Damage Tolerance and Durability of Fiber Metal Laminates For Aircraft Strucrtures-FAA-2010

DOT/FAA/AR-10/18
Damage Tolerance and Durability

Air Traffic Organization
NextGen & Operations Planning of Fiber-Metal Laminates for
Office of Research and
Technology Development
Washington, DC 20591
Aircraft Structures
October 2010
Final Report
This document is available to the U.S. public

through the National Technical Information
Services (NTIS), Springfield, Virginia 22161.
This document is also available from the Federal Aviation

Administration William J. Hughes Technical Center at
actlibrary.tc.faa.gov.
U.S. Department of Transportation

Federal Aviation Administration
NOTICE
This document is disseminated under the sponsorship of the U.S.

Department of Transportation in the interest of information exchange. The
United States Government assumes no liability for the contents or use
thereof. The United States Government does not endorse products or
manufacturers. Trade or manufacturer's names appear herein solely
because they are considered essential to the objective of this report. This
document does not constitute FAA certification policy. Consult your local
FAA aircraft certification office as to its use.
This report is available at the Federal Aviation Administration William J.

Hughes Technical Center’s Full-Text Technical Reports page:
actlibrary.tc.faa.gov in Adobe Acrobat portable document format (PDF).
Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
DOT/FAA/AR-10/18
4. Title and Subtitle 5. Report Date
DAMAGE TOLERANCE AND DURABILITY OF FIBER-METAL LAMINATES October 2010
FOR AIRCRAFT STRUCTURES
6. Performing Organization Code
7. Author(s) 8. Performing Organization Report No.

Jenn-Ming Yang, Thomas H. Hahn, Hyoungseock Seo, Po-Yu Chang, and
Po-Ching Yeh
9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)
Mechanical and Aerospace Engineering Department
Materials Science and Engineering Department
University of California at Los Angeles
Los Angeles, CA 90024-1597 11. Contract or Grant No.
04-C-AM-UCLA
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
U.S. Department of Transportation Final Report
Federal Aviation Administration
Air Traffic Organization NextGen & Operations Planning
Office of Research and Technology Development
Washington, DC 20591 14. Sponsoring Agency Code
AIR-100
15. Supplementary Notes
The Federal Aviation Administration Airport and Aircraft Safety R&D Division Technical Monitor was Curt Davies.
16. Abstract
Fiber-metal laminates are a family of hybrid metal and composite laminates for aircraft structural applications. These hybrid
laminates have the advantage of metal and fiber-reinforced composites, providing superior mechanical properties to the
conventional lamina merely consisting of fiber-reinforced lamina or monolithic aluminum alloys. As a replacement of high-
strength aluminum alloys, S2-glass fiber-reinforced aluminum laminates (GLARE) have been applied to aircraft structures for
weight reduction and improved fatigue resistance. The research and development activities have covered damage tolerance and
durability of GLARE subjected to multiple impacts and multiple-site fatigue damage.
The experiment and finite element simulation for low-velocity multiple impacts was performed on GLARE laminates in this
report. Drop-weight impact was applied twice at a variety of energy levels to inflict multiple-impact damages for aluminum
2024-T3 and GLARE 5-2/1. The numerical model using the finite element method was developed by using the commercial finite
element code ABAQUS and verified by comparing to the experimental data. The two- and three-dimensional failure criteria in
composite layer were used to predict all dynamic responses, including load-time history, maximum deflection-time history, and
damage progression. In addition, the user-defined material subroutine VUMAT was developed and linked with ABAQUS to
represent a three-dimensional progressive failure. The finite element simulation showed good agreement with experimental data
for multiple-impact behavior.
The multiple-site fatigue damage behaviors of fiber-metal laminates were investigated experimentally and analytically under
constant-amplitude fatigue loading. It was found that the presence of multiple-site fatigue cracks would accelerate the crack
growth rates in the metal layers of fiber-metal laminates as two propagating cracks approach each other. An analytical
methodology was proposed to calculate the fiber-bridging stress based on the crack-opening relations. The multiple-site crack
growth rates were predicted by an empirical Paris-type fatigue law. The effective stress-intensity factor (SIF) at a crack tip was
formulated as a function of applied SIF, crack-opening SIF, and nondimensional SIF. Weight function approach was used for
SIF calculation. The predicted crack growth was validated by the experiments.
17. Key Words 18. Distribution Statement
Fiber-metal laminates, GLARE, Multiple-site damage, This document is available to the U.S. public through the
Multiple impacts, Fatigue crack growth National Technical Information Service (NTIS), Springfield,
Virginia 22161. This document is also available from the
Federal Aviation Administration William J. Hughes Technical
Center at actlibrary.tc.faa.gov.
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 83
Form DOT F 1700.7 (8-72) Reproduction of completed page authorized
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY xiii
1. INTRODUCTION 1
2. TEST PROGRAM AND TEST MATRIX 4
2.1 Multiple-Impact Load Test 4
2.1.1 Materials 4
2.1.2 Specimen Geometry, Coordinates, and Impact Energy 5
2.1.3 Impact and Postimpact Test Article Instrumentation and Procedure 5
2.2 Multiple-Site Damage Fatigue Test 7
2.2.1 Materials 7
2.2.2 Specimen Geometry, Coordinates, and Loads 7
2.2.3 Fatigue Test Article Instrumentation and Procedure 9
3. ABAQUS FE SIMULATION FOR FMLs UNDER MULTIPLE-IMPACT

LOADS 9
3.1 General Development of FE Modeling 10

3.2 Numerical Method for GLARE Impact 13
3.2.1 Three-Dimensional Modeling Rationale 13
3.2.2 Three-Dimensional Progressive Damage Model for

Composite Materials 14
3.2.3 Energy Dissipation Due to Damage 18
4. ANALYTICAL PREDICTION MODEL OF MSD FATIGUE CRACK

GROWTH IN FIBER-METAL LAMINATES 20
4.1 Classic Lamination Theory 22

4.2 Secondary Bending Effect 23
4.3 Bridging Stress Distribution 27
4.4 Weight Function for SIF Calculation 31
4.5 Formation of Effective SIF 32
4.5.1 Far-Field SIF 33

4.5.2 Crack-Opening SIF 33
4.5.3 Nondimensional SIF 34
iii
4.6 Fatigue Crack Growth Prediction 38
4.7 Fatigue Delamination Growth Prediction 38
5. TEST RESULTS 39
5.1 Multiple-Impact Behavior of GLARE 5-2/1 Laminates 39

5.2 Multiple-Site Damage Fatigue Behavior of GLARE 3-3/2 Laminates 44
5.2.1 Through-Thickness Crack Growth 45

5.2.2 Partial-Thickness Crack Growth 50
6. COMPARISON OF TEST AND ANALYSIS OF MULTIPLE-IMPACT LOADS

AND FATIGUE CRACK GROWTH 55
6.1 Multiple-Impact Loads 55
6.1.1 Comparison of Experiment and FE Simulation 55

6.1.2 Two- and Three-Dimensional Failure Criterion Comparison 59
6.2 Multiple-Site Damage Fatigue Crack Growth 63
6.2.1 Through-Thickness Test Specimens 63

6.2.2 Partial-Thickness Test Specimens 65
7. CONCLUSIONS 66
8. REFERENCES 67
iv
LIST OF FIGURES
Figure Page
1 A Cross-Ply GLARE 3-3/2 Laminate 2
2 Bridging Mechanism in FMLs With Cracked Metals and Delamination 2
3 Schematic and Cross-Sectional Views of GLARE 5-2/1 4
4 Multiple-Site Impact Damage Specimens 5
5 Dynatup Model 8250 Instrumented Drop Weight Impact Tower 6
6 Multiple-Site Impact Damage Fatigue Test 6
7 Configuration of GLARE 3-3/2 Laminate Specimens (a) Through-Thickness MSD

and (b) Partial-Thickness MSD 8
8 Cross-Sectional View of GLARE 3-3/2 9
9 Finite Element Model Geometry and Boundary Conditions (a) Experimental and
Mesh Setup, (b) Cross View of Multiple Impacts, (c) Meshed Model of
GLARE 5-2/1, and (d) Boundary Condition 11
10 Continuum Shell Element Normal Vector Relation to Direction of Impact 14
11 Hashin Composite Failure Modes (a) Fiber Tension, (b) Fiber Compression,
(c) Matrix Tension/Shear, and (d) Matrix Compression 15
12 Composite Progressive Damage VUMAT Flowchart for the ABAQUS/Explicit

Processor 18
13 General Schematic for Energy Dissipation of a Continuum Body 19
14 Flowchart of Crack Growth Prediction Approach for GLARE Laminates 21
15 Typical Stress Distribution in the Al Layer and Curing Stress in Each Layer of GLARE
Laminate Under Cyclic Loading at Room Temperature With a Frequency of 10 Hz 23
16 A Surface-Cracked FML Subjected to Tensile Load (a) Neutral Line Shift due to
Unsymmetrical Configuration and (b) Eccentricity in Neutral Lines for FMLs 25
17 Deflection of the Neutral Line in GLARE Laminates Along Loading Direction 26
18 Bending Stress Distribution in Metal Layer of GLARE Laminates Along Loading

Direction 26
v
19 Bridging Stress in FMLs (a) Combination of Applied Load and Secondary
Bending Moment and (b) Crack Opening and Closing Profiles 27
20 Bridging Stress Distribution at the Crack Face in GLARE 3-3/2 Specimens Under
Unit Applied Load in FMLs With an Applied Stress Ratio of 0.05 31
21 Crack-Opening Stress as a Function of Stress Ratio 34
22 Effective Nondimensional SIF ( f eff = f o − fbr ) as a Function of Crack Length

(a) Through-Thickness and (b) Partial-Thickness 36
23 Effective SIF for GLARE 3-3/2 Specimens With Multiple Surface Cracks Under the
Applied Stresses of 120 MPa With an Applied Stress of 0.05 (a) Through-Thickness
and (b) Partial-Thickness 37
24 Multiple-Impact Damage of Al 2024-T3 and GLARE 5-2/1 at BVID

(8 J (2x4 J)), BVID (16 J (2x8 J)), and CVID (26 J (2x13 J)) 39
25 Load-Time Histories of Al 2024-T3 and GLARE 5-2/1 Under Multiple Impact With
Impact Energy = 8 J, Impact Energy = 16 J, and Impact Energy = 26 J 40
26 The Permanent Central Deflection as a Function of Impact Energy 41
27 Multiple-Site Impact Damage Fatigue Crack Propagation 42
28 Multiple-Site Impact Damage Fatigue Crack Growth (116.2 MPa) 42
29 Multiple-Site Impact Damage Fatigue Crack Growth 43
32 On-Going MSD Fatigue Cracks in GLARE 3-3/2 (a) Crack Growth in the Surface
Metal Layer and (b) Crack Growth in the Inner Metal Layer 45
33 The MSD Fatigue Delamination Pattern in the Fiber/Adhesive Layer at Cracks

1B and 2B for GLARE 3-3/2 46
34 Average Crack Length as a Function of Cycles for GLARE 3-3/2 Laminates at the
Applied Stress Level of 120 MPa 48
35 Average Crack Length as a Function of Cycles for GLARE 3-3/2 Laminates at the
Applied Stress Level of 100 MPa 48
36 Average Crack Growth Rate as a Function of Crack Length for GLARE 3-3/2
Laminates at the Applied Stress Level of 120 MPa 49
vi
37 Average Crack Growth Rate as a Function of Crack Length for GLARE 3-3/2
Laminates at the Applied Stress Level of 100 MPa 49
38 Crack Propagation in Surface-Cracked GLARE 3-3/2 at the Applied Stress Level

of 120 MPa 50
39 Delamination Growth in Surface-Cracked GLARE 3-3/2 Specimens, Nonlead

Delamination and Leading Delamination and Linkup 51
40 The MSD Crack Length as a Function of Cycles for Surface-Cracked GLARE 3-3/2
at the Applied Stress of 120 MPa 53
41 The MSD Crack Length as a Function of Cycles for Surface-Cracked GLARE 3-3/2
at the Applied Stress of 100 MPa 53
42 The MSD Crack Growth Rate as a Function of Crack Length at the Applied Stress
of 120 MPa for Surface-Cracked GLARE Specimen 54
43 The MSD Crack Growth Rate as a Function of Crack Length at the Applied Stress
of 100 MPa for Surface-Cracked GLARE Specimen 54
44 First Impact on GLARE 5-2/1 55
45 Second Impact on GLARE 5-2/1 56
46 Comparisons of Experimental and Simulation Results of Multiple-Impact,

Force-Time History for GLARE 5-2/1, Impact Energy (a) 8 J and (b) 16 J 57
47 Difference Between Predicted and Measured Central Displacement for

GLARE 5-2/1 Impact Energy (a) E = 8 J and (b) E = 16 J 58
48 Failure Mode Comparison for 2D and 3D Models of GLARE 5-2/1 After

First Impact E = 8 J 59

Second Impact E = 8 J 60

First Impact E = 16 J 61

Second Impact E = 16 J 62
52 The 3D ABAQUS Simulation of Multiple-Site Impact Damage on GLARE 5-2/1 63
53 Comparison of Predicted Averaged Leading Crack Growth Rates and Experimental

Results With an Applied Stress Ratio of 0.05 for GLARE 3-3/2 Through-Thickness
Test Specimens (120 MPa) 64
vii
Results With an Applied Stress Ratio of 0.05 for GLARE 3-3/2 Through-Thickness
Test Specimens (100 MPa) 64

Results With an Applied Stress Ratio of 0.05 for GLARE 3-3/2 Partial-Thickness Test
Specimens (120 MPa) 65

Specimens (100 MPa) 66
viii
LIST OF TABLES
Table Page
1 Grades of GLARE Laminates 1
2 Materials Properties of GLARE Laminates 4
3 Materials Properties Used in the FE Model 13
4 Damage and Failure Properties Used in the FE Model 13
5 Degradation Scheme for Composite Progressive Damage 20
6 Minimum Cracking Energy and Perforation Energy for GLARE Laminates 20
7 Multiple-Site Impact Damage Fatigue Crack Growth Data 44
8 Summary of Experimental Results for Through-Thickness GLARE 3-3/2 Laminates 47
9 Summary of Experimental Results for GLARE 3-3/2 Surface (Partial-Thickness)

Crack Growth 51
ix
LIST OF SYMBOLS AND ACRONYMS
Al Aluminum
σij Material stress state
εij Material strain state
εij+init Tensile strain component at failure
εij-init Compressive strain component at failure
fft Hashin fiber tension failure initiation
ffc Hashin fiber compression failure initiation
fmt Hashin matrix tension failure initiation
fmc Hashin matrix compression failure initiation
E11 Composite modulus along the fiber axis
E22 Composite modulus transverse to the fiber axis
υ12 Poisson’s ratio for the fiber-matrix direction
υ23 Poisson’s ratio for the matrix-matrix direction
G12 Shear modulus fiber-matrix direction
t Time value at the current finite element analysis increment
Δt Time step between successive increments
w i Damage propagation rate for each failure mode (i)
wi Damage state variable for each failure mode (i)
Ω0 Damage nucleation rate constant
Ω1 Damage growth rate constant
Sij Material compliance matrix
d Material degradation factor
Cij Material stiffness matrix
vi Velocity vector
Ti Surface traction vector
Fi Body force vector
ρ Density
U Internal energy per unit volume
W Internal energy
ε ij Strain rate
Sijkl Material compliance tensor
Cijkl Material stiffness tensor
Iijkl Fourth-order identity tensor
2D Two-dimensional
2D Three-dimensional
BVID Barely visible impact damage
C3D8R Hexahedral solid element
CVID Clearly visible impact damage
FML Fiber-metal laminates
FE Finite element
GLARE S2-glass fiber-reinforced aluminum laminates
x
MSD Multiple-site damage
SC8R Hexahedral shell continuum element
SIF Stress-intensity factor
VUMAT User-defined material subroutine
xi/xii
EXECUTIVE SUMMARY
Fiber-metal laminates (FML) are a family of hybrid metal and composite laminates for aircraft
structural applications. Commonly, FMLs, such as aramid-reinforced aluminum laminates and
S2-glass fiber-reinforced aluminum laminates (trade name GLARE), are composed of alternating
layers of unidirectional or bidirectional fiber-reinforced prepregs with aluminum alloy sheets.
These hybrid laminates have the advantage of metal and fiber-reinforced composites, providing
superior mechanical properties compared to the conventional lamina, which merely consists of
fiber-reinforced lamina or monolithic aluminum alloys. As a replacement of high-strength
aluminum alloys, GLARE laminates have been applied to aircraft structures, such as the Airbus
A-380, including fuselage and leading edges, for weight reduction and improved fatigue
resistance. This report summarizes a comprehensive study on the damage tolerance and
durability of GLARE subjected to impact and dynamic fatigue loading. Specifically, this study
focused on investigating the multiple impacts and multiple-site fatigue damage of GLARE
laminates.
For multiple-impact behavior, two different levels of impact energy were used to inflict barely
visible impact damage and clearly visible impact damage on the laminates. The experiment and
finite element (FE) simulation for low-velocity multiple impacts was performed on GLARE
laminate. Drop-weight impact was applied twice at a variety of energy levels to cause multiple-
impact damages for aluminum 2024-T3 and GLARE 5-2/1. The numerical model, using the FE
method, was developed by using the commercial FE code ABAQUS and verified by comparing
to the experimental data. The two-dimensional and three-dimensional (3D) failure criteria of the
composite layers were used for predicting all dynamic responses, including load-time history,
maximum deflection-time history, and damage progression. The 3D progressive failure model
was incorporated into the user-defined material subroutine (VUMAT) in the ABAQUS software
package. The FE simulation shows good agreement with experimental data for stress analysis
under multiple-impact loads.
The multiple-site fatigue damage behaviors of FMLs were investigated experimentally and
analytically under tension-tension, constant-amplitude fatigue loading. It was found that the
presence of multiple-site fatigue cracks would accelerate the crack growth rates in the metal
layers of FMLs as two propagating cracks approached each other. An analytical methodology
was proposed to calculate the fiber-bridging stress based on the concept of crack-opening
relations. The multiple-site crack growth rates were predicted by an empirical Paris-type fatigue
law in which the effective stress-intensity factor (SIF) was formulated as a function of applied
SIFs, crack-opening SIFs, and nondimensional SIFs. The predicted crack growth was validated
by constant-amplitude fatigue tests at various stress levels on multiple-site damage specimens.
xiii/xiv
1. INTRODUCTION.
Fiber-metal laminates (FML) are a family of hybrid metal and composite laminates for aircraft
structural applications. Commonly, FMLs, such as aramid-reinforced aluminum laminates and
S2-glass fiber-reinforced aluminum laminates (trade name GLARE), are composed of alternating
layers of unidirectional or bidirectional fiber-reinforced prepregs with aluminum alloy sheets.
GLARE was invented by the Delft University in the Netherlands [1 and 2]. These hybrid
laminates have the advantage of metal and fiber-reinforced composites, providing superior
mechanical properties compared to the conventional lamina, which merely consist of fiber-
reinforced lamina or monolithic aluminum alloys. As a replacement of high-strength aluminum
alloys, GLARE laminates have been applied to aircraft structures, such as the Airbus A380,
including fuselage, leading edges, etc., for weight reduction and improved fatigue resistance [3].
There are different grades of GLARE laminates, as shown in table 1. Because of different lay-up
sequence and fiber orientation, different grades of GLARE laminates have different mechanical
properties. A typical GLARE 3-3/2 laminate is shown in figure 1.
Table 1. Grades of GLARE Laminates
GLARE Metal Sheet Thickness Prepreg Orientation Main Beneficial

Grade Subgrades (mm) and Alloy in Each Fiber Layer Characteristics
1 (0.3-0.4) 7475-T761 0°/0° Fatigue, strength
yield stress
2 GLARE 2A (0.2-0.5) 2024-T3 0°/0° Fatigue, strength
GLARE 2B (0.2-0.5) 2024-T3 90°/90° Fatigue, strength
3 (0.2-0.5) 2024-T3 0°/90° Fatigue, impact
4 GLARE 4A (0.2-0.5) 2024-T3 0°/90°/0° Fatigue, strength in
0° direction
GLARE 4B (0.2-0.5) 2024-T3 90°/0°/90° Fatigue, strength in
90° direction
5 (0.2-0.5) 2024-T3 0°/90°/90°/0° Impact
6 GLARE 6A (0.2-0.5) 2024-T3 (+)45°/(-)45° Shear, off-axis
properties
GLARE 6B (0.2-0.5) 2024-T3 (-)45°/(+)45° Shear, off-axis
properties
1
Figure 1. A Cross-Ply GLARE 3-3/2 Laminate
It is well known that FMLs exhibit excellent fatigue crack growth resistance due to the fiber-
bridging mechanism, as shown in figure 2. As the crack propagates, the crack opening is
restrained by the intact fibers in the wake of the fatigue crack. This results in a reduction of
stress-intensity factor (SIF) at the crack tip and a far lower fatigue crack growth rate compared to
the monolithic aluminum alloy.
Figure 2. Bridging Mechanism in FMLs With Cracked Metals and Delamination
For aircraft composite structures, the most significant safety issues are impact resistance, damage
durability, and tolerance after single or multiple impacts. Impact damage of aircraft structures
can occur by a collision between cargo vehicles, dropped tools during maintenance, hail, bird
strikes, and lightning strikes [4]. Damage caused by impact degrades laminate stiffness and
strength, which is the load-bearing ability of the structure [5 and 6]. In general, if the impact
energy is low, only barely visible impact damage (BVID) will occur. However, if the impact is
repeated, even if it is at the level of low-impact energy, the small dent damage may accumulate.
Thus, it would lead to strength reduction in FMLs, as observed in references 7-11. However, the
information for FMLs subjected to multiple-impact loads is limited in the literature. More
studies are needed to investigate the multiple-impact scenarios in FMLs.
2
In 1988, Aloha Airline Flight 243 suffered explosive decompression when part of the fuselage
ruptured during flight due to metal fatigue and failed epoxy bonding. The multiple-site fatigue
damage in metallic airframe has received considerable attention. Since that accident, much
research has been conducted to study the multiple-site damage (MSD) problems in aircraft
structures to avoid catastrophic failure in flight [12-18]. When monolithic aluminum alloys are
subjected to cyclic fatigue load, cracks emanate around the fastener holes and propagate. A
number of neighboring cracks might coalesce to form a single dominant crack. As these fatigue
cracks are present in the fuselage of an aging aircraft, significant load redistribution will occur,
especially ahead of the dominant cracks. The interaction of fatigue cracks will increase the SIF
and enhance the crack growth rate in metallic alloys. Subsequently, the residual strength is
reduced and the integrity of structural components is affected.
In FMLs, the crack propagation and fracture characteristics are essentially different from those
of metallic alloys. When FMLs are subjected to fatigue loading, cracks initiate and progress in
the metal layer. The crack opening in the metal layers is restrained by the intact fibers in the
wake of the fatigue crack. The crack growth in the metal layer is impeded by the fiber-bridging
mechanism, resulting in a much lower crack growth rate compared to monolithic aluminum
alloy. As a result, it is anticipated that the influence of multiple-site damage in FMLs will differ
from that of a metallic structure. Also, MSD fatigue may occur only in the surface metal layers,
i.e., a crack through the thickness direction of a single aluminum surface layer. These surface
cracks may initiate from a scratch, a notch, or an impact damage site. In the FML surface crack
configuration, the fiber-bridging effect can be treated as a crack opening bridged by the intact
laminate [19]. Away from the surface metal layer and fiber/prepreg with delamination, the
remaining intact laminate is able to support the outer surface metal layer, leading to significant
reduction of stress level in the cracked metal layer. Also, though FMLs are designed
symmetrically in lay-up sequence, the presence of the surface cracks would make the laminates
unsymmetrical. Besides the fiber-bridging effect in crack growth, there is neutral line
shift/deflection from secondary bending because of unsymmetrical internal stresses in laminates
[20]. The secondary bending effect increases stresses in the layers, depending on the laminate
thickness and the crack length. If the thickness of the FMLs is large compared to crack length,
the shift of the neutral axis of a laminate induced by the cracked outer metal layer would be
small, and the corresponding secondary bending can be neglected. Concerning multiple-site
surface crack growth, there is very few data available in current literature [19-21] that studies
crack propagation, crack linkup phenomena in cracked surface metal layers of FMLs under
cyclic load. The fatigue behavior of FMLs with the presence of MSD has not been well
documented. A better understanding of the multiple-site fatigue damage behaviors in FMLs,
such as crack growth interaction, crack linkup, and delamination linkup, are structural integrity
concerns.
In this study, experiments and finite element (FE) simulations were conducted to study the
multiple-impact behavior of GLARE laminates. The commercial FE software package
ABAQUS was used with VUMAT (explicit user-defined material subroutine) to simulate the
damage progression in composite layers of GLARE laminates. The simulation results were
validated with experiments. The multiple-site fatigue damage in GLARE laminates was then
experimentally investigated under different applied loads with detailed damage characterization.
Both through-thickness and partial-thickness MSD fatigue scenarios were studied to understand
3
crack and delamination interactions. An analytical prediction model of crack growth was
developed and experimentally validated.
2. TEST PROGRAM AND TEST MATRIX.
2.1 MULTIPLE-IMPACT LOAD TEST.
2.1.1 Materials.
Aluminum (Al) 2024-T3 and GLARE 5-2/1 were used for multiple-impact experiments. Figure
3 shows the configuration of GLARE 5-2/1 laminates used in the present experimental
investigation. Figure 3(a) shows that GLARE 5-2/1 laminate consists of two layers of
Al 2024-T3 and one layer of [0°/90°/90°/0°] glass/epoxy composite, and figure 3(b) shows a
cross-sectional view of GLARE 5-2/1 in detail. The materials properties of GLARE constituents
are shown in table 2.
(a) (b)
Figure 3. (a) Schematic and (b) Cross-Sectional Views of GLARE 5-2/1
Table 2. Materials Properties of GLARE Laminates
Al 2024-T3 S2-Glass
Fiber/Epoxy
Ex (GPa) 72.4 55.5
Ey (GPa) 9.5
Gxy (GPa) 2706 5.55
σys (MPa) 369
νxy 0.33 0.3
νyx 0.33 0.0575
-6
α[1/°C] 22x10 6.1x10-6
t (mm) 0.3~0.4 0.133
4
2.1.2 Specimen Geometry, Coordinates, and Impact Energy.
The average thickness of different layers was measured using optical micrographs. For GLARE
5-2/1, the average thickness of the Al layer was 0.489 mm, and the average thickness of the
composites was 0.584 mm. The total laminate thicknesses were 1.562 mm for GLARE 5-2/1.
The average thickness of the Al 2024-T3 sheet was 1.60 mm. For multiple-impact load tests, all
specimens were cut by a water jet to a size of 76.2 x 76.2 mm. The GLARE 5-2/1 was examined
carefully for any damage caused by cutting. Diverse impact energies like 8 J (2x4J), 16 J (2x8
J), and 26 J (2x13 J) were applied by adjusting the dropping height in low-velocity, multiple-
impact tests. Two scenarios were studied for the multiple-site impact damage, as shown in
figure 4.
Figure 4. Multiple-Site Impact Damage Specimens
2.1.3 Impact and Postimpact Test Article Instrumentation and Procedure.
As shown in figure 5, low-velocity impact tests were conducted on a Dynatup® Model 8250 drop
weight impact tower at impact velocities below 5 m/s. Al 2024-T3 was used as a baseline for
comparison with GLARE laminates. The data during impact testing were collected by a PC-
based data acquisition system: GRC 930-I Dynatup with a photodiode velocity detector. After
just applying impact, the pneumatic rebound breaks are activated to push up and hold the
impactor assembly in place so that the specimen is not subjected to multiple impacts. To achieve
multiple-impact damages, the impactor was dropped twice on the same locations of Al 2024-T3
and GLARE 5-2/1. The specimens were clamped between two 114.3- by 114.3-mm-square steel
plates. The steel plates had a 31.7-mm-diameter circular opening at the center. A 12.7-mm-
diameter steel rod with a semi-spherical end, and a mass of 5.6 kg, was used as an impactor.
5
Figure 5. Dynatup Model 8250 Instrumented Drop Weight Impact Tower
As shown in figure 6, after the impact test, postfatigue tests were conducted and the crack length
was measured on each crack site. After the impact or postimpact fatigue test, characterization
was performed to investigate the state of damage. The outer Al layers of the GLARE laminate
were removed chemically, using sodium hydroxide solution, to observe the damage
characteristics in the S2-glass fiber/epoxy layer after any two cracks linked or cracks reached the
edge of the specimens.
Figure 6. Multiple-Site Impact Damage Fatigue Test
6
2.2 MULTIPLE-SITE DAMAGE FATIGUE TEST.
2.2.1 Materials.
The materials used in this study were GLARE 3-3/2 that consisted of three thin layers of
Al 2024-T3, bonded together with two thin S2-glass fiber layers in an FM®94 adhesive system.
These materials were manufactured and provided by Airbus, Germany. They were cured in an
autoclave cycle with a maximum pressure of 6 bar at a curing temperature of 120°C. The lay-up
for this grade of GLARE laminate is defined as [2024-T3/0° glass fiber/90° glass fiber/2024-
T3/90° glass fiber/0° glass fiber/2024-T3].
To avoid secondary bending effects due to unsymmetrical internal stresses present in the FMLs,
these GLARE laminates have a symmetrical lay-up. The materials properties of the constituents
of GLARE laminates are listed in table 2.
2.2.2 Specimen Geometry, Coordinates, and Loads.
The specimen dimensions were 300 mm in length and 75 mm in width. Two types of MSD were
introduced: through-thickness open holes and surface slits. The spacing in between two
neighboring open holes/slits was 30 mm, and the distance in between the top and bottom rows of
circular holes/surface slits was 100 mm. For through-thickness specimens, two rows of open
holes were carefully prepared with radii of 2.5 mm (length of 4 mm), and the starter notch was 1
mm in length. For partial-thickness specimens, two rows of surface slits were carefully prepared
with a length of 4 mm. The width of the slit was 1 mm. The configuration is shown in figure 7
for both through-thickness specimens and partial-thickness specimens. No delamination
occurred during machining, as shown in figure 8.
7
(a)
(b)
T = top B = bottom
Figure 7. Configuration of GLARE 3-3/2 Laminate Specimens (a) Through-Thickness MSD and
(b) Partial-Thickness MSD
8
Figure 8. Cross-Sectional View of GLARE 3-3/2
2.2.3 Fatigue Test Article Instrumentation and Procedure.
The constant-amplitude fatigue tests were conducted using a servohydraulic testing machine,
INSTRON™-8350. The fatigue test was done according to ASTM D 3479 [22]. All constant-
amplitude fatigue tests were conducted in tension-tension loading at a frequency of 10 Hz and a
stress ratio of R = 0.05. Tests were conducted in a sinusoidal cyclic waveform under load
control. The maximum applied stress levels were 120 MPa and 100 MPa, respectively. The
crack lengths as a function of fatigue cycles for GLARE 3-3/2 specimens were recorded through
continuous monitoring during the fatigue tests. After the fatigue tests, the chemical-mechanical
removal method was used to detect the delamination in the interface of metal/prepreg in GLARE
laminates. To investigate the fatigue cracking in the inner Al layers of GLARE 3-3/2, the
surface aluminum sheets were etched away by chemical solutions, and the composites layers
were removed mechanically.
3. ABAQUS FE SIMULATION FOR FMLs UNDER MULTIPLE-IMPACT LOADS.
The ABAQUS FE simulation work was performed to represent and predict the response
to multiple-impact behavior in GLARE laminates. The response was then compared to
Al 2024-T3.
9
3.1 GENERAL DEVELOPMENT OF FE MODELING.
The multiple impact was simulated under a diverse multiple-impact energy level using the
commercial FE code ABAQUS [23]. The applied impact energy levels in the FE model were
identical to what was used in the experiments. The FE model was simulated for BVID with 8 J
(2x4 J) and 16 J (2x8 J)) and clearly visible impact damage (CVID) with 26 J (2x13 J)). In the
FE simulation of GLARE 5-2/1, the same geometry of the specimens used in the experiments
was taken. For predicting damage caused by impact, the ABAQUS/Explicit solver was used for
GLARE laminates. The element types used in the simulation work were hexahedral solid
element (C3D8R) for the aluminum layer and hexahedral solid element (C3D8R) and hexahedral
shell continuum element (SC8R) for the composite layer with different fiber orientation.
Damage in GLARE laminate can be distinguished between damage in aluminum and in
S2-glass/epoxy layers. The aluminum layers were assumed to have elastic-plastic deformation in
compression and crack damage in tension. For damage in the fiber/prepreg, both the damage in
the epoxy matrix and in the fiber layers was considered. The first damage was caused by
cracking under tension, and the second damage was caused by fiber and matrix failure under
tensile loading [24]. The damage that occurred in the S2-glass/epoxy layer was implemented in
ABAQUS using VUMAT. The S-2 glass fiber-reinforced composite layers were modeled with
VUMAT, which was developed based on Hashin three-dimensional (3D) failure criteria [25], to
degrade the stiffness. When each node was integrated in the composite layer during multiple
impacts, VUMAT was used to calculate the stiffness matrix.
In modeling dynamic impact behavior, because of the excessive distortion that some FEs may
experience, this nonlinearity of constitutive behavior would lead to complicity in convergence or
divergence. The issue of numerical convergence needs to be addressed properly to prevent the
occurrence of divergence of numerical solution. In ABAQUS, the parameters’ “hour glassing”
are used to hasten the convergence of numerically nonlinear modeling for the reduced
integration elements, such as C3D8R and SC8R, in stress/displacement analyses [23]. To
eliminate the singularity upon the integration point, hourglass control is critical in the nonlinear
impact simulation. If the returned stress/strain value from the integration point is zero, the
degree of distorted meshes will not be able to be controlled. Numerically, the FE model will go
divergent.
Figure 9 shows the experimental setup compared to the FE mesh. Since the impact load was
only on the near-field meshed grids, the far-field meshed grids were not considered, as shown in
figure 9(a). The GLARE laminate mesh was modeled based on the real size of the impact fixture
hole. The cross and front views of the FE model are shown in figure 9(b) and (c), respectively.
As shown in the cross and front views of the multiple-impact loads, two impact loads were
applied at a different time interval. There were no interactions between these two impactors
simulated as rigid bodies in FE modeling, although they did have contact. The boundary
conditions in figure 9(d) correspond to the experimental setup. Because of the symmetry of
GLARE 5-2/1 laminates, only half the laminates were modeled with symmetrical boundary
conditions applied. All edge elements were fixed in the x, y, and z directions with zero
10
displacement. The rigid body simulated as an impactor was allowed to move only along the
z direction with zero rotation. The materials properties and damage failure criteria [24, 26, and
27] of GLARE 5-2/1 laminates used in this model are summarized in tables 3 and 4,
respectively.
(a) Experimental Setup and FE Mesh
(b) Cross View of Multiple Impacts
Figure 9. Finite Element Model Geometry and Boundary Conditions (a) Experimental and
Mesh Setup, (b) Cross View of Multiple Impacts, (c) Meshed Model of GLARE 5-2/1, and
(d) Boundary Condition
11
(c) Front View of Meshed FE Model
(d) Boundary Condition
Figure 9. Finite Element Model Geometry and Boundary Conditions (a) Experimental and
Mesh Setup, (b) Cross View of Multiple Impacts, (c) Meshed Model of GLARE 5-2/1, and
(d) Boundary Condition (Continued)
12
Table 3. Materials Properties Used in the FE Model [5 and 6]
Parameter Values, GPa

Material
E11 E22 G12 G23 ν12 ν23
Al 2024-T3 72.2 72.2 - - 0.33 0.33
S2-Glass fiber 55 9.5 5.5 3 0.33 0.33
Table 4. Damage and Failure Properties Used in the FE Model [5 and 6]
Parameter Values, MPa

Material σys Xt Xc Yt Yc Slt
Al 2024-T3 320 - - - - -
S2-Glass fiber - 2500 2000 50 150 75
3.2 NUMERICAL METHOD FOR GLARE IMPACT.
3.2.1 Three-Dimensional Modeling Rationale.
In typical FE analyses of composite structures, continuum shell elements are used to model the
composite laminate. Continuum shells combine the planar nature of traditional shells with the
3D geometry of solid elements. As such, they offer a computationally efficient method for
simulating systems that are globally 3D, but locally planar [28]. A multilayered, laminated
composite structure is a good example of such a system. Because the mechanical properties of
the composite laminate will, in general, vary between subsequent layers, the structure cannot be
modeled with a single set of 3D solid elements; each layer must be represented by a unique
element set. One layer of a unidirectional tape-based (i.e., not woven) composite typically has a
thickness on the order of hundreds of microns, whereas the structure into which it is incorporated
has geometry on the order of meters. Because of this large aspect ratio with respect to the
laminate thickness, each layer can be assumed to be in a state of plane stress, with all through-
thickness normal and shear components of the stress tensor (σ13, σ23, σ33) assumed negligible in
comparison to their in-plane counterparts (σ11, σ22, σ12). Therefore, such a system is technically
3D, but approximately planar, and can be modeled with continuum shell elements.
This planar assumption is valid for most composite structures, but breaks down in several
specific instances. The most obvious breakdown scenario is for a thick composite, in the
limiting case for which the aspect ratio of the laminate width and/or length with respect to the
thickness approaches unity. For this system, the laminate thickness is no longer negligible in
comparison to the other dimensions, and interlaminar through-thickness stress tensor
components must be considered. A second, and perhaps less obvious, scenario that prevents the
use of continuum shell elements involves the nature of the boundary conditions present in the FE
model. If the boundary conditions are applied such that nontrivial out-of-plane stresses arise in
response to their application, then the system is no longer planar and continuum shells should not
be used. Such is the case for the FE model under consideration in the present study, a rigid
impactor incident upon the laminate in a direction aligned with the plane of material isotropy. In
13
this configuration, the direction of impact is parallel to the shell element normal vector, as shown
in figure 10, and therefore will produce nontrivial through-thickness stresses. The plane stress
assumption associated with continuum shell elements is inappropriate in this instance since it is
in direct contradiction with the applied boundary conditions. Another important factor that
precludes the usage of continuum shells in an impact simulation is the formation and propagation
of delamination zones between the composite laminate. Since delamination is governed by the
out-of-plane stress states in consecutive layers, incorporation of all stress tensor components is
essential for accurate representation of postimpact delamination.
Figure 10. Continuum Shell Element Normal Vector Relation to Direction of Impact
3.2.2 Three-Dimensional Progressive Damage Model for Composite Materials.
Since shell elements are not appropriate for the impact simulation under consideration in the
present study, a vectorized VUMAT was developed to simulate progressive damage and failure
of composite materials with 3D solid elements. The methodology described herein applies to the
ABAQUS/Explicit processor, but can be adapted to any displacement-based FE code. For each
integration point in the model, the initiation of composite damage is defined using the well-
known Hashin failure criteria, expressed in terms of the strain tensor (εij) and experimentally
measured strain at failure ( εinit
ij ) in equations 1-4 [25]. These failure strain values can be
obtained from the more widely used failure stress measurements by taking the inner product with
respect to the corresponding components of the material compliance tensor. The Hashin criteria
allows for four modes of composite damage, illustrated in figure 11, with each mode represented
by its own normalized failure variable (f i).
14
(a) (b)
(c) (d)
Figure 11. Hashin Composite Failure Modes (a) Fiber Tension, (b) Fiber Compression,
(c) Matrix Tension/Shear, and (d) Matrix Compression
a. Fiber Tension Failure Mode (ε11 >0)
2
⎛ ε11 ⎞ 1
= ⎜ init ⎟ + init 2 ( ε12 + ε13 ) ≥ 1
ft 2 2
f
⎝ ε11+ ⎠ ε12 (1)
b. Fiber Compression Failure Mode (ε11 <0)
−ε11
f fc
= ≥1
ε11
init
− (2)
c. Matrix Tension/Shear Failure Mode (ε22+ ε33 >0)
1 1
f mt =
ε init 2 ( ε22 + ε33 )
2
+
ε
init 2 (ε 2
23 + ε12
2
+ ε13
2
− ε 22 ε33 ) ≥ 1
22 + 12 (3)
d. Matrix Compression Failure Mode (ε22+ ε33 <0)
1 ⎡⎛ εinit ⎤
2
⎞ 1 ⎡ ε2 + ε2 2 ⎤
f mc
= init ⎢⎜ init ⎟ − 1⎥ ( ε 22 + ε33 ) + init 2 ⎢ 22 33 + ε 223 + ε12
22 − 2
+ ε13 ⎥ ≥1
ε 22− ⎢⎝ 2ε12 ⎠ ⎥ ε12 ⎣ 2 ⎦
⎣ ⎦ (4)
When any of the above failure variables exceed unity, damage is defined to have occurred in the
composite in that specific mode. It is important to note that this failure criterion includes all
components of the strain tensor. The two-dimensional (2D) failure methodology employed by
ABAQUS can be obtained from equations 1-4 by assuming a state of plane stress, which
15
eliminates the out-of-plane shear strains (ε13 and ε23) and yields an explicit form for the out-of-
plane normal strain (ε33).
1 v
ε33 = ( σ33 − v23σ22 ) − 12 σ11
E22 E11 (5)
In equation 5, the composite is assumed to be transversely isotropic, with the plane formed by
the 2-3 axes representing the plane of isotropy. Because of the anisotropic nature of the material,
all calculations performed by VUMAT are with respect to the local coordinate system. This
provides a consistent general framework for any off-axis laminate whose local principal material
directions are not aligned with those of the global system. Once failure has occurred in any
given mode, the material response is evaluated using the damage mechanics approach proposed
by Matzenmiller, et al. [29]. It is assumed that damage to the laminate in any of the four failure
modes reduces the effective load bearing area of the composite. Since a physical reduction of
area is not practical in an FE simulation, the damage mechanics area reduction is accomplished
via modification of the material’s elastic constants using an internal damage state variable (wi).
As a consequence, damage in the material at a given point reduces the tensorial stress
components, typical of a strain-softening constitutive model. Once any Hashin failure variable,
equations 1-4, exceeds a value of one, the internal damage state of the material follows a rate-
based progression model, similar to that proposed by Iannucci, et al. [30]. The rate of
propagation of damage in the material in all four modes at the current time step (ẃi (t + Δt))
depends on the damage state variable at the previous time step (wi(t)) and the current strain state
(ε(t + Δt)), as well as constant damage rate terms due to crack nucleation and growth (Ω0 and Ω1,
respectively). The explicit form for the damage propagation rate is provided in equation 6,
where the superscript (i) corresponds to each unique Hashin failure mode in equations 1-4.
⎡⎛ ⎞ ⎤
⎢ ⎜ ε kl (t + Δt )ε kl (t + Δt ) ⎟ ⎥
w (t + Δt ) = Ω0 + Ω1w (t )
i i
−1
⎢⎜ εinit εinit (1 − wi (t ) )2 ⎟ ⎥
⎣⎝ mn nm ⎠ ⎦ (6)
In equation 6, it is important to note that the strain state at failure ( εinit

ij ) is modified by the
previous damage state (wi(t)). This ensures that the “threshold” strain value required for
composite damage decreases with increasing damage density. In other words, following the
initial onset of damage, the energy required for crack formation is inversely proportional to the
number of cracks in the material at a given time. Since the time step between successive
increments is very small (on the order of picoseconds for an impact duration of 10 milliseconds),
the degradation rate of equation 6 is assumed to have a linear form. Thus, the damage state
variable at the current time step (wi(t+Δt)) can be obtained from the previous damage state and
the damage growth rate.
wi ( t + Δt ) = wi (t ) + w i (t + Δt ) ) Δt
wi ∈ [ 0,1) (7)
16
In this case, the damage state variables must be bounded by an upper limit less than one to
prevent computational issues arising during inversion when they are used to degrade the
composite elastic constants. Damage in a fiber-related mode (f ft, f fc) results in a decrease in the
modulus in the local fiber direction (E11) along with in-plane shear modulus and Poisson’s ratio
(G12 and ν12, respectively). Conversely, damage in a matrix-specific mode (f mt, f mc) produces
degradation of all elastic constants except for the modulus in the local fiber direction. Switching
from a tensorial representation to a reduced index matrix form, the composite material
compliance (S(t)) at any time step can be found from equations 8 and 9, where the linear elastic
material response for undamaged laminate is obtained by setting all damage state variables to
zero.
⎡ 1 −1 ⎛ v12 ⎞ −1 ⎛ v12 ⎞ ⎤
0 0 0
⎢ d1 E11 d 3 ⎜⎝ E11 ⎟⎠ d3 ⎜⎝ E11 ⎟⎠ ⎥
⎢ ⎥
⎢ −1 ⎛ v12 ⎞ 1 −1 ⎛ v23 ⎞ 0 0 0 ⎥
⎢ d3 ⎜⎝ E11 ⎟⎠ d 2 E22 d 2 ⎜⎝ E22 ⎟⎠ ⎥
⎢ ⎥ (8)
⎢ −1 ⎜⎛ v12 ⎞ −1 ⎛ v23 ⎞ 1 0 0 0 ⎥
⎢ d3 ⎝ E11 ⎟⎠ d 2 ⎜⎝ E22 ⎟⎠ d 2 E22 ⎥
S (t ) = ⎢ ⎥
⎢ 0 0 0 1 0 0 ⎥
⎢ d3G12 ⎥
⎢ ⎥
⎢ 0 0 0 0 1 0 ⎥
d3G12
⎢ ⎥
⎢ 2(1 + d 2V23 ) ⎥
0 0 0 0 0
⎢
⎣ d 2 E22 ⎥⎦
d1 = (1 − w ft )(1 − w fc )
d 2 = (1 − wmt )1 − wmc ) (9)
d3 = (1 − w ft )(1 − wmt )(1 − wmc )
Note that in the above constitutive model, the transversely isotropic nature of the material is
retained following the onset of damage. Once the material compliance is established at the
current time step, the current stress state in reduced index form can be determined from
equations 6-9 through inversion. For non-zero damage state variables, it is evident that the
stresses carried by a damaged element will be smaller than those of the corresponding
undamaged element, as required by the damage mechanics assumption.
σ(t + Δt ) = C (t + Δt ) ε (t + Δt )
(10)
C (t + Δt ) = S −1 (t + Δt )
Specific implementation of the damage mechanics procedure provided in equations 1-10 in the
ABAQUS/Explicit processor is outlined via flowchart in figure 12. In this model, the
compliance of FMLs was calculated first and then converted to stiffness under undamaged
17
scenarios. The strain-based failure criteria were determined and checked at each step of
integration by updating strain and stress states. The starting point of laminates failure was
obtained through Hashin failure variables. If an element met the failure criteria, it would be
removed from the mesh.
Figure 12. Composite Progressive Damage VUMAT Flowchart for the

ABAQUS/Explicit Processor
3.2.3 Energy Dissipation Due to Damage.
As a consequence of the progressive failure methodology covered in the previous section, once
the constitutive matrices are degraded, the material cannot revert to the undamaged state. Thus,
energy is dissipated by the material due to damage formation and growth. The amount of this
dissipated energy is an important quantity in impact simulations, since it represents the ability of
the material to absorb the kinetic energy of the theoretically rigid impactor. If thermal effects on
the material are neglected, the global balance of energy for the continuum body schematically
illustrated in figure 13 can be established from the specified body forces, tractions, and the
changes in kinetic and internal energy.
⎡ ⎤
∫ v T dS + ∫ v F dV = ∂ ∂t ⎢⎣ ∫ ( 1 2 ρv v + ρU ) dV ⎥⎦
i i i i i i
∂Γ Γ Γ (11)
18
Figure 13. General Schematic for Energy Dissipation of a Continuum Body
Applying the divergence theorem to the surface integral in the above equation and assuming that
the continuum body in figure 13 is at an equilibrium state, the time-integrated internal energy
(W) can be expressed as
t
W = ∫ ρUdV = ∫ ∫ σij ε ij dV
Γ 0 Γ (12)
The energy dissipated by the continuum body due to damage formation can be determined from
equation 12 by examining the response of the continuum body to a constant strain field (ε*).
Considering two material states, one damaged and the other undamaged (denoted by “D” and
“UD,” respectively), it is evident that the stress required to deform the material by ε* will not be
the same in both cases due to the strain-softening nature of the constitutive model. The energy
dissipated by the continuum body is equal to the difference in internal energy between the two
states
t t
ΔW = WUD − WD = ∫ ∫ σUD
ij εij dV − ∫ ∫ σij ε ij dV
* D *

0 Γ 0 Γ (13)
Since the strain is required to be identical in both cases, an expression linking the damaged and
undamaged stress states can be determined from the constitutive relations
ε*ij = Sijkl
D
σklD = Sijkl
UD UD
σkl
∴σUD
kl = Cijkl Sijmn σ mn
UD D D
(14)
19
Inserting the above expression into the internal energy integral equation produces the following
result for the dissipated energy, where (Iklmn) represents the fourth-order identity tensor
t
⎡ ⎤
(
ΔW = ∫ ⎢ ∫ {Cijkl
UD D
Sijmn − I klmn } σmn
D
)
ε kl dV ⎥ (15)
0 ⎣Γ ⎦
In the practical implementation of the progressive damage procedure into the FE model, this
integral is evaluated in a piecewise fashion due to the extremely small time interval between
successive steps. The degradation scheme of the composite’s progressive damage is shown in
table 5, and the cracking energy of GLARE laminates is shown in table 6.
Table 5. Degradation Scheme for Composite Progressive Damage
Degraded Elastic Material Properties

Failure Mode E11 E22 v12 v23 G12
Fiber tension x x x
Fiber compression x x x
Matrix tension/shear x x x x
Matrix compression x x x x
Table 6. Minimum Cracking Energy and Perforation Energy for GLARE Laminates [7]
Thinness Areal Density Minimum Cracking Perforation Energy

Materials (mm) (kg/m2) Energy (J) (J)
Al 2024-T3 1.6 4.45 18.1 33.4
GLARE 5-2/1 1.562 3.74 16.3 34.5
4. ANALYTICAL PREDICTION MODEL OF MSD FATIGUE CRACK GROWTH IN

FIBER-METAL LAMINATES.
To investigate the crack growth behavior of GLARE laminates with MSD, an analytical crack
growth model was proposed to calculate the crack growth rates. A flowchart is used to describe
this prediction approach, as shown in figure 14.
20
Figure 14. Flowchart of Crack Growth Prediction Approach for GLARE Laminates
21
4.1 CLASSIC LAMINATION THEORY.
The stress levels in each layer of FMLs are calculated through the 2D class lamination theory in
x-y coordinates, which is written as [31]
⎡ σx ⎤ ⎡ ⎛ ⎡α x ⎤ ⎡α x ⎤ ⎞⎤
⎢ ⎥ ⎢ −1 ⎜⎢ ⎥ ⎟⎥
σ φ ,l = ⎢σ y ⎥ = Qφ,l ⎢Qlam σlam + ΔT ⎜ ⎢α y ⎥ − ⎢⎢ α y ⎥⎥ ⎟ ⎥ (16)
⎢ ⎥ ⎢ ⎜ ⎟⎥
⎣ τ xy ⎦ φ ,l ⎣ ⎝ ⎢⎣ 0 ⎥⎦ lam ⎢⎣ 0 ⎥⎦ ⎠ ⎦
where the stiffness matrix is expressed as
⎡ Ex ν xy E y ⎤
⎢ 0 ⎥
⎢1 −ν xyν yx 1 −ν xyν yx ⎥
⎢ Ey ⎥
Q=⎢ 0 ⎥ (17)
⎢ 1 −ν xyν yx ⎥
⎢ ⎥
⎢ sym Gxy ⎥
⎢⎣ ⎥⎦
The coordinates x and y are parallel and perpendicular to the loading direction, respectively.
When the stiffness matrix is under off-axis angle φ
Qφ = ΦQΦT (18)
where Ф is the off-axis matrix and is expressed as
⎡ cos 2 (φ) sin 2 (φ) 2 cos(φ) sin(φ) ⎤

⎢ ⎥
Φ = ⎢ sin (φ)2
cos (φ)
2
−2 cos(φ) sin(φ) ⎥ (19)
⎢ − cos(φ) sin(φ) cos(φ) sin(φ) cos 2 (φ) − sin 2 (φ) ⎥
⎣ ⎦
Using the lamination theory above, the actual stress levels in the Al and fiber layers were
obtained and plotted as a function of cycle time and are shown in figure 15. Clearly, the stress
level in the Al layer is higher than the applied stress due to the difference in the stiffness of the
S2-glass/prepreg and metal layers. Due to the mismatch of thermal expansion coefficients in
different constituents, the curing stress would also raise the stress level in the Al layer.
22
260
240
220
200
180
160
140
Stress level (MPa)

120
100
80
60
40
20
applied stress
0
stress in Al
-20
curing stress in Al
-40 curing stress in 0 ply
-60 curing stress in 90 ply
-80
-100
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Time (sec)
Figure 15. Typical Stress Distribution in Al Layer and Curing Stress in Each Layer of GLARE
Laminate Under Cyclic Loading at Room Temperature With a Frequency of 10 Hz (The
maximum applied stress was 160 MPa with a stress ratio of 0.1 [32].)
4.2 SECONDARY BENDING EFFECT.
A secondary bending will occur in a laminate if the configuration is unsymmetrical, which

results in a shift of the neutral line. The thickness of a GLARE 3-3/2 laminate is thin, and this
allows a beam theory to be used to analyze the stress variation in a laminate caused by the
secondary bending. If the secondary bending is absent, the neutral axis will not be displaced and
no additional stress will be incurred. In the presence of the eccentricity of neutral lines, as
shown in figure 16, the neutral axis shifts and results in the introduction of bending stress. To
calculate the magnitude of bending stress, the displacement of the neutral line needs to be
obtained first for further consideration of stress level variation in a laminate. Using the neutral
line model [33], the bending moment, Mx, is expressed as
Mx = P·z (20)
where P is the applied force in Newtons, and w is the displacement in unit of mm. The symbol z
denotes the thickness of a laminate. By using the beam theory for a thin laminate, the bending
moment can be written as
d 2w
Mx = E*I 2 (21)
dx
23
For plane bending, E* = E/(1-ν²). A laminate is divided into surface metal cracked (Lcr, i = 1)
and uncracked (Lun, i = 2) parts, respectively. Then, equating equations 5 and 6, the differential
equation yields
(d ²wi/dxi²)–(P/E*I)i wi = 0, i = 1,2 (22)
where E is the stiffness, and I is the moment inertia of a laminate. The general solution for the
displacement of the neutral line is expressed as
wi = sinh(ηixi) + cosh(ηixi) (23)
and
ηi = P E * I )i (24)
For surface-cracked GLARE 3-3/2 laminates, the boundary conditions are defined as
x1 = 0, w1 = 0 (25)
w1(x1 = L1) = w2(x2 = 0) + e, e = 0.5 tAl (26)
(e = eccentricity of the neutral line in cracked and uncracked parts)
dw1(x1 = L1)/d x1= d w2(x2 = 0)/dx2 (27)
x2= L2, d w2(x2 = L2)/ dx2 = 0 (28)
Hence, the bending stress in the thickness direction of a laminate can be obtained with
σb = (Mw/I)z = (d ²w/ dx²)E*z (29)
Using equations 8 through 21, the actual stress level in layer l of a laminate is able to be obtained
as a combination of applied stress and bending stress. That is
σb = σl,lam + σb(x) (30)
24
(a)
(b)
Figure 16. A Surface-Cracked FML Subjected to Tensile Load (a) Neutral Line Shift due to
Unsymmetrical Configuration and (b) Eccentricity in Neutral Lines for FMLs
The deflection along the loading direction for a different lay-up sequence of GLARE 3 laminates
is shown in figure 17. As the thickness of laminates increases, the magnitude of deflection
decreases significantly. Clearly, the secondary bending would slightly change the neutral line
position in GLARE 3-3/2 (Al/0/90/Al/0/90/Al) laminates because of the neutral line’s
eccentricity. As the crack advances, the deflection of the neutral line would gradually increase.
The calculated bending stresses for GLARE 3-3/2, GLARE 3-4/3, and GLARE 3-5/4 were
plotted as a function of crack length and are shown in figure 18. The maximum bending stress is
around 15 MPa, 4 MPa, and 2 MPa for GLARE 3-3/2, 4/3, and 5/4, respectively. Hence, the
secondary bending effect is not very significant to affect the stress level in the surface-cracked
metal layer of GLARE 3-4/3 laminates and up.
25
Figure 17. Deflection of the Neutral Line in GLARE Laminates Along Loading Direction
Figure 18. Bending Stress Distribution in Metal Layer of GLARE Laminates Along
Loading Direction
The maximum bending stress occurs at the surface metal layer at the delamination tip along the
loading direction. The averaged bending stress through-the-thickness direction is used in certain
layers.
26
4.3 BRIDGING STRESS DISTRIBUTION.
In a GLARE laminate with MSD, the crack growth is impeded by the fiber-bridging mechanism
so that the crack opening in the metal layers is restrained by the fiber-bridging stress. Guo and
Wu established a theoretical prediction model for crack growth for center-cracked FMLs [34 and
35]. The configuration of the bridging stresses along the crack length for a center-cracked
geometry in FMLs is shown in figure 19(a), which includes the crack opening contour under the
maximum applied load with the corresponding closing bridging stress acting on the delamination
boundary. Based on the crack displacement relation, Guo and Wu [34 and 35] incorporated the
bridging traction into a theoretical prediction model for crack growth in GLARE laminates. This
relation had been adopted by other authors in crack growth modeling of GLARE and
boron/glass/Al FMLs [36-38]. With the presence of secondary bending, the bending moment
needs to be considered as shown in figure 19(b) for bridging stress calculation.
(a)
(b)
Figure 19. Bridging Stress in FMLs (a) Combination of Applied Load and Secondary Bending
Moment and (b) Crack Opening and Closing Profiles
27
The crack opening in the Al layers of FMLs can be written as the crack opening due to remote
applied stresses in the Al layers and the crack closing bridging stresses as result of the intact
fiber layers [35]. That is
u fml , Al = u∞ ( x) − ubr ( x) (31)
The effects of crack opening and closing bridging stress balance the deformation of fibers and
adhesives. In other words, the crack opening in the Al layers is identical to the elongation and
deformation of the fiber/prepreg layers in the crack opening area [35]. That is
u fml , Al = δ f ( x) + δ pp ( x) + δ Al (32)
where δf and δpp denote the deformation in the fiber and adhesive layers, respectively, and u∞ and
ubr denote the crack-opening displacement due to the remote applied stress and fiber-bridging
stress, respectively. δ Al is the metal deformation, which is not considered owing to its negligible
value. The subscripts f and Al represent the fiber and Al layers, respectively.
In a specimen containing multiple center cracks, the crack-opening displacement in the metal
layer caused by uniform applied stress and secondary bending under plane stress condition is
expressed in reference 39 as
σ Al
u∞ ( a , x ) = 2 a2 − x2 (33)
E Al
where x is the distance from the crack center, and EAl is the elastic modulus of the Al layer. The
stress level in the Al layer is determined by the applied stress and secondary bending. If the
secondary bending effect is considered, the bending stress should be superimposed.
The crack-opening displacement, ubr, caused by the fiber-bridging stresses can be calculated
through a series of crack-opening displacements induced by the point loads acting on position xj
of the crack flanks, as shown in figure 19(b), that is
a
ubr = ∫ u ( x, x j )dx j for 0 ≤ x ≤ a (34)
0
The crack-opening displacement u(x,xj) caused by a point load acting on arbitrary positions is
expressed as [39 and 40], if x < xj
⎛ 1 ⎞
⎜ a 2
− x 2 (1 +ν )b 2 ⎟
a 2 − x 2j
4 2
u ( x, x j ) = ⎜ tanh
−1 j
+ σ ( x j )dx j
2 ⎟ br
(35)
πE ⎜ a 2 − x 2 + b 2 x 2j − x 2 + b 2 a − x +b ⎟
2 2
⎝ ⎠
28
If x > xj, then replace x = x and xj = x in equation 35, where E is the elastic modulus, ν is the
Poisson ratio of laminate, and s is the spacing between two open holes.
Concerning the fiber/layer deformation, it should be noted that a superposition is used for
deformation induced by the stresses in the fibers and by the bridging stresses. In the presence of
delamination, the elongation of the fiber layer over the crack opening area is expressed as
σ f + σbr
δ f ( x) = ε f ( x)b( x) = b( x ) (36)
Ef
where εf and Ef are the strain and elastic modulus of the fiber layers, and σf and σbr are the stress
level and bridging stress in the fiber layers, respectively. b(x) is an arbitrary shape function for
delamination. The parabolic and triangular shape functions are expressed as
(a − x)
b( x) = bD
(a − r ) (37)
and
( a − x)
b( x) = bD (38)
(a − r )
where bD is the initial delaminating length at the notch root.
The shear deformation as a function of the shear stress at the Al/prepreg interface is expressed as
τf
δ pp = γt f = ( )t f (39)
Gf
Based on this expression and with linear shear deformation analysis, the prepreg shear
deformation in FMLs with cross-ply prepreg layers is derived and expressed as [41]
tf Gf 1 1
δ pp = Ccorr σ Al t Al (∑ )×( + ) (40)
Gf tf 2nAl t Al E Al ∑ n f t f E f
where the subscripts Al and f denote aluminum and fiber layer, respectively. The symbol t
represents the thickness of each layer. If the small delamination length is considered, a
correction factor, Ccorr, needs to be used, as presented in reference 40.
29
Similar to a governing equation in reference 34, the fiber-bridging stress σbr is derived as
σ br = M −j 1 N
(41)
where
σf
N = u∞ ( x) − δ pp ( x) − b( x)
Ef
(42)
and
ubr ( xi , x j )Δx j b( xi )
Mj =∑ − δ (i , j )
σ br ( x j ) Ef
(43)
⎧1,if i ≠ j
δ(i, j ) = ⎨ Kronector factor (44)
⎩0,if i = j
This governing equation can be solved using any programming tool. The bridging traction acts
on the surface of the metal layer to impede the crack growth, and the bridging stress at the crack
surface of the Al layer, σbr,Al, is obtained using the relationship of force balance in the FMLs.
Thus, the bridging stress at the crack face of the metal layers in a cross-ply FML is expressed as
n0, f t 0, f + n90, f t90, f

σ br , Al = ( ) M −j 1 N (45)
nAl t Al
In the application of weight function for the SIF calculation, the stress distribution on the crack
face can be known by using the same governing equation. The metal-bridging stress at the crack
surface under the unit applied load in the laminate as a function of crack length was plotted and
is shown in figure 20. Observably, the bridging stress is close to uniform along the crack length,
and the highest value is at the crack tips.
30
Figure 20. Bridging Stress Distribution at the Crack Face in GLARE 3-3/2 Specimens Under
Unit Applied Load in FMLs With an Applied Stress Ratio of 0.05
4.4 WEIGHT FUNCTION FOR SIF CALCULATION.
In the presence of multiple-site fatigue damage, the geometry and configuration need to be taken
into account while calculating the far-field SIF in the metal layer of GLARE laminates. Using
the weight function approach, the effects of geometry and configuration of the test specimens
can be properly incorporated. The direction of crack propagation is perpendicular to the loading
direction, therefore, only mode I SIF needs to be considered.
For one-dimensional, through-thickness crack propagation, the SIFs associated with symmetric
mode I loading can be calculated by applying the concept of weight function. That is [42]
a
K = ∫ σ ( x)m(a, x)dx (46)
0
where σ(x) is the stress distribution on the crack face, and m(a,x) is the weight function. In
equation 21, the SIF can be expressed as [42]
K = f σ πa (47)
31
where f is the nondimensional SIF, and a is the crack length starting from an open hole. The
weight function for a one-dimensional crack can be expressed as [42]
E' ∂ur (a, x)

m( a, x ) = (48)
f rσ π a ∂a
where E ' is the elastic modulus, ur (a, x) is the displacement of the crack face relative to the
plane of symmetry, and fr is the reference SIF. To apply this relation, the distributions of the
applied load have to be calculated from the relation between the load and the crack-opening
displacement. For convenience of calculation, normalization is applied to crack length a,
coordinates x, and crack-opening displacement u with respect to s. That is
a x
a= and x =
s s
4.5 FORMATION OF EFFECTIVE SIF.
As FMLs are subjected to cyclic load, crack propagation is driven by the far-field applied load
and retarded by the bridging stress on the delamination boundary. For a crack to continue to
propagate, the driving force needs to overcome residual strength-induced crack closure while the
crack growth is bridged by the bridging mechanism. The effective SIFs of FMLs at the crack
tips need to incorporate the effects from applied load, crack closure, and bridging stresses. Also,
the presence of residual stress in FMLs needs to be taken into account in fatigue crack growth in
terms of the crack opening stress.
Following the definition of effective SIF used for FMLs in reference 34, the effective SIF of an
FML can be expressed as
K tip = ( K I , Al − K op , Al ) f eff (49)
where KI,Al is the mode I SIF in the Al layer, Kop,Al is the crack opening SIF in the Al layer, and
feff is the effective nondimensional SIF. The effective nondimensional SIF is defined as
f eff = ( f o − f br ) = f o (1 − β f ) (50)
and
fo
βf = (51)
f br
32
where fo is attributed to the unit applied load, and fbr is attributed to the corresponding bridging
stress. The nondimensional βf is the fiber-bridging factor. Equating these equations, the
effective SIF at the crack tips can be rewritten as
K tip = ( K I , Al − K op , Al )(1 − β f ) f o (52)
A nondimensional fiber-bridging factor, βf, of zero is the crack growth behavior for monolithic
Al alloy, and fo is a geometrical correction factor. As the bridging effect increases, as described
in βf, the effective SIF is modulated by the fiber-bridging factor. As a result, the crack growth
rate is reduced by the fiber-bridging mechanism in FMLs. If the nondimensional bridging factor,
βf, is equal to one, then the crack growth in FMLs is fully bridged. Each item in the effective
SIF will be discussed in the following sections.
4.5.1 Far-Field SIF.
For specimens subjected to unidirectional fatigue load, the crack path in the metal layer is
perpendicular to the loading direction, only mode I SIF is considered. Taking into account the
effect of bending and applied stress in the metal layer, superposition is used to calculate the
actual stress level in the metal layer. Therefore, the model I SIF in the Al layer is expressed
as [39]
K I , Al = (σ Al + σ Alb ) π a (53)
where σAl is the stress level in the Al layer, σbAl is the stress in the Al layer caused by secondary
bending effect, and a is the total crack length in FMLs. If there is no secondary bending, then
the bending stress is zero. Different scenarios of crack interaction due to the presence of MSD in
Al alloy can be found in reference 12.
4.5.2 Crack-Opening SIF.
The crack-opening SIF is expressed as [21]
K op = σ Al ,op πa (54)
where σAl,op is the crack-opening stress in the Al layer of FMLs. It should be noted that the
empirical crack-opening stress is a function of stress ratio. The stress ratio used to obtain the
crack-opening stress is the actual stress ratio in the metal layer of FMLs. In other words, the
actual stress ratio in the Al layer is applied to obtain the crack-opening stress in Al alloy.
The crack-opening stress for the Al alloy was plotted as a function of stress ratio and is shown in
figure 21 [43]. A plain stress condition (constraint factor c = 1) is considered due to the thin
metal layer used in GLARE laminates.
33
1.0
0.9 c=1
c=1.5
0.8
c=2
0.7
0.6
σ op 0.5
σ max 0.4
0.3
0.2
0.1
0.0
-1 0 1
R
Figure 21. Crack-Opening Stress as a Function of Stress Ratio [43]
4.5.3 Nondimensional SIF.
Analytically, the nondimensional SIFs, fo, is expressed as [42]
a mo (a, x)
f o = ∫ σo ( x) dx (55)
ao
πa
where σn(x) is the stress on the crack face induced by the unit applied load and secondary
bending, and σ is the scaling factor. The weight function is expressed as
a
mo (a, x) = 2 (56)
π(a − x 2 )
2
and the stress on crack face is expressed as
σ Al x
σ o ( a, x ) = (57)
x 2 − ao2
34
The nondimensional SIF, fbr, is expressed as [42]
a mbr (a, x)
fbr = ∫ σ n ,br , Al ( x) dx (58)
0
πa
where σn,br,Al (x) is the bridging stress on the crack face of the metal layer in FMLs under a unit-
applied load. For a periodic array of collinear cracks in a sheet under applied load, the crack-
opening displacement leads to the following exact weight function [42]
2 πa
tan( )
πx πa 2
mbr (a, x) = cos( ) (59)
2 sin 2 ( ) − sin 2 ( πx )
πa
2 2
The dimensionless bridging SIF was plotted as a function of crack length and is shown in figure
22. Clearly, there is a transition in the dimensionless bridging factor. Physically, it indicates
that the influence of the fiber-bridging effect would lower the SIF, leading to an approximately
steady-state crack growth after a certain number of cycles. Compared to the far-field and
bridging SIFs, the effective SIF was plotted as a function of crack length and is shown in
figure 23.
35
Nondimensional SIF
(a)
Nondimensional SIF
(b)
Figure 22. Effective Nondimensional SIF (feff = fo − fbr) as a Function of Crack Length
(a) Through-Thickness and (b) Partial-Thickness
36
(a)
(b)
Figure 23. Effective SIF for GLARE 3-3/2 Specimens With Multiple Surface Cracks Under the
Applied Stresses of 120 MPa With an Applied Stress of 0.05 (a) Through-Thickness and
(b) Partial-Thickness
37
4.6 FATIGUE CRACK GROWTH PREDICTION.
In GLARE laminates, the constituent metal layer is Al 2024-T3, and it is assumed that the
fatigue law can be used for predicting crack growth rate (da). The empirical Paris-Walker
fatigue equation (mm/cycle) is expressed as [36]
da ng 1
= C g1 ⎡⎣ ΔK tip ⎤⎦ (60)
dN
The constituent metal of hybrid FMLs is Al 2024-T3, which corresponds to

C g = 1.27 × 10−11 and ng = 2.94 . The effective SIF is in the unit of MPa mm .
4.7 FATIGUE DELAMINATION GROWTH PREDICTION.
For FMLs, the delamination growth (mm/cycle) (db) can be expressed as a function of the strain
energy release rate [44]. That is
db
− Cb (Δ g ) nb = Cb ( g max − g min ) nb (61)
dN
The GLARE empirical coefficients Cb and nb are 0.05 and 7.5, respectively [44], and the strain
energy release rate is expressed as [44 and 45]
σ2fml E f ,0 E f ,90
g= [ γ 2 (nAl − 1) − λ 2 nAl t Al + n f ,0t f ,0 ( γ 2 − λ 2 ) + n f ,90t f ,90 ( γ 2 − λ 2 )] (62)
2ni E Al E Al E Al
where ni is the number of interfaces, and
t fml
γ= (63)
E f ,0 E f ,90
(ni − 1)t Al + n f ,0t f ,0 + n f ,90t f ,90
E Al E Al
t fml
λ= (64)
E f ,0 E f ,90
nAl t Al + n f ,0t f ,0 + n f ,90t f ,90
E Al E Al
As observed in this study, there is a mixed-mode involvement of delamination growth near the
stage of delamination linkup. In this study, only mode II delamination growth was considered.
Also, it should be noted that the secondary bending is not considered in delamination growth.
38
5. TEST RESULTS.
5.1 MULTIPLE-IMPACT BEHAVIOR OF GLARE 5-2/1 LAMINATES.
The multiple-impact loads were applied to GLARE 5-2/1 by the weight drop. The impact loads
were repeated twice to observe the accumulated damage in GLARE 5-2/1 laminates. Three
levels of impact energy were set in the experiment: 8 J, 16 J, and 26 J, respectively. If there was
immediate metal cracking after weight drop, the impact energy would not increase. The
maximum impact energy applied was 26 J, which would cause a permanent metal opening in the
GLARE 5-2/1 laminates.
Figure 24(a) and (b) show multiple-impact dent damage (which is BVID) in Al 2024-T3 and
GLARE 5-2/1 in the top and bottom figures, respectively. The crack multiple-impact damage
(which is CVID) in Al 2024-T3 and GLARE 5-2/1 is shown in figure 24(c). The CVID in
GLARE 5-2/1 occurred in the outer Al layer on the nonimpacted side along the 0° and 90° fiber
direction.
Figure 24. Multiple-Impact Damage of Al 2024-T3 (Top) and GLARE 5-2/1 (Bottom) at
(a) BVID (8 J (2x4 J)), (b) BVID (16 J (2x8 J)), and (c) CVID (26 J (2x13 J))
Figure 25 shows experimental results for the load-time history of Al 2024-T3 and GLARE 5-2/1.
After the first impact of 8 J and 16 J, Al 2024-T3 and GLARE 5-2/1 experience the plastic
deformation. After the second impact for Al 2024-T3 and GLARE 5-2/1, 8 J shows the same
trend as the first impact. However, the 16-J GLARE 5-2/1 shows a small, sharp load drop. This
discrete load drop is believed to indicate the delamination and failure of the composite layer [4].
At a multiple-impact energy of 26 J, after the second impact in Al 2024-T3 and GLARE 5-2/1,
the load was dramatically reduced, indicating the occurrence of CVID. The specific energy to
create a dent and crack damage in the outer Al layer for Al 2024-T3 is somewhat higher than
GLARE 5-2/1. It comes from the effects of thickness and minimum cracking energy. As shown
39
in reference 4, the minimum cracking energy of Al 2024-T3 is somewhat higher than GLARE 5-
2/1. As the relationship of the area density and impact energy is considered, GLARE 5-2/1
shows a better impact performance than Al 2024-T3.
Multiple-Impact Energy: E = 8 J (2x4 J)

5
Al2024: First impact
Al2024: Second impact
Impact Force (kN)

4 GLARE 5-2/1: First impact
GLARE 5-2/1: Second impact
0
0 1 2 3 4 5 6 7 8
Time (msec)
(a)
7
Al 2024: First impact
6 Al 2024: Second impact
Impact Force (kN)
GLARE 5-2/1: First impact

5 GLARE 5-2/1: Second impact
0
0 1 2 3 4 5 6 7
Time (msec)
(b)
8
7 Al 2024: Second impact
Impact Force (kN)

6
GLARE 5-2/1: Second impact
5
0
0 1 2 3 4 5 6 7
Time (msec)
(c)
Figure 25. Load-Time Histories of Al 2024-T3 and GLARE 5-2/1 Under Multiple Impact With
(a) Impact Energy = 8 J, (b) Impact Energy = 16 J, and (c) Impact Energy = 26 J
40
Figure 26 shows the permanent central displacement of Al 2024-T3 and GLARE 5-2/1 as a
function of impact energy after low-velocity impact. After the first impact, GLARE 5-2/1 has
almost the same dent depth as Al 2024-T3. However, after the second impact, GLARE 5-2/1
shows a larger dent depth than Al 2024-T3. The total dent depth may be affected by the dent
depth after the second impact. Finally, the difference of total dent depth between GLARE 5-2/1
and Al 2024-T3 is 10%.
12
Al 2024: Second impact
10
Permanent Deflection (mm)
Al 2024: Total deflection

8 GLARE 5-2/1: Second impact

GLARE 5-2/1: Total deflection
0
0 5 10 15 20 25 30
Impact Energy (J)
Figure 26. The Permanent Central Deflection as a Function of Impact Energy
Figure 27 shows the multiple-site impact damage fatigue crack growth on a variety of GLARE
5-2/1 specimens at different fatigue cycles. Cracks may initiate from the edge or from the
damage area and travel along the crack path. Crack lengths were carefully measured at specific
fatigue cycles.
Figures 28-31 show the crack length as a function of cycles under different applied stresses. The
applied stresses corresponded to the postimpact tensile strength. In this experiment, the site
location is indicated by the number. The letters “R” and “L” indicates whether the fatigue crack
is on the right or left of the impact site. The letters “d” and “e” represent the damage area and
edge crack initiation sites, respectively. For example, “1Le” represents the fatigue crack that is
located on the left-hand side of impact 1, and this crack initiates from the edge of the sample.
Table 7 is a summary of the crack growth data.
41
Figure 27. Multiple-Site Impact Damage Fatigue Crack Propagation
Crack Length a (mm)
Figure 28. Multiple-Site Impact Damage Fatigue Crack Growth (116.2 MPa)
42
Crack Length a (mm)
Figure 29. Multiple-Site Impact Damage Fatigue Crack Growth

Crack Length a (mm)
43
Crack Length a (mm)
Table 7. Multiple-Site Impact Damage Fatigue Crack Growth Data
No./Cycle 1E 1L 1R 2L 2R 2E 3E 3L 3R 4L 4R 4E
55202 14.34 0 6.93 0 0 7.09 13.67 4.27 3.49 3.49 11.04 12.02
60548 20.72 0 16.93 5.11 0.75 11.64 17.03 7.79 13.68 13.27 11.88 12.72
Unit of measurement: mm
5.2 MULTIPLE-SITE DAMAGE FATIGUE BEHAVIOR OF GLARE 3-3/2 LAMINATES.
To study MSD fatigue crack growth behavior, two scenarios were considered in GLARE 3-3/2
laminates. (1) Through-thickness crack growth, which means that all cracks will initiate and
propagate in all the metal layers of FMLs. (2) The partial-thickness crack growth, which means
that cracks only initiate and propagate in the surface metal layer. For surface crack growth in
FMLs, due to the unsymmetry of crack growth, the secondary bending effect needs to be
addressed properly. All specimens were tested under tension-tension, constant-amplitude fatigue
loading.
44
5.2.1 Through-Thickness Crack Growth.
In the presence of a through-thickness starter notch, the cracks initiated from the notch tip in the
direction perpendicular to the applied load in all cases, as shown in figure 32. Cracks initially
grew away in the stable growth phase from the open hole/start notch at zero angles and then
continued to grow without any change in crack trajectories, i.e., a mode I SIF-dominant crack
growth in the metal layer. As two cracks propagated toward each other and reached the region
of linkup, they bypassed each other, leading to a formation of an eye-shaped region, instead of
having a direct crack linkup (figure 32). At this juncture, the crack growth was unstable, and the
crack path angle changed toward the opposite crack trajectory. The crack path deviated and did
not remain 0°. This phenomenon is explained as two propagating cracks in the vicinity of a
crack linkup mutually influencing each other, and the crack trajectory is determined by a
function of mode I and mode II SIF.
Figure 32. On-Going MSD Fatigue Cracks in GLARE 3-3/2 (a) Crack Growth in the Surface
Metal Layer and (b) Crack Growth in the Inner Metal Layer (A close-up of the crack bypass
is shown on the right.)
Furthermore, it was observed that the crack length in the inner metal layer was almost identical
to the crack length in the outer metal layer, as shown in figure 32(b), after removing the surface
metal layers. Similarly, an eye-shaped region was observed in the inner Al layer of GLARE 3-
3/2 laminates. Compared to specimens with a single through-thickness notch, this observation
was different since the direction of the crack path remained the same as the cracks propagated
through the metal layers of the laminate.
Meanwhile, the delamination zones in the fiber/adhesive layer appeared to avoid linkup as
they grew toward each other, as shown in figure 33. The stable delamination growth is a
45
mode II-dominant fracture behavior; however, it became a mixed mode (mode I and mode II)
delamination behavior as two delaminations approached the region of linkup and deviated from
the original center line. In other words, the shear stress induced by cyclic fatigue loading did not
affect the direction of the mode II-dominant delamination growth until it fell into the eye-shaped
crack formation region. Ultimately, the two propagating delaminations emerged to be
approximately a rectangular shape. A brief summary of the experimental results are listed in
table 8.
Figure 33. The MSD Fatigue Delamination Pattern in the Fiber/Adhesive Layer at Cracks
1B and 2B for GLARE 3-3/2
The crack lengths as a function of fatigue cycles for GLARE 3-3/2 specimens under maximum
applied stresses of 120 and 100 MPa were recorded through continuous monitoring during the
fatigue tests. For the ease of investigation in MSD crack growth, all the cracks in the upper and
bottom rows in table 8 were categorized as lead cracks and nonlead cracks, and the total average
crack length was used as a base line.
The multiple-site crack propagations as a function of fatigue cycles under different applied
stresses are shown in figures 34 and 35, respectively. The lead and nonlead cracks do not have
the same crack length, which indicates that the cracks interacted with one another. The crack
growth rates as a function of crack length were plotted and are shown in figures 36 and 37 for
the applied stresses of 120 and 100 MPa, respectively. The initial fast crack growth phase was
one in which insufficient fiber bridging occurred to transfer enough load from the Al sheet to
reduce the effective SIF at the crack tip. Initially, the crack growth rate decreased, and then
grew progressively. Unlike a single-notched specimen, the crack growth rate was approximately
a constant due to the absence of crack interaction. This correspondence to the fiber-bridging
mechanism is dependent on the delamination profiles proposed by Guo and Wu [34 and 35] and
the damage-tolerant characteristics of fiber-metal laminates. As the bridging SIF becomes
predominant, the crack growth reaches a steady state.
46
Table 8. Summary of Experimental Results for Through-Thickness GLARE 3-3/2 Laminates
Applied Stress Level = 120 MPa, Applied Stress Ratio = 0.05

Lead Crack
Starter Total Crack Averaged Crack (linkup)/
Panel Notch Length Final Cycles Growth Rates Nonlead
ID (mm) (mm) (N) (mm/cycle) Crack
T1 1 11.430 70,000 1.63E-04 Lead
T2 1 11.430 70,000 1.63E-04 Lead
T3 1 11.430 70,000 1.63E-04 Lead
T4 1 11.430 70,000 1.63E-04 Lead
B1 1 9.144 70,000 1.31E-04 Nonlead
B2 1 9.144 70,000 1.31E-04 Nonlead
B3 1 9.144 70,000 1.31E-04 Nonlead
B4 1 9.144 70,000 1.31E-04 Nonlead
T1 1 7.620 145,984 5.21975E-05 Nonlead
T2 1 7.620 145,984 5.21975E-05 Nonlead
T3 1 11.430 145,984 7.82963E-05 Lead
T4 1 11.430 145,984 7.82963E-05 Lead
B1 1 11.430 145,984 7.82963E-05 Lead
B2 1 11.430 145,984 7.82963E-05 Lead
B3 1 8.890 145,984 6.08971E-05 Nonlead
B4 1 8.890 145,984 6.08971E-05 Nonlead
47
Crack Growth a (mm)
N (Cycles)
Figure 34. Average Crack Length as a Function of Cycles for GLARE 3-3/2 Laminates at the
Applied Stress Level of 120 MPa
Crack Growth a (mm)
N (Cycles)
Figure 35. Average Crack Length as a Function of Cycles for GLARE 3-3/2 Laminates at the
Applied Stress Level of 100 MPa
48
Da/dN (mm/cycle)
Crack Growth a (mm)
Figure 36. Average Crack Growth Rate as a Function of Crack Length for GLARE 3-3/2
Laminates at the Applied Stress Level of 120 MPa
Da/dN (mm/cycle)
Crack Growth a (mm)
Figure 37. Average Crack Growth Rate as a Function of Crack Length for GLARE 3-3/2
Laminates at the Applied Stress Level of 100 MPa
The presence of multiple-site fatigue cracks would accelerate the crack growth rates, as shown in
figures 36 and 37. Therefore, as two propagating cracks approached each other, the crack
growth rates increased gradually due to the effect of crack interaction. Ultimately, the crack
growth rates reached peak values when the cracks linked up near the eye-shaped region.
49
In MSD crack growth, since the Al sheet had a higher modulus, it picked up more load than the
S2-glass/prepreg; this higher percentage of overall load translated to higher stress in the metal
layers to drive the cracks. Cracks began to propagate and then reached approximately constant.
Some cracks grew faster and some were relatively slower. Ultimately, the final leading crack
lengths were significantly greater than the average crack lengths.
5.2.2 Partial-Thickness Crack Growth.
To evaluate the fatigue behavior of multiple surface cracks, all cracks were divided into two
categories, leading cracks (lead crack) and nonleading (nonlead) cracks. Experimentally, it was
observed that with the presence of multiple surface slits under fatigue loading, cracks in all cases
initiated at locations on the edge of the surface slits and propagated approximately perpendicular
to the direction of the applied load. The cracks continued to grow away perpendicularly to the
loading direction. These cracks did not initiate at the same time nor did they propagate with the
same growth rates. Figure 38 shows the nonlead and lead cracks. As the propagating cracks
were approaching each other at the vicinity of the crack fronts, they had a direct crack linkup,
leading to a straight line formation instead of bypassing each other at the applied stress level of
120 MPa, as shown at the bottom of figure 38. In other words, mode I fracture behavior
dominates in surface crack growth as cracks linkup. Figure 39 shows the corresponding nonlead
delamination and lead delamination. Similarly, lead delaminations formed a direct linkup at the
vicinity of the crack tips at the applied stress of 120 MPa.
The surface crack lengths were measured during the fatigue loading. For the ease of comparing
the deviation of crack growth rates, the averaged lead crack and averaged nonlead crack growth
rates were used with the averaged total crack growth rate as a base line. A brief summary of the
experimental results are listed in table 9.
Figure 38. Crack Propagation in Surface-Cracked GLARE 3-3/2 at the Applied Stress Level of
120 MPa (Top: nonlead cracks and Bottom: lead cracks and crack linkups)
50
Figure 39. Delamination Growth in Surface-Cracked GLARE 3-3/2 Specimens, Nonlead
Delamination (top) and Leading Delamination and Linkup (bottom)
Table 9. Summary of Experimental Results for GLARE 3-3/2 Surface (Partial-Thickness)

Crack Growth

Total Crack Averaged Crack Lead Crack
Panel Half Slit Length Final Cycles Growth Rates (linkup)/Nonlead
T1 2 13.335 570,000 2.33947E-05 Lead
T2 2 12.828 570,000 2.25053E-05 Lead
T3 2 11.303 570,000 1.98298E-05 Nonlead
T4 2 10.922 570,000 1.91614E-05 Nonlead
B1 2 12.7))) 570,000 2.22807E-05 Nonlead
B2 2 10.922 570,000 1.91614E-05 Nonlead
B3 2 10.795 570,000 1.89386E-05 Nonlead
B4 2 9.652 570,000 1.69333E-05 Nonlead
51
Table 9. Summary of Experimental Results for GLARE 3-3/2 Surface (Partial-Thickness)
Crack Growth (Continued)

Total Crack Averaged Crack Lead Crack
Panel Half Slit Length Final Cycles Growth Rates (linkup)/Nonlead
T1 2 5.08 750,000 6.77333E-06 Nonlead
T2 2 3.556 750,000 4.74133E-06 Nonlead
T3 2 5.08 750,000 6.77333E-06 Nonlead
T4 2 3.302 750,000 4.40267E-06 Nonlead
B1 2 6.6096 750,000 8.8128E-06 Lead
B2 2 5.08 750,000 6.77333E-06 Nonlead
B3 2 2.794 750,000 3.72533E-06 Nonlead
B4 2 2.794 750,000 3.72533E-06 Nonlead
The averaged crack lengths as a function of fatigue cycles under the maximum applied stress of
120 and 100 MPa were plotted and are shown in figures 40 and 41, respectively. Under the same
fatigue cycles, the lead and nonlead cracks did not reach the same length, and crack growth
deviation existed. Especially for cracks propagated at the applied stress of 100 MPa, the crack
length deviation was obvious. The averaged crack growth rates as a function of crack length are
shown in figures 42 and 43 for the corresponding applied stress levels. The crack growth
deviation between the lead and nonlead cracks is small at the applied stress of 120 MPa, and the
crack growth deviation is obvious at the applied stress of 100 MPa.
In FMLs, crack growth rate decelerated once the crack reached a certain length. When the
cracked metal layer becomes more compliant, it will carry less load locally, which is the source
of fatigue crack growth retardation. That is, there is load transfer to the fiber layers through the
crack flanks and at the crack tips. Since the aluminum sheet has a higher modulus, it picks up
more load than the S2-glass/prepreg. This higher percentage of overall load translates to higher
stress in the metal layers to drive the cracks. On the experimental observation, surface crack
growth was much slower compared to through-thickness crack growth. The further reduced
crack growth rates in the surface-cracked GLARE laminates were attributed to the extent of
greater bridging traction, provided by the intact laminates, and to the extent of higher residual
strength. These surface-cracked specimens are more damage-tolerant than through-thickness
GLARE laminates.
52
Figure 40. The MSD Crack Length as a Function of Cycles for Surface-Cracked GLARE 3-3/2
at the Applied Stress of 120 MPa
Figure 41. The MSD Crack Length as a Function of Cycles for Surface-Cracked GLARE 3-3/2
at the Applied Stress of 100 MPa
53
Figure 42. The MSD Crack Growth Rate as a Function of Crack Length at the Applied Stress of
120 MPa for Surface-Cracked GLARE Specimen
Figure 43. The MSD Crack Growth Rate as a Function of Crack Length at the Applied Stress of
100 MPa for Surface-Cracked GLARE Specimen
54
6. COMPARISON OF TEST AND ANALYSIS OF MULTIPLE-IMPACT LOADS AND
FATIGUE CRACK GROWTH.
6.1 MULTIPLE-IMPACT LOADS.
6.1.1 Comparison of Experiment and FE Simulation.
In modeling multiple-site impacts on GLARE 5-2/1, the impact loads were repeated twice on the
same position of the laminates. The double-impact procedure is shown in figures 44 and 45 for
the first and second impact load, respectively. As the first impact load was applied to the
GLARE 5-2/1 laminates, the second impact load was immediately applied. This procedure is
identical to the experimental setup of multiple-impact loads. It was expected that the application
of the second impact load would cause the GLARE laminates to experience larger deformation
than the single impact case, as was observed in the experiments.
Figure 44. First Impact on GLARE 5-2/1
Literature proved that in the case of multiple-impact loads, the Al layer absorbs most of the
energy instead of the composite layers [3 and 4] for the first impact. If the impact energy is
relatively low for both 8-J and 16-J impact energy, the GLARE 5-2/1 Al layer experienced
plastic deformation to absorb the impact energy.
55
Figure 45. Second Impact on GLARE 5-2/1
Experimentally, the response of the impact force as a function of time was recorded
automatically. Numerically, a reference point was set on the head of the drop weight to extract
the response of the impact load. That is, the maximum impact response was used to compare to
the experimental data. As shown in figure 46, the peak impact force was plotted as a function of
time from the results of the 2D and 3D FE models, respectively. For ease of comparison, the
nonfailure mode was used as a base line. Two sets of impact energies were studied and applied
to GLARE 5-2/1 laminates, 8 J (2x4 J) and 16 J (2x8 J), respectively. In the FE simulation, only
the cases with low-velocity impact load were considered. Hence, the 26-J impact energy, which
would cause permanent metal cracking observed in experiments, was not within this scope of
this study.
For the 8-J impact energy case, the load-time response in simulation was similar in trend during
the first and second impact. This indicated that the failure criteria applied to the composite
layers in both the 2D and 3D FE models did not critically affect the stiffness degradation in the
GLARE laminates. Although the repeated load was applied to the GLARE laminates, only
plastic deformation in the metal layers would occur as long as the impact energy was low. This
can be proved by experimental observation or FE modeling. However, for the 16-J impact
energy case, experimentally, dent damage was observed. Following the second impact load,
damage accumulated and caused failure in both the aluminum and composite layers. In the FE
simulation, it was observed that composite layers began to degrade either in the 2D or 3D model
with Hashin failure criteria [20]. Clearly, both the 2D and 3D failure criteria models are capable
of capturing the impact response of the laminate materials. The prediction model with nonfailure
criteria would overestimate the impact force and time history compared to the experimental data
or to the FE model with failure criteria setup. Therefore, it is important that the failure criteria
be used to simulate damage progression in stiffness or compliance degradation.
56
Multiple Impact: E = 8 J (2x4 J)
6
Experiment-first impact
Experiment-second impact
No failure model-first impact
5
No failure model-second impact
Impact Force (kN) 2D failure model-first impact
2D failure model-second impact
4 3D failure model-first impact
0
0 1 2 3 4 5 6 7
Time (msec)
(a)
Multiple Impact: E = 16 J (2x8 J)

9
8
No failure model-first impact
No failure model-second impact
Impact Force (kN)

5
0
0 1 2 3 4 5 6 7
Time (msec)
(b)
Figure 46. Comparisons of Experimental and Simulation Results of Multiple-Impact, Force-

Time History for GLARE 5-2/1, Impact Energy (a) 8 J and (b) 16 J
57
Figure 47 shows the central displacement changes at different levels of impact energy. Because
of the fully fixed boundary condition, the FE model of GLARE 5-2/1 is less deflected than the
experiment after the first and second impact. There is no difference in the central displacement
in the 2D and 3D failure models.
Multiple Impact E = 8 J (2x4 J)

0
0 1 2 3 4 5 6 7
-1
Deflection (mm)
-2
-3 Experiment-first impact
2D failure model-first impact
-4 2D failure model-second impact
-5
Time(msec)
(a)
Multiple Impact E = 16 J (2x8 J)
0
0 1 2 3 4 5 6 7
-1
-2
Deflection (mm)
-3
-4
-5
-6 3D failure model-first impact
-7
Time (msec)
(b)
Figure 47. Difference Between Predicted and Measured Central Displacement for GLARE 5-2/1
Impact Energy (a) E = 8 J and (b) E = 16 J
58
6.1.2 Two- and Three-Dimensional Failure Criterion Comparison.
As mentioned, since the composite layer is so thin, 2D failure criterion works well under plane
stress assumption problems. Thus, the 3D stress tensor <σ13, σ23, σ33> and the strain tensor
<ε13, ε23, ε33> in the dynamic problem cannot be ignored because of three-directional large
deformation. Even though 2D failure criteria may work in composite layers using the continuum
shell element, 3D failure criteria, including three-directional stress and strain parts, is believed to
compensate for the weak points for impact simulation.
A series of damage failure patterns in the 2D and 3D FE models were put together for
comparison. These damage areas were compared based on the time step in simulation. Only
failed regions were marked as red and the others were marked as white. In other words, the
stress levels were purposely removed.
Figures 48 through 51 show the Hashin failure result in the GLARE 5-2/1 composite layers with
different fiber orientations [0°/90°/90°/0°] and different impact energies. The damage
progression induced by the first and second impact loads are also shown in these figures.
2D Failure Model 3D Failure Model
(a) (c)
Fiber tension failure at time: (a) 0.392 msec and (c) 0.392 msec
.
(b) (d)
Fiber tension failure at time: (b) 3.66 msec and (d) 3.66 msec
Figure 48. Failure Mode Comparison for 2D (left) and 3D (right) Models of GLARE 5-2/1 After
First Impact E = 8 J
59
.
(a) (c)
(b) (d)
Second Impact E = 8 J
60
(a) (c)
.
(b) (d)
First Impact E = 16 J
61
2D Failure model 3D Failure model
.
(a) (c)
(b) (d)
Second Impact E = 16 J
The 2D failure damage models are shown in figures 48(a) and (b), 49(a) and (b), 50(a) and (b),
and 51(a) and (b) at multiple-impact energy 8 and 16 J. The 3D failure criterion models are
shown in figures 48(c) and (d), 49(c) and (d), 50(c) and (d), and 51(c) and (d). The damage area
is realized by a state variable value greater than or equal to one that corresponds to a point at
which the failure criteria value has been exceeded in the GLARE 5-2/1 composite layer. As
shown in each figure, the FE model with 3D failure criteria clearly has a greater number of
integration points marked as damaged areas when compared to the 2D failure criteria model.
Figure 52 shows the damage state in the GLARE 5-2/1 metal layer with Hashin failure criteria in
composite layers. The differences between single- and multiple-impact loads are also shown in
figure 52. It is known that compressive stress is exhibited on the nonimpacted side of the
GLARE 5-2/1 specimens. The compressive stress on the impacted Al layer is small compared to
the nonimpacted side. The colored damage is realized by a state variable value greater than or
equal to one that corresponds to a point at which the failure criteria value has been exceeded.
The colored areas display the failure or damage of the Al metal layer. The 3D FE model with
failure criteria has more integration points that show damage initiation/progression when
compared to the previous 2D model with failure criteria.
62
Figure 52. The 3D ABAQUS Simulation of Multiple-Site Impact Damage on GLARE 5-2/1
6.2 MULTIPLE-SITE DAMAGE FATIGUE CRACK GROWTH.
6.2.1 Through-Thickness Test Specimens.
To predict the average crack growth rates, the proposed methodology was used. At the initial
stage, the crack growth of FMLs was faster due to the increasing fiber-bridging factor, followed
by a steady-state crack growth due to the flat-out of fiber-bridging factor. In most MSD, the
greatest concern is leading cracks; the predicted leading crack growth rates under different
applied stress levels were averaged and plotted, as shown in figures 53 and 54, respectively.
The predicted crack growth rates using the proposed methodology were calculated under
different applied stress levels. At the initial stage of crack growth, the far-field SIF dominates,
so the crack growth is relatively faster than in the steady-state region. As the bridging SIF
becomes predominant, the crack growth reaches a steady state. The presence of multiple-site
fatigue cracks would accelerate the crack growth rates as two cracks approach each other.
Therefore, as the propagating cracks approach each other, the crack growth rates increased
gradually due to the effect of crack interactions. Ultimately, the crack growth rates reached the
peak values when the cracks linked up. The predicted results under different maximum applied
stress are in good agreement with the experimental results.
63
Figure 53. Comparison of Predicted Averaged Leading Crack Growth Rates and Experimental
Results With an Applied Stress Ratio of 0.05 for GLARE 3-3/2 Through-Thickness Test
Specimens (120 MPa)
Results With an Applied Stress Ratio of 0.05 for GLARE 3-3/2 Through-Thickness Test
Specimens (100 MPa)
64
6.2.2 Partial-Thickness Test Specimens.
The averaged crack lengths as a function of fatigue cycles under the maximum applied stress of
120 and 100 MPa were studied. Under the same fatigue cycles, the lead and nonlead cracks did
not reach the same length, and crack growth deviation existed. Especially for cracks propagated
at the applied stress of 100 MPa, the crack length deviation was obvious.
In FMLs, crack growth rate decelerated once the crack reached a certain length. When the
cracked metal layer becomes more compliant, it will carry less load locally, which is the source
of fatigue crack growth retardation. That is, there is load transferring to the fiber layers through
the crack flanks and at the crack tips. Since the aluminum sheet has a higher modulus, it picks
up more load than the S2-glass/prepreg. This higher percentage of overall load translates to
higher stress in the metal layers to drive the cracks. On the experimental observation, surface
crack growth is much slower compared to through-thickness crack growth. The further reduced
crack growth rates in the surface-cracked GLARE laminates were attributed to the extent of
greater bridging traction, provided by the intact laminates, and to the extent of higher residual
strength. These surface-cracked specimens are more damage-tolerant than through-thickness
GLARE 3-3/2 laminates.
The predicted results of averaged leading surface crack growth for GLARE 3-3/2 laminate were
plotted in figures 55 and 56 at the applied stress levels of 120 and 100 MPa, respectively, along
with the experimental results. The predicted crack growth rates agreed well with the
experimental results of averaged crack growth rates, though discrepancy still existed for mainly
nonlead cracks.
Specimens (120 MPa)
65
Specimens (100 MPa)
7. CONCLUSIONS.
The low-velocity, multiple-impact behavior for 2024-T3 aluminum alloy and S2-glass fiber-
reinforced aluminum laminates (trade name GLARE) GLARE 5-2/1 was investigated through
experiment and finite element (FE) simulation. Three different multiple-impact energies of 8 J
(2x4 J), 16 J (2x8 J), and 26 J (2x13 J) were applied on aluminum alloy and GLARE 5-2/1 to
induce two different levels of damage (barely visible impact damage (BVID) and clearly visible
impact damage (CVID)). As in single-impact cases, the BVID of 8 J and 16 J introduced plastic
deformation dents on the outer aluminum layer. The CVID of 26 J was characterized by cracks
coming out from the plastic indentation in the fiber direction.
The FE analyses were performed on GLARE 5-2/1 for two different multiple-impact energies
(8 J and 16 J). The 26-J impact energy caused metal cracking and was not studied in FE
modeling in the scope of low-velocity multiple impacts. The FE simulation showed good
agreement with experiment for impact force-time history curve. A three-dimensional failure
criteria model was compared to a two-dimensional failure criteria model. Both failure models
correlated well with the experimental results. But in the impact force-time history curve case,
the two-dimensional failure model was better than the three-dimensional failure model to predict
peak impact force in multiple-impact behavior. The damage progression size in the three-
dimensional failure criteria model was bigger than the two-dimensional criteria failure model.
The FE simulation results were able to predict where the crack can occur in post-multiple-impact
fatigue behavior from the results of stress distribution on the impacted and nonimpacted
aluminum layer of GLARE laminates.
66
The crack growth behavior of a fiber-metal laminate with multiple-site damage (MSD) has been
investigated experimentally and analytically. To understand the surface crack growth in GLARE
3-3/2 laminates, the surface-bridging mechanism and the secondary bending effect were proven
through data analysis and experimental work, which supported the arguments. It was concluded
that in fiber-metal laminates with through-thickness MSD, as the fatigue cracks emanated from
the open holes and propagated, the crack growth rate was faster with the presence of MSD
cracks as compared to the case without the interaction of MSD cracks. The propagating cracks
tended to bypass each other and formed an eye-shaped region as two propagating cracks
approached each other before linkup, mode I fracture behavior dominates and results in crack
linkup at the vicinity of the crack tips. The crack growth acceleration was not observed at the
vicinity of crack linkup in experiments with multiple surface slits. As surface cracks linkup in
the metal layer, delaminations in the interface of metal/prepreg linkup with each other
simultaneously. The fatigue crack resistance in surface crack growth was better than in the
through-thickness crack growth. The proposed methodology for predicting the crack growth
rates of GLARE laminates with multiple-site fatigue damage was validated with experiments
under constant-amplitude fatigue load at a variety of stress levels.
8. REFERENCES.
1. Vogelesang, L.B., Marissen, R., and Schijve, J., “A New Fatigue Resistant Material:
Aramid Reinforced Aluminum Laminate (ARALL),” Proceedings of the 11th ICAF
Symposium, Noordwijkerhout, NLR, 1981, pp. 3.4/1-3.4/39.
2. Vlot, A. and Gunnink, J.W., eds., Fibre Metal Laminates-An Introduction, Kluwer
Academic Publisher, USA, 2001.
3. Wu, G. and Yang, J.-M., “The Mechanical Behavior of GLARE Laminates for Aircraft
Structures,” JOM, Vol. 57, No. 1, January 2005, pp.72-79.
4. Wu, G. and Yang, J.-M., “The Impact Properties and Damage Tolerance and of
Bidirectionally Reinforced Fiber Metal Laminates,” Journal of Material Science, Vol. 42,
No. 3, 2007, pp. 948-957.
5. Cantwell, W.J. and Morton, J., “The Impact Resistance of Composite Materials—A
Review,” Composites, Vol. 22, Issue 5, 1991, pp. 347-362.
6. Found, M.S. and Howard, I.C., “Single and Multiple Impact Behaviour of a CFRP
Laminate,” Composite Structures, Vol. 32, 1995, pp. 159-163.
7. Rotem, A., “The Strength of Laminated Composite Materials Under Repeated Impact
Loading,” Journal of Composites Technology and Research, Vol. 10, No. 2, 1988,
pp. 19-27.
8. Harris, B., Chen, A.S., Coleman, S.L., and Moore, R.J., “Residual Strength and
Toughness of Damaged Composites,” Journal of Material Science, Vol. 26, 1991, pp.
307-320.
67
9. Mittelman, A., “Low-Energy Repetitive Impact in Carbon-Epoxy Composites,” Journal
of Material Science, Vol. 27, 1992, pp. 2458-2462.
10. Jones, R., “Residual Strength of Composites With Multiple Impact Damage,” Composite
Structures, Vol. 28, 1994, pp. 347-356.
11. Al-Hassani, S.T.S., “Numerical Simulation of Multiple Shot Impact,” The 7th
International Conference on Shot Peening, Institute of Precision Mechanics, Warsaw,
Poland, 1999.
12. Partl, O. and Schijve, J., “Multiple-Site Damage in 2024-T3 Alloy Sheet,” International
Journal of Fatigue, Vol. 15, No. 4, 1993, pp. 293-299.
13. Jones, R., Molent, L., and Pitt, S., “Study of Multi-Site Damage of Fuselage Lap Joints,”
Theoretical and Applied Fracture Mechanics, Vol. 32, 1999, pp. 81-100.
14. Park, J.H., Singh, R., Pyo, C.R., and Atluri, S.N., “Integrity of Aircraft Structural
Elements With Multi-Site Fatigue Damage,” Engineering Fracture Mechanics, Vol. 51
Issue 3, 1995, pp. 361-380.
15. Schijve, J., “Multiple-Site Damage in Aircraft Fuselage Structures,” Fatigue Fracture
Engineering Material Structure, Vol. 18, No. 3, 1995, pp. 329-344.
16. Pitt, S. and Jones, R., “Multiple-Site and Widespread Fatigue Damage in Aging
Aircraft,” Engineering Failure Analysis, Vol. 4, No. 4, 1997, pp. 237-257.
17. Schijve, J., “Fatigue of Aircraft Materials and Structures,” International Journal of
Fatigue, Vol. 16, No. 1, 1994, pp. 21-32.
18. Schijve, J., “Prediction of Fatigue,” Japan Society of Mechanical Engineers International
Journal Series I—Solid Mechanics, Strength of Materials, Vol. 34, No. 3, 1991, pp.
269-280.
19. Alderliesten, R.C. and Homan, J.J., “Fatigue Crack Growth Behaviour of Surface Cracks
in GLARE, Advances in Damage Mechanics,” Fatigue Damage of Materials:
Experiment and Analysis, Wessex Institute of Technology Press, United Kingdom, 2003,
pp. 213-222.
20. Schijve, J., Campoli, G., and Monaco, A., “Fatigue of Structures and Secondary Bending
in Structural Elements,” International Journal of Fatigue, Vol. 31, 2009, pp.1111–1123.
21. Guo, Y.J. and Wu, X.R., “A Phenomenological Model for Predicting Crack Growth in
Fiber-Reinforced Metal Laminates under Constant-Amplitude Loading,” Composites
Science and Technology, Vol. 59, 1999, pp. 1825–31.
68
22. ASTM D 3479, “Standard Test Method for Tension-Tension Fatigue of Polymer Matrix
Composite Materials,” American Society for Testing and Materials (ASTM) Standards
and Literature References for Composite Materials, ASTM, pp. 67-70.
23. “A Finite Element System,” ABAQUS User Manual, Ver. 6.7, Dassault Systems Inc.,
2008.
24. Linde, P., Pleitner, J., De Boer, H., and Carmone, C., “Modeling and Simulation of Fiber
Metal Laminates,” ABAQUS User’s Conference, Boston, Massachusetts, 2004.
25. Hashin, Z., “Failure Criteria for Unidirectional Fiber Composites,” Journal of Applied
Mechanics, Vol. 47, 1980, pp. 329-335.
26. DeVries, T.J., “Blunt and Sharp Notch Behavior of GLARE Laminate,” Ph.D.
Dissertation, Delft University Press, 2001.
27. Hengenbeek, M., VanHengel, C., Bosker, O.J., and Vermeerne, C.A.J.R., “Static
Properties of Fibre Metal Laminates,” Applied Composite Materials, Vol. 10, 2003, pp.
207-222.
28. Nguyen, M.Q., Elder, D.J., Bayandor, J., Thomson, R.S., and Scott, M.L., “A Review of
Explicit Finite Element Software for Composite Impact Analysis,” Journal of Composite
Materials, Vol. 39, No. 4, 2005, pp. 375-386.
29. Matzenmiller, A., Lubliner, J., and Taylor, R.L., “A Constitutive Model for Anisotropic
Damage in Fiber Composites,” Mechanics of Materials, Vol. 20, No. 2, 1995, pp.
125-152.
30. Iannucci, L., Dechaene, R., Willows, M., and Degrieck, J., “A Failure Model for the
Analysis of Thin Woven Glass Composite Structures Under Impact Loadings,” Computer
and Structures, Vol. 79, 2001, pp. 785-799.
31. Chang, P.Y., Yeh, P.C., and Yang, J.M., “Fatigue Crack Initiation in Hybrid
Boron/Glass/Aluminum Fiber Metal Laminates,” Materials Science and Engineering A,
Vol. 496, 2008, pp. 273–280.
32. Chang, P.Y., Yang, J.M., Seo, H., and Hahn, H.T., “Off-Axis Fatigue Cracking
Behaviour of Notched Fiber Metal Laminates,” Fatigue and Fracture of Engineering
Materials and Structures, Vol. 30, 2007, pp.158-1171.
33. Schijve, J., Campoli, G., and Monaco, A., “Fatigue of Structures and Secondary Bending
in Structural Elements,” International Journal of Fatigue, Vol. 31, 2009, pp.1111–1123.
34. Guo, Y.J. and Wu, X.R., “A Theoretical Model for Predicting Fatigue Crack Growth
Rates in Fibre-Reinforced Metal Laminates,” Fatigue Fracture Engineering Material
Structure, Vol. 21, 1998, pp. 1133-45.
69
35. Guo, Y.J. and Wu, X.R., “Bridging Stress Distribution in Center-Cracked Fiber
Reinforced Metal Laminates: Modeling and Experiment,” Engineering Fracture
Mechanics, Vol. 63, 1999, pp. 147-63.
36. Alderliesten, R.C., “Analytical Prediction Model for Fatigue Crack Propagation and
Delamination Growth in GLARE,” International Journal of Fatigue, Vol. 29, 2007, pp.
628-646.
37. Chang, P.Y. and Yang, J.M., “Modeling of Fatigue Crack Growth in Notched Fiber
Metal Laminates,” International Journal of Fatigue, Vol. 30, 2008, pp. 2165–2174.
38. Chang, P.Y., Yeh, P.C., and Yang, J.M., “Fatigue Crack Growth in Hybrid
Boron/Glass/Aluminum Fibre Metal Laminates,” Fatigue Fracture Engineering Material
Structure, Vol. 30, 2008, pp. 989-1003.
39. Tada, H., Paris, P.C., and Irwin, G.R., The Stress Analysis of Cracks Handbook, 3rd ed.,
The American Society of Mechanical Engineers, 2000.
40. Alderliesten, R.C., “Fatigue Crack Propagation and Delamination Growth in GLARE,”
Ph.D. Thesis, Delft University of Technology, Delft, Netherlands, 2005.
41. Alderliesten, R.C., Campoli, G., and Benedictus, R., “Modeling Cyclic Shear
Deformation of Fibre/Epoxy Layers in Fibre Metal Laminates,” Composites Science and
Technology, Vol. 67, Issues 11-12, 2007, pp. 2545-2555.
42. Wu, X.R. and Carlsson, A.J., “Weight Functions and Stress Intensity Factor Solutions,”
Pergamon Press, 1991.
43. Newman, J.C., “A Crack Opening Stress Equation for Fatigue Crack Growth,”
International Journal of Fracture, Vol. 24, 1984, pp. R131-R135.
44. Alderliesten, R.C., “Application of the Energy Release Rate Approach for Delamination
Growth in GLARE,” Engineering Fracture Mechanics, Vol. 73, No. 6, 2006, pp. 697-
709.
45. Suiker, A.S.J. and Fleck, N.A., “Crack Tunneling and Plane-Strain Delamination in
Layered Solids,” International Journal of Fracture, Vol. 125, 2004, pp. 1-32.
70

Damage Tolerance and Durability of Fiber Metal Laminates For Aircraft Strucrtures-FAA-2010

Uploaded by

Copyright:

Available Formats

Damage Tolerance and Durability of Fiber Metal Laminates For Aircraft Strucrtures-FAA-2010

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Damage Tolerance and Durability of Fiber Metal Laminates For Aircraft Strucrtures-FAA-2010

Uploaded by

Copyright:

Available Formats

DOT/FAA/AR-10/18

Damage Tolerance and Durability

This document is available to the U.S. public

This document is also available from the Federal Aviation

U.S. Department of Transportation

This document is disseminated under the sponsorship of the U.S.

This report is available at the Federal Aviation Administration William J.

7. Author(s) 8. Performing Organization Report No.

EXECUTIVE SUMMARY xiii

2. TEST PROGRAM AND TEST MATRIX 4

2.1 Multiple-Impact Load Test 4

2.2 Multiple-Site Damage Fatigue Test 7

3. ABAQUS FE SIMULATION FOR FMLs UNDER MULTIPLE-IMPACT

3.1 General Development of FE Modeling 10

3.2.1 Three-Dimensional Modeling Rationale 13

3.2.2 Three-Dimensional Progressive Damage Model for

3.2.3 Energy Dissipation Due to Damage 18

4. ANALYTICAL PREDICTION MODEL OF MSD FATIGUE CRACK

4.1 Classic Lamination Theory 22

4.5.1 Far-Field SIF 33

5.1 Multiple-Impact Behavior of GLARE 5-2/1 Laminates 39

5.2.1 Through-Thickness Crack Growth 45

6. COMPARISON OF TEST AND ANALYSIS OF MULTIPLE-IMPACT LOADS

6.1 Multiple-Impact Loads 55

6.1.1 Comparison of Experiment and FE Simulation 55

6.2 Multiple-Site Damage Fatigue Crack Growth 63

6.2.1 Through-Thickness Test Specimens 63

1 A Cross-Ply GLARE 3-3/2 Laminate 2

2 Bridging Mechanism in FMLs With Cracked Metals and Delamination 2

3 Schematic and Cross-Sectional Views of GLARE 5-2/1 4

4 Multiple-Site Impact Damage Specimens 5

5 Dynatup Model 8250 Instrumented Drop Weight Impact Tower 6

6 Multiple-Site Impact Damage Fatigue Test 6

7 Configuration of GLARE 3-3/2 Laminate Specimens (a) Through-Thickness MSD

8 Cross-Sectional View of GLARE 3-3/2 9

10 Continuum Shell Element Normal Vector Relation to Direction of Impact 14

12 Composite Progressive Damage VUMAT Flowchart for the ABAQUS/Explicit

13 General Schematic for Energy Dissipation of a Continuum Body 19

14 Flowchart of Crack Growth Prediction Approach for GLARE Laminates 21

17 Deflection of the Neutral Line in GLARE Laminates Along Loading Direction 26

18 Bending Stress Distribution in Metal Layer of GLARE Laminates Along Loading

21 Crack-Opening Stress as a Function of Stress Ratio 34

22 Effective Nondimensional SIF ( f eff = f o − fbr ) as a Function of Crack Length

24 Multiple-Impact Damage of Al 2024-T3 and GLARE 5-2/1 at BVID

26 The Permanent Central Deflection as a Function of Impact Energy 41

27 Multiple-Site Impact Damage Fatigue Crack Propagation 42

28 Multiple-Site Impact Damage Fatigue Crack Growth (116.2 MPa) 42

29 Multiple-Site Impact Damage Fatigue Crack Growth 43

30 Multiple-Site Impact Damage Fatigue Crack Growth (174.3 MPa) 43

31 Multiple-Site Impact Damage Fatigue Crack Growth (174.3 MPa) 44

33 The MSD Fatigue Delamination Pattern in the Fiber/Adhesive Layer at Cracks

38 Crack Propagation in Surface-Cracked GLARE 3-3/2 at the Applied Stress Level

39 Delamination Growth in Surface-Cracked GLARE 3-3/2 Specimens, Nonlead

44 First Impact on GLARE 5-2/1 55

45 Second Impact on GLARE 5-2/1 56

46 Comparisons of Experimental and Simulation Results of Multiple-Impact,

47 Difference Between Predicted and Measured Central Displacement for

48 Failure Mode Comparison for 2D and 3D Models of GLARE 5-2/1 After

49 Failure Mode Comparison for 2D and 3D Models of GLARE 5-2/1 After