Professional Documents
Culture Documents
of vibratory stresses at various low and high frequency vibration modes under either
non-rotating or rotating conditions.
Similar to the blade failure, the structural reliability and safety of the adhesive bonded joint, one of the most commonly used structural joint designs in the
aerospace industry, is also a serious concern of the aircraft design community. Adhesive joints easily become weaker due to environmental degradation and/or improper
manufacturing procedures. This often reduces structural durability and reliability
signicantly. This motivates us to develop a new nite element tool/procedure
for assessing the interfacial disbonding mechanics of the single-lap joint with various imperfectly-bonded conditions in order to predict the adhesive bonded joint's
strength more precisely during its service period. According to these conclusions,
a new three-dimensional graphic mesh has been created to display the maximum
stress variations under dierent amounts and sizes of disbonded area. This new
procedure can be used as a basis for the development of a bonded joint reliability
prediction method and accept/reject inspection criteria.
Approved by
Adviser
Aeronautical and
Astronautical Engineering
Graduate Program
c Copyright by
Hsin-Yi Yen
2000
ABSTRACT
Most load-carrying structural systems such as gas turbine engines are usually
under severe operating conditions. This type of structure demands durability, high
reliability, light weight, and high performance. In turn, as it has been reported, a
number of structural failures have occurred in aircraft engines during development
testing and operational service. In order to prevent failures of turbine engines, it
becomes necessary to attenuate the turbine blade vibration to an acceptable level.
To achieve this goal, the blade has to be provided with higher damping, either
externally or internally. The objective of this study is to explore the feasibility of
using a stress dependent magnetomechanical surface coating material for enhancing
high damping capacity on turbine blades.
In the beginning, an analytical procedure is developed for modeling the dynamic behavior of beams and blades coated with a plasma sprayed iron-chromium
based coating. The stress or strain dependent damping capability of the coating
material is evaluated experimentally and further quantied with a probabilistic distribution function. The equations of motion and associated boundary conditions
are derived for uniform coated beams and blades. Later, the resulting equation of
motion is solved for beams and blades with a free surface coating layer by a closed
form procedure and nite element approximation. The results show that a 4% of
blade thickness free surface magnetomechanical coating layer can have a signicant
ii
under dierent amounts and sizes of disbonded area. This new procedure can be
used as a basis for the development of a bonded joint reliability prediction method
and accept/reject inspection criteria.
iv
ACKNOWLEDGMENTS
I would like to acknowledge my adviser, Dr. M. -H Herman Shen for his guidance,
encouragement, support, and care during the course of this dissertation. As my
adviser, his suggestions and observations contributed immensely to the success of
this work.
I would like to acknowledge Dr. Gerald M. Gregorek, Dr. Rama K. Yedavalli,
and Dr. Chia-Hsiang Menq for serving on my dissertation committee and for various
suggestions for this dissertation.
Additionally, I also would like to thank Ms. Sandy Rhoads for helping my
paper works , and thank Mr. Tommy George for his valuable suggestions in this
dissertation.
Finally, I thank my parents for their support and encouragement and all my
friends at The Ohio State University for their assistance during the past four years.
vi
VITA
H. -Y. Yen and M. -H. Herman Shen \Passive Vibration Suppression of Turbine
Blades Using Magnetomechanical Coating". Journal of Sound and Vibration, under
revision process.
H. -Y. Yen and M. -H. Herman Shen \Development of a New Modeling Procedures
for the Imperfectly-Bonded Interface in Single-Lap Joints". International Journal
of Adhesion and Adhesives, under review.
H. -Y. Yen and M. -H. Herman Shen \A New Finite Element Modeling Technique
for Disbonded Adhesive Joints". to appear in 2000 ASME International Mechanical
Engineering Congress and Exposition, November 2000.
vii
H. -Y. Yen and M. -H. Herman Shen \Development of Passive Turbine Blade
Damper Using Magnetomechanical Coating". International Gas Turbine & Aeroengine Congress & Exhibition, 2000-GT-0366, May 2000.
H. -Y. Yen and M. -H. Herman Shen \Magnetomechanical Coating Applications in
Vibration Suppression of Turbine Blades". 5th National Turbine Engine High Cycle
Fatigue Conference, March 2000.
H. -Y. Yen and M. -H. Herman Shen \Passive Vibration Suppression of Turbine
Blades using Magnetomechanical Coating". 4th National Turbine Engine High
Cycle Fatigue Conference, February 1999.
H. -Y. Yen and M. -H. Herman Shen \Passive Vibration Suppression of Beams Using
Magnetomechanical Coating". Vibration and Noise Control, DE-Vol. 97/DSC-Vol.
65, ASME, 1998.
H. -Y. Yen and M. -H. Herman Shen \The Eects of Fatigue Cracks on Free
Torsional Vibration of Shafts". International Gas Turbine & Aeroengine Congress
& Exhibition, 97-GT-249, June, 1997.
FIELDS OF STUDY
viii
TABLE OF CONTENTS
Page
Abstract . . . . . .
Dedication . . . . .
Acknowledgments .
Vita . . . . . . . .
List of Tables . . .
List of Figures . .
Chapters:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Problem Review and Literature Search . . . . . . . . . . . . . .
1.2.1 Vibration Suppression of Turbine Blades . . . . . . . . .
1.2.2 The Modeling Technique of the Bonded Joint . . . . . .
1.3 Problem Approach . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Magnetomechanical Coating on Turbine Blade . . . . . .
1.3.2 The Modeling Technique for Imperfectly-Bonded Joints .
. ii
. v
. vi
. vii
. xii
. xiii
.
.
.
.
.
.
.
.
1
1
7
7
12
15
15
16
19
19
22
30
ix
37
38
42
49
61
73
100
101
105
109
109
111
120
6. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Appendices:
A. Collocation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
B. The Abbreviated MSC/NASTRAN Routine for the Cantilevered Blade
with the Magnetomechanical Coating . . . . . . . . . . . . . . . . . . . 140
x
xi
LIST OF TABLES
Table
Page
xii
LIST OF FIGURES
Figure
Page
1.1 Air Force Engine Fatal Accident Rate Caused by Engine Failures. .
1.2 High Cycle Fatigue Failure in the Turbine Engine Blade [22]. . . . .
3.16 The von Mises Stress in the Second Stripe Mode Vibration of the
Coated Cantilevered Blade . . . . . . . . . . . . . . . . . . . . . . 68
3.17 The von Mises Stress in the Third Stripe Mode Vibration of the
Cantilevered Blade . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.18 The von Mises Stress in the Third Stripe Mode Vibration of the
Coated Cantilevered Blade . . . . . . . . . . . . . . . . . . . . . . 70
3.19 The von Mises Stress in the Fourth Stripe Mode Vibration of the
Cantilevered Blade . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.20 The von Mises Stress in the Fourth Stripe Mode Vibration of the
Coated Cantilevered Blade . . . . . . . . . . . . . . . . . . . . . . 72
3.21 The Rotating Diagram of the Coated Blades . . . . . . . . . . . . . 76
3.22 The Third Stripe Mode Shape of the Cantilevered Blade about Rx
under 10000 rpm Rotation . . . . . . . . . . . . . . . . . . . . . . . 77
3.23 The Fourth Stripe Mode Shape of the Cantilevered Blade under 10000
rpm Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.24 The von Mises Stress of the Fourth Stripe Mode of the Cantilevered
Blade Rotating about Rx under 10000 rpm . . . . . . . . . . . . . . 79
3.25 The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Rx under 10000 rpm with 2% Thickness Coating Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.26 The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Rx under 10000 rpm with 4% Thickness Coating Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.27 The von Mises Stress of the Fourth Stripe Mode of the Cantilevered
Blade Rotating about Ry under 10000 rpm . . . . . . . . . . . . . . 82
3.28 The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Ry under 10000 rpm with 2% Thickness Coating Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
xv
3.29 The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Ry under 10000 rpm with 4% Thickness Coating Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.1 The Geometry and Finite Element Mesh of the Bonded Joints . . . 90
4.2 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Linear Geometry (Uniform Loads)(GR: Goland
and Reissner; OP: Oplinger; HS: Hart-Smith; FEM: Finite Element
Method )(y=0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Nonlinear Geometry (Uniform Loads)(y=0)(GR:
Goland and Reissner; OP: Oplinger; HS: Hart-Smith; FEM: Finite
Element Method) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.4 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Nonlinear Geometry (Single Loads)(y=0)(GR:
Goland and Reissner; OP: Oplinger; HS: Hart-Smith; FEM: Finite
Element Method ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Nonlinear Geometry (Uniform Loads)(y=0) . . . 94
4.6 Adhesive Stress Distributions Obtained from the Present Finite Element Model for Double-Strap Joint (OP: Oplinger; HS: Hart-Smith;
FEM: Finite Element Method) . . . . . . . . . . . . . . . . . . . . . 98
4.7 Adhesive Stress Distributions Obtained from the Present Finite Element Model Compared with TJOINTNL Program (OP: Oplinger;
HS: Hart-Smith; FEM: Finite Element Method) . . . . . . . . . . . 99
5.1 The Geometry and Finite Element Model of the Imperfectly-Bonded
Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Perfectly-Bonded Interface (Uniform Loads) . . . 116
xvi
5.3 Adhesive Stress Distributions Obtained from the Present Finite Element Model with 20% (Ad=0.2) Imperfectly-Bonded Interface by
Using the Spring Element (Uniform Loads) . . . . . . . . . . . . . . 117
5.4 Adhesive Stress Distributions Obtained from the Present Finite Element Model with 20% (Ad=0.2) Imperfectly-Bonded Interface by
Using the Line Element (Uniform Loads) . . . . . . . . . . . . . . . 118
5.5 Adhesive Stress Distributions Obtained from the Present Finite Element Model with Various Disbonded Areas by Using the Line Element.(30%
(Ad=0.3) near the End of Overlap Decreases to 0% (Ad =0) in the
Center) (Uniform Loads) . . . . . . . . . . . . . . . . . . . . . . . . 119
5.6 The Strength of the Interface with Various Disbonded Island Sizes . 121
5.7 Maximum Peel Stress Distributions at the Upper Level (y = t ) of
the Bonded Interface along the Percentage of Disbonded Area and
Disbonded Island Size (inches) . . . . . . . . . . . . . . . . . . . . . 128
2
xvii
LIST OF SYMBOLS
Ad
E
Eab
Ec
Es
Ea
Eb
Fint
F (x)
FN
Ft
G
I
Is
Kj
K
K 0
Kt
Kn
Kr
Ks
M
N
Uc
Ud
Uj
a
bl
b
2Cb
2d
db
h
l
lb
m
s
t
ux
u
u; v; w
y1 ; y2
K
Uc
b
c
s
i
j
loc
l
cr
i
a
b
x
!
t
Vibratory Strain
Average Internal Strain
Shear Stress of the Adhesive Bonded Joint
Eective Poisson's Ratio
Poisson's Ratio for Adherends
Poisson's Ratio for Adhesives
Shear Deformation of the Bonded Interface
Free Vibration Natural Frequency
xx
CHAPTER 1
INTRODUCTION
1.1 Background
Many load-carrying structural systems such as aircraft gas turbine engines are
typically under severe operating conditions. This type of structure demands durability, high reliability, light weight, and high performance. Traditionally, lifetime
failure-free design criteria based on the Goodman Diagram and Miner's Rule have
been adopted by the aircraft engine design community for ensuring safety of critical
structural components.
These design criteria, design guides, or design codes are often established using
the results of a simple deterministic analysis procedure without taking into account
information such as degradation of material properties, scatter in test data, previous successful design experience, and uncertainties inherent in real world operating
conditions. In turn, as has been reported, a number of structural failures have
occurred in aircraft engines during development testing and operational service.
Among those failures, the Sioux City incident is one of the most famous examples.
In 1989, United Airlines
ight 232 enrout from Denver to Chicago crash landed
in Sioux City, Iowa, due to the defected fan disk failure of its No. 2 engine. In
1
Figure 1.1: Air Force Engine Fatal Accident Rate Caused by Engine Failures.
this incident, the failed disk broke the turbine blades which rendered all three hydraulic systems on board inoperative. This made the DC-10 almost uncontrollable
and resulted in a crash landing killing 111 of 296 passengers and crew members.
These incidents triggered an awareness of the fact that although current aircraft
engine critical structural components satisfy the lifetime failure-free design criteria
they sometimes fail, blading systems in particular. Although the Sioux City incident was caused by a material defect during the manufacturing process, high cycle
fatigue of the blades in jet engines is often the major concern in aviation safety,
2
especially in high performance military jet, as shown in Fig. 1.1. High cycle fatigue directly causes blade cracking, which increases maintenance and inspection
costs, reduces operational readiness, and sometimes even results in the loss of the
aircraft and crew members on broad. In the past one year alone, the U. S. Air
Force spent roughly one third of its total maintenance expenses on the high cycle
fatigue incidents [25]. According to their records, there are at least two F-16 ghter
crashes related to high cycle fatigue incidents within the last twelve months. One
was caused by a catastrophic failure in high pressure turbine assembly when two
turbine blades separated due to the high cycle fatigue, as shown in Fig. 1.2. The
other incident was caused by rst-stage compressor failure when one compressor
blade broke away due to high cycle fatigue. In this incident, an ultrasound inspection test had been performed on that particular engine, but still failed to detect
the developing cracks. Until recently, a newly designed U.S. navy F/A-18E Super
Hornet ghter was grounded due to the compressor blade failure of its new 22; 000
lb thrust GE F414 engine. After a full examination, fatigue cracks were found near
the tip of two compressor blades. A root cause analysis conducted by the U.S.
Navy pinpointed the cause of this cracking problem which was primarily due to the
high cycle fatigue. This cracking problem is directly caused by the extreme maneuvering the aircraft involved during its regular testing procedure. Unfortunately,
extreme maneuvering is one of the activities that military ghters cannot avoid
during daily training missions. This makes avoiding high cycle fatigue problems the
priority mission in the new integrated high performance turbine engine technology
(IHPTET) for the next generation Joint Strike Fighter (JSF). This project has been
corresponded by all the current major engine research groups, including Air Force,
3
Figure 1.2: High Cycle Fatigue Failure in the Turbine Engine Blade [22].
Navy, NASA, GE Aircraft Engine, Allison Advanced Development, Pratt-Whitney,
Williams International, Teledyne Ryan Aeronautical, Allied Signal Aerospace, and
a few smaller companies as well as other government agencies. Therefore, preventing turbomachinery blade failures caused by high cycle fatigue is one of the major
objectives of current aircraft engine design and in-service maintenance.
Similar to the blade failure, the structural reliability and safety of the environmentally degraded adhesive bonded joint, one of the most commonly used structural
joint designs in the aerospace industry, is also a serious concern of the aircraft design
4
community. The bonded joints have been widely used in Boeing 737, 747, 757, and
767s as bonded tear strap designs, shown in Fig. 1.3 and Fig. 1.4, or in common
airplane fuselage windows and doors as bonded doubler designs.
adhesive joint's lifetime tremendously. Recently, the French Civil Aviation Directorate issued a warning about a potential problem related to the weaker strength
of adhesive bonds in A-320 series and A330-200 composite vertical stabilizers. This
alarmed the aerospace industry of the potential reliability and safety problem of
adhesive bonded joints. Unfortunately, the current non-destructive inspection technologies, such as ultrasonic wave and X-ray, have diculties to monitor the quality
of the bonded joints. Recently, the composite honeycomb liquid hydrogen tank of
the NASA X-33 delaminated during testing when the adhesive bonds appeared to
be in perfect condition under ultrasonic test. This leads the urgency to develop a
method or technology in order to monitor and detect the imperfectly-bonded (poor)
interface of the adhesive bonding joints.
1.2 Problem Review and Literature Search
1.2.1 Vibration Suppression of Turbine Blades
These failures of turbine blades generally result from high vibratory or alternating stresses in various stage turbomachinery blades during resonant response.
Resonant response occurs, in a non-uniform
ow eld operating environment when
the excitation frequency from unsteady aerodynamic loading is coincident with any
blade's natural frequency. Hence, an adequate blade design would exist if all the resonant conditions could be avoided. Realistically, it is impossible due to the fact that
insucient real world loading information is available at the time of the analysis
and design.
To prevent blade failure, the maximum stress of the excited resonant response
needs to be attenuated to an acceptable level through the control of blade vibration.
7
As an example in Fig. 1.5 obtained from Pratt-Whitney, the engine life cycle drops
tremendously when the maximum vibratory stress slightly increases [1]. Among
the solutions to reduce blade vibratory stress, adding damping is the most attractive and achievable method. In the last thirty years, several investigators have
presented dierent approaches to suppress blade vibratory stress by providing additional damping through blade dampers. For example, dry friction dampers [15]
which include blade-to-ground, blade-to-blade, and shroud dampers are the most
common vibration suppression devices employed by aircraft engine designers, shown
in Fig. 1.6.
F-100-PW-2200 Engine
Stator Row
Friction Damper
blades no longer have the shrouds in order to improve their aerodynamic performance, which even further reduces the installation location of traditional friction
dampers. At the same time, the small displacements of the high frequency vibration modes make aerodamping also relatively small. Therefore, since the structural
damping from the dry friction dampers as well as the aerodamping are negligible
at high frequency vibration modes such as second stripe, third stripe, and fourth
stripe vibration modes, the remaining damping of the blades is solely from the energy dissipation from the material. Unfortunately, modern low aspect ratio blades
frequently vibrate like this plate-like stripe modes resulting in higher modal density
in the engine operating range. Consequently, the low material damping results in
high vibratory stress, increased failure risk, and signicantly reduced reliability and
safety [29]. This has motivated the recent research activities for the development
of high frequency dampers.
Recently, numerous studies [14, 19] have been undertaken regarding the integration of viscoelastic damping materials into rotating blades for the purpose of
reducing vibratory stresses in high frequency modes. In this viscoelastic patch, the
additional vibratory energy dissipation is accomplished through high internal friction of the viscoelastic material patches inserted into milled cavities, which are then
sealed with a coversheet to maintain the structural integrity and the original airfoil
contour. However, this internal constrained layer damping treatments are not be
widely used because the viscoelastic material creep under the high loads and its
temperature limitation.
Recently, another interesting magnetic vibration suppression system has been
introduced by Homan [18]. This magnetic damping system contains three groups
10
of electromagnets embedded near the blade tip. Each electromagnet is powered and
controlled by one central electronics module. The module can turn on the magnetic
elds when the rotor is at a known critical speed or passing through a vibration
mode. By powering the electromagnets, the energy from the blade vibration can
be dissipated through the eddy currents caused by the additional magnetic elds
from the electromagnets. In addition to the temperature limitation problem, adding
damping patches to blades with either internal cavities or internal electromagnets
also has manufacturing and durability considerations, especially with the cooling
lm design in the modern turbine blade. To avoid those diculties, a surface high
damping layer, such as a high damping coating layer, is still likely to be more
practical.
Cross, Lull, Newman, and Cavanagh [10] presented an alternative damping
method, using several candidate materials as graded plasma coating on aluminum
blades for accomplishing higher structural damping. Their test data demonstrated
the use of three to six layers of coating, such as magnesium aluminate, molybdenum, and Hastelloy-X, to enhance the structural damping of the blades. The
coating consists of an outermost layer portion formed of an oxdie ceramic or refractory carbide an intermediate portion formed of a mixture of one of the above
alloys, and the material forming the outermost layer. However, the graded plasma
coating materials introduced in this study are used only on the aluminum blades,
and it is still questionable about their high damping capacity when operating under
high temperatures and high rotating speed environments. It becomes necessary to
consider another alternative coating method to suppress the blade vibration.
11
The bonded adhesive joint is one of the most commonly used structural joint
congurations. Therefore, the stress analysis of the adhesive joint, especially for
the single-lap joint, has been in development for more than a half century. Goland
and Reissner [13] developed a stress analysis which models the single-lap joints
as bending plates. In their analysis, the stress distribution of a single-lap joint
was formulated in accordance with the joint geometries and material properties
of the adherend and adhesive. Based on Goland and Reissner's assumption, the
eccentricity of the loading path creates a bending moment, the so-called joint edge
moment. This bending moment has a major eect on the stress distributions of
the overlap of the adhesive joint. Furthermore, Goland and Reissner acknowledged
the critical role of the transverse normal (peel) stress near the ends of the adhesive
layer.
Later, Hart-Smith [16] improved the Goland and Reissner analysis by treating
the adherends as elastic beams, and calculated the stress distributions for linear elastic and elastic-plastic adhesives. Therefore, the Hart-Smith analysis is not restricted
by the lumped overlap which is assumed in the Goland and Ressiner analysis, and
provides more accurate linear elastic or elastic-plastic solutions. Recently, Oplinger
[26] proposed a more realistic layered beam model by considering the in
uence of
the large de
ection of the overlap, so called the geometrical nonlinear eect.
In order to further understand the stress distribution along the thickness of
the adhesive and to ensure the boundary conditions are satised at the free ends,
several researchers have proposed two-dimensional analytical solutions. Chen and
12
Cheng [8] rst computed the stress distribution at the overlap of the single-lap
joint by using the variational principle of complementary energy. In their study, a
more realistic stress distribution in the adhesives was obtained. However, due to
the calculation diculties in Chen and Cheng's analysis, the longitudinal normal
stresses were assumed to be zero and the moment and shear force at the end of
the overlap are treated as external forces obtained by the Goland and Reissner
formulations.
In addition to Chen and Cheng's work, Barthelemy, Kamat, and Brinson [4]
develop a two dimensional nite element program called STAP to analyze a singlelap joint. After testing several dierent two-dimensional shell elements, an eightnoded isoparametric element was chosen to be used in their program, and proved to
be relatively accurate in comparison with their experimental data. Later, Lang and
Mallick [20] also created a two-dimensional nite element analysis to investigate
the recessed bonded joints. In their investigation, the stress distributions at the
adhesive mid-thickness and interface were determined for joints with various levels
of recessing, and compared to a continuous single-lap joint solution. More recently,
Tong [33] developed another two-dimensional nite element procedure to simulate
the relationship between surface displacement and adhesive peel stress in bonded
double-lap joints. However, in spite of tremendous numbers of papers related to the
perfectly-bonded structure analysis, the mechanics of imperfectly-bonded joints,
the common practical situation, are still not been studied. Therefore, the aerospace
and airline industry has begun to question the imperfectly-bonded joints' safety and
reliability.
13
First in 1984, Bail and Thompson [3] developed a quasi-static model for the
ultrasonic transmission and re
ection at the imperfectly-bonded interface. In their
work, the imperfectly-bonded interface is assumed to be represented by a distributed
spring. This spring stiness has been derived from the solutions of fracture mechanics [32] for the elastic displacement of materials containing cracks and inclusions under static loads. Later, Margetan, Thompson, and Gray [23] extended this
quasi-static spring model to represent the ultrasonic re
ectivity of an imperfectlybonded interface as a function of frequency and angle of incidence. Those results
were then incorporated in a new model for the corner re
ection from a diusionbonded joint between two plates. In their study, the distributed spring model used
for modeling the cracks can represent the imperfectly-bonded interface relatively
well. In addition, Lavrentyev and Rokhlin [21] further applied this distributed
spring model in several adhesive bonded joint non-destructive evaluations to simulate the environmental degradation. The distributed spring model in Lavrentyev
and Rokhlin's research has been used as the transverse spring boundary conditions
in order to simulate the ultrasonic wave re
ection and the frequency shift caused by
the imperfectly-bonded interface lled by absorbed water. Unfortunately, the previous three studies are restricted to using the distributed spring model to simulate
the ultrasonic characterization and interactions, such as ultrasonic amplitude reduction and re
ection frequency shift of the imperfectly-bonded interface. In turn,
the stress and strength changes in the disbonded joints caused by the environmental degradation still can not be predicted, which increase the safety concerns and
inspection diculties.
14
In this study, the objective is to explore the feasibility of using a magnetomechanical surface coating material [7] for enhancing the high frequency damping of
turbine blades. The additional damping is achieved through internal friction via
the magneto-elastic eects caused by stress or strain-induced irreversible movement
of magnetic domain walls. According to the domain theory [9], every ferromagnetic
material consists of so-called domain walls which are more or less randomly oriented
in an unmagnetized material. The domain walls of the coating material rotate and
generate a higher magnetic eld strength under external loading, because the domain walls tend to align themselves in the direction of the stress or stain. As the
loading is removed, the domain walls rotate to a dierent pattern which corresponds
to a lower magnetic eld strength. This movement of the domain walls results in an
irreversible magnetostrictive strain. During this process, it has been observed and
shown to produce a signicant amount of magnetomechanical hysteresis energy loss
as a result of the magnetomechanical eect.
The strain or stress level dependence of the magnetomechanical damping capacity or loss factor has been experimentally determined in Chapter 2, which in general
is to linearly reach a pronounced peak and then decay slowly to its initial value as
the stress or strain increases. A theoretical foundation is laid down to characterize the energy dissipation due to the irreversible movement of magnetic domain
boundaries of the coating material. Following the energy dissipation result, an analytical model for a uniform beam coated with a plasma sprayed magnetomechanical
15
Hart-Smith [16], and Oplinger [26]. The optimal nite element mesh was then used
as a basis in Chapter 5 for the development of a modeling technique for predicting
the eects of disbonded interfaces in the adhesive joints.
In Chapter 5, the disbonded interface modeling technique is accomplished by
introducing a customized spring element as the connections between the adhesive
bonds and adherends to simulate the weaker imperfectly-bonded interface. The
stiness of the spring elements represent the strength of the imperfectly-bonded
interface and depend on the percentage of disbonded area to the overall interface
area under a xed individual disbonded island size.
Additionally, in this study, a new one-dimensional six-noded line element has
also been developed and integrated into the procedure for assessing the damage
progress of the imperfectly-bonded interface. Similar to the spring element model,
the line element also simulates the strength changes of the disbonded interface by
modifying the stiness matrices in the nite element procedure. The advantage of
this line element approach is that the imperfectly-bonded joint problems can be
solved directly with the modied stiness matrices without adding more boundary
conditions like the distributed spring element procedure. This improvement with
the line element model provides a better ability to predict the disbonding mechanics
than the spring model.
By using the previous results, a three-dimensional graphic mesh has been created
to represent the maximum stresses of the imperfectly-bonded interface change under
dierent disbonded island size and disbonded area in the last part of Chapter 5.
The purpose of this mesh is to understand the eects of the disbonded island size
and the disbonded area on the quality of imperfectly-bonded joints. Furthermore,
17
by using this information, the safety and reliability of imperfectly-bonded joints can
be determined during the environmental degradation. Therefore, these results can
be used as a design or inspection criteria to predict the bonded joint reliability and
create an accept/reject inspection procedure for the bonded joints.
18
CHAPTER 2
PASSIVE VIBRATION SUPPRESSION WITH
MAGNETOMECHANICAL COATING MATERIAL
2.1 A Brief Review of the Energy Dissipation Mechanism
Demagnetized
Crystal Axes
Magnetic Field
if
< cr
(2.1)
Uc = Kloc
if
> cr
(2.2)
loc
20
4 )e
i
(2.3)
2
i
ZVZ
0
0
Kl [ l e
i
2l
i
]dl dV
(2.4)
Z
0
Ki f1 e
2
i
(1 + 2 + 2 )gdV
2
(2.5)
e
i
2s
(1 + 2s + 2s )g
2
=
c =
KE
[1 e
i
2s
(2.6)
(1 + 2s + 2s )]=s
2
(2.7)
(2.8)
(2.9)
where E is the Young's modulus and Uc is the vibratory energy density of the
coating material.
Eq. 2.9 shows the loss factor or the energy dissipation density per cycle of the
coating material is independent of the vibratory frequency.
2.2 Damping Measurement of the Coating Material
The above energy dissipation concept is applied to examine the eects of the
coating damping on forced response of beams and blades. Initially, as shown in
22
Fig. 2.8, simply supported and cantilevered beams with a coating thickness of 10%
of the beam's thickness are considered, where the coating layer is located on the
top surface of the beams. The bending rigidity, EI (1 + i), of the coated beam is
calculated by solving the Ross-Kerwin-Unger equation [28]
EI (1 + i ) = Es Is f(1 + is ) + e2 h32 (1 + ic )
where
e2 =
Ec
;
Es
h2 =
h
d
(2.10)
(2.11)
and h is the thickness of the coating layer, d is the thickness of the beam, and
Es Is (1 + is ) is the bending rigidity of the uncoated portion of the beam.
This RKU analysis was developed and solved using sinusoidal expansions for
the modes of vibration for simple beams and plates. The analysis also assumes
rigid connections between the coating layer and the beam itself. In this research,
the RKU equation has been successfully applied to predict the performance of the
unconstrained-layer coating damping treatment.
To further measure the damping properties from the magnetomechanical coating, a damping test has been accomplished. In accordance with the ASTM standard
E756-93 [12], a Ti-6Al-4V beam and a steel beam were machined having dimensions
7:087in 0:3937in 0:079in. A sinusoidal signal produced by a Siglab 20-22 digital
signal processor and amplied by a 75V power amplier (Bruel and Kjaer Power
Amplier 2706) was used to drive a dynamic shaker (Bruel and Kjaer Vibration Excitor 4809), exciting the clamped edge of the beams. The excitation frequency was
23
the rst four vibration modes of the beams were calculated using the half power
point method. This method measures the bandwidth of resonant frequencies where
p
the amplitudes of response are 1= 2 times the maximum response located at the
resonant frequencies of the system. Then, the loss factor can be written as the
bandwidth of the resonant frequency divided by the resonant frequency. Following
this procedure, the loss factor of the steel beam, as shown in Fig. 2.2, and Ti-6Al-4V
beam were measured as (approximately) 0.003 and 0.002, respectively. The steel
beam result is presented in Fig. 2.4.
25
0.03
0.025
Voltage
0.02
0.015
0.01
0.005
0
220
225
230
235
240
Frequency (Hz)
245
250
255
260
based on the de
ection at the end of the beam as measured by the laser vibrometer.
By adjusting the input velocity of the dynamic shaker, the damping loss factor was
determined over a wide range of strain levels (0 4 10 ) from the frequency
response function of the coated beam using the half power point method at the
rst vibration mode. Then the damping loss factor c, as shown in Fig. 2.9, of the
coating layer was calculated using the Ross-Kerwin-Ungar equation Eq. 2.10.
4
Laser Vibrometer
Accelerometer
Beam
Shaker
Pre Amplifier
Computer
Power Amplifier
SigLab Digital
Signal Processor
27
Figure 2.6: The Two Steel Beams Coated with the Magnetomechanical Coating and
One Magnetomechanical Material Beam Before Machining.
response can be used to evaluate the damping properties, only the fundamental (1st
mode) frequency was considered here for simplicity.
Due to the diculties of hysteresis loop measurement, it is almost impossible to
directly create the hysteresis loop for this magnetomechanical coating material in
order to obtain magnetic parameters K from the hysteresis loop's slope. So, an
alternative way has been used in this research. In the beginning, the dependence
of the damping loss factor on the strain amplitude is obtained from the previous
experiment, shown in Fig. 2.9. Then, the least-squares t between the testing
28
Figure 2.7: The Steel Beam Coated with One Magnetomechanical Coating Layer
data and the loss factor determined the damping properties (K) and the internal
stress to be 0:36, and 2:5 10 psi, respectively. Since the damping properties
are observed to be stationary at high strain range, only a strain amplitude up to
4 10 was considered in the least squares procedure. A least-squares procedure
that leads to better agreement for high strains could have been performed, but then
a discrepancy would occur for small strains. Thus, the least-squares t chosen here
is believed to be a good compromise because it leads to excellent results for strains
up to 4 10 , while large strains are of no practical importance since noticeable
magnetomechanical damping would only occur at lower strain amplitude.
3
29
h
b
2d
l
Cantilevered Beam
h
b
2d
l
Coating Layer
Beam
2d
The energy dissipation density per cycle of the coating can be characterized by
a single function dened in Eq. 2.6. Using this function, the energy dissipation
through the coating layer can be expressed as:
30
0.35
Test Results
Curve Fit
0.3
0.25
0.2
0.15
0.1
0.05
cr/Ec
0
0.5
1.5
2
strain amplitude
2.5
3.5
4
4
x 10
Uc =
Z lZ
0
d+h
Ki [1 e
2s
(1 + 2s + 2s )]dzdx
2
(2.12)
By using the beam theory, the strain of the beam can be rewritten as:
= zw
00
(2.13)
where z represents the distance from the neutral axis of the beam.
So, substituting Eq. 2.13 into Eq. 2.12, Eq. 2.12 can be rewritten as follows:
Uc =
Z lZ
0
d+h
d
Ki [1 e
00
2 zw
i
zw00
(1 + 2
i
31
zw00 2
+ 2( ) )]dzdx
i
(2.14)
(d+h)w
00
i
dw
2
i
(2.15)
In accordance with the proposed free layer coating concept, the thickness of the
layer should be relatively small in comparison with the thickness of the beam or
blade. By using this thin coating layer assumption, the lower surface strain of the
coating layer is equal or similar to the upper beam surface, which can be written as
= (d)w00 . Therefore, the stresses or strains are assumed to be the same through
the coating thickness. Also, this assumption leads that the overall thickness of the
coated blade 2d + h can be replaced by the thickness of the uncoated ones 2d during
the following computation. Hence,
e
(d+h)w
i
00
e
(d)w
00
(2.16)
i
Z
0
KEc i h[1 + (2 +
(h + 2d)w00 )e
i
2dw
i
00
]dx
(2.17)
Using classical variational principles such as the Hu-Washizu principle [34], the
entire energy expression for the coated beam problem has been formulated by taking
into account the energy dissipation in the coating layer. This yields the following
functional:
Zt Zl1
= t f [ 2 EI (1 + is)w00 + K 0Eci h[1 + (2+
32
2
=
l
0
2dw
00
i
(
h 2d) 2(h + 2d)dw00 00
]w e
+
i
2dw
00
i
i 2
(2.19)
Z
0
(2hd + 4d )w00 ]e
2
i
2dw
i
00
00
b(2d + h)w
_ w_ gdx = 0
33
(2
hd + 4d )w00
]
6d)
2
(2.20)
2dw
00
i
jl b(2d + h)ww
_ jl
0
l
0
f[EI (1 + is)w000
(2.21)
2dw
i
i
= EI (1 + is )w w
jl EI (1 + is)w000 w jl
0
00
(2
hd + 4d )w00
0
6d)
]
e
w jl
i
000
000
4
hdw
8
dw
4(
h + 2d)d w00 w000
0
K Ec h[
+ +
]
2
+K 0Ech[(h
2dw
i
00
2dw
i
jl b(2d + h)ww
_ jl + fEI (1 + is)wiv
0
K 0 Ec h[
i
3
+ 162d w000 + 4hd wiv
i
i
i
8d
i
wiv ]e
00
2dw
i
8hd
i
+ b(2d + h)wgwdx = 0
0002
(2.22)
Therefore, for arbitrary independent variations, the equation of motion is obtained for the coated beams:
EI (1 + is )wiv + K 0 Ec h[
8(h + 2d)d
3i
00 0002
ww
2i
! 2 b(2d + h)w = 0
h)
wiv ]e
2dw
00
i
(2.23)
EI (1 + is )w w
34
jl = 0
0
(2.24)
000
EI (1 + is )w w jl0 = 0
2(
h + 4d)dw00
6d)
]e
i
h) 000 4(h + 2d)d w00 w000
w +
]e
K 0 Ec h[(h
4d(2d
K 0 Ec h[
2dw
i
00
i
i
(2.25)
w
2dw
i
00
jl = 0
0
b(2d + h)ww
_ jl0= 0
jl = 0
0
(2.26)
(2.27)
(2.28)
The above equation of motion shows that damping eect of the magnetomechanical coating layer is dependent on three issues: the magnetic strength of the
coating material (K 0 and i ), the thickness of the coating layer (h), and the thickness of the beam (2d). Therefore, the beam coated with the thicker coating layer
did have a larger eect than the thinner coating layer on suppressing the vibration.
Additionally, K 0 and i are related to the properties of the ferromagnetic material. In this research, the magnetomechanical coating has a unique large number
of K 0, which made the hysteresis loop wider than the regular ferromagnetic material. This creates the unusual high damping properties for this magnetomechanical
coating material. Finally, the coating layer damping properties are dependent on
the vibratory strain or stress in the beam. This is related to the thickness of the
beam, 2d, contained in those energy dissipation terms. In spite of its complicated
formula, the equation of motion obviously represents the unique movement of the
magnetomechanical coating material.
Clearly, if there is no coating layer, the coating parameters K 0 are zero. This
leads to the second, third, fourth , and fth terms of Eq. 2.23 being zero. Then,
Eq. 2.23 can be rewritten as follows:
EI (1 + is )wiv
! 2 b(2d + h)w = 0
35
(2.29)
Therefore, the complex equation of motion in Eq. 2.23 reduces to the simple
uniform Bernoulli-Euler beam equation.
36
CHAPTER 3
APPLICATION TO BEAMS AND BLADES
The eects of coating, 10% of the beam thickness, on the forced response of
simply supported and cantilevered beams have been studied. Results have been obtained for the lowest three vibration modes by a closed form procedure [11] using the
collocation method [2], described in Appendix A, and the nite element approach.
In the nite element approach, there are two alternative ways which have been
chosen to model the coated layer. At rst, four-node isoparametric elements have
been used to simulate the simply supported or cantilevered beam with the coating layer. By using Eq. 2.10, each individual element has its own loss factor and
material properties. During this nite element calculation, the simply supported
or cantilevered beam is treated as a composite beam containing twenty dierent
materials. Each material has its own individual loss factor. Another approach used
in this study is to create another customized four-node isoparametric element to
represent the coating layer, which is connected to the eight-node solid elements of
the regular steel beam, shown in Fig. 3.1. Unlike regular four-node shell elements,
four-node isoparametric elements in this research need to have a customized stress
or strain dependent loss factor and their neutral axis is located on the bottom of the
element, which is connected to the solid elements, in order to correct the moment of
37
inertia error during the numerical computation. Due to the similarity of these two
nite element results, only the results from the rst approach have been presented
in the beams' cases.
The nite element approach has also been applied to examine the forced response of coated blades as shown in Fig. 3.6. Approximately ten thousand four-node
isoparametric and eight-node solid elements have been used to model the coating
layer and the blade. The number of the elements is chosen to capture the eect
of the coating on low frequency bending and torsion modes as well as on high frequency stripe modes in the frequency range of 3K-20K Hz. To cover the
uctuating
vibratory stress pattern, the coating is applied to the entire pressure side of the
blade and modeled with two-dimensional four-node isoparametric elements, which
is similar to the coated beam case. It was demonstrated numerically that the nite
element mesh shown in Fig. 3.6 gives a nearly optimal result for the present problem. Results for both low frequency bending and torsion modes and high frequency
stripe modes have been obtained and evaluated. Additionally, later in this section,
the eect of high rotating speed also will be considered in the coated cantilevered
blades for high stripe mode vibrations.
3.1 Eect of Coating on Forced Response of Simply Supported Beams
Magnetomechanical
Coating
4-Node
Isoparametric Elements
Steel Beam or
Ti-6Al-4V Blade
8-Node
Solid Elements
Figure 3.1: The Finite Element Model for the Blades Containing the Coating Layer
Since the harmonic loading is applied at the mid-span of the beam, no even numbered mode vibrations appeared in the results for simply supported coated beams.
As shown in Fig. 3.2, the rst and the third forced vibration mode responses
of a simply supported beam with a coating layer were computed and compared
to the responses from the uncoated beam. The results show that the mode shape
of the simply supported uncoated beam is a sine wave function, as expected, but,
surprisingly, the mode shape of the coated beam is also close to a sine wave function.
This can be explained by the eigenvalue from the equation of motion, Eq. 2.23,
which did not suer a tremendous change by adding four more energy dissipation
39
terms for the magnetomechanical coating , because K and i are still relatively
small compared to the Young's modulus, E, and the moment of inertia, I, from the
potential energy of the coated beams. Therefore, the only major dierence between
the mode shapes of the uncoated and coated beams are their amplitudes.
From these simply supported beam results, the maximum displacement reduction in the third mode is signicant, as much as 62% from the nite element approximation and 55% from the closed-form solution. However, the rst mode shape
reduction is relatively low (29:8% in closed-form solution and 22:8% by nite element approximation), as shown in Table 3.1. This result can be explained by noting
that a high strain level (as high as 0:0033) is produced in a major portion of the
beam during the rst mode vibration. Consequently, lower damping (0:004 0:068
in loss factor) was produced from the coating layer in the rst mode.
Vibration Mode 1st Mode
Uncoated
0.272
Coated (FEM)
0.191
Reduction in % 29.8%
Coated (CF)
0.210
Reduction in % 22.8%
3rd Mode
0.0029
0.0013
55%
0.0011
62%
Table 3.1: The Maximum De
ection (in) of Simply Supported Beams. (FEM: Finite
Element Results ; CF: Closed-Form Results)
The mode shapes obtained by the closed form approach and nite element approximation are compared in Fig. 3.2. One can observe that the agreement between
closed form solution and the nite element result is reasonable. Most importantly,
40
both results indicate that the 10% coating layer is capable of suppressing vibratory
energy and consequently reduces the stress. In addition, these results also demonstrate that the strain patterns of the rst and third modes trigger the irreversible
movement of magnetic domain boundaries of the coating material.
As expected, the stress distributions, as shown in Fig. 3.3, are also sine wave
functions, but with ninety degree phase angle dierences from the previous deformation functions. This is consistent with simple beam theory. In turn, the coating
layer is able to reduce the maximum stresses by 32% for the rst mode vibration.
It follows that the eect of the coating layer is heavily dependent on the stress or
strain in the simply supported beam.
On the other hand, since the majority portion of the beam under the third mode
vibration is experiencing strain in the range of 0 1:3 10 , a high damping (as
high as 0:2872 in loss factor) is therefore achieved in the third mode. Similar to
the displacement, the coating layer has been able to reduce the maximum stress in
third mode vibration by 63% from the nite element approximation, as shown in
Table 3.2 and Fig. 3.3.
4
3rd Mode
4.3
1.6
63%
Table 3.2: The Maximum Stress (ksi) of Simply Supported Beams. (FEM: Finite
Element Results)
41
The eects of the coating layer on the rst three modes of vibration of cantilevered beams were studied. Similar to the simply supported beam case, a relatively small eect of the coating layer on the rst mode response is observed. Again,
this is caused by the large strain during the rst bending vibration mode. With
the exception of a small portion of the beam near the free end, the majority of the
cantilevered beam has a strain range larger than 4 10 . This high strain range
directly re
ects the relatively low damping performance from the magnetomechanical coating layer when strains are above 4 10 . In the Table 3.3 and Fig. 3.4,
the maximum deformation from the rst mode vibration is only reduced by 22:8%
from closed-form solution and by 29:8% from nite element approximation.
Then, as expected, a much larger eect is achieved in the second and third
vibration modes, with 90:3% and 84% reduction in the maximum displacement by
nite element approximation and 87% and 76% by closed-form solution, respectively,
as shown in the Table 3.3.
The maximum strain of the second and third modes are 3:50 10 and 1:67
10 , respectively, which produce far greater loss factors of 0:3095 and 0:2501. This
indicates that the coating layer is able to suppress the vibratory energy in these
strain ranges for both simply supported and cantilevered beams. Also, the second mode results showed incredibly close agreement between analytical and nite
element results, with nite element results only 5% lower than analytical results,
as shown in Fig. 3.4. In addition, the displacement of the coated beam is highly
damped at the second mode vibration.
42
4
2nd Mode
0.062
0.008
87%
0.006
91%
3rd Mode
0.0031
0.0007
76%
0.0005
84%
Table 3.3: The Maximum Deformation (in) of Cantilevered Beams. (FEM: Finite
Element Results ; CF: Closed-Form Results)
Vibration Mode 1st Mode
Uncoated
66
Coated (FEM)
54
Reduction in %
18%
2nd Mode
20.4
4.7
77%
3rd Mode
4.8
2.0
58%
Table 3.4: The Maximum Stress (ksi) of Cantilevered Beams. (FEM: Finite Element
Results)
Finally, as shown in Table 3.4 and Fig. 3.5, the vibratory stress of the cantilevered coated beams appear to have a similar response to their deformation results. The addition of the coating layer has only reduced by 18% the maximum
stress in the rst mode, because its strain is much higher than the best coating
layer damping performance range. However, it is a totally dierent story for the
second and third modes. As mentioned in the previous paragraph, the strain in
the second mode and third mode is located within the range 0 4 10 . So,
4
43
as expected, the magnetomechanical coating has a large eect on the stress reductions, which are 77% and 58% in the second and third mode resonant vibration,
respectively. The results can be compared in Table 3.4 clearly.
44
0.35
uncoatedFEM
coatedFEM
coatedCF
Displacement (in)
0.3
0.25
0.2
0.15
0.1
0.05
0
0.5
1.5
2.5
3.5
x 10
Displacement (in)
2
1
0
uncoatedFEM
coatedFEM
coatedCF
1
2
3
0.5
1.5
2
Length (in)
2.5
3.5
Figure 3.2: First and Third Mode Shapes of Uncoated and Coated Simply Supported
Beams
45
1.5
x 10
uncoatedFEM
coatedFEM
Stress (psi)
1
0.5
0
0.5
1
1.5
0.5
1.5
Stress (psi)
5000
2.5
3.5
2.5
3.5
uncoatedFEM
coatedFEM
5000
0.5
1.5
2
Length (in)
Figure 3.3: Stress Distributions in the First and Third Modes of Uncoated and
Coated Simply Supported Beams
46
Displacement (in)
1.5
uncoatedFEM
coatedFEM
coatedCF
1
0.5
0
Displacement (in)
0.1
0.5
1.5
2.5
3.5
1.5
2.5
3.5
1.5
2
Length (in)
2.5
3.5
uncoatedFEM
coatedFEM
coatedCF
0.05
0
0.05
0.1
Displacement (in)
0
3
x 10
0.5
uncoatedFEM
coatedFEM
coatedCF
2
0
2
4
0.5
Figure 3.4: First Three Mode Shapes of Uncoated and Coated Cantilevered Beams
47
Stress (psi)
10
coatedFEM
uncoatedFEM
5
0
5
2
Stress (psi)
x 10
0
4
x 10
0.5
1.5
2.5
3.5
0
2
4
coatedFEM
uncoatedFEM
0
0.5
1.5
2.5
3.5
Stress (psi)
5000
coatedFEM
uncoatedFEM
0
5000
0.5
1.5
2
Length (in)
2.5
3.5
Figure 3.5: Stress Distributions in the First Three Modes of Uncoated and Coated
Cantilevered Beams
48
A case study was also conducted to illustrate a general procedure for enhancing
damping capability of gas turbine blades using magnetomechanical coating. The
gas turbine blade, shown in Fig. 3.6, is made of Ti-6Al-4V, having dimensions: 3.0
inches wide 7.0 inches long and 0:4 inches average thickness. The aspect ratio
of the blade (1.75) was higher than some modern blade designs, but resulted in
frequencies and mode shapes, especially in high frequency stripe modes, similar to
a modern blade. Also, the thickness of the coating is 0:016 inches, which is about
4% of the blade's average thickness. The coating layer is applied on the pressure
side of the blade. A NASTRAN routine used in this simulation has been presented
in Appendix B.
Even though adding the coating layer on the pressure side of blade can reduce
displacement signicantly, reducing the high stress for high cycle fatigue prevention
is a more important issue. Therefore, only stress distributions of the blade are
presented in this section. The Goodman Diagram of Ti-6Al-4V in Fig. 3.7 was
obtained from the study by Maxwell and Nicholas [24]. These data are used to
form the failure criteria or failure boundary in the traditional high cycle fatigue
turbine blading system design procedure. This design process usually consists of
a structural dynamics analysis to determine natural frequencies and mode shapes
at certain operating speeds and a stress analysis to calculate the dynamic stress
distribution for identifying the maximum vibratory stress location or region under
a series of given excitations.
49
Once the maximum stresses for each vibration mode have been determined, high
cycle fatigue assessment can be achieved by measuring the margin between the
maximum vibratory stress and the material fatigue capability which is a straight
line drawn between the mean ultimate strength at zero vibratory stress and mean
fatigue strength at 10 cycles (or innite life), as shown in Fig. 3.7. Therefore,
the rst case study was conducted to illustrate how to attenuate the vibration
stresses in the rst three vibration modes of the cantilevered blade through the
magnetomechanical coating, which was presented in the previous section.
Consider the case of the blade under a concentrated external harmonic excitation
force applied at the tip portion of the blade. The magnitudes of the excitation
forces may be adjusted to achieve a stress ratio equal to 1 and a strain range of
0 4 10 . The rst three modes forced responses were computed for the coating
strain-dependent loss factors obtained from the Eq. 2.10. The von Mises stress
distributions of the rst three mode vibrations excited by 200lb force are shown in
Figs. 3.83.13.
One should note that the maximum vibratory stress for mode one, as expected
for this bending mode, occurs near the trailing and clamped edges of the blade and
shows signicant reduction in comparison with uncoated blades. The maximum von
Mises stress dropped from originally 428ksi to 52:3ksi, which is more than 87%, at
the rst mode frequency, shown in Fig. 3.8. Actually, the von Mises stress of 428ksi
is already over the yielding point of the Ti-6Al-4V. Therefore, it appears that the
original turbine blade model without the coating layer has actually failed at the
rst resonant frequency 304.8Hz. According to this gure, the trailing edge of this
turbine blade will be the rst one to break apart, similar to Fig. 1.2. However, the
7
50
maximum von Mises stress of the coated blade is still in the safe range, and far
below the yielding point. This demonstrates that the magnetomechanical coating
layer has the ability to increase the design boundary of turbine blades, or it can
improve an original poor design such that it is acceptable. Fig. 3.8 and Fig. 3.9 also
display that the maximum von Mises stress actually shifts from the trailing edge
to the leading edge. This leads that the maximum stress also can be moved from
certain areas of the turbine blade by adding this coating layer. According to these
observations, the possibility of the magnetomechanical coating improving turbine
blade safety and reliability has been further enhanced.
The further investigation of the drop in von Mises stress at the second mode has
been shown in Fig. 3.10 3.11. By comparing these two gures, the maximum von
Mises stress drop a relatively small 36% from 16:9ksi to 10:8ksi. Once again, the
4% thickness coating layer appears to successfully suppress vibration. In the second
mode vibration, the stress distribution is also slightly changed by adding the coating layer. After adding the coating layer, the maximum von Mises stress is located
on not only the trailing edge, but also at the blade base. This development causes
the stresses to concentrate on two dierent regions of the blade, both with a relatively smaller amplitude, instead of one region with a much larger stress amplitude.
Finally, the third mode results are displayed in Figs. 3.123.13. The additional
coating layer makes the maximum von Mises stress decrease 28%, which drop from
70:5ksi to 50:6ksi. Similarly to the previous second mode, the reduction is much
smaller than the rst mode, because the strain in the rst mode is distributed most
in a particular strain range (1 4 10 ), leading to high damping contribution
4
51
from the coating layer. Also, observing these rst three mode vibrations, the resonant frequencies appear to increase after the magnetomechanical coating has been
applied to the blade surface. This can be explained as the magnetomechanical coating material has higher Young's modulus than Ti-6Al-4V. So, the coated blade has
a higher stiness than the uncoated blade.
Several interesting observations can be made from the above discussions. First,
it is believed that no matter what the mode shape is, the coating layer can still
eciently reduce stresses of the blades when strains are near the range of 1 4
10 . This is because the energy dissipation from the magnetomechanical coating is
dependent on strain, but independent of the mode shape at the resonant frequencies.
Second, the stress distribution can be changed by adding this coating layer. Even
though the change is small, it is sometimes enough to avoid stress concentrations
in critical areas, such as the trailing and the leading edge of turbine blades.
4
52
53
54
Figure 3.8: The von Mises Stress in the First Mode Vibration of the Uncoated
Cantilevered Blade
55
Figure 3.9: The von Mises Stress in the First Mode Vibration of the Coated Cantilevered Blade
56
Figure 3.10: The von Mises Stress in the Second Mode Vibration of the Uncoated
Cantilevered Blade
57
Figure 3.11: The von Mises Stress in the Second Mode Vibration of the Coated
Cantilevered Blade
58
Figure 3.12: The von Mises Stress in the Third Mode Vibration of the Uncoated
Cantilevered Blade
59
Figure 3.13: The von Mises Stress in the Third Mode Vibration of the Coated
Cantilevered Blade
60
In addition to the low frequency rst three mode vibrations, another case study
was also conducted to illustrate a general procedure for enhancing damping capability of gas turbine blades using a magnetomechanical coating for high frequency
stripe mode vibration. The goal in this case study is to investigate the feasibility of
using the magnetomechanical coating for enhancing damping at high frequencies, as
shown in Figs. 3.143.20. As mentioned in the introduction, this plate-like stripe
mode has low or sometimes no movement near the base of blade. This leads to
traditional friction dampers having no ability to dissipate the energy. Also, aerodamping has very little eect on this high frequency vibration, due to their low
deformations. These realities cause current technologies to have diculty achieving
the desired stress reduction. Unfortunately, the high frequency stripe modes are the
typical causes of the actual hardware high cycle fatigue. It is necessary to develop
an alternative method, such as magnetomechanical coating, to avoid the high cycle
fatigue problem, which has been addressed in Chapter 1.
The gas turbine blade is made of Ti-6Al-4V, having the same dimensions as the
previous low frequency simulation model of Fig. 3.6. The aspect ratio of the blade
is still 1:75. Also, as indicated in the previous section, this blade model is similar
to modern blade behavior in the high frequency stripe vibration modes. Again, the
pressure side of the blade has been coated with the magnetomechanical coating,
which is about 4% of the blade's average thickness.
As mentioned in the previous section, even though adding the coating layer on
the pressure side of blade can reduce displacement signicantly, eliminating the
61
high stress for high cycle fatigue prevention is the major concern in high cycle
fatigue problems, which is the purpose of this research. Therefore, only the stress
distributions of the second, third, and fourth stripe modes of coated and uncoated
beams are presented and examined.
Therefore, a case study was conducted to illustrate how to attenuate the vibration stresses in the three dierent stripe high frequency vibration modes using the
magnetomechanical coating damping.
Consider the case of the blade under a similar concentrated external harmonic
excitation force applied at the tip portion of the blade, as used in the previous
section. Once again, the magnitudes of the excitation force may be adjusted to
achieve a stress ratio equal to 1 and a strain range of 0 4 10 . The von
Mises stress distributions of coated blades for the second, third, and fourth stripe
modes excited by a 50lb force on the central free end of the blades are computed
and compared to those of the uncoated blades in Figs. 3.153.20. Observe that
the nite element results conrm the previous experimental predictions that high
vibratory stresses will be produced at high stripe modes.
One should note that in Fig. 3.16 the maximum vibratory stress, as expected
in this second stripe mode, occurs near the clamped edges of the blade and shows
signicant reduction in comparison with the uncoated blade in Fig. 3.15. The drop
in the maximum von Mises stress is from 94:4ksi to 77:8ksi, which is about 17:6%.
Comparing Fig. 3.15 to Fig. 3.16, the stress concentrations are on two regions near
the clamped edge, where the color is red and yellow. Among these regions, the
uncoated blade appears have a larger size than the coated blade. Actually, the
stress concentrated areas of the blade coated with the magnetomechanical coating
4
62
are almost only displayed as a smaller yellow area. This is because the magnetomechanical coating successfully reduced the von Mises stress of this second stripe mode,
and makes the stress concentrated regions become much smaller. However, unlike
the stress amplitude, the original stress distribution was not changed by adding this
magnetomechanical coating.
Similar to the second stripe mode, the magnetomechanical coating was able to
suppress the fourth stripe mode vibration by 9:4%. In this fourth stripe mode,
the maximum von Mises stress has been reduced from 22:4ksi to 20:3ksi, shown
in Figs. 3.193.20. These two gures show a typical example of the high stripe
mode problem. In these two gures, almost no von Mises stress appears near the
clamped edge of blade. However, the maximum von Mises stress can be found near
the tip of blade. This indicates the stress of this stripe mode at 18956Hz resonant
frequency cannot be reduced by the traditional friction dampers. These two gures
also clearly display the larger von Mises stresses only located on the six ellipses
shape regions. With the exception of these six areas, the von Mises stresses are
either small or zero. This results in the diculties of energy dissipation through
the traditional approach.
Compared to the 9:4% 17:6% reduction at the second and fourth stripe mode,
Figs. 3.173.18 present a far greater drop of 42:9% occurring in the third stripe
mode. The maximum von Mises stress of the coated blade was 12:4ksi, which
dropped from 21:7ksi for the uncoated blade. Similar to the beam case, the differences in the stress reduction can be explained by noting that an adequate (in
the sense of inducing coating damping) strain level or pattern is produced in the
majority of the blade during the third stripe mode vibration. Consequently, higher
63
damping is therefore produced from the coating layer. By observing the Figs. 3.17
3.18, the stress level in those stress concentrated regions near the tip of blade is obviously suppressed after the blade has been coated with the magnetomechanical
coating, where the color is changed from red-yellow (20ksi) to yellow-green (10ksi).
In accordance with the stress patterns presented in Figs. 3.153.20, the maximum stress of the second, third, and fourth stripe vibration modes have been
reduced between 9:4% and 42:9% by using only a 4% blade's thickness magnetomechanical coating. This leads to the fact that even though the maximum stress of
the blade can be as high as 94; 400 psi, the majority portion of the blade as shown
in Fig. 3.15 responds to a much lower vibratory stress level ranging from 400-6000
psi, which in fact is an ideal stress level for triggering the irreversible movement of
magnetic domain boundaries of the coating material. Therefore, even though the
highest strain at one particular resonant mode vibration sometimes is higher than
4 10 , which is over the primary target range for this magnetomechanical coating
material, the coating material is still able to perform well in some portions of the
blade, where the strain level still appears on the range of 1 4 10 . This means
the maximum strain or stress of the coated blade is still lower than the original
blade by adding this magnetomechanical coating.
Several interesting observations can be made from the above discussions. First,
the maximum vibratory stress of the coated blade can be signicantly reduced at
a specic mode and reasonably suppressed at the other modes through the magnetomechanical coating under a certain loading condition. Second, the strain range
or loading condition becomes a major factor on the magnetomechanical coating
damping capacity for a specic vibration mode. Third, after simulating the low
4
64
65
Figure 3.14: The Third Stripe Mode Shape of the Cantilevered Blade
66
Figure 3.15: The von Mises Stress in the Second Stripe Mode Vibration of the
Cantilevered Blade
67
Figure 3.16: The von Mises Stress in the Second Stripe Mode Vibration of the
Coated Cantilevered Blade
68
Figure 3.17: The von Mises Stress in the Third Stripe Mode Vibration of the Cantilevered Blade
69
Figure 3.18: The von Mises Stress in the Third Stripe Mode Vibration of the Coated
Cantilevered Blade
70
Figure 3.19: The von Mises Stress in the Fourth Stripe Mode Vibration of the
Cantilevered Blade
71
Figure 3.20: The von Mises Stress in the Fourth Stripe Mode Vibration of the
Coated Cantilevered Blade
72
A case study was performed to investigate the feasibility of using two dierent
thicknesses of magnetomechanical coating on the rotating cantilevered blade. Both
2% and 4% of the blade thickness were chosen as the magnetomechanical coating
thickness in this study. Since turbine blades typically rotate at high speed, the
eects of the coating damping capability is also investigated in this study. It is
well known that rotating the blade creates another new source of force, the centrifugal force, as presented in Appendix C. Such force produces a non-negligible
amount of stress or strain along the length of the blade, and directly in
uences the
damping capacity of the magnetomechanical coating. Figs. 3.223.29 are presented
the fourth stripe mode results of the simulated turbine blade for which traditional
dampers have some diculties to achieve the desired stress reduction.
In this section, the stiening eect of centrifugal loading has been considered
during the forced response simulation, as shown in Appendix D, in order to demonstrate the ability of magnetomechanical coating to enhance damping of the turbine
blade during its high rotating speed operations. Due to the complexity of the real
world twisted blade conguration, two dierent blade sitting angles (0o and 90o),
as shown in Fig. 3.21 by rotation Rx and Ry , were investigated.
In Figs. 3.243.26, the turbine blade with 0o sitting angle, modeled by rotation
Rx in Fig. 3.21, under 10; 000 rpm has been investigated. Similar to the non-rotating
cases, a maximum 50lb concentrated external harmonic excitation force has been
applied at the tip portion of the blade in the rotating blade cases. However, unlike
the previous cases, the magnitude of the overall stresses or strains are expected to be
73
much larger due to the extra centrifugal force and Coriolis acceleration. Comparing
Fig. 3.14 to Fig. 3.22 clearly shows that the third stripe mode deformation and the
resonant frequency are increased by considering the rotating motion, even though
the mode shapes are similar.
One fourth stripe mode vibration is presented in this section, shown in Fig. 3.23.
The maximum von Mises stress at this fourth stripe mode, as expected, occurs near
the tip edges of the blade and shows signicant reduction in comparison with uncoated blades, shown in Fig. 3.24, in either 2% thickness coating case, shown in
Fig. 3.25, or 4% thickness coating case, shown in Fig. 3.26. The drop of the maximum von Mises stress is 6:5% by adding 2% thickness of the coating layer, which
drop from 26:2ksi in the uncoated blade to 24:5ksi in the coated blade. When the
coating layer thickness becomes 4% of the overall blade thickness, the maximum von
Mises stress decreases from 26:2ksi to 22:5ksi. The reduction of the maximum von
Mises stress increases to 14:1%. However, the drop is still slightly higher than the
9:4% in the non-rotating case. Like the previous non-rotating case, the dierences
in stress reduction can be explained by noting that an adequate (in the sense of
inducing coating damping) strain level or pattern is produced in the majority of
the blade during the fourth stripe mode vibration. Another important observation
needs to be noted from these three gures. The thickness of the magnetomechanical
coatings did aect their performance. By cutting the coating thickness to half, the
coating layer only makes the maximum von Mises stress drop 6:5% from 14:1%.
On the other hand, reducing by 6:5% the overall maximum stress is still considered
acceptable. The purpose of simulating the blades with these two dierent coating
thicknesses is to demonstrate that the design engineers would be able to optimize
74
their design by adding the proper thickness of coating to meet their reliability and
safety requirements. More precisely, by knowing the limitation of the maximum
stress during the blade design process, the engineers can design the coating layer
as thin as possible, such that it is enough to suppress the stress to an acceptable
level, without adversely increasing the overall blade weight and reducing their performance and fuel eciency.
Consequently, there are a couple of observations that can be made of these results. First, the thicker coating layer produced the higher damping, as expected.
Second, the addition of centrifugal force in
uenced the performance of the coating layer. However, the coating layer is still able to suppress the vibration in a
reasonable manner.
As mentioned in the previous section, the turbine blade in the real world usually
has a twisted shape and is rotated with a particular angle to the blade surface in
order to have the best aerodynamic performance. However, as previously noted the
damping performance of the magnetomechanical coating is sensitive to the stress
in the blade. Therefore, an additional study about the magnetomechanical coating damping capacity with 90o blade sitting angle was investigated. As shown in
Figs.3.273.29, for the fourth stripe mode vibration the von Mises stress has been
reduced 5:1% by the 2% thickness of the coating layer or 13:5% by the 4% thickness of the coating layer, because the average strain is not located in the highest
damping range of the magnetomechanical coating. These results show the magnetomechanical coating has the ability to enhance the turbine blade damping under
dierent sitting angles and amplitudes of the centrifugal force, even though the
stress distributions and the resonant frequencies are changed.
75
Blade
Rotor
Shaft
Z
Ry
Rx
X
76
Figure 3.22: The Third Stripe Mode Shape of the Cantilevered Blade about Rx
under 10000 rpm Rotation
77
Figure 3.23: The Fourth Stripe Mode Shape of the Cantilevered Blade under 10000
rpm Rotation
78
Figure 3.24: The von Mises Stress of the Fourth Stripe Mode of the Cantilevered
Blade Rotating about Rx under 10000 rpm
79
Figure 3.25: The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Rx under 10000 rpm with 2% Thickness Coating
Layer
80
Figure 3.26: The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Rx under 10000 rpm with 4% Thickness Coating
Layer
81
Figure 3.27: The von Mises Stress of the Fourth Stripe Mode of the Cantilevered
Blade Rotating about Ry under 10000 rpm
82
Figure 3.28: The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Ry under 10000 rpm with 2% Thickness Coating
Layer
83
Figure 3.29: The von Mises Stress of the Fourth Stripe Mode of the Coated Cantilevered Blade Rotating about Ry under 10000 rpm with 4% Thickness Coating
Layer
84
CHAPTER 4
DEVELOPMENT OF A FINITE ELEMENT
PROCEDURE FOR ADHESIVE BONDED JOINTS
The main objective of this chapter is to develop a numerical tool capable of predicting stress and strain distribution in an adhesive bonded joint. Traditionally, the
closed form analytical methods, as mentioned in the introduction section, successfully present the stress or strain distribution for a small number of simple bonded
structures. However, the classical methods have diculties to further explore the
more complicated adhesive bonded joints. In turn, numerical methods, especially
the nite element procedure, have been developed to simulate a wide range of dierent adhesive bonded joints by using a set of given parameters, such as the adherend
thickness, adhesive thickness, joint width, and overlap length. In this chapter, a
two-dimensional nite element procedure has been developed for the adherends and
adhesive layer of the single-lap and double-strap joints. The results from this newly
developed nite element procedure are then compared to the earlier, classical, stress
analyses of adhesive joints, such as Goland and Reissner, Hart-Smith, and Oplinger.
Based on this nite element procedure, a modied procedure has been developed to
predict the stress distributions of the more complicated imperfectly-bonded joint in
the next chapter.
85
The stress distribution along the adhesive bonded layer of a single-lap joint,
one of the most commonly used bonding congurations, has been computed by
using a two-dimensional nite element procedure. The geometrical and material
parameters of a single-lap joint have been chosen for this model development, as
shown in Fig. 4.1, where lb is the length of the outer adherend, 2Cb is the length
of the overlap, t is the thickness of the adhesive layer, and b is the thickness of
the adherend. The elastic modulus and Possion's ratio of adherend are Ea and a .
Respectively, Eb and b represent the elastic modulus and Poisson's ratio for the
adhesive layer. In this study, the widely used FM73 adhesive has been chosen for
the bonding layer. In the nite element analysis, ten elements were used across
the thickness of each adherend and four across the thickness of the adhesive layer.
Three hundred elements were used along the length of the bonded overlap and forty
elements along the length of the unbonded overlap. Due to the stress singularities
near the ends of the overlap, the number of elements was tripled in these areas in
order to have more accurate solutions, as shown in Fig. 4.1. The aspect ratio of
elements in the adhesive layer was 3. Since this conguration has four elements
across the thickness, the stress variations across the thickness of the adhesive layer
can be demonstrated. In addition, Quad-8 elements have been successfully used to
simulate the high stress and strain gradients near the end of the overlaps.
Due to the stress symmetry along the mid-span of the adhesive layer in the singlelap joint, only one half of the mid-span peel and shear stress distributions from the
two-dimensional nite element procedure are presented in this section. The results
86
show the peel and shear stress distributions under 1000lbs of uniform tensile loading
or a single concentrated tensile load, compared to three analytical results, Goland
and Reissner (dash line), Hart-Smith (dash-dotted line), and Oplinger (dotted line),
shown in Figs. 4.24.4.
In Fig. 4.2, a linear geometrical nite element procedure is compared to the
three theoretical solutions of the single-lap joint. This linear geometrical procedure
only considers the small de
ection of bonded joints. It is assumed that the bonded
structure still has the same shape during the loading process. In this rst comparison, all three theoretical solutions, Goland and Reissner, Hart-Smith, and the
Oplinger approach, show the stresses slowly decrease from the center of the joint
overlap to near the end, but sharply increase near the free edges of the overlap. The
comparison between the three theoretical solutions and the nite element analysis
show the Goland and Reissner result is the closest to the nite element result, because Goland and Reissner consider the geometrical linearity like this nite element
procedure. The other two results appear to have both lower peel and shear stresses
than the nite element results. However, the nite element procedure also show
both peel and shear stresses suddenly decreasing to near zero while approaching
the free edge instead, which satises the boundary conditions on the free edges. It
is believed that the classical theoretical results are not able to represent this stress
singularity because of their one-dimensional approach. Excluding the results near
these edges, generally, the four analyses show the stress distributions are symmetric,
and have lower stresses near the middle of the adhesive layer.
Considering nonlinear material and geometrical properties, shown in Fig. 4.3,
the stress distributions from the two-dimensional nite element procedure is slightly
87
lower than the previous linear solution. With the exception of the stress singularities
near the free edge of the overlap, this improvement made the peel stress results from
the nite element procedure agree well with the Hart-Smith and Oplinger results,
because of their geometrical nonlinear assumptions. However, the shear stress from
the nite element results shown in Fig. 4.3 is slightly higher than desired. This
requires further investigation into the possible cause of this shear stress dierence
between the theoretical solutions and the nite element results.
In order to have more accurate comparisons between the results, a single concentrated load has been applied to the single-lap joint to replace the original uniform
load conguration used in the nite element procedure, because of the single load
conguration used in the one-dimensional theoretical solutions. The nite element
result is expected to have much closer shear stress results in comparison with the
three theoretical results after a 1000lbs single load is applied to the middle level of
one end of the adherend. The results, shown in Fig. 4.4, show that the dierent
loading conguration did cause the maximum peel and shear stress distributions to
further drop about 5%. Both peel and shear stresses in this single load case agree
well with the theoretical results and nite element analysis, with the exception of
the stress singularities near the free edge of the overlap. This proves the accuracy
and eciency of this newly created nite element procedure.
Following the above successful comparison, the same nite element procedure
is used as a base tool for the imperfectly-bonded adhesive bonding research in the
next chapter. However, there is one more issue which needs to be addressed. Even
though these results better match the theoretical solutions, the two-dimensional
88
nite element procedure for the rest of this research will still use the uniform loading
conguration to avoid the stress concentration near the end of the joint lap.
Now, Fig. 4.5 demonstrates how powerful this two-dimensional nite element
procedure is. Because it contains four two-dimensional Quad-8 elements through
the adhesive bonding thickness, the peel and shear stress distributions can not
only show the results along the length of the overlap, but also show the peel and
shear stress variations along the adhesive bond thickness. This new information lets
engineers understand the peel and shear stress change inside the adhesive bonded
layer. In Fig. 4.5, the top (y = t ) surface of the adhesive bonded layer, presented as
a solid line, appears to have its maximum peel and shear stresses located near the
left end of the overlap, and another concentration of peel and shear stresses near
the right end. Obviously, those stresses near the left end are larger than those near
the right end. It is believed to be caused by the external uniform loads from the
upper adherend pulling the adhesive bonds from the left side. At the same time,
the lower adherend has been constrained by the right edge of its lap. This actually
leads to the maximum shear and peel stresses being located on either the upper left
or lower right corner of the adhesive bonded layer. According to this observation,
the bottom surface (y = t ) of the adhesive bonded layer has its maximum peel
and shear stress, shown as a dot line, near the right end of the overlap, and there is
another high stress region near the left end, but smaller than the one at the right
end. Comparing the peel stress to the shear stress, the shear stress distribution
reaches its local maximum value just slightly inside the left edge, and drops nearly
to zero at the left free edge of top surface, caused by the stress singularity, but the
peel stress has its local maximum value at the left edge. Therefore, the peel stress
2
89
y
lb
Cb
Cb
lb
x
Pb
b
b/2
2Cb
lb
x
t
Figure 4.1: The Geometry and Finite Element Mesh of the Bonded Joints
distributions at the top surface are no longer symmetric, where the free edge side has
nearly no stress and the side connected to the joint has the maximum peel stress.
The bottom surface stress distributions are asymmetric to the top surface stress
distributions. Instead of the asymmetry between these top and bottom surface
results, the middle level of the shear and peel stress distributions appear to be a
symmetrical curve between the right and left end of the overlap.
90
0.8
GR
OP
HS
FEM
0.4
xy
/P
0.6
0.2
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
0.8
GR
OP
HS
FEM
0.6
y/P
0.4
0.2
0
0.2
0.05
0.1
Cb
Figure 4.2: Adhesive Stress Distributions Obtained from the Present Finite Element Model with Linear Geometry (Uniform Loads)(GR: Goland and Reissner;
OP: Oplinger; HS: Hart-Smith; FEM: Finite Element Method )(y=0)
91
0.8
GR
OP
HS
FEM
0.4
xy
/P
0.6
0.2
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
0.8
GR
OP
HS
FEM
0.6
y/P
0.4
0.2
0
0.2
0.05
0.1
Cb
Figure 4.3: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with Nonlinear Geometry (Uniform Loads)(y=0)(GR: Goland and Reissner;
OP: Oplinger; HS: Hart-Smith; FEM: Finite Element Method)
92
0.8
GR
OP
HS
FEM
0.4
xy
/P
0.6
0.2
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
0.8
GR
OP
HS
FEM
0.6
y/P
0.4
0.2
0
0.2
0.05
0.1
Cb
Figure 4.4: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with Nonlinear Geometry (Single Loads)(y=0)(GR: Goland and Reissner;
OP: Oplinger; HS: Hart-Smith; FEM: Finite Element Method )
93
0.7
y=t/2
y=0
y=t/2
0.6
0.4
xy
/P
0.5
0.3
0.2
0.1
0
0.25
0.2
0.15
0.1
0.05
0
Cb
0.05
0.1
0.15
0.2
0.25
0.05
0.1
0.15
0.2
0.25
y=t/2
y=0
y=t/2
0.8
y/P
0.6
0.4
0.2
0
0.2
0.25
0.2
0.15
0.1
0.05
0
Cb
Figure 4.5: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with Nonlinear Geometry (Uniform Loads)(y=0)
94
of the overlap. However, the shear stress near both ends of the double-strap joint
only slightly increase initially, then drop to zero at the free edge of the adhesive
bonding layer, instead of the stress singularities in the single-lap joint case. This
is because the loads are distributed through two overlaps in the double-strap joint.
Unlike the shear stress, the maximum peel stresses are located near the left edge
of this overlap, regardless of their interface locations. Then, the peel stress almost
constantly decreases to zero near the right end of the overlap. Additionally, at
the right end of the overlap, the peel stress becomes compressive at the bottom
interface, but it is still tensional stress at the top interface. This is believed to be
caused by the large de
ection occurring in the double-strap patch that connects the
two adherends.
In order to further evaluate the accuracy of this double-strap joint result, the
stress distributions at the middle level of the adhesive layer from the nite element results have been compared to those from the theoretical results developed by
Oplinger [27], shown in Fig. 4.7. Oplinger's results have been computed by using
the Visual Basic based TJOINTNL program. Fig. 4.7 shows that the nite element
results agree well with Oplinger's theoretical solutions in both peel and shear stress
distribution, with the exception of the peel and shear stresses near both ends of the
overlap. Because of Oplinger's one-dimensional computation, this theoretical result
is not decreasing to zero near the free edge. However, the nite element result does
appear to be zero, which satises the boundary conditions. Excluding the results
near the edges, both methods yield consistent shear and peel stress distributions.
After testing two dierent congurations, a good comparison with theoretical
results has increased condence in the shear and peel stress distributions predicted
96
97
1.5
y=t/2
y=0
y=t/2
xy
/P
0.5
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
1.5
1
y/P
0.5
0
0.5
y=t/2
y=0
y=t/2
1
1.5
0.05
0.1
Cb
Figure 4.6: Adhesive Stress Distributions Obtained from the Present Finite Element
Model for Double-Strap Joint (OP: Oplinger; HS: Hart-Smith; FEM: Finite Element
Method)
98
1.5
OP
FEM
xy
/P
0.5
0.05
0.1
0.15
0.2
0.25
Cb
0.6
OP
FEM
0.4
y/P
0.2
0
0.2
0.4
0.05
0.1
0.15
0.2
0.25
Cb
Figure 4.7: Adhesive Stress Distributions Obtained from the Present Finite Element
Model Compared with TJOINTNL Program (OP: Oplinger; HS: Hart-Smith; FEM:
Finite Element Method)
99
CHAPTER 5
DEVELOPMENT OF IMPERFECTLY-BONDED
INTERFACE MODELS
The computational model developed during the last chapter has proved to be
accurate and reliable. Based on these two-dimensional nite element models, the
newly developed nite element procedure is now extended to consider the environmental degradation of adhesive joints. Among those adhesive bonded joints
damaged by environmental degradation, the imperfectly-bonded joint is one of the
most common and serious defects. Furthermore, it is also the most dicult structural problem to be detected by current non-destructive inspection techniques. The
main objective of this chapter is to predict the peel and shear stress distributions
of the imperfectly-bonded joint. To represent the change in the mechanics of the
imperfectly-bonded interface, a distributed spring element model has been created
to simulate this imperfectly-bonded interface. Later, improving this distributed
spring model, a newly developed one-dimensional line element also will be introduced in this chapter. By using these simulations as a base, a more detailed inspection or design criteria can be created in order to detect the failure of the adhesive
bonded interface due to environmental degradation.
100
)=
m! 2
(5.1)
2 [Uj (0 ) + Uj (0 )]
where j = x; y; z directions, Uj represents the j th component of a particle's displacement from its equilibrium position, j is the j th component of the force per
unit area exerted on the bonded area, m is the mass of the adhesive bonded layer,
! the natural frequency, and the interface is located at y = 0.
By using this relationship from Eq. 5.1, Baik and Thompson [3] developed a
distributed spring model to simulate this extra deformation due to the imperfectlybonded interface, where the spring stiness per unit area, K , is a function of the
topography of the partially contacting surface. The idea is that the spring stiness is
101
innity when two objects are connected by a rigid connection, such as the perfectlybonded interface, but it becomes zero when there is no connection, like the totally
disbonded interface. So, Eq. 5.1 can be rewritten as:
1 [ (0 ) (0 )] = K [U (0 ) + U (0 )]
(5.2)
j
j
j
j
2 j
where Kj is the stiness of the interface.
With the exception of when the adherend and adhesive are directly connected,
there is a virtual spring between the adherend and adhesive to simulate the imperfectlybonded interface. This spring model was proved successful in predicting the frequency dependence of re
ection and transmission in ultrasonic measurements of
partially contacting interfaces in adhesive bonded joints [3]. Furthermore, Margetan, Thompson, and Gray [23] extended this spring model to diusion bonds,
which can predict the dependence of bond re
ectivity upon both angle of incidence
and ultrasonic polarization. Recently, Lavrentyev and Rokhlin [21] described the
degradation as an interfacial disbond lled by an absorbed
uid, which is also modeled by the transverse spring boundary conditions, with the complex spring stiness
representing the quality of the bond. In their research, the disbond due to the degradation can be characterized by slip boundary conditions, where the normal spring
stiness, dened as Kn0 , is innity and only the shear spring stiness, dened as
Kt0 , need be determined in the spring model [23].
Therefore, the next step is to determine the spring stiness Kt . At rst, the
imperfectly-bonded interface between adherend and adhesive might be assumed
to be a complex pattern of cracks and contacts. It is dicult to nd the exact
stress distribution in the imperfectly-bonded region. An alternative approach is to
102
+
(5.3)
2EaEb
Eb (1 b ) + Ea (1
(5.4)
ab
where is the shear stress, is the Poisson's ratio, is the deformation, a is the
crack length, and bl is the interface length.
The eective Young's modulus Eab in Eq. 5.3 can be written as:
Eab =
a2 )
where Ea and Eb represented the Young's modulus from two side of the crack.
The equilibrium stiness of an imperfectly-bonded interface can be computed
by the potential energy of the imperfectly-bonded layer. According to Tada, Paris,
and Irwin's formula [32], the stiness Kt0 can be written as:
Kt0 =
a
Eab 1
ln
(
sec
4 (1 2) b
2b )
(5.5)
disbonding is low, the cracks are nearly noninteracting. Therefore, Kt can be calculated by assuming the strength per unit length of a cylindrical unit cell containing
a single centered circular crack multiplied by the overlap length and divided by
number of the spring elements as follows:
2 C E 1
b
ab
1=2
Kt =
2 4N 1 2 a (1:2997Ad
0
:1491Ad=
1:8687ln(1 + Ad= ) 0:4199ln(1
0:0295Ad +
=
1 + Ad
1 2
+0:1407Ad
5=2
1 2
1 2
A1d=2 ))
(5.6)
where Ad = a ab is the disbonded area and N is the number of the spring elements.
The model of the imperfectly-bonded interface has been developed as an array
of very thin interfacial disbonds lled by water, where the shear direction stiness
of the spring has been determined by the area ratio between disbonded and bonded
area as Ad, so-called the disbonded area, in Eq. 5.6. By using this relationship,
a new two-dimensional nite element procedure has been developed to simulate
the imperfectly-bonded interface by adding the spring elements as the boundary
conditions between adherend and adhesive bonded layers. The shear stiness of this
customized spring element is Kt , and the stiness of the spring element for the other
ve degrees of freedom are assumed to be innity, and therefore are assumed to be
constrained perfectly in this nite element procedure. In this procedure, one eightnoded isoparametric shell element from the adherend layer has been connected with
three spring elements to another eight-noded isoparametric shell element from the
adhesive bonded layer, as shown in Fig. 5.1. There are, in total, 601 spring elements
connecting these two interfaces to simulate the imperfectly-bonded interface. The
equilibrium stiness of each spring elements is equal to the Kt in Eq. 5.6.
104
(
)2
4 2
After the spring model simulated the imperfectly-bonded interface in the previous section, the spring model appears to have some disadvantages. First, the spring
element has a particular eect which stores more potential energy in the heavy
loading regions than the relatively light loading regions. This causes the displacements to be similar along the adhesive bonding side of the spring elements, just
like a car's suspension system. Also, each spring element acts individually according to the external shear or peel force between the two nodes it is connected to,
which doesn't happen in the real imperfectly-bonded interface. In reality, the occurrence of one small disbonded area will lead to the stress changes in the surrounding
bonded region. Second, the spring elements only connects the nodes between the
adherends and adhesive layer, which causes discontinuities when the environmental
degradation only appears in some portions of the interface. For example, consider if
the imperfectly-bonded interface only happened on one end of the spring element,
while the interface was still perfectly-bonded on the other end. This will cause
the distributed spring element model to show a discontinuous stress distribution,
which does not represent the reality of the imperfectly-bonded interface. In order to
modify the limitations of the spring element and evaluate more complex degradation processes between the adherend and adhesive in an accurate manner, a newly
developed one-dimensional line element model has been proposed in this section to
improve the simulation of the imperfectly-bonded interface.
105
Similar to the circular cracks case [23], the imperfectly-bonded interface leads
to additional deformations between the adherend and adhesive for axial, rotation,
and shear deformation.
From Eq. 5.2, the stiness Kj represents the j th component strength of the
imperfectly-bonded interface, which becomes weaker as the disbonded area of interface increases. In this study, K is associated with the normal (peel) stresses, similar
to Kt in the spring model, which tend to open the interface, and is innity to ensure
the continuity of displacement in the normal direction for the water lled disbonded
interface, according to Lavrentyev and Rokhlin [21]. K is associated with the the
shear stresses , similar to the Kn in the spring model, which depend on the crack
area Ad . K is not considered in this research due to those two-dimensional element
models.
Unlike the spring element model using the boundary conditions to connect the
imperfectly-bonded interface, the line element model uses a constitutive relation
between the adherend and adhesive surfaces as follows:
3
"
FN
Ft
"
K3
#"
U3
U2
= 0 K
(5.7)
where FN is normal force per unit length, Ft is the shear force per unit length, U
is the normal strain, and U is the shear strain.
From Eq. 5.1 and Eq. 5.7, the virtual work of the line element can be constructed
as:
"
#
Zy
F
N
= [U U ] F dy
(5.8)
y
t
2
where y and y are the coordinates of two ends of the line element.
1
106
By substituting Eq. 5.7 into the force FN and Ft , the Eq. 5.8 can be rewritten
as:
=
x1
[U
U2 ]
"
K3
#"
#
U3 dy
U2
(5.9)
0 K
Using a two-point Gaussian integration, the element stiness matrices of Eq. 5.9
can be formulated.
In order to determine the stiness of the imperfectly-bonded interface, the deformation of the interface needs to be calculated rst. This is done in a manner
similar to the previous spring element.
x2
2 d E 1
b
ab
K2 =
(1
:2997A1d=2
2
2 8 1 a
1=2
1 2
A1d=2 ))
(5.10)
This newly developed line element which has six nodes, shown in Fig. 5.1, where
nodes 1, 2, and 3 are located on the adhesive layer, and nodes 4, 5, and 6 located
on the adherend has been used to simulate the stiness of the interface. So, the
Quad-8 shell element joins the 6-node line element along their common intersection
of nodes 4, 5, and 6. Because the imperfectly-bonded interface is virtually innitely
thin, the coordinates of nodes 4, 5, and 6 actually coincide with nodes 1, 2, and 3,
respectively. By using the appropriate conditions calculated by the above Eq. 5.10
107
and Eq. 5.11, the displacement between the interface connected nodes 1, 2, and 3
and the interface connected nodes 4, 5, and 6 can be represented as:
U3
U2
=
=
uy (y = 0+ ) uy (y = 0
)
x(y = 0 ) x (y = 0 )
+
(5.12)
(5.13)
where uy and x are the normal and shear displacements interpolated from the
nodes 1, 2, and 3 from the lower interface as well as nodes 4, 5, and 6 from the
upper interface.
After using the nite element procedure, the generalized strains and stresses
were recovered at the integration points.
Compared to the spring element, the one-dimensional line element actually
modies the nite element stiness matrices of the adhesive joint to represent the
imperfectly-bonded interface movement, which is dierent than the spring element
procedure which uses the boundary conditions to simulate the imperfectly-bonded
interface. This modication allows the line element to better simulate the interface under more complicated water diusion conditions. For example, when the
interface degradation occurs only in the area near the end of overlap of the adhesive joint, the stress distributions computed by the spring element procedure have
some discontinuties in both peel stress and shear stress. However, by solving the
same problem, the line element procedure can avoid those stress discontinuities by
allowing the stiness of the interface to changes along the interface. Unlike the
spring elements only connecting the nodal points, this new line element procedure
connects the entire Quad-8 elements along the imperfectly-bonded interface from
108
the adherends and adhesive through the 6-noded line elements. This prevents the
unrealistic stress concentrations of the spring element analysis.
5.3 Results and Discussion
The spring element and the new one-dimensional line element developed in the
previous section were applied to examine the peel and shear stress distributions
of a single-lap joint with one imperfectly-bonded interface. According to these
results, the new procedure can be further compared to the nite element results
from the previous perfectly-bonded joint in order to study the change of the stress
distribution during the environmental degradation.
5.3.1 The Distributed Spring Element
Due to computing expense, only the single-lap joints with shorter 0:25in overlap
and 0:01in thick FM73 adhesive bonds, and 0:1in thick aluminum adherends were
investigated in this research. Two joints with 0% imperfectly-bonded, or perfectlybonded, and 20% uniform imperfectly-bonded area, Ad , occurring at the upper
interface of the adhesive bonded layer have been evaluated by using the distributed
spring elements, as shown in Fig. 5.2 and Fig. 5.3, respectively. The imperfectlybonded interface in this section was assumed to contain only one disbonded island
size, 0:01 inch. Each joint was loaded by a 1000lbs static uniform loading at the left
end of the joint, and constrained at the right end of the joint. Since this research
focuses on stress changes during the environmental degradation of the adhesive
bonded layer, the peel and shear stress distributions in the upper (y = t), middle
(y = 0), and lower surfaces (y = t) of the adhesive layer are presented.
1
2
1
2
109
The change in the stress distributions after the degradation can be found by
comparing Fig. 5.2 and Fig. 5.3. The peel and shear stresses increase about 23%
near the middle of the overlap area after the environmental degradation occurred as
shown in Fig. 5.3, but decrease 15% near both ends of the overlap area, where the
stress singularities are located. These changes in stress distributions are believed to
be caused by the weaker strength of the imperfectly-bonded layer no longer resisting
the high stresses near the end of the overlap, and shifting the loads to the middle of
the overlap. This activity is even more obvious in the top interface than the bottom
interface, where the maximum shear stress increases 12% more than the bottom
interface. The bottom interface between adhesive and adherend in this simulation
is assumed to be still perfectly-bonded. In addition to the stresses increasing at the
middle of the overlap, the imperfectly-bonded interface made the peel stress and
shear stress distributions become almost a straight line except near both ends of the
overlap after the environmental degradation. This situation is caused by the spring
element storing more potential energy in high stress regions near the end of the
overlap than the other regions which make the deformations in the adhesive bonded
layer equal along the interface, even though the shear or normal forces applied to
the spring elements in the adherend are dierent. However, although the spring
element is questionable to present accurate results for the stresses near the middle
of the interface, it still provides a relatively reasonable stress reduction near the
ends of the overlap, as expected. Since neither the peel nor the shear stresses in the
middle of overlap are higher than the stresses near both ends of overlap, only the
stress singularities (high stress regions) near the ends of the overlap take a major
role in the creep and the failure of the bonded joint. Actually, this is also the place
110
that kissing bond usually rst occurred. This makes the spring element model a
useful tool for the bonded joint reliability prediction.
5.3.2 The One-Dimensional Line Element
In order to improve the previous distributed spring element model, a 20% uniform disbonded area at the imperfectly-bonded upper interface of the adhesive layer
has been recalculated once again by using the newly developed one-dimensional line
element procedure. Similar to the previous spring element case, the individual disbonded island sizes are assumed to be 0:01 inch. The results, shown in Fig. 5.4,
appear to be pretty close to the previous spring element's results. However, the peel
stresses at the middle of the overlap are obviously lower than the spring element
case. This made the peel stress distribution nonlinear. Actually, it is more like the
curved line from the perfectly-bonded interface's results, as shown in Fig. 5.4, but in
a smaller scale. Similar results happened with the shear stress. The middle surface
shear stress distribution near the middle of overlap computed by the line element
is 6% lower than the spring element. Also, the upper and bottom interfaces have
higher shear stress than the spring element case. The explanation is that the line
element utilizes the whole element stiness matrix directly, which is more accurate
than the spring element. Another issue is that the loads are not shifted to high
stress regions in the line element model, unlike what occurred in the spring element
model. This made the shear and peel stresses near the middle of the interface lower
and more accurate. Clearly, the newly developed line element procedure is believed
to have better ability to simulate the imperfectly-bonded interface not only at the
end of the overlap, but across the whole interface. Further evaluating the stress
111
through the adhesive thickness, Fig. 5.4 shows that the upper interface has larger
eects from the imperfectly-bonded interface than the bottom interface. From this
interesting observation, the in
uence from the upper imperfectly-bonded interface
rst changes the stress distribution of the upper interface, obviously, while the same
time, the maximum peel and shear stresses in the middle interface also begin to drop
but much slowly. However, the stress distribution at the bottom interface which
is away from the imperfectly-bonded interface almost have no eect. In order to
further study this observation, it becomes necessary to investigate the imperfectlybonded joints under a more complicated situation, such as one containing a dierent
disbonded area percentage in the imperfectly-bonded interface.
So, another case of a single-lap joint with a non-uniform imperfectly-bonded
area at the upper interface has been computed by using the one-dimensional line
element procedure, shown in Fig. 5.5. In this more complex case, the upper interface degradation near the end of the overlap is assumed to reach as much as 30%
disbonded area. The disbonded area decreases while moving close to the middle of
the overlap. Finally, the environmental degradation near the middle of the overlap
is assumed not to have occurred, corresponding to 0% disbonded area. Or it is
still perfectly-bonded. The individual island size is assumed to have a 0:01 inch
diameter throughout the whole adhesive bonded layer. This leads to an interface
containing imperfectly-bonded areas as well as perfectly-bonded areas. This interface happens in the real world particularly when liquid lls the disbonded island [21]
during the environmental degradation. The one-dimensional line element simulates
this imperfectly-bonded interface with lower K and K values (corresponding to
30% disbonded area in this case) near the end, then slowly increases to a higher K
2
112
113
only the line element model has the ability to simulate the stress distributions of
the non-uniform imperfectly-bonded interface, unlike the stress discontinuties from
the spring element model.
114
lb
2Cb
lb
t
b
Pb
l
b
= 10 , b = 8
t
cb
Imperfectly-Bonded
Interface
a
Spring Model of Interface
Ad =
( a b) 2
4a 2
2Cb
lb
x
Pb
6
2
b
l
= 10 , b = 8
t
cb
Imperfectly-Bonded
Interface
3
Line Element of Interface
Figure 5.1: The Geometry and Finite Element Model of the Imperfectly-Bonded
Joints
115
1.5
y=t/2
y=0
y=t/2
xy/P
0.5
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
4
y=t/2
y=0
y=t/2
/P
2
1
0
1
0.05
0.1
Cb
Figure 5.2: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with Perfectly-Bonded Interface (Uniform Loads)
116
1.5
y=t/2
y=0
y=t/2
xy/P
0.5
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
3
y=t/2
y=0
y=t/2
/P
1
0
1
2
0.05
0.1
Cb
Figure 5.3: Adhesive Stress Distributions Obtained from the Present Finite Element Model with 20% (Ad =0.2) Imperfectly-Bonded Interface by Using the Spring
Element (Uniform Loads)
117
1.5
y=t/2
y=0
y=t/2
xy/P
0.5
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
4
y=t/2
y=0
y=t/2
/P
2
1
0
1
0.05
0.1
Cb
Figure 5.4: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with 20% (Ad=0.2) Imperfectly-Bonded Interface by Using the Line Element
(Uniform Loads)
118
1.5
y=t/2
y=0
y=t/2
xy/P
0.5
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
Cb
4
y=t/2
y=0
y=t/2
/P
2
1
0
1
0.05
0.1
Cb
Figure 5.5: Adhesive Stress Distributions Obtained from the Present Finite Element
Model with Various Disbonded Areas by Using the Line Element.(30% (Ad =0.3)
near the End of Overlap Decreases to 0% (Ad=0) in the Center) (Uniform Loads)
119
According to the previous analysis, the strength of the imperfectly-bonded interface is highly dependent on the percentage of the disbonded area. However, the
strength of the imperfectly-bonded interface is also aected by another issue, the
disbonded size, so-called the island size. In this section, the in
uence from the disbonded island sizes is considered for the peel and shear stress distributions of the
imperfectly-bonded interface.
By using the relationship from Eq. 5.6, there are two major issues that need to
be considered in predicting the failure of the adhesive bonded joints. First, obviously, increasing the percentage of the disbonded area weakens the adhesive bonded
interface. Second, increasing the size of each individual disbonded island also weakens the strength of the interface as well, as shown in Fig. 5.6. By using these two
conclusions, a newly developed three-dimensional graphic mesh has been created to
represent the maximum peel and shear stresses of the imperfectly-bonded interface
under dierent disbonded island sizes and dierent percentages of disbonded areas.
This new three-dimensional information can be used to determine when the quality
of the bonded interface is no longer safe to resist the loads. Furthermore, it also
can be used as a base for the development of the bonded joint reliability prediction
method and accept/reject inspection criteria.
In Figs. 5.75.9, the maximum peel stresses of the imperfectly-bonded interface
with dierent percentages of the disbonded area and dierent individual disbonded
island sizes have been presented. In the rst gure, Fig. 5.7, it shows the maximum
peel stress at the upper level (y = t ) of the adhesive layer where the imperfect
120
2
High Strength
Small Disbonded Islands
(Ad=0.25)
Low Strength
Large Disbonded Islands
(Ad=0.25)
Figure 5.6: The Strength of the Interface with Various Disbonded Island Sizes
bond occurred. As presented in this gure, the maximum peel stress at this upper
interface of the imperfectly-bonded interface becomes larger during the early stage
of the environmental degradation. This is the result of the weaker strength of the
upper interface due to either the increasing disbonded area or larger individual
disbonded island size. As discussed in the previous section, this weaker interface
creates additional deformation on the overlap of the single-lap joint which directly
result in the higher strain or stress near the end of the interface. This action leads
121
to the maximum peel stresses of the upper surface to begin to slowly increase when
the interface begins to weaken, shown in the upper corner of the Fig. 5.7. Following
this early stage, the maximum peel stresses begin to drop while the disbonded area
or the disbonded island size keeps increasing. This can be named as the second
stage, shown in the lower corner of the Fig. 5.7. In this stage, the upper interface
becomes much weaker and the de
ection of the interface is so large that the original
mechanics of the bonded joint has been changed, and the overlap of the single-lap
joint no longer can be treated under the assumptions of linear geometry. Now,
because of the large de
ection eect involved in this peel stress distribution, it
leads the peel stress of the interface to rearrange from both ends of the overlap,
where the high stress concentration regions were located, to other portions of the
interface. Simplifying this activity, it can be said that the peel stresses or strains
begin to shift away from the ends of the overlap. This caused the decreasing of the
maximum peel stress, but the increasing of the minimum peel stress at the same
time. Observing Fig. 5.7, it clearly reconrms this assumption that the maximum
peel stress increase slightly about 7% in the beginning, when disbonded area is
small, then drop tremendously after the disbonded area passes 10% with 1 inch
diameter disbonded island or 78% with 0:1 inches diameter disbonded island.
However, it also leads to another issue in this research that disbonded island
size plays a major role in the strength of the imperfectly-bonded interface. In
Fig. 5.7, the maximum peel stress changes little under 50% disbonded area with
the average 0:01 inch diameter disbonded island size, but the maximum peel stress
becomes only one half of its original value under the same disbonded area with the
average 1 inch diameter disbonded size. It appears that the interface under the
122
same disbonded area containing the larger size of the disbonded islands and fewer
numbers of disbonded islands has much weaker strength, compared to the small
size of the disbonded island or large numbers of the disbonded islands. Therefore,
future inspection criteria for bonded joints needs to consider not only the amount
of disbonded area, but also the size of each individual disbonded island.
In order to further investigate the in
uence of the disbonded island size, the
maximum peel stresses located at the middle and bottom levels of the adhesive
bonded layer containing an upper imperfectly-bonded layer in the single-lap joint
are also investigated in this section. The Fig. 5.8 shows the maximum peel stresses
at the middle level of the interface (y = 0). In this gure, similar to the upper level,
the maximum peel stresses initially increase when the interface becomes weaker,
then later drop when the interface keeps growing weaker. Unlike the upper level
peel stresses, the maximum peel stresses do not begin to drop until the disbonded
area reaches 50% with 0:1 inches diameter disbonded islands or 18% with 1 inch
diameter disbonded islands. This can be explained since the maximum peel stresses
of the middle level did not reach as high as those at the upper interface during the
early stage of the imperfectly-bonded interface degradation process. As mentioned
in the previous section, the maximum peel stress of the adhesive is either in the
upper right or the lower left corner of the adhesive bond of the single-lap joint. So,
in this case, the peel stress at the upper right corner is the rst one to reach a
point which the deformation is so large that the peel stress has been redistributed
to other areas within the interface, which includes the middle level of the interface.
Therefore, the peel stresses in the middle level is still increasing at this stage. The
middle level even needs to resist the load coming from the weakened upper interface.
123
Later in the higher percentage of disbonded area or larger disbonded size, 21% with
1 inch diameter island or 87% with 0:1 inches diameter island, the deformation in
the middle level also reaches the point that the right end of the middle level adhesive
layer is no longer able to resist the loading, and begin to transfer to the other regions
of the middle level and lower interface. It also appears that the imperfectly-bonded
interface has more in
uence in the stress distribution at the upper layer than the
middle layer which is away from the imperfectly-bonded interface.
In order to further conrm this theory, the lower level of the adhesive bonded
layer has been presented in Fig. 5.9. Similar to the middle level case, the maximum
peel stresses at the lower level of interface showed the maximum peel stresses do not
drop until reaching 40% disbonded area with the 1 inch diameter disbonded island
or 90% with the 0:1 inches diameter disbonded island. Therefore, the maximum
peel stresses of the lower level of the overlap do not have much in
uence from the
weaker upper interface, compared to the previous upper and middle level cases. This
is because the bottom interface is even further away from the imperfectly-bonded
interface. Even with the 1 inch diameter disbonded island size, the maximum peel
stresses at the bottom interface increase only 3:8% at the 46% disbonded area,
compared to the stresses at the perfectly-bonded interface, then slowly decrease
2:2% before reaching 80% disbonded area.
In addition to the maximum peel stresses, the maximum shear stresses in the
interface also have been presented in Figs. 5.105.12. Similar to the previous peel
stress cases, the maximum shear stresses become higher while the disbonded area
or disbonded island size is increasing. Then, they begin to drop after reaching an
ultimate point like the previous peel stress case. Once again, the disbonded island
124
size still heavily in
uences on the results of the maximum shear stress. Fig. 5.10
displays the maximum shear stresses at the upper level of the interface. This gure
clearly shows that the maximum shear stress initially increases then slightly drops
when the disbonded island size is small (0:01 in). But unlike the peel stress case, the
maximum shear stresses almost immediately decreases shortly after the disbonded
interface occurs when the disbonded island size is larger than 0:05in. This is caused
by the imperfectly-bonded interface having an even weaker bond to resist the shear
stress, since the shear stiness is smaller than the normal stiness. This can be
seen in Eq. 5.11 where Kt is lower than Kn. Also, the tension applied at the left
end of the lap causes much higher shear stress than peel stress in the adhesive
bonded layer. Therefore, the extra shear deformation already changes the shear
stress distribution, even in the small disbonded area or disbonded island size cases.
In this gure, the maximum shear stresses drop 36% before the disbonded area
reaches 55% with 0:1 inch diameter disbonded island or 25% with 1 inch diameter
disbonded island size.
The maximum shear stresses at the middle level of the interface have not changed
too much during the upper interface degradation process, shown in Fig. 5.11. The
change of the maximum shear stresses at the middle level of the interface is similar
to the peel stress results, but on a smaller scale. The shear stress increases 2%
when the disbonded area reaches 35% with 1 inch diameter disbonded island or
80% with 0:1 inches diameter disbonded island. Then, it decreases back to almost
its original amount when the imperfectly-bonded interface with 80% disbonded
area increases its island size to 1 inch diameter. This indicates that the in
uence
from the imperfectly-bonded interface is not so obvious for those areas which are
125
away from the imperfectly-bonded interface within the adhesive bonded layer. The
next gure, Fig. 5.12, presents the maximum shear stresses at the lower level of the
interface. The maximum shear stresses of this level increase like the previous results
when the disbonded area or disbonded island size is increasing, but do not decrease
while the disbonded area or island size keeps increasing. Actually, the results show
the maximum shear stresses keep increasing up to 16% at small disbonded island
size (0:01in) or up to 11% at large disbonded island size (1:0in) before reaching
80% disbonded area. The reason is that the lower level of the interface has little
additional deformation caused by the imperfectly-bonded interface, as discussed
previously. Because it is away from the imperfectly-bonded interface, the maximum
shear stresses at the lower level are not aected by the environmental degradation
of the upper interface.
According to these maximum stress results, the strength of the imperfectlybonded interface is very sensitive to each individual disbonded island size. Therefore, the imperfectly-bonded interface with smaller disbonded island size can sustain
a higher percentage of the disbonded area than those with larger disbonded islands.
Using this conclusion, a possible design or inspection criteria of imperfectly-bonded
interfaces can be proposed by considering both its disbonded area and each individual disbonded island size. For this criteria, the author assumes that when the additional deformation in a particular portion of structure caused by the imperfectlybonded interface begins to change the original stress distribution of the structure,
it means the object is not longer reliable and safe. In this research, this is considered to happen when the maximum peel and shear stresses begin to drop. In
other words, when the peel or shear stress of the single-lap joint begins shifting.
126
This point can be located in the three-dimensional graphic meshes when the slope
of the mesh is equal to zero. Using the previous assumptions, it is believed that the
joints' reliability and safety is questionable when all three levels (y = t ; 0; t ) of
either maximum peel or shear stresses begin to decrease or the slope of the mesh is
zero or becomes negative. However, a more detailed denition of this criteria can
be made after the tension tests of the various imperfectly-bonded joints have been
conducted in the future. The ultimate goal of this criteria is to provide the eld
engineers a procedure to acknowledge the reliability of imperfectly-bonded singlelap joints by evaluating the average disbonded island size and the percentage of the
disbonded area. Furthermore, based on the material properties and geometry of
the structures, more detailed and accurate three-dimensional graphical information
can be customized for each individual adhesive material as a design or inspection
criteria of the bonded structures in the future.
2
127
4
3
2
1
0
0
0
0.2
0.4
0.2
0.6
0.4
0.8
0.6
0.8
Disbond %
Figure 5.7: Maximum Peel Stress Distributions at the Upper Level (y = t ) of the
Bonded Interface along the Percentage of Disbonded Area and Disbonded Island
Size (inches)
2
128
1.2
1
0.8
0.6
0.4
0.2
0
0.1
0.2
0.2
0.3
0.4
0.4
0.6
0.5
0.6
0.8
0.7
0.8
Disbond %
Figure 5.8: Maximum Peel Stress Distributions at the Middle Level (y = 0) of the
Bonded Interface along the Percentage of Disbonded Area and Disbonded Island
Size (inches)
129
3.8
3.6
3.4
0
3.2
0
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
Disbond %
Figure 5.9: Maximum Peel Stress Distributions at the Lower Level (y = t ) of the
Bonded Interface along the Percentage of Disbonded Area and Disbonded Island
Size (inches)
2
130
1.4
1.2
Stress Ratio
xy
/P
1
0.8
0.6
0.4
0.2
0
0
0
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
Disbond %
Figure 5.10: Maximum Shear Stress Distributions at the Upper Level (y = t ) of the
Bonded Interface along the Percentage of Disbonded Area and Disbonded Island
Size (inches)
2
131
0.61
0.6
Stress Ratio
xy
/P
0.59
0.58
0.57
0.56
0.55
0.54
0.53
0
0.2
0.4
0.2
0.6
0.4
0.8
0.6
0.8
Disbond %
132
0.95
Stress Ratio
xy
/P
0.9
0.85
0.8
0.75
0
0.7
0
0.2
0.4
0.2
0.6
0.4
0.8
0.6
0.8
Disbond %
133
CHAPTER 6
CONCLUSIONS
A new passive damping concept based on magneto-elastic eects from magnetomechanical coating materials is proposed for vibratory stress and amplitude
reduction on vibrating beams and blades. A model for the vibration of beams
and blades containing a magnetic coating layer is presented. This is based on the
magneto-elastic theory [30, 31] developed to formulate the mechanical energy dissipation from the irreversible movement of magnetic domain boundaries of the coating
material.
The steady state responses of coated beams and blades are computed under a
harmonic excitation force. The eect of the magnetic coating layer has been successfully examined and compared in this forced vibration analysis. The closed-form
solutions and the nite element results show a signicant vibratory stress and amplitude reduction of the vibrating beams and blades with one coating layer. The
eects of the coating layer on the vibratory energy dissipation are found to be sensitive to the level of the strain or stress in the coating layer and independent of
the vibration frequency. Additionally, several case studies have been presented to
demonstrate the ability of this magnetomechanical coating with various thicknesses
to suppress vibration in the high rotating speed environment. The results proved
134
that the maximum von Mises stress can be reduced 8 36% by adding one coating
layer. The present model could be extended to real turbine blades with twisted
non-uniform cross-sections and to also account for the eects of aerodynamic, and
thermal loading. Furthermore, the in
uence of the dierent host materials, such
as nonisotropic materials, on the stress-dependent damping capacity of the magnetomechanical coating materials will need to be investigated in order to improve
the accuracy of the simulation results. Also, the thermal coecient of expansion
and the creep eect of this magnetomechanical coating material need to be further
studied to insure the reliability of the coating material during the high temperature
and rotating speed operations in the turbine engine.
Like the blade vibration, another structural problem, the imperfectly-bonded
adhesive joint, that seriously aects aircraft safety and reliability has been further
investigated. A two-dimensional nite element procedure for double-strap joints
and single-lap joints was developed by using the Quad-8 isoparametric shell element.
Due to the stress singularities near the ends of the overlap, the number of elements
was tripled in order to have more accurate solutions. The comparison between this
two-dimensional nite element procedure and Goland and Reissner's, Hart-Smith's,
and Oplinger's analytical solutions for the single-lap joint and Oplinger's analytical
solution for the double-strap joint are presented and proved the accuracy of this
new nite element procedure.
The further development of a new two-dimensional nite element procedure to
simulate the environmental degradation of the bonded layer has been successful.
Based on Baik and Thompson [3], Margetan, Thompson, and Gray [23], and Laventyev and Rokhlin's [21] previous research, a nite element procedure, using the
135
distributed spring elements, has been created and shows the peel and shear stress
reduction at both ends of the overlap of the single-lap joint after the environmental degradation occurred. The spring model is later improved and replaced by a
newly developed one-dimensional line element. These line elements successfully
improve the simulation of the imperfectly-bonded interface between the adherend
and the adhesive. This element is also able to predict the shear stress as well
as the peel stress due to the environmental degradation in a more accurate manner. Also, the line element was extended to predict the partial imperfectly-bonded
interface. From these results, newly developed three-dimensional graphic design diagrams have been created as a possible design or inspection criteria for the adhesive
bonded joints. The results show the strength of the imperfectly-bonded adhesive
joint is highly dependent on not only the percentage of disbonded area it contains,
but also each individual disbonded island size. In the future, after conducting more
tension tests of the adhesive joints, the information for the stress propagation from
the imperfectly-bonded interface can be determined to create a more reliable design
and inspection criteria. This information can be further used as a foundation to
detect and avoid the premature failure of bonded joints and improve the reliability
and quality of bonded joints. Also, similar procedures can be repeated to create
the design or inspection criteria for the dierent adhesive joint structures.
136
APPENDIX A
Collocation method
nx
nx
nx
)
+
a1 cos(
)
+
a2 sinh(
2l
2l
2l )
+a3cosh( nx
2l )
(A.1)
Also, the equation of motion for the simply-supported beam, Eq. 2.23, can be
written as:
f (x; w(x)) = EI (1 + is )wiv +
0002
! 2 b(2d + h)w
137
+ 4hd wiv ]e
i
2dw
00
i
(A.2)
(A.3)
Or actually using Eq A.1 as the P (x). In order to determine P (x), the expression
is considered as:
e(x) = f (x; p(x)) F (x)
(A.4)
minf
l
0
[e(x)] dxg
2
(A.5)
Due to the nonlinear properties of Eq. 2.23, Eq. A.2 is too complicated to nd
its global minimum position. So, rather than using Eq. A.2, the method of collocation proceeds by asking that p(x) be chosen so that e(ti ) = 0 at certain selected
"collocation points", t , t , t , ... , tn . Instead of minimizing Eq. A.2, the solution
can be computed by solving n + 1 equations as follows:
0
e(t0 )
= 0
e(t ) = 0
1
::::::
138
e(tn )
= 0
(A.6)
139
APPENDIX B
The Abbreviated MSC/NASTRAN Routine for the
Cantilevered Blade with the Magnetomechanical Coating
The following abbreviated routine is written in the popular nite element package MSC/NASTRAN to evaluate the magnetomechanical coating eect on a cantilevered blade.
SOL 103
TIME = 60
CEND
TITLE = Turbine Blade
SUBTITLE = Magnetomechanical Coating
$ GLOBAL CASE
SPC = 1
DISPLACEMENT = ALL
STRESS = ALL
DLOAD = 10
FREQUENCY = 17
METHOD = 88
BEGIN BULK
PARAM AUTOSPC YES
PARAM POST -2
EIGR, 88, SINV, 17000., 18000.
RLOAD1, 10, 22, 32
DAREA, 22, 6, 2, 200.
TABLED1, 32 +
+ 2., 1., 16., 1., ENDT
GRID, 1, 0, 3.00000, -2.E-17, 7.00000,, 0
GRID, 2, 0, 3.03271, -2.5E-2, 7.00000,, 0
GRID, 3, 0, 3.05428, -6.0E-2, 7.00000,, 0
140
141
1,
1,
1,
1,
1,
137,
161,
185,
209,
233,
123456,
123456,
123456,
123456,
123456,
0.00000
0.00000
0.00000
0.00000
0.00000
142
7.43E-02
1.35E-02
5.10E-03
2.70E-03
1.70E-03
1.30E-03
1.00E-03
9.00E-04
8.00E-04
8.00E-04
8.00E-04
8.00E-04
0.00500
143
APPENDIX C
Dynamics of the Rotating System
The rotating turbine blade in this research is modeled under the assumption that
the coordinate system used in the description of the blade will rotate at a constant
speed about a xed axis. Because the rotating blade is accelerating relative to a
previous stationary inertial blade, the inertial or mass impedance cannot be directly
calculated in the rotating blade. The inertial impedance in the rotating blade must
rst be determined in the stationary blade, and then transformed to the rotating
blade. Following the procedures in two advanced dynamics books [5] and [17], the
blade coordinates from the stationary system rst need to be transferred to the
rotating system in order to simulate the rotating blade, which can be written as:
8
9 2
9
38
>
< u(t)r >
= 6 cos
t sin
t 0 7 >
< u(t)s >
=
v
(
t
)
sin
t
cos
t
0
v
(
t
)
=
(C.1)
4
5
r
s
>
>
>
: w(t)r >
;
:
;
0
0 1
w(t)s
where
is the rotating speed, u, v, and w are the coordinates from the threedimensional system, r represents the rotating system, and s represents the stationary system.
By using the exponential function to replace the cos
t and sin
t, Eq. C.1 is
rewritten as:
8
9 0
9
2
3
2
3 2
31 8
>
< u(t)r >
= B 1 i 6 1 i 0 7 1 i 6 1 i 0 7 6 0 0 0 7C >
< u(t)s >
=
v
(
t
)
i
1
0
i
1
0
0
0
0
v
(
t
)
=
e
+
e
+
(C.2)
@
4
5
4
5
4
5
A
s
>
>
>
2 000 2
: w(t)rr >
;
:
;
0 00
001
w(t)s
144
where is
t.
Dening three matrices in Eq. C.2 as the transformation matrix as R , R , and
R , the original vector from the stationary system also can be presented in terms of
the vector from the new rotating system by multiplying the inverse transformation
matrix on both sides of the Eq. C.2, shown as follows:
8
9
8
9
>
>
u
(
t
)
r >
< u(t)s >
= 1 i
<
1 e i [R ] + [R ]) v(t) =
v
(
t
)
=
(
e
[
R
]
+
(C.3)
>
>
2
: w(t)ss >
; 2
: w(t)rr >
;
In order to simplify the problem to show the inertial forces, only the vector
u(t) has been used in the derivative process. First, taking the second derivative of
Eq. C.3 with respect to time we obtain
d [u(t)s ]
1 ei [R ] + 1 e i [R ] + [R ]) d [u(t)r ]
=
(
dt
2
2
dt
+i2
( 12 ei [R ] 12 e i [R ]) d[udt(t)r ]
( 21 ei [R ] + 12 e i [R ])[u(t)r ] (C.4)
1
Re(i[R ])
2
Re([R ])[u(t)r ]
dt
dt
dt
2
(C.6)
acceleration. This new equation represents the inertial system acceleration in terms
of rotating system coordinates. In addition, the inertial force can be written as:
!
d[u(t)r ]
d [u(t)r ]
F (t) = [M ]
4
Re(i[R ])
2
Re([R ])[u(t) ] (C.7)
2
in
dt
dt
where F (t)in is the inertial force, and M is the mass of the turbine blade.
In Eq. C.7, the rst term obviously represents the blade acceleration. However,
the second and third terms are caused by the rotating movement. In the nite
element method, these additional terms in Eq. C.7 are always present on the rotating
blade as an additional stiness-like term, so called the spin stiening eect, written
as follows:
K = dFduin
(C.8)
The general stiness matrix can be rewritten as:
Kr = Ks + K
(C.9)
where Kr is the stiness matrix for the rotating system, and Ks is the stiness
matrix for the stationary system.
The forced vibration problem is terms of the nodal displacements u is expressed
as
[Kr ]u ! [M ]u = F
(C.10)
2
146
APPENDIX D
DMAP Routine for the Centrifugal Force on the Rotating
System
A DMAP code [6] has been customized for calculating the additional centrifugal
forces for rotating cantilevered blades in a nite element package, MSC/NASTRAN.
echooff
COMPILE SESTATIC
alter 'enddo.*hsflag'
TYPE DB SLIST
TYPE PARM,,I,N,lpflg1a=0,NOUP
TYPE PARM,,I,N,SCNDRY=+1,EXTRN=+1
TYPE PARM,,I,Y,DEPEND= 0
TYPE PARM,,RS,N, OMEGA= 0.
TYPE PARM,,CS,N, SCALE
FILE HISTS=APPEND
MATGEN ,/TMPMGG/1/2
MATGEN ,/TMPBGG/1/2
MATGEN ,/TMPKGG/1/2 RESTART TMPMGG,,/HISTS
RESTART TMPBGG,,/HISTS
RESTART TMPKGG,,/HISTS
SEID=-1
DO WHILE ( lpflg1a >= 0 )
IF ( RSONLY ) THEN
lpflg1a=-1
NOUP=-1
SEID=0
ELSE
SEP2DR SLIST,EMAP//S,N,SEID/S,N,PEID//S,N,lpflg1a/////
S,N,NOUP/S,N,SCNDRY/S,N,EXTRN/S,N,NOMR/'ALL'
ENDIF
MESSAGE //' '/
147
148
EQUIVX BJJTEMP/BJJ/ALWAYS
EQUIVX BJJ/BJJTEMP/NEVER
MESSAGE //' '/
MESSAGE //' USER INFORMATION: SPEED DEPENDENT TERMS'/
' ADDED'/
MESSAGE //' TO STRUCTURAL MATRICES'/
MESSAGE //' '/
ELSE IF ( (DEPEND < 0) AND (OMEGA = 0.) ) THEN
MESSAGE //' '/
MESSAGE //' USER WARNING: SPEED DEPENDENCE REQUESTED,'/
MESSAGE //' BUT NO RFORCE SPECIFIED.'/
MESSAGE //' DIFFERENTIAL STIFFNESS OR'/
MESSAGE //' GYROSCOPIC TERMS WERE NOT '/
MESSAGE //' ADDED TO STRUCTURAL MATRICES.'/
MESSAGE //' '/
ELSE
ADD KJJ,KJJGYRO/KJJTEMP/
EQUIVX KJJTEMP/KJJ/ALWAYS
EQUIVX KJJ/KJJTEMP/NEVER
ADD BJJ,BRJJ/BJJTEMP/
EQUIVX BJJTEMP/BJJ/ALWAYS
EQUIVX BJJ/BJJTEMP/NEVER
MESSAGE //' '/
MESSAGE //' USER INFORMATION: CONSTANT SPEED TERMS'/
' ADDED'/
MESSAGE //' TO STRUCTURAL MATRICES'/
MESSAGE //' '/
ENDIF
ENDIF
ENDIF
RESTART ,,,HISTS/TLIST/TRUE
ENDDO
lpflg1a>=0 COMPILE CORICEN
SUBDMAP CORICEN
SLT,BGPDT,EQEXIN,CSTM,SIL,MJJ/KCEN,BCOR/OMEGA
TYPE
PARM,,I,N,SLTRECNO T Y P E P ARM; ; RS; N; OM E GA
P ARM; ; C S; N; N 1C; N 2C; N 3C TYPE PARM,,CS,N,OMEGAC
SLTRECNO= 1 DO WHILE ( SLTRECNO > 0 )
PARAML SLT//'DTI'/S,N,SLTRECNO/1//S,N,LTYPE/
IF ( (SLTRECNO <> -1 ) AND ( LTYPE = 10 ) ) THEN
PARAML SLT//'DTI'/SLTRECNO/4//S,N,CID/
PARAML SLT//'DTI'/SLTRECNO/5/S,N,RPT/
PARAML SLT//'DTI'/SLTRECNO/6/S,N,N1/
PARAML SLT//'DTI'/SLTRECNO/7/S,N,N2/
PARAML SLT//'DTI'/SLTRECNO/8/S,N,N3/
SLTRECNO = -2
ENDIF
SLTRECNO = SLTRECNO + 1
ENDDO
IF ( RPT = 0. ) THEN
149
OMEGA= 0.
RETURN
ENDIF
MATGEN ,/I1X1/1/1/
MATGEN ,/I6X6/1/6/
DIAGONAL MJJ/MJJC/'COLUMN'
IF (N1 <> 0.) THEN
MATGEN ,/PVC6/6/6/1/1/4
MATGEN ,/PVR6/6/6/2/1/3
MERGE ,,,,I1X1,PVC6,PVR6/ONE6X6/1/0/1
MPYAD ONE6X6,I6X6,ONE6X6/ROTX/1/-1
DELETE /PVC6,PVR6,ONE6X6,,/
N1C = CMPLX(N1,0.)
ENDIF
MATGEN ,/PVC6/6/6/2/1/3
MATGEN ,/PVR6/6/6/0/1/5
MERGE ,,,,I1X1,PVC6,PVR6/ONE6X6/1/0/1
MPYAD ONE6X6,I6X6,ONE6X6/ROTY/1/-1
DELETE /PVC6,PVR6,ONE6X6,,/
N2C = CMPLX(N2,0.)
ENDIF
MATGEN ,/PVC6/6/6/0/1/5
MATGEN ,/PVR6/6/6/1/1/4
MERGE ,,,,I1X1,PVC6,PVR6/ONE6X6/1/0/1
MPYAD ONE6X6,I6X6,ONE6X6/ROTZ/1/-1
DELETE /PVC6,PVR6,ONE6X6,,/
N3C = CMPLX(N3,0.)
ENDIF
ADD5 ROTX,ROTY,ROTZ,,/ROTTOT/N1C/N2C/N3C
ADD ROTTOT,/OMEGA/OMEGAC
MATMOD OMEGA,SIL,,,,/OMEGAGG,/5
PARTN OMEGAGG,MJJC,/,,,OMEGAXX/
PARAML CSTM//'PRESENCE'////S,N,NOGTOB=0
IF (NOGTOB < 0) THEN
DIAGONAL MJJ/GTOBY/'SQUARE'/0.
ELSE
MATMOD CSTM,SIL,BGPDT,,,/GTOBY,/5//-1
ENDIF
PARTN GTOBY,MJJC,/,,,GTOB/
IF (CID = 0) THEN
EQUIVX GTOB/CTOG/ALWAYS
ELSE
MATMOD CSTM,SIL,BGPDT,,,/CTOBX,/5//CID
PARTN CTOBX,MJJC,/,,,CTOB/
MPYAD CTOB,GTOB,/CTOG/1
ENDIF
SMPYAD CTOG,OMEGAXX,CTOG,,,/OMEGAYY/3////1
MERGE ,,,,OMEGAYY,MJJC,/OMEGAGGX/
FORCE MATRIX
MPYADMJJ,OMEGAGGX,/CJJ1
MPYAD OMEGAGGX,MJJ,/CJJ2
150
ADD CJJ1,CJJ2/BCOR
CETRIPETAL FORCE MATRIX
SMPYAD OMEGAGGX,MJJ,OMEGAGGX,,,/KCEN/3////////6
RETURN
END
echoon
151
BIBLIOGRAPHY
[1] D. L. Anton. \Simultaneous Fretting and Fatigue of = Ti-6Al-4V". Proceedings (CD) of the 4th National Turbine Engine High Cycle Fatigue Conference,
1999.
[2] U. M. Ascher, R. M. M. Mattheij, and R. D. Russell. Numerical Solution of
Boundary Value Problems for Ordinary Dierential Equations. Prentice-Hall
Inc., 1988.
[3] J. M. Baik and R. B. Thompson. \Ultrasonic Scattering from Imperfect Interfaces: A Quasi-Static Model". Journal of Nondestructive Evaluation, 4:177{
195, 1984.
[4] B. M. Barthelemy, M. P. Kamat, and H. F. Brinson. Finite Element Analysis
of Bonded Joints. Oce of Naval Research No. N0014-82-K-0185, 1984.
[5] F. P. Beer and E. R. Johnston, Jr. Vector Mechanics for Engineers - Dynamics.
McGraw-Hill International Group, 1988.
[6] D. Bella, M. Reymond, and Editors. DMAP Module Dictionary. The MacnealSchwendler Corp., 1994.
[7] R. M. Bozorth. Ferromagnetism. D. Van Nostrand Co., 1951.
[8] D. Chen and S. Cheng. \An Analysis of Adhesive-Bonded Single-Lap Joints".
Journal of Applied Mechanics, 50:109{115, 1983.
[9] A. W. Cochardt. \The Origin of Damping in High-Strength Ferromagnetic
Alloys". Journal of Applied Mechanics, 20:196{200, 1953.
[10] K. R. Cross, R. L. Newman W. R. Lull, and J. R. Cavanagh. \Potential of
Graded Coatings in Vibration Damping". An Engineering Note, Journal of
Aircraft, 10:685{687, 1973.
[11] E. J. Doedel, A. R. Champneys, T. F. Fairgrieve, Y. A. Kuznetsov, B. Sandstede, and X. Wang. Continuation and Bifurcation Software for Ordinary Differential Equations. Canada, 1998. (AUTO97).
152
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
1993.
M. Goland and E. Reissner. \The Stresses in Cemented Joints". Journal of
Applied Mechanics, 11:A17{A27, 1944.
R. W. Gordon and J. J. Hollkamp. \Internal Damping Treatment for Gas
Turbine Blades". AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, VI:442{451, 1997.
J. H. Grin. \Friction Damping of Resonant Stresses in Gas Turbine Engine
Airfoils". International Journal of Turbo and Jet Engine, 7:297{307, 1990.
L. J. Hart-Smith. \Adhesive-Bonded Single-Lap Joints". NASA Report, CR112236, 1973.
D. N. Herting. MSC/NASTRAN Advanced Dynamic Analysis User's Guide.
The Macneal-Schwendler Corp., 1997.
Jay Homan. \Magnetic Damping System to Limit Blade Tip Vibrations in
Turbomachines". U. S. Patent, 5,490,759, 1996.
R. Kielb and etal. \Advanced Damping Systems for Fan and Compressor
Blisks". Proceedings (CD) of the 4th National Turbine Engine High Cycle
Fatigue Conference, 1999.
T. P. Lang and P. K. Mallick. \The Eect of Recessing on the Stresses in
Adhesively Bonded Single-Lap Joints". International Journal of Adhesion and
Adhesives, 19(1):257{271, 1999.
A. I. Lavrentyev and S. I. Rokhlin. \Models for Ultrasonic Characterization
of Environmental Degradation of Interfaces in Adhesive Joints". Journal of
Applied Physics, 76(8):4643{4650, 1994.
Kaushik A. Lyer and Shanker Mall. \Fretting Fatigue of Ti-6Al-4V Under
Variable-Amplitude Loading". Proceedings (CD) of the 4th National Turbine
Engine High Cycle Fatigue Conference, 1999.
F. J. Margetan, R. B. Thompson, and T. A. Gray. \Interfacial Spring Model for
Ultrasonic Interactions with Imperfect Interfaces: Theory of Oblique Incidence
and Application to Diusion-Bonded Butt Joints". Journal of Nondestructive
Evaluation, 7:131{152, 1988.
153
[24] D.C. Maxwell and T. Nicholas. \A Rapid Method for Generation of a Haigh
Diagram for High Cycle Fatigue". Fatigue and Fracture Mechanics, 29:ASTM
STP{1321, 1998.
[25] David M. North. Aerospace Daily. McGraw-Hill Co., 1999.
[26] D. W. Oplinger. \Eects of Adherend De
ections in Single-Lap Joints". International Journal of Solid Structures, 31(18):2565{2587, 1991.
[27] D. W. Oplinger. TJOINTNL: For Stresses in Bonded Joints with Tapered
Adherends and Nonlinear Adhesive Response, 1995.
[28] D. Ross, E. E. Ungar, and E. M. Kerwin, Jr. \Damping of Plate Flexural
Vibrations by Means of Viscoelastic Laminate". Structural Damping, ASME,
New York, pages 49{88, 1959.
[29] M. H. Herman Shen. \Reliability Assessment of High Cycle Fatigue Design of
Gas Turbine Blades Using the Probabilistic Goodman Diagram". International
Journal of Fatigue, 21:699{708, 1999.
[30] G. W. Smith and J. R. Birchak. \Eect of Internal Stress Distribution on
Magnetomechanical Damping". Journal of Applied Physics, 39(5):2311{2316,
1968.
[31] G. W. Smith and J. R. Birchak. \Internal Stress Distribution Theory of Magnetomechanical Hysteresis-An Extension to Include Eects of Magnetic Field
and Applied Stress". Journal of Applied Physics, 40(13):5174{5178, 1969.
[32] H. Tada, P. Paris, and G. Irwin. The Stress Analysis of Cracks Handbook. Del
Research Corporation, St. Louis, MO., 1973.
[33] L. Tong, A. Sheppard, and D. Kelly. \Relationship Between Surface Displacement and Adhesive Peel Stress in Bonded Double Lap Joints". International
Journal of Adhesion and Adhesives, 5(1):43{48, 1995.
[34] K. Washizu. Variational Methods in Elasticity and Plasticity. Pergamon Press,
1982.
154