 2 a t . %T i
4 2 a t . %T i
(
D

D
0
)
/
D
0
(
%
)
T i me ( s e c o n d )
Figure 3. Changes in reflected spot relative spacing from
Si(100) beam substrates during TiC:H deposition for
several TiC:H coatings of various Ti compositions.
T,
c
t
c  s
c
Y
T
K
6
2
s
t
s
Y
∆ ∆ · ∆ α
29
2002 ME Graduate Student Conference
April 13, 2002
Manufacturing and analysis of a LIGA heat exchanger for the surface
of a cylinder: a cooling simulation of the leading edge region of a
turbine blade
Christophe Marques
Ph.D. Candidate
Faculty Advisor: Dr. Kevin Kelly
ABSTRACT
The presence a pin fins array in a channel has the ability
to enhance heat transfer greatly. That concept is already
used in the internal cooling channel present in the trailing
edge of some turbine blades. The reasons why fins enhance
heat transfer are numerous: i) their presence increases the
wetted surface area available to heat transfer ii) they
promote higher turbulence levels by breaking up boundary
layers which causes mixing and creates high convective
thermal coefficients iii) the wetted surface area associated
with the fins is characterized by high convective heat
transfer coefficients associated with the cylinder in cross
flow configuration.
The interest of this study is to use our LIGA experience
to build a micro heat exchanger (a shroud supported by
micro pin fins) around a cylinder, as a simulation of the
leading edge region of a turbine blade. Such a heat
exchanger can be an alternative to the existing techniques in
use for turbine blade cooling.
A parallel micro heat exchanger with a staggered array
of pin fins (Figure1) was manufactured and is being tested in
order to measure the performance with respect to the
Reynolds number of the coolant. The heat exchanger is a
simple 2 x 2 inch flat channel heat exchanger with a
staggered array of micro pin fins 500 µm high, 500 µm
diameter and 1250 µm apart in the downstream direction
made using a LIGA process. This pure Nickel heat exchanger
is tested using an apparatus allowing a uniform heat flux on
the top plate and an air coolant flow between the plates. The
Reynolds number of the coolant ranges from 5000 to 20000.
A similar experiment with a smooth channel (parallel plates
with a gap of 500 µm) is also performed in order to validate
the testing apparatus.
The comparison of the results to a parallel plate heat
exchanger without pin fins can demonstrate the expected
benefits of the micro heat exchanger.
The heat exchanger was customized to fit the shape of a
cylinder where the flow conditions approximate those at the
leading edge of a turbine blade. A new modified LIGA
process was developed to build a shroud around a ¾ inch
cylinder (Figure2). The shroud is supported by an array with
dimensions similar to those used for the flat heat exchanger.
This heat exchanger and a testing facility were built. In the
testing procedure, the heat exchanger is exposed to burner
rig high temperature gases while internal cooling airflow is
provided. The performance versus the Reynolds number is
determined. The micro heat exchanger can be compared to
other turbine blade cooling schemes.
Experimental results are not available at the present time
but will be showed on the day of the conference.
FIGURES
Figure 1. SEM of the completed heat exchanger
30
Figure 2: Cross section of heat exchanger on stainless
steel tube
ACKNOWLEDGMENTS
The author would like to thank the Defense Advanced
Research Project Agency (DARPA) for providing the bulk of
the funding for this research. The Center for Advanced
Microstructures and Devices (CAMD), which receives
funding from the State of Louisiana, was also indispensable
in the LIGA process.
REFERENCES
1. E. W. Becker, W. Ehrfeld, P. Hagmann, A. Maner and D.
Munchmeyer, “Fabrication of microstructures with high
aspect ratios and great structural heights by synchrotron
radiation lithography, galvanoforming and plastic
moulding (LIGA process)”, Microelectronic Engineering
4, p 3556, 1986.
2. P. Bley, W. Menz, W. Bacher, K. Feit, M. Harmening, H.
Hein, J. Mohr, W. Schomburg and W. Stark, “Application
of the LIGA Process in Fabrication of Threedimensional
Mechanical Microstructures, Microprocess 91, 1991
International Microprocess Conference, p 384 389,
Kanazawa, Japan, 1991.
3. M.K. Chyu, “Heat Transfer and Pressure Drop for Short
PinFin Arrays with PinEndwall Fillet”, Journal of Heat
Transfer, 1990, Vol. 112, pp. 926932.
4. M.K. Chyu, Y.C. Hsing, T.IP. Shih, V. Natarajan, “Heat
Transfer Contributions of Pins and Endwall in PinFin
Arrays: Effects of Thermal Boundary Condition
Modeling”, Journal of Turbomachinery, April 1999, Vol.
121, pp. 257263.
5. V. Galhotra, C. Marques, Y. Desta, K. Kelly, M. Despa, A.
Pendse, J. Collier, “Fabrication of Mold Inserts Using a
Modified LIGA procedure”, SPIE Proceedings
Micromachining and Microfabrication Process
Technology II, October 1415 Austin TX, Vol. 2879, p 168
173, 1996.
6. D.K. Hennecke, “Turbine Blade Cooling in Aero Engines
Some New Results, Future Trends and Research
Requirements”, pp. 116.
7. K. Kelly, J. Rogers, C. Marques, V. Galhotras, Y. Desta, M.
Murphy, M. Despa, “High Aspect Ratio Microstructures
and their Use in Heat Transfer Applications”, submitted
to the “Journal of heat transfer” in September 1996.
8. S.C. Lau, J.C. Han, Y.S. Kim, “Turbulent Heat Transfer
and Friction in Pin Fin Channels With Lateral Flow
Ejection”, Journal of Heat Transfer, Feb 1989, Vol. 111, pp.
5158.
9. C. Marques, “Heat Transfer of a MicrostructureCovered
Cylinder in Crossflow”, M.S. Thesis, Louisiana State
University, Baton Rouge, August 1997.
10. C. Marques, Y. Desta, J. Rogers, M. Murphy and K. Kelly,
“Fabrication of HighAspectRatio Microstructures on
Planar and Nonplanar Surfaces Using a Modified LIGA
Process”, Journal of MEMS, Dec 1997, pp. 329336
11. D.E. Metzger, C.S. Fan, S.W. Haley, “Effects of Pin Shape
and Array Orientation on Heat Transfer and Pressure
Loss in Pin Fin Arrays”, Journal of Engineering for Gas
Turbines and Power, Jan 1984, Vol. 106, pp. 252257.
12. R.J. Moffat, “Turbine Blade Cooling”, Presented at
Symposium on Heat Transfer in Rotating Machinery, Apr
28 May 3, 1985, Hawai.
13. D.A. Olson, “Heat Transfer in Thin, Compact Heat
Exchangers With Circular, Rectangular, or PinFin Flow
Passages”, Journal of Heat Transfer, May 1992, Vol. 114,
pp. 373382.
14. S. Thakur, J. Wright, W. Shyy, “Convective Film Cooling
Over a Representative Turbine Blade LeadingEdge”,
International Journal of Heat and Mass Transfer, 1999,
Vol. 42, pp. 22692285.
15. G.J. VanFossen, “Heat Transfer Coefficients for
Staggered Arrays of Short Pin Fins”, Journal of
Engineering for Power, Vol. 104, 1982, pp. 268274.
16. A. Zukauskas “Heat Transfer from Tubes in Cross Flow”,
Advances in Heat transfer, Vol. 8, pp.116133.
31
2002 ME Graduate Student Conference
April 13, 2002
MASS/HEAT TRANSFER IN DIMPLED TURBINEBLADE COOLANT PASSAGES
Fuguo Zhou
M.S. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
To increase thermal efficiency, gas turbine blades are
being designed to operate at increasingly higher inlet
temperatures. This requires the development of more
effective internal and external blade cooling strategies.
Internal cooling is generally enhanced using ribturbulators
1
3
that provide additional surface area for heat transfer,
interrupt the development of the boundary layer, and
increase turbulence levels.
In recent years, the concept of using indented (dimpled)
surface has attracted attention due to the high heat transfer
enhancement and lower pressure loss penalty associated
with dimples. Russian investigators performed many of the
early studies regarding the effect of dimples on heat transfer
and flow structure
46
. Chyu, et al.
7
reported an experimental
study with hemispherical and teardrop shaped dimples. The
dimples were imprinted on a heated plate with staggered
arrays. Their results showed that everywhere on the dimpled
surface the heat transfer coefficients were significantly
higher than the values on smooth wall. Over a range of
Reynolds numbers, the overall heat transfer rate is about 2.5
times smooth surfaces values, while the pressure loss is
about only half of the values caused by rib turbulators.
Similar observations have been made by Mahmood et al.
89
who have reported heat transfer enhancements of the order
of 22.5 with dimples.
The above studies have been reported for stationary
coolant channels. In the present study, we examine the effect
of dimples in rotating coolant channels. The naphthalene
sublimation method is employed in this study.
Measurements were made in a twopass rotating coolant
channel facility. The test plates have hollow recesses into
which naphthalene can be poured. The experiments
themselves consist of measuring the naphthalene surface
contour (with a profilometer) before and after the
experiments, and converting the sublimation depth into a
mass transfer rate or Sherwood number.
The present work focuses attention on mass/heat
transfer measurements with imprinted arrays of circular
dimples on the test plates. The coolant flow passage is
0.0254m square. The leading and trailing surfaces are
dimpled, while the side walls are kept smooth.
Measurements are made at a Reynolds number of 21,000 and
for Rotation number of 0 and 0.2.
Figures 1 and 2 show the spanwiseaveraged Sherwood
number ratio along the leading, trailing and side walls of
both the inlet and outlet duct. The measurements indicate
that dimples enhance surface mass/heat transfer. This
enhancement is stronger in the inlet duct (Figure 1) than in
the outlet duct (Figure 2). As seen in Figure 1, peak
mass/heat transfer occurs immediately downstream the
dimples (Sh/Sh
0
=5 in the inlet duct), while the minimum
mass/heat transfer occurs in the dimple region itself
(Sh/Sh
0
=2 in the inlet duct). Higher mass/heat transfer is also
observed along the lateral edges of the dimple. The location
of the Sherwood number peaks (Figure 3) suggest the
existence of streamwise vortical structures generated from
the leading and lateral edges of the dimples.
In Figures 1 and 2, it can be seen that heat/mass transfer
is higher along the trailing wall in the inlet duct and along
the leading wall in the outlet duct. This represents the effect
of Coriolis induced secondary flows. These flows are seen
not to alter the basic heat transfer distributions in the
vicinity of the dimples.
The effect of dimple shapes was also investigated in the
present work. Four types of dimple shapes (square, triangle,
circle and teardrop) were studied in a stationary channel,
with the goal of determining the dimple geometry with the
highest mass/heat transfer enhancement. Sherwood
numbers are obtained both inside and around the dimples at
a Reynolds number of 21,000., and are shown in Figures 4
and 5 for a circular and tearshaped dimple. The area
averaged Sherwood number around the dimple increases
from 30% for circular dimples to 53% for tearshaped
dimples.
32
FIGURES AND TABLES
Sh/Sh0 Distributions in the Dimpled
Inlet(Re=21,000, Ro=0.2)
0
1
2
3
4
5
6
7
8
0 2 4 6 8 10
X/D
S
h
/
S
h
0
Leading Wall
Trailing Walls
Side Wall
Fig.1 Spanwise Averaged Sh/Sh0 Distributions in Inlet
Sh/Sh0 Distributions in the Dimpled
Outlet(Re=21,000, Ro=0.2)
0
1
2
3
4
5
6
0 2 4 6 8 10
X/D
S
h
/
S
h
0
Leading Wall
Trailing Wall
Side Wall
Fig.2: Streamwise Average Sh/Sh0 Distributions in Outlet
Fig.3 Sh/Sh0 Contours in the Fully Developed Region on
Dimpled Walls (trailing wall in the inlet, Re=21,000
Ro=0.2)
Fig. 4 Sherwood Number Contours around and inside
Circular Dimples (Re = 21,000, Ro = 0)
Fig. 5 Sherwood Number Contours around and inside
teardrop Dimples (Re = 21,000, Ro = 0)
REFERENCES
1. HAN, J.C., ZhANG, Y.M., and LEE, C.P., 1992, Journal of
Turbomachinery, Vol. 114, pp. 872880.
2. MYRUM, T., ACHARYA, S., SINHA, S., and QIU, X.,
1996, Journal of Heat Transfer, Vol. 118, pp. 294300.
3. HIBBS, R., ACHARYA, S., CHEN, Y., NIKITOPOULOS,
D., and MYRUM, T., 1998, Journal of Turbomachinery,
Vol. 120, No. 4, pp. 724734.
4. GLEZER, B., MOON, H. K., KERREBROCK, J., BONES,
J., and GUENETTE, G., 1998, ASME Paper 98GT214.
5. HEDLUND, C. R., LIGRANI, P. M., MOON, H. K., and
GLEZER, B., 1998, ASME Paper 98GT466.
6. SCHUKIN, A. V., KOSLOV, A. P., and AGACHEV, R. S.,
1995, ASME Paper No. 95GT59, ASME 40th Intl. Gas
Tutrbine and Aero Congress, Houston.
7. CHYU, M. K., YU, Y., DING, H., DOWNS, J. P., and
SOECHTING, F., 1997, ASME Paper No. 97GT437,
ASME 42nd Intl. Gas Turbine and Aero Congress,
Orlando.
8. LIN, Y. L., SHIH, T. I. P., and CHYU, M. K., 1999, ASME
99GT263, ASME Turbo Expo, 1999, Indianapolis
X/D
Y
/
D
8 8.25 8.5 8.75
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
6.01
5.74
5.21
4.94
4.14
3.61
3.45
3.34
3.06
2.97
2.78
2.58
2.56
2.54
2.42
2.41
2.31
2.01
1.74
1.47
1.21
0.94
S h/S h0
F ul l y Deve l op pe d Reg io n o n the T rai l in g Wa ll in th e I nl et(Re =2 1, 00 0, Ro= 0. 2 ) 09 Au g 20 00 F ul l y Deve l op pe d Reg io n o n the T rai l in g Wa ll in th e I nl et(Re =2 1, 00 0, Ro= 0. 2 ) 09 Au g 20 00
X/D
Y
/
D
8.5 9 9.5
0.4
0.2
0
0.2
0.4
2 .13
1 .99
1 .88
1 .80
1 .77
1 .72
1 .66
1 .63
1 .60
1 .56
1 .55
1 .54
1 .48
1 .44
1 .43
1 .38
1 .35
1 .34
1 .30
1 .21
1 .13
1 .04
0 .96
0 .88
X/D
Y
/D
8.5 9 9.5
0.4
0.2
0
0.2
0.4
2. 16
2. 06
1. 95
1. 90
1. 83
1. 81
1. 71
1. 68
1. 63
1. 59
1. 57
1. 49
1. 47
1. 41
1. 35
1. 34
1. 30
1. 29
1. 26
1. 23
1. 12
1. 00
0. 88
33
2002 ME Graduate Student Conference
April 13, 2002
ON THERMAL GALLING OF OSCILLATING PINBUSHING ASSEMBLY
Haribabu Krishnamurthy
M.S. Candidate
Faculty Advisor : Dr. Michael M Khonsari
ABSTRACT
Temperature induced galling in oscillating pinbushing
mechanisms subjected to heavy loads is known to be the
cause of failure of many industrial machinery. Yet,
surprisingly, very little information is available to shed light
on this important failure mechanism. An examination of the
open literature reveals that the published work on the
thermal characteristics of pinbushing type configuration
has been limited to unidirectional rotation. For example,
Gecim and Winer
1
developed an analytical method for
predicting the surface temperature distribution of a shaft
rotating at very high speed assuming the system to be at
quasisteady state. In the case of rotating shaft, the
temperature distribution will be uniform over its entire
circumference. In contrast, when either the bushing or the
shaft is subjected to oscillatory motion, the generation of
heat is concentrated primarily over the contact angle, which
is dependent on the load and material properties. The
severity of the temperature is a function of the oscillatory
speed, load and the angle of oscillation. The prediction of
thermally induced galling failure requires a transient heat
transfer analysis with consideration of thermomechanical
interaction of the contacting surfaces.
Hazlett and Khonsari
2
performed one of the first
comprehensive finite element formulations to predict the
failure of a journal bearing due to Thermally Induced Seizure
(TIS). More recently Krithivasan and Khonsari
3
have
developed design equations to predict the seizure time, as a
function of operating conditions where in temperature is the
only contribution to the seizure of the pin. All of these
analyses are restricted to constant, unidirectional speeds.
In the present work we develop an analysis for the
prediction of the thermal behavior of oscillatory mechanism.
The objective of this study is to develop appropriate data
that can be used at the design stage to guard against the
occurrence of galling.
The method of solution is using the finite element
modeling technique. A commercial FEM solver called
FlexPDE
4
is utilized for this purpose. Referring to Figures 1
and 2, the governing equations of the system is the transient
heat equation without internal heat generation in cylindrical
coordinates. The boundary conditions are time dependent
and take into account the applied flux in the bite angle
region as well as heat loss by convection within the pin
bushing clearance and on the outer circumference of the
bushing.
The governing heat equation and the related boundary
conditions are:
t
T T
r r
T
r r
T
∂
∂
·
∂
∂
+
∂
∂
+
∂
∂
α φ
1 1 1
2
2
2 2
2
At the outer surface of the pin:
( ) ( ) [ ] T h t f q t f
r
T
k
1
1 − + ·
∂
∂ #
for 0 ≤ φ ≤ 2θ and
T h
r
T
k
1
− ·
∂
∂
for 2θ ≤ φ ≤ 2π
At the inner surface of the bushing:
q
r
T
k ·
∂
∂ #
for 0≤ φ ≤ θ
b
and T h
r
T
k
1
− ·
∂
∂
for θ
b
≤ φ ≤ 2π
At the outer surface of the bushing:
T h
r
T
k
2
− ·
∂
∂
for 0 ≤ φ ≤ 2π
# To model temperature and slope continuity in the contact
region, this particular boundary condition is coupled with
pin and bushing. Therefore, flux need not be applied on both
the pin and bushing, and it is sufficient to apply this
condition on either shaft or bush and coupled with the
other.
The application of combined flux and convection on the
pin due to oscillatory nature described in the boundary
conditions is accomplished as per the illustration in Figure
2(b). The pin is divided into five sectors and on each sector,
the boundary condition is defined by a Fourier series
function f(t). At any instant of time only either flux is applied
or convection is applied on any particular sector. The idea of
applying the time dependent boundary condition originated
from the work of Glass and Ozisik
5
who determined the
transient temperature distribution in a semi infinite medium
due to an onoff heat flux.
To begin with, a dimensional case is solved for with the
following parameters,
34
Geometry: Pin radius, R
1
=0.05 m; Bushing inner radius, R
2
=0.05025 m; Bushing outer radius, R
3
=0.1 m; Length, L= 0.1
m.
Operating Conditions: Load, W = 25000 lbs = 111.2 kN;
Oscillation speed, u = 0.08 m/s; Angular velocity, ω = 1.6
rad/s = 91.67 deg/s; Oscillation angle, θ = t15 deg from mean
position (i.e. total of 30 degrees).
Material Properties: Both pin and bushing is made of
carbon steel. Thermal conductivity, k = 52 W/mK; Thermal
diffusivity, α = 1 x 10
5
m
2
/s.
Heat Transfer: Convective coefficient, h
1
= 80 W/m
2
K
(within the clearance) and h
2
= 80 W/m
2
K (outside the
Bushing).
Analysis: Using the Hertzian contact theory, the bite angle
is predicted, to be: θ
b
= 25 deg.
Total frictional heat flux, q = f.W.u/ A
c
A
c
= R
1
.θ
b
.L = 2.279 x10
3
m
2
and friction coefficient, f = 0.1.
Therefore, the total frictional heat flux, q = 386801.74 W/m
2
.
The time for one complete oscillation (one cycle) for the
above case is 0.654 sec. Under these specifications, the
steady state maximum rise in temperature of 180
0
C is
attained around 9000 cycles, which is shown in Figure 3.
Work is in progress:
• To verify the validity of time dependent boundary
conditions applied in FlexPDE, an analytical solution will
obtained for the temperature distribution for the pin
alone subjected to oscillating flux on its circumference
for a particular oscillating angle using Duhamel’s
approach.
• To determine the galling time, (the time taken for failure,
where in the temperatures attain a value such that the pin
and bush tend to weld to each other).
• Realistic prediction of convection coefficient in the
clearance is needed for an accurate heat transfer
analysis.
• Thermal expansion of the pin and the bush are to be
considered to calculate the instantaneous heat sources
that occur from the interaction of the expanded pin and
the bushing. These effects contribute in addition to the
existing heat flux.
• Effect of protective coating on the pin surface requires
multilayer heat transfer analysis.
REFERENCES
1. Gecim B. and Winer W.O.’ Steady temperature in a
rotating cylinder subject to surface heating and
convective cooling. ASME Journal of Tribology, 1984,
106, 12027.
2. Hazlett T.L. and Khonsari M.M.’ Finite element model of
journal bearing undergoing rapid thermally induced
seizure. Tribology International, 1992a, 25, No.3, 177
82.
3. Krithivasan R. and Khonsari M.M., Thermally induced
seizures in journal bearings, 2002 ME Graduate Student
Conference, LSU, Baton Rouge, Louisiana.
4. FlexPDE Users Manual, PDE Solutions.
5. Glass.D.E and Ozisik M.N., Transient temperature
resulting from periodic onoff heat flux for a semiinfinite
mediumASME Journal of Heat transfer, 1988, 110, 250
52.
FIGURES AND TABLES
Figure 1 Finite Element model generated by FlexPDE. The
clearance is scaled by a factor of about 300 for clarity
(a) (b)
Figure 2 Implementation of boundary condition on pin
Figure 3 Variation of Maximum Temperature rise with
respect to Number of Oscillation cycles
h
2
h
1
35
2002 ME Graduate Student Conference
April 13, 2002
ACTIVE MAGTNETIC BEARING CONTROL WITH ZERO STEADYSTATE POWER LOSS
Nader Motee
M.S. Candidate
Faculty Advisor: Dr. Marcio S. de Queiroz
ABSTRACT
Active magnetic bearings (AMBs) are experiencing an
increased use in many rotating machinery applications (e.g.,
compressors, milling spindles, flywheels, etc.) as an
alternative to conventional mechanical bearings. Due to the
noncontact nature of the magnetic bearings and rotor,
AMBs have the unique ability to suspend loads with no
friction, operate at higher speeds, and operate under
environmental constraints that prohibit the use of
lubrication. Furthermore, since AMBs can be actively
controlled, they offer other potential advantages over
conventional mechanical bearings, such as eliminating
vibration through active damping, adjusting the stiffness of
the suspending load, or providing an automatic rotor
balancing capability. See [3] and the references therein for
previous work on AMB control. Typically, an AMB is
operated by introducing a “high”, constant bias magnetic
flux (or electrical current) in each electromagnet. This
conservative practice allows the system to be modeled by a
controllable linear system, thus, facilitating the application
of standard linear control design techniques (see, for
example, [2,6]). On the other hand, the bias flux increases coil
and journal power losses which may cause rotor heating and
affect the machine efficiency [4]. For example, ohmic (i.e.,
coil) losses, which arise from the flow of current in the
electromagnet coils, are proportional to square of the flux.
While lowering or eliminating the bias flux will result in
reduced ohmic power losses, it will also enhance the AMB
nonlinearities and may lead to a control singularity as was
shown in [7]. As is clear from this discussion, the design of
AMB controllers that reduce ohmic power losses is a
challenging problem. Constant lowbias nonlinear
controllers using the integrator backstepping technique
were proposed in [7,8] with no discussion about their
implication on power losses. In [4], a gainscheduled linear
controller was developed with a low bias current. A lowbias
control scheme was recently proposed in [11] using the
smallgain theorem. Zerobias control approaches can be
found in [1,5,12]. Unfortunately, these zerobias results have
the common drawback of potentially producing unbounded
voltage control inputs. In this paper, we propose a solution
to the problem of AMB control with reduced ohmic power
losses. Specifically, our goal is to design a control law with
the following three attributes: regulates the rotor position to
zero, eliminates the steadystate bias flux, and contains no
control singularities. The starting point for the proposed
control law is the lowbias nonlinear control structure
presented in [7]. We modify the controller of [7] by
introducing a timevarying bias flux in the form of an
exponentially decaying function. In the steady state, this
will produce zero bias flux, and hence, reduce the ohmic
power losses. The proposed controller is shown to ensure
the exponential stability of the closedloop system with the
rate of convergence dependent on control gains. This fact
facilitates the avoidance of control singularities by allowing
us to force the rotor position and velocity to converge to
zero faster than the bias flux. The basic premise behind the
proposed control/bias flux design can be physically
explained as follows. After the rotor is centered in the
bearing system, there is no need for the AMB
electromagnets to apply forces on the rotor; thus, the
electromagnets can be deenergized by turning off the bias
flux. We will address the abovedescribed control problem
for two different modes of operation of the AMB system.
Initially, we consider the standard AMB mode of operation
where all electromagnets are active at any given time. We
will refer to this operating mode as “fullflux” (FF). Next, we
will consider the mode of operation commonly known as
“complementaryflux” (CF) [12] where only one
electromagnet along each direction is active at any given
time. As explained in [12], this reduces the total flux in the
AMB system (and hence, helps reduce the ohmic power
losses) since the electromagnets do not produce
counteracting forces. Unlike what is normally presented in
the literature, we provide a rigorous treatment of the CF
mode of operation inclusive of a stability analysis, switching
strategy, and circuit implementation
36
ACKNOWLEDGMENTS
This work was supported in part by the Louisiana State
University Council on Research.
REFERENCES
1. Charara and B. Caron, “Magnetic Bearings: Comparison
Between Linear and Nonlinear Functioning,” Proc. Int.
Symp. Magnetic Bearings, pp. 451460, Alexandria, VA,
Jul. 1992.M
2. . Fujita, K. Hatake, and F. Matsumura, “Loop Shaping
Based Robust Control of a Magnetic Bearing,” IEEE
Control Syst. Mag., Vol. 13, No. 4, pp. 5765, Aug. 1993
3. IEEE Transactions on Control Systems Technology,
Special Issue on Magnetic Bearing Control, Vol. 4, No.
5, Sept. 1996
4. C. Knospe and C. Yang, “GainScheduled Control of a
Magnetic Bearing with Low Bias Flux,” Proc. Conf.
Decision and Control, pp. 418423, San Diego, CA, Dec.
1997
5. J. Levine, J. Lottin, and J.C. Ponsart, “A Nonlinear
Approach to the Control of Magnetic Bearings,” IEEE
Trans. Control Syst. Tech., Vol. 4, No. 5, pp. 524544,
Sept. 1996.
6. A.M. Mohamed and F.P. Emad, “Conical Magnetic
Bearings with Radial and Thrust Control,” IEEE Trans.
Automatic Control, Vol. 37, No. 12, pp. 18591868, Dec.
1992
7. M.S. de Queiroz and D.M. Dawson, “Nonlinear Control
of Active Magnetic Bearings: A Backstepping
Approach,” IEEE Trans. Control Syst. Tech., Vol. 4, No.
5, pp. 545552, Sept. 1996
8. M.S. de Queiroz, D.M. Dawson, and H. Canbolat, “A
BacksteppingType Controller for a 6DOF Magnetic
Bearing System,” Proc. IEEE Conf. Decision and
Control, pp. 33703375, Kobe, Japan, Dec. 1996
9. M.S. de Queiroz, D.M. Dawson, S. Nagarkatti, and F.
Zhang, LyapunovBased Control of Mechanical
Systems, Cambridge, MA: Birkhauser, 2000
10. J. Slotine and W. Li, Applied Nonlinear Control,
Englewood Cliff, NJ: Prentice Hall Co., 1991
11. Tsiotras and E. Velenis, “LowBias Control of AMB's
Subject to Saturation Constraints,” Proc. IEEE Conf.
Control Appl., pp. 138143, Ancorage, AK, Sept. 2000
12. P. Tsiotras, B. Wilson, and R. Bartlett, “Control of a
ZeroBias Magnetic Bearing Using Control Lyapunov
Functions,” Proc. IEEE Conf. Decision and Control,
pp. 40484053, Sydney, Australia, Dec. 2000
13. H.H. Woodson and J.R. Melcher, Electromechanical
Dynamics  Part I: Discrete Systems, New York, NY:
John Wiley & Sons, 1968.
37
2002 ME Graduate Student Conference
April 13, 2002
PRECIPITATE GROWTH FEATURES IN THE DUPLEX SIZE
γ ′ DISTRIBUTION IN THE SUPERALLOY IN738LC
Indranil Roy
Ph.D. Candidate
Faculty Advisor: Dr. A. Raman
ABSTRACT
IN738LC is a modern, nickelbased superalloy utilized at
high temperatures in aggressive environments. Durability of
this superalloy is dependent on the strengthening by g ¢
precipitates. Like other superalloys, IN738LC owes this
exceptional strength to its FCC Nirich matrix strengthened
by the g ¢ Ni3 (Al, Ti) precipitate phase with L12 superlattice
structure. The volume fraction of this precipitate phase in
this superalloy is about 40 to 43 %.
Microstructure control and stabilization is very
necessary for the effective utilization of the superalloy
IN738LC at high temperatures. Developing suitable
processes to reduce the grain size of the parent phase and
preventing the grain growth and precipitate growth to
improve the mechanical properties of the material are
important in industrial applications. The current study
focuses on the stability of the duplex size g ¢ precipitate
microstructure in the alloy.
A standard heat treatment that is generally applied in
the industry to IN738LC is a solution treatment at 1120 °C /
2h / WQ and a subsequent aging treatment at 850 °C / 24h /
FC. This treatment gives the duplex size distribution for the
g ¢ precipitates in the matrix. A microstructure with fine g ¢
precipitates develops if solutionizing is carried out at 1200
°C / 2 h / AAC. Agings at lower temperatures after 1200 °C /
4 h / AAC or WQ conditions lead to precipitate growth.
Aging at 1120 °C / 24 hours / WQ produces coarse size g ¢
precipitates. Further heating of the above specimens at 1140
°C / 6 hours / WQ produces a duplex microstructure1).
Increasing the holding times from 1 hour to 100 hours shows
joining of the fine particles to form bigger particles, leading
to distinct raft patterns. Two different g ¢ precipitate growth
processes were observed: merging of smaller precipitates to
produce larger ones (in duplex precipitatesize
microstructures) and growth through solute absorption from
the matrix (Oswald Ripening)2).
The growth characteristics of the g ¢ precipitate in their
duplex size distribution (fine and medium sizes) are not
known. The current research takes up this issue and
analyzes the simultaneous growth features of both the fine
and the medium size precipitate particles in the duplex size
microstructure when the alloy samples are heated for
different lengths of time at temperatures in the range of 800
°C to 1100 °C.
The experimental procedure is as detailed. Small pieces
of the alloy IN738LC, 4 mm thick and 6 mm edge length were
wrapped in stainless steel foils and sealed in silica tubes in
vacuum and solution treated at 1200 °C for 4 hours. The
tubes were quenched in cold water and then reheated at
1120 °C for 24 hours resulting in a coarse precipitate
microstructure. The temperature was then raised to 1140 °C
and the coarse precipitates were allowed to dissolve partially
into the matrix. Water quenching the capsules after 6 hours
at 1140 °C yielded the starting g ¢ precipitate duplex size
distribution.
These samples were later individually sealed under
vacuum in silica capsules and heat treated at 800 °C, 850 °C,
900 °C, 980 °C, 1040 °C and 1100 °C for periods of 1, 3, 7, 13,
25, 50, 100 and 200 hours. Six to eight capsules, each
carrying one sample, were put in the furnace at one of the
chosen temperatures. The samples were individually
removed, after specified time periods at the given
temperature, and quenched in cold water. Cold water
quenching retained the precipitate grain microstructure after
the given reheating.
For precipitate microstructure studies the samples were
ground using grinding paper with coarse (120) to fine (600)
grit size and polished on rotating wheels with fine alumina
powder of size down to 0.05 micron. After cleaning they were
etched with a solution of composition 33 % HNO3 + 33 %
Acetic acid + 33 % H2O + 1 % HF. JEOL 840A Scanning
Electron Microscope equipped with an ultrathin window
EDS was used to characterize the size and morphology of
the precipitates.
Microstructures were digitally recorded using a
Macintosh Quadra data acquisition and control system and
analyzed for precipitate grain size later. Precipitates in
selected microstructures were also analyzed for composition
38
using Energy Dispersive Spectrometer (EDS) attached to the
SEM. Magnification is maintained at 8000 X to facilitate
direct comparison.
The Salient Results are as follows.
(i) Both the fine and medium size g ¢ precipitates grow with
time at given temperature (e.g. 1040 °C) for which
microstructural changes are illustrated with time – Fig 1.
(ii) With increasing temperature growth rate increases for
constant time – Fig 2 shows micrographs at different
temperatures for 25 hours of heating time.
(iii) From the sizes of precipitates obtained from different
microstructures, the molar activation energy for the
growth process for both the fine and the coarse
precipitates were calculated using the formula:
) ( exp
0 0
RT
Q
K K where Kt d d
n n
t
− · · −
(iv) Size data obtained are plotted in Fig 3 for different times
at 1040 °C and in Fig 4 for the 25 hours treatment at
different temperatures.
(iv) From these data, activation energies for growth of fine
and coarse precipitates were obtained. Their values are:
_______ and _____ respectively (from data for
different temperatures for 25 hours, Fig 3) in the range
_______ to ______ °C and ______ to _______ in the
range _______ to _______ °C. The corresponding
activation energies at 1040 °C are _______, ______,
respectively for fine and coarse precipitates in the time
range 13 to 50 hours.
FIGURES AND TABLES
1 hour 3 hours 7 hours
13 hours 25 hours 50 hours
Figure 1: Microstructures of duplex IN738LC aged for
different times at 1040 °C /WQ
800 C 850 C 900 C
980 C 1040 C 1100 C
Figure 2: Microstructures of duplex IN738LC aged at
different temperatures for 25 hours/ WQ
Grain Size (Mu m) Vs Aging time (Hours)
0.22
0.25
0.29
0.35
0.55
0.48
0.53
0.58
0.64
0.7
0.74
0.42
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60
Aging Time (Hours)
G
r
a
i
n
S
i
z
e
(
M
u
m
)
Fine
Coarse
Figure 3: Plot of precipitate sizes Vs time at 1040 °C
Grain Size at 25 hours at diff.
Temperatures
0 . 5
0 . 7 3
0.12
0 . 2 1
0 . 2 8
0 . 3 6
0 . 4 2
0 . 7
0 . 6 4
0 . 5 5
0 . 4 8
0 . 3 8
0
0.2
0.4
0.6
0.8
1
750 800 850 900 950 1000 1050 1100 1150
Temper at ur es Deg C
Fine
Coarse
Figure 4: Plot of precipitate sizes Vs temperature for 25
hours aging treatment.
ACKNOWLEDGMENTS
Support from NASA through DGAP grant sponsored
through the LA Board of Regents and LaSPACE consortium
is appreciated.
REFERENCES
1. Ercan Balikci, PhD Dissertation, p 51, LSU, Baton
Rouge, LA, May 1998.
2. Ercan Balikci, A. Raman, R. Mirshams, Met. and
Materials Transactions A, 28 A (10/1997) 19932003.
3. R.E. Reed – Hill and R Abbaschian: Physical Metallurgy
Principles, 3rd ed., 1992, pp – 256261.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.