Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
by
Auburn, Alabama
May 9, 2011
Approved by
foam core architectures are studied. The syntactic foam core is made by dispersing
hollow glass microballoons in epoxy matrix at different volume fractions ranging from
20%-40%. The face sheets of the sandwich structures are made of thin AL6061 sheets.
Three types of sandwich structures, first one with a regular syntactic foam core
(identified as IPC), and third one with a syntactic foam core with graded face sheet
(identified as SFS-b) are studied. Static three-point bend tests are carried out on all three
types of sandwich structures and load-deflection responses are measured. The three
failure, nonlinearity of the flexural response, and strain energy absorbed. The global
map 2D deformations and strains. The measured normal and shear strain fields and
optical microscopy are used to discern failure mechanisms of the three architectures. The
SFS sandwiches fail predominantly due to face-core debonding with a large scatter in the
peak load and deflection at failure. They also show very limited nonlinearity in their
the deflection at failure with a slight reduction in peak load and significant nonlinearity in
ii
debonding between the phases of the interpenetrating core as well as face sheet yielding.
The SFS-b sandwiches, on the other hand, show substantial improvement in both peak
load and deflection at failure along with significant nonlinearity. This is attributed to a
gradual shear strain variation at the graded face-core interface, unlike the SFS
The superior material under quasi-static loading based on load, deflection, and
energy metrics was the syntactic foam core sandwiches with graded face sheets (SFS-b).
The only time it might not be considered the optimal choice might be when considering
the IPC core sandwiches for their ability to use the metallic foam network to hold the
Low velocity impact tests were also conducted on select IPC and SFS-b
case, as well as initiate dynamic characterization research. Also, the feasibility of using
DIC method in conjunction with high-speed digital photography to study impact behavior
of sandwich structures is illustrated. Contrary to the quasi-static case, IPC foam core
absorption. The presence of the interpenetrating phases seems to have a positive effect
iii
Acknowledgements
I would first of all like to thank my advisor, Dr. Hareesh Tippur, for his expertise
and guidance throughout my graduate and undergraduate careers. His style of managing
his graduate students allowed me the freedom to complete this work in a manner that was
satisfactory to us both. Thanks also to Dr. Jeffrey Suhling and Dr. Ruel (Tony) Overfelt
for their input and willingness to serve as my committee members. I would like to thank
the U.S. Army Research Office for funding this work through grants W911NF-08-1-0285
SMART Scholarship. This wonderful program provided me with both generous funding
for my graduate education and a civil service job in the Air Force once my degree is
My time in the Failure Characterization and Optical Techniques Lab under Dr.
Tippur was aided and made easier by my colleagues. Thanks to Dr. Dongyeon Lee,
Kailash Jajam, Chandru Periasamy, Vinod Kushvaha, Robert Bedsole, and Rahul Jhaver
for their help along the way and for helping make the day to day life in the lab a
Finally, thanks to my family and friends. Your unwavering support and patience
were both crucial and greatly appreciated. As Warren Buffet said, “Beware of geeks
bearing formulas.”
iv
Table of Contents
Abstract............................................................................................................................ ii
Acknowledgements......................................................................................................... iv
List of Tables.................................................................................................................. xv
1. Introduction...................................................................................................... 1
1.6 Objectives............................................................................................................ 14
v
2.2.2 Syntactic Foam Core Specimen Preparation............................................. 20
2.2.4 Syntactic Foam Core with Graded Face Sheets Specimen Preparation..... 24
3.3 Summary............................................................................................................. 46
4.3 Summary............................................................................................................. 69
5.3 Summary............................................................................................................. 91
vi
6.4.3 Syntactic Foam (SFS-b) Results.............................................................. 123
7. Conclusions............................................................................................................. 144
References..................................................................................................................... 156
vii
List of Figures
Figure 1.4: (a) Cancellous Bone and (b) Duocel® Aluminum Foam…………….. 3
Figure 2.6: Single Cast SFS-b Mold: (1) Mold, (2) AL6061 Face Sheet, and (3)
Brazed Aluminum Foam …………………………………………….. 24
Figure 2.7: Syntactic Foam Core Sandwich Structure with Graded Face Sheets
(SFS-b)……………………………………………………………….. 25
viii
Figure 3.4: Overlay of Load-Deflection Data for 20ppi Aluminum Foam Core
Sandwich……………………………………………………………… 30
Figure 3.10: Normalized Measured Peak Load for Sandwich Structures with SF
Foam Core with Different Volume Fractions…………………………. 37
Figure 3.12: Energy Absorbed for Sandwich Structures with SF Foam Core……… 40
Figure 3.13: Specific Energy Absorbed for Sandwich Structures with SF Foam
Core…………………………………………………………………… 41
Figure 3.14: (a) Failed Syntactic Foam Core Sandwich with (b) Highlighted
Face-core Debond …………………………………………………….. 42
Figure 4.2: Overlay of Measured Load-Deflection Data for Three Different IPC
Core Sandwich Beams (IPC20)……………………………………….. 50
ix
Figure 4.3: Overlay of Measured Load-Deflection Data for Three Different IPC
Core Sandwich Beams (IPC30)……………………………………….. 50
Figure 4.4: Overlay of Measured Load-Deflection Data for Two Different IPC
Core Sandwich Beams (IPC40)……………………………………….. 51
Figure 4.5: Typical IPC Load-Deflection Curve: Point A – Transition Point, Point
B – Crack Formation, Point C – Loss of Load Bearing Capacity ……. 53
Figure 4.10: Normalized Measured Peak Load for Sandwich Structures with IPC
Foam Core with Different Volume Fraction of SF in the IPC ……….. 57
Figure 4.14: Energy Absorbed for Sandwich Structures with IPC Foam Core…….. 61
Figure 4.15: Specific Energy Absorbed for Sandwich Structures with IPC Foam
Core…………………………………………………………………… 63
Figure 4.16: (a) Failed IPC Foam Core Sandwich with (b) Yielded Bottom Face... 64
Figure 4.17: Magnified Images of Crack Surfaces Near the Lower Face Sheet…... 65
x
Figure 4.22: Overlay of Measured Load-Displacement of Sandwich Structures with
IPC Foam Core with Different Vf of Microballoons in the SF……….. 69
Figure 4.23: Illustration of IPC (a) Before and (b) After Interphase Separation
between Syntactic Foam and Aluminum Ligaments …………………. 70
Figure 5.1: Syntactic Foam Core Sandwich Structure with Graded Face Sheets
(SFS-b)………………………………………………………………… 73
Figure 5.3: Overlay of Measured Load-Deflection Data for Four Different SF Core
Sandwich Beams with Graded Face Sheets (SFSb30)………………… 74
Figure 5.4: Overlay of Measured Load-Deflection Data for Four Different SF Core
Sandwich Beams with Graded Face Sheets (SFSb40)………………… 75
Figure 5.9: Normalized Measured Peak Load for Sandwich Structures with
Syntactic Foam Core and Graded Face Sheets ……………………….. 80
Figure 5.13: Energy Absorbed for Sandwich Structures with Syntactic Foam Core
and Graded Face Sheets……………………………………………….. 84
Figure 5.14: Specific Energy Absorbed for Sandwich Structures with Syntactic
Foam Core and Graded Face Sheets………………………………….. 86
xi
Figure 5.15: Failed SFS-b Specimen with (a) Side, (b) Top Face, and (c) Bottom
Face Views ……………………………………………………………. 87
Figure 5.16: Magnified Images (a), (b) of Failure Surfaces and (c) From Below …. 88
Figure 6.7: PMMA Strain (εx) Variation Along the Height of the Beam………… 102
Figure 6.8: PMMA Shear Strain (εxy) Variation Along the Height of the Beam … 103
xii
Figure 6.17: ARAMIS™ Results for SFS40 (FLEX-24, εxy)………………………. 112
Figure 6.36: Schematic of Impact Test Setup: (1) high-speed digital camera,
(2) drop tower impactor tup, (3) delay generator, (4) specimen,
(5) light sources, (6) DAQ – drop tower, (7) DAQ – camera,
(8) lamp control unit, and (9) drop tower controller………………… 134
Figure 6.37: Impact Test Setup: (1) high-speed camera, (2) drop tower impactor
tup, (3) specimen, and (4) light source ……………………………… 135
xiii
Figure 6.38: Camera Viewing Area………………………………………………... 136
Figure 7.1: Normalized Measured Peak Load for Sandwich Structures with SF
and IPC Foam Cores…………………………………………………. 147
Figure 7.5: Energy Absorbed for Sandwich Structures with SF and IPC Foam
Cores ………………………………………………………………… 149
Figure 7.6: Specific Energy Absorbed for Sandwich Structures with SF and IPC
Foam Cores ………………………………………………………….. 149
Figure 7.7: 20% Vf: Strain (εx) Variation Along the Height of the Sandwich
Core………………………………………………………………….. 151
Figure 7.8: 20% Vf: Shear Strain (εxy) Variation Along the Height of the Sandwich
Core …………………………………………………………………. 152
xiv
List of Tables
xv
CHAPTER 1
INTRODUCTION
plates that form the edges or faces of cells. The simplest of all cellular solids is a two-
dimensional array of regular polygons that look like the cells created by honeybees. For
this reason they are often referred to as honeycombs. More commonly seen are three-
dimensionally packed polyhedra which are known as foams. If the cells connect through
open faces with the cell edges that are solid, the foam is said to be open-celled. On the
other hand, if the cell faces are also solid, sealing off each cell from its neighbors, the
foam is said to be closed-celled [1]. Figures 1.1-1.3 show images of these three foam
types.
http://materiali.matech.it/matech/materiali/images/okpic/CP2079.jpg)
1
Figure 1.2 Open-Celled Foam (Ref: http://www.grantadesign.com/images/foam.gif)
Schaeume/ALPORAS_AC/ALPORAS.JPG)
Many structural materials found in nature are cellular solids: cork, sponge, coral,
wood, and cancellous bone. Wood is still the most widely used structural material.
Increasingly, man-made foams and honeycombs are being used in structural applications.
Considering their lightweight and stiff nature, their most obvious use today is in
sandwich panels. From the World War II planes that used panels made from thin
plywood skins bonded to balsa wood cores to the planes of today that use fiberglass or
2
carbon fiber composite skins with cores of aluminum or paper-resin honeycombs or rigid
polymer foams, sandwich panels provide excellent specific bending stiffness and
strength. Sandwich panels are also found in nature. The cuttlebone of a cuttlefish is a
multi-layer sandwich panel. Some types of leaves are structured much in the same
manner as sandwich panels. Even the human skull is composed of two layers of hard
bone separated by a lightweight core of cancellous bone. Figure 1.4 shows the cancellous
bone found in the human skull compared with Duocel® open-celled aluminum foam.
(a) (b)
Figure 1.4 (a) Cancellous Bone and (b) Duocel® Aluminum Foam (Ref:
http://www.ergaerospace.com/foamproperties/aluminumproperties.htm)
density, the density of the foam divided by the density of the solid of which it is
comprised.
dispersing hollow microballoons into a metal, polymer, or ceramic matrix. The presence
3
of the microballoons leads to improved properties including buoyancy, lower density, and
higher strength.
Sandwich panels consist of two stiff, strong skins separated by a lightweight core.
Separating the skins, or face sheets, increases the moment of inertia of the panel with
little penalty in terms of weight. This produces an efficient structure for withstanding
bending and buckling loads. For this reason sandwich panels are often used in situations
where weight limitation is critical, such as in aircraft structures. The conventional way of
stiffening panels are with stringers: strips attached between the faces with profiles such as
a Z, a tee, or a top hat. In general, stringer stiffened panels offer greater specific stiffness
(stiffness per unit weight). The drawbacks, however, are anisotropy and cost. Metal or
isotropy as well as the ability to easily and cheaply mold complex lightweight structures
attachment or honeycomb stiffened sandwich panels can offer the best specific stiffness.
This applies only in the direction the panel has been stiffened, though. In contrast, for
loads about which less information is known or that have comparable magnitudes in
multi-axial (even changing) directions, the isotropic nature of foam cored sandwich
4
Figure 1.5 Sandwich Panels with Metallic Foam Core (Ref: http://www.agstaron.com)
materials that has the potential for outstanding multifunctional properties. In traditional
composites, discrete reinforcing phases such as whiskers, fibers, or dispersed particles are
introduced into a matrix to improve or alter the resulting properties of the matrix. In an
IPC, constituent phases are continuous and three dimensionally interconnected. In their
consisting of a matrix with reinforcing phases that do not experience such complete
an IPC.
5
An advantage of IPCs over traditional reinforced composites is that higher
volume fractions of the reinforcing or secondary phase may be introduced more easily
into an IPC. Furthermore, each phase is mechanically constrained by the other resulting
in a better structural response. Each phase of an IPC contributes its property to the
provides strength, the other might enhance stiffness or thermal stability. Among the
potential mechanical benefits of IPCs, the ones regarding fracture and energy dissipation
characteristics are of interest. In traditional composites with aligned fibers, stiffness and
strength advantages are limited to the fiber direction as crack propagation along the fibers
the phases in an IPC could stave off failure effectively while offering macro scale
isotropy.
The work done by Wu et al. [2] illustrates improved mechanical properties and
impact resistance of sandwich structures with graphite/epoxy face sheets and aluminum
honeycomb core by first filling the honeycomb core with rigid polyurethane foam. Low
velocity impact tests have revealed an improvement in impact resistance, including less
frequent debonding of the face sheet and localized core crushing. Zarei et al. [3]
attempted to study and optimize bending behavior due to impact loading of foam-filled
beams. Impact tests were conducted on aluminum tubes in order to simulate car bumpers
6
during crashes. An attempt was made to optimize the tube energy absorption and tube
design to include a metallic foam core consisting of Alporas aluminum foam. Energy
absorption increased between 2.5%-5.2% once the foam filler was added. Specific
energy absorbed also increased; however, it was less pronounced. Seitzberger et al. [4]
examined the effects of foam filling on axially loaded tubular columns made of steel.
Different shapes, such as square, hexagonal, and octagonal, as well as different filling
configurations (single tube with no foam filling, two tubes with foam in between, single
completely filled tube) were tested. For the square, fully filled tubes, experiments
Belingardi et al. [5]. Bending tests failed starting with 45 degree shear cracks in the core
followed by face sheet debonding. The modulus and Poisson’s ratio gathered from
bending tests were approximately 25 percent below the values from tensile tests, likely
due to the dual tensile and compressive nature of loading. Tests conducted on the
sandwich structure showed no strain-rate effects over the range of test conditions
investigated in the study. The impact response of this foam-based sandwich structure
rates of strain.
Cenosphere fly ash (waste product from burning coal) microballoons and
aluminum were used to create syntactic foam via pressure infiltration technique by Zhang
et al. [6]. Compression tests revealed energy absorption capacity to be 27 MJ·m-3 at 47%
strain with a peak stress of 73 MPa. Metal-matrix IPCs were created by Zhou et al. [7].
Using volatile agents, an aluminum matrix was created with open porosity as high as
7
83%. The aluminum matrix interpenetrating phase composites reinforced with
continuous Al2O3-TiC ceramic phase were successfully fabricated by the vacuum high
One work by Styles et al. [8] investigates the strain distribution and failure
mechanisms associated with aluminum foam or polymer foam sandwich structures under
four-point bending. Real-time strain data was collected using a 3D optical technique
(ARAMIS™, GOM mbH, Germany). It was determined that the energy absorbed to
yield load for aluminum foam was approximately two times that of the polymer, and the
total energy absorption of aluminum was about 3.5 times greater. In a separate work by
Styles et al. [9], core thickness effects on the flexural behavior of aluminum foam
sandwich structures were examined. In sandwich material with Alporas foam as the core,
the max core stress and max facing stresses decreased as core thickness increased,
possibly due to diminishing coupling between core and facing. Core shear stress,
however, remained nearly the same (approximately 1.75 MPa). The failure of the core
became more dominant as the core thickness increased. McKown and Mines [10] studied
static and impact behavior of metal foam cored sandwich beams. The sandwich material
consisted of Alporas core with glass fiber polypropylene matrix composite (Plytron) face
sheets. In quasi-static tests, the shear effects in the core increased as the total specimen
specimen of length=230 mm) for the two test types were similar, indicating that strain
rate effects essentially cancel out the core failure initiation at weak points in the structure.
8
The work done by McCormack et al. [11] illustrated the different modes of
sandwich beam failure under quasi-static bending. Among face yielding, face wrinkling,
core yield, indentation, and delamination, the most common failure modes observed were
core yield and face yield. Chen et al. [12] studied the plastic collapse of sandwich beams
with Alporas™ foam cores. Four-point bending allowed the competing failure modes of
face yield, core shear and indentation to be separated physically along the beam. Face
yield occurred between the inner rollers, core shear occurred between the inner and outer
rollers, and indentation was triggered directly beneath the rollers. The collapse
mechanism maps were generated as a tool to predict failure type/mode, and criteria for
polypropylene IPC material. Strength and flexural modulus of the IPC were compared to
that of polypropylene. The IPC was stiffer than both constituents, and increased with
decreasing pore size. When 20 ppi aluminum net was used, an increase in strength of 3-
interspersed into alumina. For 12% aluminum (closest to the material presented in this
study) the bending strength increased from about 205 to 740 MPa, a 2.5 times increase.
The Young’s modulus increased from about 205 to 330 GPa, an increase of 60%. Chang
in bending strength from approximately 5 MPa to 250 MPa once a metallic phase was
9
introduced to a foam of 15% relative density. Also, a decreasing trend in IPC strength
This section puts forth the theory associated with bending of sandwich structures
applicable to linear elastic cases. It will be shown in later chapters that both IPC foam
core sandwiches and syntactic foam core sandwiches with graded face sheets (see
Chapter 2 for material descriptions) may be initially linear elastic, but does not remain so
for the entirety of its tests. This section is comprised of such theory associated with
Equations based on classical beam theory are provided by Allen [16] and are
given below for sandwich structures. Consider a sandwich beam subjected to three-point
y y
z
x
10
Let the flexural rigidity be denoted by EI. For stacked composite beams (e.g.
sandwich structures), the flexural rigidity is the sum of the flexural rigidities of the two
faces and the core. It is generally represented by the symbol D and evaluated as follows:
bc3
Dcore = Ec ,
12
bt 3 btd 2
D faces = 2 E f + Ef ,
12 4
bt 3 btd 2 bc3
D = Dcore + D faces = E f + Ef + Ec ,
6 2 12
where the subscripts f and c represent the face sheet and core, respectively. Similarly, the
bd 2
( AG ) eq = Gc .
c
where A and G represent the cross section area and shear modulus, respectively. These
rigidities are used to calculate the displacement at the midspan of a (linear elastic) beam
as:
PL3 PL
δ= + .
48 D 4( AG )eq
11
PL3
The first term of the deflection equation, , represents the contribution due to
48 D
PL
bending. The second term, , represents the contribution due to shear.
4( AG )eq
My c h h c
σf =− Ef ≤ y≤ ; − ≤ y≤− ,
D 2 2 2 2
My c c
σc = − Ec − ≤ y ≤ ,
D 2 2
Px L
2 , 0≤x≤
2
where moment is given by, M = ,
P ( x − L ) ,
L
<x≤L
2 2
and D represents the flexural rigidity. (Consult Figure 1.7 for definitions of the
expression for shear stress at depth, y, below the centroid in the cross-section as:
QS
τ= ,
Ib
where Q is the first moment of area, I is the second moment of area about the centroid, S
is the shear force, and b is the beam width. For a compound beam, differing moduli for
different elements in the cross-section must be accounted for using the modification:
S
τ=
Db
∑ (QE ) ,
12
where ∑ (QE ) represents the sum of the products of the first moments of area (Q) and
moduli (E), and D is the flexural rigidity. For the sandwich construction presented here,
btd Ec b c c
∑ (QE ) = E f
2
+
2 2
+ y − y .
2
S td Ec c 2
τ= E f + − y 2 ,
D 2 2 4
P L
2 , 0 ≤ x ≤ 2
where the shear force is given by, S = .
− P , L < x ≤ L
2 2
The following are definitions of the symbols used in this and subsequent chapters:
b = specimen width
c = core thickness
δ = deflection at mid-span
D = flexural rigidity
13
l = total specimen length
L = support span
P = applied load
σf = facing stress
t = facing thickness
1.6 Objectives
sandwich structure with epoxy-based syntactic foam (SF) core and graded face sheets and
an SF-based IPC foam core sandwich for improving the flexural response relative to that
of a similar sandwich with a simple SF core. The specific goals for the work are:
• Develop a method for processing sandwich structures with syntactic foam and
• Obtain both global and local flexural response data of SF and IPC core sandwich
structures.
• Study the effect of hollow filler volume fraction in the syntactic foam (20%, 30%,
14
• Study the possibility of improving load-deflection and strain energy absorption
interconnected metallic foam phase to the core or (b) tailoring the face sheets to
• For both quasi-static and impact loading cases, demonstrate ability to make local
sandwich structures.
This thesis is divided into seven chapters, including this one. The first chapter
describes the materials to be investigated, motivation for the research, and previous work
by others in related areas. Material description and preparation of the syntactic foam core
and IPC foam core sandwiches are discussed in Chapter 2. The flexural characterizations
of unfilled aluminum foam core sandwiches and syntactic foam core sandwiches are
syntactic foam core sandwiches with graded face sheets. Chapter 6 is concerned with the
digital image correlation method used to analyze full-field deformations. Finally Chapter
7 presents a comparative summary and conclusions to be drawn from this work. There
are appendices that show global/local measurement comparisons and images related to
15
CHAPTER 2
Sandwich structures with aluminum faces and syntactic foam (SF) or SF-based
IPC foam core are processed in this work. Pressureless infiltration techniques are
employed to create the syntactic foam-filled IPC. The IPC foam is created by
introducing uncured syntactic foam into aluminum sandwich performs and curing the SF
in situ.
The syntactic foam used in this work is made of epoxy matrix and hollow glass
microballoon filler. Physical and elastic properties of the constituents of the syntactic
foam used are shown in Table 2.1. The epoxy used was Epo-Thin®, a low viscosity resin
(supplied by Buehler, Inc., USA). The advantage of the Epo-Thin® epoxy is its low
viscosity, necessary to fill small holes and voids between microballoons. The hollow
diameter ~65 µm and wall thickness ~550 nm. The microballoons used are the most
16
economical hollow ceramic fillers produced by 3M Corp. and advantages of their use
include the ability to create a good bond between epoxy and glass and it having one of
The open cell aluminum foam was used as the scaffold for infusing SF into the
core of the IPC sandwiches. The scaffold material was Duocel® foam obtained from
ERG Materials and Aerospace Corp. Its properties are shown in Table 2.2. Mean cell
size for 20 ppi material is ~1.02mm (0.040”). Ligament diameter is ~0.2mm (0.008”).
The open-cell scaffold is made from annealed 6101 aluminum, and the face sheets
of the sandwich structures are made from annealed 6061 aluminum. Relevant elastic
properties for these two alloys are the same, Young’s modulus of 69 GPa (10 Msi) and
17
yield strength of 552 MPa (8 ksi). Figure 2.1 shows the open-celled aluminum foam
sandwich scaffold, and Figure 2.2 shows an enlarged view of the open-cell Duocel®
foam.
http://www.ergaerospace.com/foamproperties/introduction.htm)
The IPC core sandwiches were produced by introducing uncured syntactic foam
into the aluminum foam core scaffold and allowing the SF to cure in situ. From the
results and conclusions from Jhaver [17] which demonstrate IPC foam superiority over
epoxy-based syntactic foam in compression, it was thought that using IPC as the core
18
material in the sandwiches would produce an improved response compared to sandwiches
with an SF core.
The specimens used in this work were made using castable two-part thermosetting
polymer. Given the simplicity of geometry, the specimen was cast into a near net shape
with minor finish machining, reducing the processing time required for individual
aluminum block. A mold cavity was then produced by pouring silicone rubber around
the blank to create a negative of the specimen. Specimens were then created by casting
The specific rubber product used for creating the mold was PlatSil® 73 Series
from PolyTek Development Corporation. This is a flexible rubber mold material that has
high tear strength and can be vulcanized at room temperature. It has low shrinkage,
resulting in very good dimensional stability. To create the mold, a cardboard barrier was
constructed and placed on a glass substrate. The specimen blank was placed within the
cardboard barrier and the silicone rubber mixture was poured over the blank and allowed
to cure for 30 hours. After curing, the cardboard barrier and the specimen blank were
19
2.2.2 Syntactic Foam Core Specimen Preparation
Epoxy-based syntactic foams with varying volume fractions (20%, 30%, and
40%) of hollow soda-lime glass microballoons were produced. For a given test
specimen, the general preparation sequence was as follows. First, the desired quantities
of resin, hardener, and glass particles were measured out. Next, the resin and hardener
were mixed slowly until the mixture appeared homogeneous. Upon achieving visible
homogeneity, the glass particles were added to the mixture and stirred slowly until the
mixture was once again homogeneous. Once the mixture was visibly homogeneous, it
was then placed into a vacuum chamber. A vacuum pump was used to bring the chamber
down to a pressure of approximately -80 kPa (gage). This pressure was maintained for
approximately 4-5 minutes and then released, returning the mixture to atmospheric
pressure. This process of vacuuming and releasing was repeated in order to bring trapped
air to the surface (typically 3 times). This ensured full degassing of the mixture. After
the vacuum/release process, air bubbles were skimmed from the top of the mixing
container. During degassing, the mixture began to separate slightly into epoxy and glass
particle constituents. It was briefly and slowly stirred once more until fully mixed before
pouring into the mold. A photograph of the mold is shown in Figure 2.3 along with a
20
Figure 2.3 Silicone Specimen Mold and Demolded Specimen
All castings for this work were allowed to cure at room temperature for 48 hours
prior to removal from the mold. Upon completion of curing, the specimens were finish
machined using a mill. Machining was done no less than 2 days after initial casting, and
To produce the Syntactic Foam Sandwich (SFS) specimens, face sheets were
punched out of a 0.813mm thick sheet of aluminum (AL6061) and machined to precise
size. They were sanded on one side with coarse grit sandpaper then cleaned off with
acetone in order to create a rougher surface for bonding. Once the core of syntactic foam
was fully cured and machined, each face sheet was separately adhered to the core using a
syntactic foam mixture of the same volume fraction. After curing, the specimen was
once again machined to remove excess material from the face sheet application process.
Figure 2.4 shows an example of the SFS sandwich structure. Though there is some minor
variation, the nominal specimen size for quasi-static testing was 127mm x 25mm x 20mm
(5” length x 1” width x 0.8” height). Each face sheet had a nominal thickness of
21
0.813mm, giving a core thickness of ~18.4mm. The nominal specimen size used for
dynamic testing was 127mm x 25mm x 18.5mm (5” length x 1” width x 0.73” height).
0.813mm thk
SF core
Aluminum
face sheet
The open cell aluminum foam sandwich material was used as the IPC scaffold.
The foam was cut into slightly oversized pieces compared to desired final specimen size.
The face sheet thickness for all specimens was in an as-received state of 0.813mm ±
0.05mm (0.032” ± 0.002”). The aluminum was first cleaned with acetone, then coated
with silane to act as a wetting agent and to promote adhesion between the aluminum
scaffold and syntactic foam. Jhaver [17] explored and confirmed the benefit of the silane
coating prior to introducing the syntactic foam. The silane used was A-1100 Amino
GE Silicones. A 5% silane solution was created (remaining 95% portion of mixture was
9 parts ethanol to 1 part deionized, filtered water) to coat the sample per the work of
22
Sadler and Vecere [18]. The solution was allowed to rest for one hour after mixing
before being used. After the coating, the aluminum was allowed to sit and dry for 12
hours. The exterior/outside of the face sheets were coated with a releasing agent to easily
be able to remove excess syntactic foam after curing. The syntactic foam was prepared in
the same manner as outlined in Section 2.2.2 and poured into the mold, and the aluminum
scaffold was gently pushed down into the pool of uncured syntactic foam.
As before, all castings for this work were allowed to cure at room temperature for
48 hours prior to removal from the mold. Upon completion of curing, the specimens
were finish machined using a mill. Testing occurred no less than 7 days after the initial
casting. A sample of the IPC foam core sandwich structure is shown in Figure 2.5. Note
that the regular diagonal marks easily observed on the aluminum are tool marks.
Syntactic foam
Aluminum foam
ligaments
AL6061
face sheet
23
2.2.4 Syntactic Foam Core with Graded Face Sheets Specimen Preparation
The graded face sheets (see Section 4.3 and Chapter 5) were obtained from the
same supplier as the aluminum foam core sandwiches. The face sheets had a ~2 mm
layer of aluminum foam brazed to the aluminum face sheets on one side. The face sheets
were cleaned with acetone, then coated with silane in the method described in the
previous section. The exterior/outside of the face sheets were coated with a mold
releasing agent, then placed into the rubber mold as shown in Figure 2.6. The syntactic
foam was prepared, poured into the mold, and allowed to cure in situ. Figure 2.7 shows
an image of the syntactic foam core sandwich with graded face sheets (SFS-b). Note that
the regular diagonal marks easily observed on the face sheets are tool marks.
1 2 3
Figure 2.6 Single Cast SFS-b Mold: (1) Mold, (2) AL6061 Face Sheet, and (3) Brazed
Aluminum Foam
24
Aluminum foam
SF core
AL6061
face sheet
} ”graded” region
Figure 2.7 Syntactic Foam Core Sandwich Structure with Graded Face sheets (SFS-b)
speckle pattern was created on one side of the specimen. The side to which the pattern
was applied was sanded using a fine grit paper and cleaned off using acetone. Once the
specimen surface was clean, alternating coats of white and black spray paint were applied
until a suitable pattern was created (approximately 4 thin coats each). After applying the
speckle pattern, the paint was allowed to dry for 24 hours before testing. Figure 2.8
shows an image of a specimen with the speckle pattern used for obtaining optical
measurements. Approximate speckle size was ~90 µm. Note that the larger black
lines/dots on the specimen are made intentionally for spatial and scaling purposes.
25
loading point
line of symmetry
x mm right support
Figure 2.8 Typical Decorated Random Speckle Pattern for Optical Measurements
26
CHAPTER 3
bending (TPB). Although the basis of the TPB tests was ASTM C393 [19], the specimen
size guidelines could not be strictly adhered due to material availability and cost. Instead,
the test specimens resemble short beams. Again, the nominal specimen size for all quasi-
static test specimens was 127mm x 25mm x 20mm (5” length x 1” width x 0.8” height).
An MTS QTest100 (10kN load cell) universal loading machine was used for
sampling rate of 5-10 points per second. The beams were centrally positioned on roller
supports, also of diameter 10 mm. The test arrangement is shown in Figure 3.1.
27
CROSSHEAD BEAM
LIGHT LIGHT
SOURCE SOURCE
LOADING
FIXTURE
CAMERA
LENS
In order to collect gray scale data of the speckle pattern for an open-cell foam, the
pores had to be filled with a material that offers negligible reinforcement. Common
expanding insulating foam was used to cover one side (face) of the foam, creating a more
solid and even surface on which speckles could be painted. After the foam had time to
cure, it was sanded down to the surface of the metal foam. The speckle pattern was then
applied as detailed in Section 2.2.4. It was determined through a series of tests that the
insulating foam would not affect the material behavior because its load bearing or
28
For quasi-static tests, a camera was set up to capture the right half of the
specimen, between the loading and support rollers. Due to the symmetry of the beam
about the loading axis, one half beam images were captured for subsequent deformation
only one digital camera was required to capture the specimen deformation during the
tests. The camera used for quasi-static testing was a Nikon D100 SLR digital camera
fitted with a 200mm lens. Its maximum image capture size is 3008 x 2000 pixels, but for
this work a spatial resolution of 1504 x 1000 pixels was chosen. The camera was
connected to and controlled by a computer using the Nikon Capture software (Version
4.0). The camera was set approximately two feet from the specimen, and white light
illumination sources were placed next to the specimen. The response of the material was
expected to be symmetric about the loading roller; therefore the images encompass the
right half of the beam only, from just outside the loading roller to just outside the support
roller. The camera was set to automatically record images once every 2-3 seconds and
store them on the computer for later analysis. The arrangement for the quasi-static tests
29
TPB test digital controls camera & stores
camera images
Illumination
sources
The results from the tests on the unfilled aluminum foam core sandwiches are
500
400
FLEX-32 (ALS)
Load (N)
FLEX-34 (ALS)
200
Figure 3.4 Overlay of Load-Deflection Data for 20ppi Aluminum Foam Core Sandwich
30
Table 3.1 Aluminum Foam Sandwich Test Results
Crosshead
Specimen Speed Plateau Peak
Number Type (in/min) Load (N) Load (N)
32 ALS 0.025 391 395
33 ALS 0.10 342 346
34 ALS 0.10 391 399
NUM OF SAMPLES 3 3
AVERAGE 375 380
STANDARD DEVIATION 28 30
The unfilled aluminum foam sandwiches behave in a linear fashion at the very
beginning before the onset of nonlinearity due to core indentation and crushing. The load
remains nearly constant throughout the core crushing at the so-called “plateau load” until
the core is free of pores and densification begins (not seen in the displacement range
used). Figure 3.5 shows a typical stress-strain response for aluminum foam and
illustrates the plateau and densification regions. For the case of a TPB test, these
crushing and densification phenomena are observed below the load roller. When in the
plateau region, the aluminum foam has tremendous energy absorption capacity,
continuing to consume energy with increasing strain (or deflection) at a nearly constant
load. The average plateau load for the aluminum foam core sandwich is 375 N ± 28 N.
31
Plateau Region (Core Crushing) Densification Region
The quasi-static TPB test setup for the syntactic foam core sandwich (identified as
SFS) specimens is the same as that detailed in Section 3.1. In order to remove any slack
from the loading fixtures, a pre-load of ~130 N was applied. After this pre-load was
applied, the load was zeroed so that the pre-loaded state is the reference (undeformed)
state. With failure on the order of 5,000 N, the pre-load amount considered was
negligible.
The measured load-deflection responses from the tests are shown in Figures 3.6 to
3.8, and the results are summarized in Tables 3.2 to 3.4. Note that the stair-case pattern
observed in the test data is attributed to the data acquisition rate and software and not to
the mechanical response of the sandwich structure. The different materials with 20%,
32
30%, and 40% volume fractions of microballoons in the SF core will henceforth be
8000
7000
6000
FLEX-37 (SFS20)
FLEX-38 (SFS20)
Load (N)
5000
FLEX-45 (SFS20)
FLEX-46 (SFS20)
4000
FLEX-47 (SFS20)
FLEX-48 (SFS20)
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
8000
7000
6000
FLEX-39 (SFS30)
Load (N)
FLEX-41 (SFS30)
4000
FLEX-42 (SFS30)
FLEX-43 (SFS30)
3000
FLEX-44 (SFS30)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
33
9000 Material: SFS40
Temperature: 70°F
Overlay
8000
7000
6000
FLEX-23 (SFS40)
Load (N)
5000
FLEX-24 (SFS40)
4000
FLEX-36 (SFS40)
3000
2000
1000
Debond initiation
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm)
37 SFS 20 0.025 4.78 68.5 6626 1.24
38 SFS 20 0.025 2.17 31.4 4997 0.81
45 SFS 20 0.025 5.92 84.3 7098 1.42
46 SFS 20 0.025 6.18 88.4 7155 1.45
47 SFS 20 0.025 1.72 24.9 4422 0.71
48 SFS 20 0.025 7.57 108.6 7568 1.68
NUM OF SAMPLES 6 6 6 6
AVERAGE 4.73 67.7 6311 1.22
STANDARD DEVIATION 2.33 33.2 1289 0.38
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm)
39 SFS 30 0.025 3.53 57.1 5526 1.24
40 SFS 30 0.025 3.47 55.7 5550 1.17
41 SFS 30 0.025 4.14 65.9 5815 1.37
42 SFS 30 0.025 4.08 64.9 5758 1.27
43 SFS 30 0.025 3.29 52.5 5405 1.14
44 SFS 30 0.025 3.57 57.5 5571 1.22
NUM OF SAMPLES 6 6 6 6
AVERAGE 3.68 58.9 5604 1.24
STANDARD DEVIATION 0.35 5.3 154 0.08
34
Table 3.4 TPB SFS40 Test Results
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm)
36 SFS 40 0.025 2.78 48.5 5126 0.97
23 SFS 40 0.025 2.35 40.5 3598 1.24
24 SFS 40 0.025 1.76 30.6 5440 0.76
NUM OF SAMPLES 3 3 3 3
AVERAGE 2.30 39.9 4722 0.99
STANDARD DEVIATION 0.51 9.0 985 0.24
The flexural response of all SFS sandwiches showed a linear elastic behavior
Once this debond occurs, the majority of load is carried by the core causing it to fail soon
thereafter. To give a visual of the debond failure, Figure 3.9 shows an image captured
soon after the debond occurred during a test. (Analyzed digital images of this specimen
Face-core debonding
35
The metrics used to characterize the syntactic foam core sandwiches were the
measured peak load, the deflection measured at failure, the total strain energy absorbed,
and the specific strain energy absorbed (which takes into account the weight difference
associated with the different volume fractions tested). Both the peak load and deflection
at failure were normalized by values from the behavior of unfilled aluminum foam
sandwiches.
The average value for the peak load decreases with increasing volume fraction.
The average for SFS20 is 6,311 N, for SFS30 is 5,604 N, and for SFS40 is 4,722 N.
separated by an empty core. Practically, the unfilled 20ppi aluminum foam core
sandwich was used to normalize the data. To compare to the behavior of the aluminum
foam core sandwich, the average load at failure for the SFS sandwich is normalized by
the plateau load experienced by the aluminum foam core sandwich (390 N). The average
normalized load at failure for SFS20 is 16.14, for SFS30 is 14.33, and for SFS40 is 12.08.
In Figure 3.10 these values are shown as a histogram. This comparison is not exact. The
plateau load for the aluminum foam core sandwich is the load at which the stress remains
nearly constant while able to withstand a large increase in strain. It is this load that is
typically used as the desired design load. (See Figure 3.5 for a typical stress-strain
response of this type.) The SFS sandwiches do not experience a similar near constant
load, but rather a steady increase (hardening behavior) until failure. This load at failure,
when compared to the plateau load, is twelve to sixteen times higher than the design load
for unfilled aluminum foam core sandwiches. Both the plateau load for the unfilled
36
aluminum foam core and the peak load for the SFS sandwiches can be used as the design
load, so, while inexact, the comparison is justified. The SFS sandwiches show superior
load bearing capacity compared to their unfilled aluminum foam counterpart. The large
error seen in the SFS20 and SFS40 samples is caused by specimens that prematurely
failed to due an especially weak face-core bond, resulting in lowering all reported
average values.
25
SFS
20
Peak Load/ALS Plateau Load
15
10
0
20% 30% 40%
Volume Fraction
Figure 3.10 Normalized Measured Peak Load for Sandwich Structures with SF Foam
The average value for the mid-span deflection at failure decreases from SFS30 to
However, without the two poorly bonded specimens in the SFS20 group that failed
relatively earlier, the trend would have again decreased with increasing volume fraction.
37
The average for SFS20 is 1.22 mm, for SFS30 is 1.24 mm, and for SFS40 is 0.99 mm.
To compare with the behavior of the aluminum foam core sandwich, the average
deflection at failure for the SFS sandwich is normalized by the deflection at which the
unfilled aluminum foam core sandwich behavior deviates from linearity (deflection at
transition, 0.2mm). This, again, is not an exact comparison since we do not have a
deflection at failure for comparison in the unfilled aluminum foam sandwiches. The
average normalized deflection at failure for SFS20 is 6.29, for SFS30 is 6.38, and for
SFS40 is 5.11. In Figure 3.11 these normalized deflection values are shown as a
histogram. Once again, the large error in the SFS20 and SFS40 samples is due to
9.0
SFS
8.0
7.0
6.0
δFailure/δALS, nonlinear
5.0
4.0
3.0
2.0
1.0
0.0
20% 30% 40%
Volume Fraction
38
Cellular foams are used in packing and automotive applications because of their
ability to dissipate strain energy. Using this lightweight sandwich material could
hopefully offer improved mechanical characteristics while still maintaining a low weight.
The energy absorbed during flexural experiments by a specimen was calculated using the
area under the load-deflection curve from the measured data shown in Figures 3.6 to 3.8.
That is, the area below the load-deflection responses were computed as,
δ
U = ∫ P (δ ) d δ ,
0
The average values for the strain energy absorbed by different SFS samples are
plotted in Figure 3.12 as a histogram. The average for SFS20 is 4.73 J, for SFS30 is 3.68
J, and for SFS40 is 2.30 J. Unfilled aluminum foam core sandwiches have a great
capacity for energy absorption, but the cost is that they are unable to bear very much
load. With these SFS sandwiches, an effort was made to understand how much energy
39
7.0
SFS
6.0
5.0
3.0
2.0
1.0
0.0
20% 30% 40%
Volume Fraction
Figure 3.12 Energy Absorbed for Sandwich Structures with SF Foam Core
In order to make comparisons for varying volume fractions (and thus the
difference in weight), specific energy (J/kg) was calculated for each specimen. The
energy absorbed was divided by the mass of the particular specimen determined by the
equation,
The densities for the SF core and aluminum faces are shown in Table 3.5.
40
Table 3.5 Material Densities
Density
Material
(kg/m3)
SF20 893.0
SF30 797.0
SF40 701.0
Al 6061 (face) 2700.0
Once again we see the trend of the average value for the specific energy absorbed
decreasing with increasing volume fraction. The average for SFS20 is 67.7 J/kg, for
SFS30 is 58.9 J/kg, and for SFS40 is 39.9 J/kg. Figure 3.13 shows these values as a
120
SFS
100
Specific Energy Absorbed (J/kg)
80
60
40
20
0
20% 30% 40%
Volume Fraction
Figure 3.13 Specific Energy Absorbed for Sandwich Structures with SF Foam Core
41
As stated earlier, the SFS sandwich failures manifested in face-core debonding.
Figure 3.14 gives a visual of a failed SFS sandwich. Note the lack of any residual SF
remaining on the interior portions of the face sheets in image (b), highlighting the poor
bond strength and ease of separation between the core and face.
(a) (b)
Figure 3.14 (a) Failed Syntactic Foam Core Sandwich with (b) Highlighted Face-core
Debond
The overall behavior of SFS sandwich structure is essentially linear elastic almost
until failure. The response of a typical specimen at each volume fraction was compared
to the theoretical calculations from Section 1.5.1 [16]. The load histories are used to
properties used for the SF material measured by Jhaver [17] under compression were
derived from the stress-strain responses shown in Figure 3.15. The values are shown in
Table 3.6. (This was an approximation made deliberately even though sandwich beams
experience both compressive and tensile stresses, and tensile stress-strain responses could
42
be different from the compression responses.) There is good agreement between the
theoretical and experimental mid-span displacements for all volume fractions, seen in the
comparisons shown in Figure 3.16 to 3.18. Finally, Figure 3.19 shows a load-
displacement overlay of the three specimens used as examples, with each containing a
Elastic Modulus
Material
(MPa)
SF20 1595
SF30 1448
SF40 1261
43
Material: SFS20
6000 Temperature: 70°F
Specimen: FLEX-38
Cast: 12Jul2010
Test: 14Sept2010
Rate: 0.025 in/min
5000
4000
Load (N)
Crosshead
3000
Elastic
Prediction
2000
1000
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Displacement (mm)
Material: SFS30
6000 Temperature: 70°F
Specimen: FLEX-39
Cast: 04Oct2010
Test: 26Oct2010
Rate: 0.025 in/min
5000
4000
Load (N)
Crosshead
3000
Elastic
Prediction
2000
1000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Displacement (mm)
44
Material: SFS40
4500 Temperature: 70°F
Specimen: FLEX-24
Cast: 18May2010
4000 Test: 15Jul2010
Rate: 0.025 in/min
3500
3000
Load (N)
2500
Crosshead
2000
Elastic
Prediction
1500
1000
500
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Displacement (mm)
8000
7000 20%
40%
3000
FLEX-24 (SFS40)
2000 SFS40
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
45
3.3 Summary
debonding, shown in Figure 3.20 where part of the top face has separated from the core.
The separation is clean, with no SF residue on the debonded portion of the face sheet.
Once this debond occurs, the majority of load is transferred to the core causing it
to fail soon thereafter. The idea of the sandwich structure is for both the core and faces to
contribute to the overall behavior. After the debond, the material no longer acted as a
lightweight sandwich composite, but rather a syntactic foam beam with aluminum plates
on either side. This suggests that overcoming the shear stress jump at the face-core
interface could prevent face-core debonding. Perhaps the addition of a metal foam matrix
to the core would serve to prevent debonding as well as improve the deflections (and thus
strains) experienced through the synergistic nature of an IPC foam core. This would also
46
serve to increase the energy absorbing capacity. The metal ligaments could serve as a
47
CHAPTER 4
SANDWICH STRUCTURES
foam interpenetrating core architecture. It was thought that the introduction of the
metallic foam scaffold into the core would improve the flexural response synergistically,
retaining the strength of the syntactic foam and gaining ductility from the aluminum
foam. The experimental characterization of these structures in bending are carried out
using the same setup described for the SFS sandwiches in Chapter 3. In this case, the
sandwich scaffold was an unfilled aluminum foam sandwich into which uncured
syntactic foam was introduced and allowed to cure in situ. An image of an IPC foam
48
Syntactic foam
Aluminum foam
ligaments
AL6061
face sheet
interpenetrating core sandwich (identified as IPC) structures are shown in Figures 4.2 to
4.4, and the results compiled from these tests are summarized in Tables 4.1 to 4.3. The
different materials with 20%, 30%, and 40% volume fractions of microballoons in the
IPC core will henceforth be referred to as IPC20, IPC30, and IPC40, respectively.
49
9000 Material: IPC20
Temperature: 70°F
Overlay
8000
7000
6000
FLEX-9 (IPC20)
Load (N)
5000
FLEX-10 (IPC20)
4000
FLEX-11 (IPC20)
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 4.2 Overlay of Measured Load-Deflection Data for Three Different IPC Core
8000
7000
6000
FLEX-12 (IPC30)
Load (N)
5000
FLEX-13 (IPC30)
4000
FLEX-16 (IPC30)
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 4.3 Overlay of Measured Load-Deflection Data for Three Different IPC Core
50
9000 Material: IPC40
Temperature: 70°F
Overlay
8000
7000
6000
FLEX-14 (IPC40)
Load (N)
5000
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 4.4 Overlay of Measured Load-Deflection Data for Two Different IPC Core
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
9 IPC 20 0.025 7.34 83.3 5366 1.96 1744 0.30
10 IPC 20 0.025 6.34 80.7 4950 1.83 1472 0.28
11 IPC 20 0.025 7.76 89.7 5880 2.08 1846 0.28
NUM OF SAMPLES 3 3 3 3 3 3
AVERAGE 7.15 84.6 5399 1.96 1687 0.29
STANDARD DEVIATION 0.73 4.6 466 0.13 193 0.01
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
12 IPC 30 0.025 7.46 87.9 5850 1.98 1922 0.30
13 IPC 30 0.025 7.71 93.4 5205 2.18 1557 0.28
16 IPC 30 0.025 8.49 98.9 5682 2.24 1726 0.30
NUM OF SAMPLES 3 3 3 3 3 3
AVERAGE 7.89 93.4 5579 2.13 1735 0.30
STANDARD DEVIATION 0.54 5.5 334 0.13 183 0.01
51
Table 4.3 TPB IPC40 Test Results
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
14 IPC 40 0.025 5.94 89.6 4445 1.91 1521 0.30
15 IPC 40 0.025 5.11 71.3 4128 1.60 1757 0.25
17 IPC 40 0.025 3.67 64.3 4360 1.70 3861* 1.24*
NUM OF SAMPLES 3 3 3 3 2 2
AVERAGE 4.91 75.1 4311 1.74 1639 0.28
STANDARD DEVIATION 1.15 13.1 164 0.16 167 0.04
*Due to loading/reloading during testing, values may not be indicative of true material behavior
In the case of the SFS sandwiches, the load-deflection behavior was essentially
linear elastic in nature (see Chapter 3). However, for the IPC sandwiches it shows
“transition point” in the subsequent discussion. To define the transition point for
analysis, a line was fit approximately through the data before the transition and another
was fit through the data after transition. The point where these two fitted lines intersect is
identified as the transition point. Both the load and deflection at the transition point were
recorded from the graphs. The aluminum foam core sandwiches exhibit a similar trend of
transition from linear to nonlinear behavior. However, unlike with the aluminum foam
core (which has a similar value for the transition deflection), the IPC sandwiches are able
to support additional load while continuing to deform. For design purposes, this
transition point can be treated similar to a yield point. At transition, certain parts of the
The plots and pictures shown below in Figures 4.5 to 4.9 illustrate the progression
of the test at important points of interest. The first photo is the undeformed specimen
used as a reference image for the particular test. The second photo is at point A (in the
plot in Figure 4.5), the transition point, where the behavior deviates from its initial linear
52
elastic response. The third photo is when a visible crack first appears in the core. There
is a short time after the initial crack that the specimen still has the ability to bear some
reduced amount of load. The fourth photo is at the very end of the test when no load
bearing capacity remains in the sample. It should be noted that although no load bearing
capacity remains, the structure itself is still mostly intact. This is unlike the SFS
Material: IPC20
6000 Temperature: 70°F
Specimen: FLEX-10
Cast: 30Sept2009
Test: 21Apr2010
Rate: 0.025 in/min
5000
4000
Load (N)
3000
Crosshead
2000
B
1000
A C
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
Figure 4.5 Typical IPC Load-Deflection Curve: Point A – Transition Point, Point B –
53
Figure 4.6 Undeformed, P=0 N
54
Figure 4.8 Point B: Crack Formation, P=3905 N
55
The same metrics used to characterize the SFS sandwiches were used for IPC
sandwiches (measured peak load, deflection measured at failure, total strain energy
absorbed, and specific strain energy absorbed). In addition, two measured values at the
transition point (load and deflection) were used for the IPC sandwiches to identify the
nonlinear behavior. Both the load and deflection measurements were normalized by
The average value for the peak load is the highest for IPC30, with the ones for
IPC20 and IPC40 being successively lower. The average load at failure for IPC20 is
5,399 N, for IPC30 is 5,579 N, and for IPC40 is 4,311 N. To compare to the behavior of
the unfilled aluminum foam core sandwich, the average load at failure for the IPC core
sandwich is normalized by the plateau load experienced by the aluminum foam core
sandwich (390 N). The average normalized load at failure for IPC20 is 13.81, for IPC30
is 14.27, and for IPC40 is 11.03. In Figure 4.10 these values are shown as a histogram.
The plateau load for the aluminum foam core sandwich is the load at which the stress
remains nearly constant while able to withstand a large increase in strain. It is this load
that is typically used as the desired design load. (See Figure 3.5 for a typical stress-strain
response of this type.) The IPC sandwiches do not experience a similar near constant
load, but rather a steady increase (hardening behavior) until failure. This load at failure,
when compared to the plateau load, is eleven to fourteen times higher than the design
load for unfilled aluminum foam core sandwiches. The IPC sandwiches show superior
56
16
IPC
14
12
0
20% 30% 40%
Volume Fraction
Figure 4.10 Normalized Measured Peak Load for Sandwich Structures with IPC Foam
Core
The average value for the mid-span deflection at failure exhibits the same trend
seen for the peak load in the case of the IPC sandwiches, decreasing from IPC30 to
IPC20 to IPC40. The average for IPC20 is 1.96 mm, for IPC30 is 2.13 mm, and for
IPC40 is 1.74 mm. To compare to the behavior of the aluminum foam core sandwich, the
average deflection at failure for the IPC sandwich is normalized by the deflection at
which the unfilled aluminum foam core sandwich behavior deviates from linearity
(deflection at transition). This, again, is not an exact comparison since we do not have a
deflection at failure for comparison in the unfilled aluminum foam sandwiches. The
average normalized deflection at failure for IPC20 is 10.09, for IPC30 is 11.01, and for
57
14.0
IPC
12.0
10.0
δFailure/δALS, nonlinear
8.0
6.0
4.0
2.0
0.0
20% 30% 40%
Volume Fraction
The average value for the load at the transition point for IPC20 is 1,687 N, for
IPC30 is 1,735 N, and for IPC40 is 1,639 N. As with the peak load, to compare to the
behavior of the aluminum foam core sandwich, the average transition load for the IPC
core sandwich is normalized by the plateau load experienced by the aluminum foam core
sandwich. The average normalized transition load for IPC20 is 4.32, for IPC30 is 4.44,
and for IPC40 is 4.19. In Figure 4.12 these values are shown as a histogram. This is the
best comparison that can be made between IPC foam and aluminum foam core
sandwiches, as both have a transition point. With the IPC foam core sandwich having a
transition load of around 4 times greater than unfilled aluminum foam sandwiches and a
similar transition deflection, the point at which plasticity begins to set in for the IPC
sandwiches is greatly improved. Also, this observation makes intuitive sense. The IPC
58
core sandwiches experienced a transition load four times higher than aluminum foam
sandwiches with a comparable transition deflection, implying that IPC foam core
sandwiches are stiffer than unfilled aluminum foam core sandwiches, which is known to
within the experimental scatter, although IPC30 specimens show a slightly higher value
relative to the two others. This suggests that microcrack formation in the tensile region
of the core is likely responsible for the early transition point in IPC foam core
sandwiches.
IPC
5
Transition Load/ALS Plateau Load
0
20% 30% 40%
Volume Fraction
Figure 4.12 Normalized Measured Load at Transition Point for Sandwich Structures
The average for mid-span deflection at the transition point for IPC20 is 0.29 mm,
for IPC30 is 0.30 mm, and for IPC40 is 0.28 mm. To compare with the behavior of the
59
aluminum foam core sandwich, the average transition deflection for the IPC core
sandwich is normalized by the deflection at which the aluminum foam core sandwich
behavior deviates from linearity (0.2mm). The average normalized transition deflection
for IPC20 is 1.49, for IPC30 is 1.53, and for IPC40 is 1.44. In Figure 4.13 these values
are shown as a histogram. As in the transition load plots, these are essentially the same
for all volume fractions of the SF in the IPC core, showing a ~50% improvement over
aluminum foam core sandwiches. With the increased load bearing capacity, higher
transition load, and a small improvement in transition deflection, the IPC foam core
sandwiches are not only similarly stiff, but much stronger than the unfilled aluminum
1.8
IPC
1.6
1.4
1.2
δTransition/δALS, nonlinear
1.0
0.8
0.6
0.4
0.2
0.0
20% 30% 40%
Volume Fraction
60
Briefly comparing the IPC foam core sandwiches to the SFS sandwiches, the IPC
seems to be the better core material. The normalized peak load showed a small reduction
of 14.5% for IPC20, 0.4% for IPC30, and 8.7% for IPC40. The deflection at failure,
improvement of 60.4% for IPC20, 72.6% for IPC30, and 75.2% for IPC40.
As before, the strain energy (U) absorbed by a specimen was calculated using the
area under the load-deflection curve from the measured data. That is,
δ
U = ∫ P (δ ) d δ .
0
The average values for the strain energy absorbed are plotted in Figure 4.14 as
histograms. The average for IPC20 is 7.15 J, for IPC30 is 7.89 J, and for IPC40 is 4.91 J.
9.0
IPC
8.0
7.0
6.0
Energy Absorbed (J)
5.0
4.0
3.0
2.0
1.0
0.0
20% 30% 40%
Volume Fraction
Figure 4.14 Energy Absorbed for Sandwich Structures with IPC Foam Core
61
In order to make comparisons for varying volume fractions, specific energy (J/kg)
was calculated for each specimen. The strain energy absorbed was divided by the mass
The density for the IPC core was determined using a 93% syntactic foam / 7%
aluminum foam composition. The density of the IPC foam core for a particular volume
The densities for the IPC core are shown in Table 4.4. (Refer back to Table 3.5
Density
Material
(kg/m3)
IPC20 1019.5
IPC30 930.2
IPC40 840.9
The results for specific strain energy are shown in Figure 4.15. Again the same
trend is observed in the IPC sandwiches when comparing specific energy absorption.
62
The average value for the specific energy absorbed is highest for the 30% volume
fraction. The IPC20 and IPC40 material values are both less than the IPC30, indicating
an optimum value for the IPC core somewhere around 30% volume fraction of
microballoons in SF. The average for IPC20 is 84.6 J/kg, for IPC30 is 93.4 J/kg, and for
IPC40 is 75.1 J/kg. Comparing the specific strain energies of IPC foam and syntactic
foam core sandwiches (which accounts for the weight difference in adding the aluminum
foam in the core), IPC20 had a 25% improvement over its syntactic foam core
counterpart, IPC30 had a 59% improvement, and IPC40 had an 88% improvement. With
IPC foam core outperforming syntactic foam core for all volume fractions, it can be
concluded that IPC foam core sandwiches are superior to a similar configuration with
120
IPC
100
Specific Energy Absorbed (J/kg)
80
60
40
20
0
20% 30% 40%
Volume Fraction
Figure 4.15 Specific Energy Absorbed for Sandwich Structures with IPC Foam Core
63
Failure observed in IPC core sandwich structures typically manifested in tensile
failure of the core and lower face sheet. The failure would begin with a crack in the core
in the tensile region. The bottom aluminum face would then yield in the area below the
loading roller. This would lead to a core to face debond in the yielded area, and finally a
crack propagating from the debond area upwards towards the compression zone of the
core. Due to the spontaneous nature of failure initiation, it could not be confirmed in
static tests if a local face-core debonding preceded and contributed to the core crack
initiation or vice versa. However, from observing damaged specimens and from impact
tests (see Section 6.5), it is believed that core crack initiation in the tensile region was
likely the first event leading to failure. Figure 4.16 shows images of a failed IPC foam
core sandwich. Note that the bottom face sheet has undergone yielding underneath the
location of the loading roller. The whole structure remains intact more than those with
other core types because of the metallic ligaments acting as bridging agents. Figure 4.17
shows some magnified images of the crack. In image (a), slight face-core and interphase
separation are observed as well as a slight crack in the face moving outward from the
core interface. In image (b) an aluminum ligament has yielded and failed in the tensile
zone. Image (c) shows some of the aluminum ligaments that have separated from the SF,
(a) (b)
Figure 4.16 (a) Failed IPC Foam Core Sandwich with (b) Yielded Bottom Face
64
(a) (b) (c)
Figure 4.17 Magnified Images of Crack Surfaces Near the Lower Face Sheet
the transition point, the behavior deviates from the initial linear elastic region to the
compared to the theoretical mid-span displacement calculation from Section 1.5.1 [16].
The load histories are used to calculate the theoretical displacements, along with
displacements, the quasi-static elastic properties used for the IPC material measured by
Jhaver [17] under compression were derived from the stress-strain responses shown in
Figure 4.18. The values are shown in Table 4.5. These comparisons of predicted elastic
behavior and experimental mid-span deflection are shown in Figure 4.19 to 4.21.
65
Figure 4.18 Compressive Response of IPC [17]
Elastic Modulus
Material
(MPa)
IPC20 2123
IPC30 1852
IPC40 1702
Material: IPC20
6000 Temperature: 70°F
Specimen: FLEX-10
Cast: 30Sept2009
Test: 21Apr2010
Rate: 0.025 in/min
5000
4000
Load (N)
Measured
3000
Elastic
Prediction
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
66
Material: IPC30
6000 Temperature: 70°F
Specimen: FLEX-16
Cast: 15Apr2010
Test: 09Jun2010
Rate: 0.025 in/min
5000
4000
Load (N)
Measured
3000
Elastic
Prediction
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
Material: IPC40
5000 Temperature: 70°F
Specimen: FLEX-14
Cast: 07Oct2009
Test: 21Apr2010
Rate: 0.025 in/min
4000
3000
Load (N)
Measured
2000 Elastic
Prediction
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
67
The theoretical and experimental values deviate from each other very early on for
all volume fractions of microballoons in the SF. The predictions exhibit greater stiffness
than measurements. Note that since the theoretical response is only for the linear elastic
case, the response after the transition point cannot be compared. Furthermore, the
deviations in the elastic region are possibly due to the fact that the elastic modulus for
IPC foam core is assumed to be the same in tension as in compression. The elastic
stiffness of IPC foam under tension is likely lower than the compression counterpart due
to relatively weak bonding between aluminum ligaments and the SF foam. Furthermore,
the lower stiffness of the structure from early on in the loading history suggests that SF
and aluminum ligaments debond in the tensile region quite early on, causing the
nonlinear behavior. Yet, the global integrity of the core continues until much later on in
the loading history. Figure 4.22 shows a load-displacement overlay of the three
specimens used as examples, with each containing a different volume fraction (Vf) of
microballoons in the SF. The transition point is virtually the same for all three cases.
After the transition, nonlinear behavior is similar across the three different cases, with the
30% Vf having both the greatest energy and specific energy absorbtion capacity. The
68
9000 Material: IPC
Temperature: 70°F
Overlay
8000
7000 20%
6000 IPC30
FLEX-10 (IPC20)
30%
Load (N)
5000
IPC20
FLEX-16 (IPC30)
4000
IPC40
40%
3000
FLEX-14 (IPC40)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
4.3 Summary
While the IPC foam core sandwiches experienced slightly lower load levels than
their syntactic foam core counterparts (comparisons are discussed in Chapter 7), the
deflection at failure was greatly improved, thus improving the overall strain energy
absorption. There was also the new addition of plastic nonlinear behavior in the IPC core
sandwiches that was not seen in the SFS sandwiches. Also of note is that these sandwich
beams did not fail abruptly, but rather developed a core crack and face sheet yielding.
The aluminum foam ligaments acted as bridges holding the SF and hence the sandwich
together, even when it could no longer take any significant load. This could be important
from an overall structural aspect, as one would not want a structure to catastrophically
69
fail after, say, being struck by a projectile or flying debris. It would be better to have a
weakened area on an intact structure than to have a catastrophic rupture in that structure.
Considering that part of the reason for IPC core sandwich failure resides in the
IPC itself and that the sandwiches with SF core typically failed due to face-core
debonding, it was considered that if we were somehow able to improve the face-core
bonding while using syntactic foam in the core, a superior material might emerge. Using
syntactic foam in the core rather than IPC would prevent the interphase separation that
Figure 4.23.
(a) (b)
Figure 4.23 Illustration of IPC (a) Before and (b) After Interphase Separation between
In order to achieve this improved bonding, similar aluminum face sheets were
used with one important difference: each face sheet had ~2mm of brazed aluminum foam
on one side. This narrow (thin) layer of aluminum foam would act as an extremely
roughed surface for bonding, acting like little fingers gripping the core and hopefully
70
preventing a premature face-core shear failure allowing the SF to perform its role of
keeping the two faces apart for effective flexural performance. This new type of face
sheet is referred to as a “graded” face sheet and will be discussed in the next chapter.
71
CHAPTER 5
This part of the research explored sandwich structures with SF foam core and
graded (or tailored) face sheets. Experimental characterization in bending was carried
out using the same setup described for structures with syntactic foam core and IPC foam
The 0.813mm thick face sheets had a ~2 mm layer of brazed aluminum foam on
one side to act as a “roughened face” that would more readily dig into the SF core to
transfer the load throughout the specimen. This new type of face sheet would also
hopefully prevent the face-core debond that caused premature failure seen in the earlier
syntactic foam core sandwiches. Additionally, there would not be a great amount of
metallic foam, as the failure mechanisms in the IPC foam core stemmed from the
interphase separation between the SF and aluminum ligaments. Figure 5.1 shows an
image of the syntactic foam core sandwich with graded face sheets (identified as SFS-b
hereafter). Note that the regular diagonal marks easily observed on the face sheets are
tool marks.
72
0.813mm
Aluminum foam
SF core
AL6061
~2mm face sheet
Figure 5.1 Syntactic Foam Core Sandwich Structure with Graded Face Sheets (SFS-b)
Flexural tests similar to those described in the previous chapters were carried out
for syntactic foam core sandwiches with graded face sheets and load-deflection responses
were measured. The tests were aimed towards investigating how a tailored or graded
face sheet could prevent premature shear separation between the face and the core. The
load-deflection responses for SFS-b sandwiches are shown in Figures 5.2 to 5.4, and the
results compiled from these tests are summarized in Tables 5.1 to 5.3. The different
materials with graded face sheets and 20%, 30%, and 40% volume fractions (Vf) of
significance. The first type of syntactic foam sandwich face sheet (sanded)/SF core
combination is identified simply as “SFS”, and the second type of face sheet (graded)/SF
73
9000 Material: SFSb20
Temperature: 70°F
Overlay
8000
7000
6000
FLEX-57
(SFSb20)
Load (N)
5000
FLEX-58
(SFSb20)
4000
3000 FLEX-61
(SFSb20)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 5.2 Overlay of Measured Load-Deflection Data for Three Different SF Core
8000
7000
6000 FLEX-55
(SFSb30)
Load (N)
5000 FLEX-56
(SFSb30)
4000 FLEX-59
(SFSb30)
3000 FLEX-60
(SFSb30)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 5.3 Overlay of Measured Load-Deflection Data for Four Different SF Core
74
9000 Material: SFSb40
Temperature: 70°F
Overlay
8000
7000
6000 FLEX-63
(SFSb40)
Load (N)
5000 FLEX-64
(SFSb40)
4000 FLEX-65
(SFSb40)
3000 FLEX-66
(SFSb40)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
Figure 5.4 Overlay of Measured Load-Deflection Data for Four Different SF Core
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
57 SFSb 20 0.025 11.59 145.4 8173 2.18 5876 0.95
58 SFSb 20 0.025 11.57 144.7 8156 2.16 5551 0.85
61 SFSb 20 0.025 11.47 143.1 8027 2.18 5960 0.98
NUM OF SAMPLES 3 3 3 3 3 3
AVERAGE 11.54 144.4 8119 2.18 5796 0.93
STANDARD DEVIATION 0.06 1.2 80 0.01 216 0.07
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
55 SFSb 30 0.025 11.16 152.5 7581 2.21 5885 0.98
56 SFSb 30 0.025 11.85 160.8 7469 2.41 5618 1.03
59 SFSb 30 0.025 12.21 167.6 7570 2.39 5738 0.98
60 SFSb 30 0.025 12.81 175.2 7505 2.49 5658 0.98
NUM OF SAMPLES 4 4 4 4 4 4
AVERAGE 12.01 164.0 7531 2.37 5725 0.99
STANDARD DEVIATION 0.69 9.7 54 0.12 118 0.03
75
Table 5.3 TPB SFSb40 Test Results
Crosshead
Specimen Speed Energy Specific Energy Peak Deflection Transition Transition
Number Type VF(%) (in/min) Absorbed (J) Absorbed (J/kg) Load (N) at Failure (mm) Load (N) Deflection (mm)
63 SFSb 40 0.025 10.57 157.3 6916 2.26 5137 0.89
64 SFSb 40 0.025 11.22 168.4 6328 2.44 4964 0.91
65 SFSb 40 0.025 11.04 166.4 6738 2.36 5222 0.91
66 SFSb 40 0.025 10.76 163.2 6479 2.34 5035 0.91
NUM OF SAMPLES 4 4 4 4 4 4
AVERAGE 10.90 163.8 6615 2.35 5090 0.91
STANDARD DEVIATION 0.29 4.8 263 0.07 113 0.01
In the case of the SFS sandwiches, the load-deflection behavior was essentially
linear elastic in nature (see Chapter 3) with face-core debonding leading to premature
ultimate failure. Moreover there was also substantial scatter in peak load and deflection
at failure. However, the SFS-b sandwiches clearly do not show any face-core debonding,
primarily attributed to the graded/tailored faces penetrating the core over a small spatial
distance adjacent to the face sheet, muting the intensity of shear stresses. This allows the
sandwich structure in general and the SF core in particular to contribute to the overall
5.2 to 5.4. Similar to the case with IPC core sandwiches, there is a kink or a “knee” in
each plot (the transition point) beyond which noticeable nonlinearity prevails up to
ultimate failure. To find the transition point, a line was fit through the data before
transition and another was fit through the data after transition. The point where these two
fitted lines intersect is identified as the transition point. Both the load and deflection at
the transition point were recorded from the graphs. The aluminum foam core sandwiches
exhibit a similar trend of transition from linear to nonlinear behavior. However, unlike
the unfilled aluminum foam core sandwiches, the SFS-b sandwiches are able to absorb
76
The plots and pictures shown below in Figures 5.5 to 5.8 illustrate the progression
of the test at important points of interest. The first photo is the undeformed specimen
used as a reference image for the particular test. The second photo is at point A (in the
plot in Figure 5.5), the transition point, where the behavior deviates from its initial linear
elastic response. The third photo is at the very end of the test once the sample has
fractured. (Analyzed images of this specimen (FLEX-59, SFSb30) are shown in Section
6.4.3).
Material: SFSb30
9000 Temperature: 70°F
Specimen: FLEX-59
Cast: 25Jan2011
8000 Test: 02Feb2011
Rate: 0.025 in/min
7000
6000
B
Load (N)
5000
4000 Measured
A
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
Point B – Failure
77
Figure 5.6 Undeformed, P=0 N
78
Figure 5.8 Point B: Failure
The failure load is the highest for SFSb20, with the ones for SFSb30 and SFSb40
being successively lower. This is a similar trend to that of the SFS material, which had
decreasing peak load averages with increasing volume fraction. The average load at
failure for SFSb20 is 8,119 N, for SFSb30 is 7,531 N, and for IPC40 is 6,615 N. To
compare with the behavior of the aluminum foam core sandwich, the average load at
failure for the SFS-b sandwich is normalized by the plateau load experienced by the
unfilled aluminum foam core sandwich. The average normalized load at failure for
SFSb20 is 20.76, for SFSb30 is 19.26, and for SFSb40 is 16.92. In Figure 5.9 these
values are shown as a histogram. The plateau load for the aluminum foam core sandwich
is the load at which the stress remains nearly constant while able to withstand a large
increase in strain. It is this load that is typically used as the desired design load. Refer to
Figure 3.5 for a typical stress-strain response of this type. The SFS-b sandwiches do not
experience a similar near constant load, but rather a steady increase (hardening behavior)
79
until failure. This load at failure, when compared to the plateau load, is seventeen to
twenty-one times higher than the design load for unfilled aluminum foam core
sandwiches. The SFS-b sandwiches show superior load bearing capacity compared to
their unfilled aluminum foam counterpart. SFS-b sandwiches also outperform both SFS
25
SFSb
20
Peak Load/ALS Plateau Load
15
10
0
20% 30% 40%
Volume Fraction
Figure 5.9 Normalized Measured Peak Load for Sandwich Structures with Syntactic
The average value for the mid-span deflection at failure for SFSb20 is 2.18 mm,
for SFSb30 is 2.37 mm, and for SFSb40 is 2.35 mm. To compare with the behavior of
the aluminum foam core sandwich, the average deflection at failure for the SFS-b
sandwich is normalized by the deflection at which the aluminum foam core sandwich
behavior deviates from linearity. The average normalized deflection at failure for
80
SFSb20 is 11.23, for SFSb30 is 12.26, and for SFSb40 is 12.13. In Figure 5.10 these
values are shown as a histogram. Note that within the experimental scatter, there is no
improvement between the 30% and 40% volume fractions although the latter is lighter
overall.
14.0
SFSb
12.0
10.0
δFailure/δALS, nonlinear
8.0
6.0
4.0
2.0
0.0
20% 30% 40%
Volume Fraction
Sandwich Structures with Syntactic Foam Core and Graded Face Sheets
The average value for the transition load for SFSb20 is 5,796 N, for SFSb30 is
5,725 N, and for SFSb40 is 5,090 N. As with the peak load, to compare to the behavior
of the aluminum foam core sandwich, the average transition load for the SFS-b sandwich
is normalized by the plateau load experienced by the aluminum foam core sandwich. The
average normalized transition load for SFSb20 is 14.82, for SFSb30 is 14.64, and for
SFSb40 is 13.02. In Figure 5.11 these values are shown as a histogram. This is the best
81
comparison that can be made between syntactic foam and unfilled aluminum foam core
sandwiches, as both have a transition point. With the SFS-b sandwich having a transition
load of around thirteen to fifteen times greater than unfilled aluminum foam sandwiches
and a similar transition deflection, the point at which plasticity begins to set in for the
in the transition load values within the experimental scatter between SFSb20 and
SFSb30, with SFSb40 specimens showing a slightly lower value relative to the two
others. This suggests that face sheet yielding is likely responsible for the transition point
in SFS-b sandwiches as the metallic ligaments of the graded face sheet help prevent shear
18
SFSb
16
Transition Load/ALS Plateau Load
14
12
10
0
20% 30% 40%
Volume Fraction
Figure 5.11 Normalized Measured Load at Transition Point for Sandwich Structures
82
The average value for the transition deflection for SFSb20 is 0.93 mm, for
SFSb30 is 0.99 mm, and for SFSb40 is 0.91 mm. For a comparison, the average
transition deflection for the SFS-b sandwich is normalized by the deflection at which the
aluminum foam core sandwich behavior deviates from linearity. The average normalized
transition deflection for SFSb20 is 4.79, for SFSb30 is 5.11, and for SFSb40 is 4.69. In
Figure 5.12 these values are shown as a histogram. With a four and a half to five times
improvement over unfilled aluminum foam core sandwiches, SFS-b sandwiches are able
to remain in the linear elastic regime much longer before the onset of plasticity.
6.0
SFSb
5.0
4.0
δTransition/δALS, nonlinear
3.0
2.0
1.0
0.0
20% 30% 40%
Volume Fraction
Point for Sandwich Structures with Syntactic Foam Core and Graded Face Sheets
As before, the strain energy (U) absorbed by a specimen was calculated using the
area under the load-deflection curve from the measured data. That is,
83
δ
U = ∫ P (δ ) d δ .
0
The average values for the strain energy absorbed are plotted in Figure 5.13 as a
histogram. The average for SFSb20 is 11.54 J, for SFSb30 is 12.01 J, and for SFSb40 is
10.90 J.
14.0
SFSb
12.0
10.0
Energy Absorbed (J)
8.0
6.0
4.0
2.0
0.0
20% 30% 40%
Volume Fraction
Figure 5.13 Energy Absorbed for Sandwich Structures with Syntactic Foam Core and
specific energy (J/kg) was calculated for each specimen. The strain energy absorbed was
divided by the mass of the particular specimen. The mass was determined by,
84
mtot = ρcoreVcore + ρ facesV faces .
The details of material densities used are in Table 4.5. The results for specific
strain energy are shown in Figure 5.14. The average for SFSb20 is 144.4 J/kg, for
SFSb30 is 164.0 J/kg, and for SFSb40 is 163.8 J/kg. Comparing the specific strain
energies of SFS-b and IPC sandwiches (which accounts for the weight difference in
removing the aluminum foam in the core), SFSb20 had a 71% improvement over its IPC
counterpart, SFSb30 had a 76% improvement, and SFSb40 had a 118% improvement.
With syntactic foam core outperforming IPC foam core for all volume fractions as well as
being lighter with the removal of the aluminum foam matrix, it can be concluded that
SFS-b sandwiches are far superior to a similar configuration with IPC foam core in their
energy absorbing capabilities. It is now apparent that the best core type is syntactic foam
rather than IPC. However, the face sheets simply need to be tailored to penetrate the SF
core in the face-core region to reap the full benefits of the face and of the core by
preventing premature face-core shear failure yet allowing effective load transfer between
85
200
SFSb
180
160
120
100
80
60
40
20
0
20% 30% 40%
Volume Fraction
Figure 5.14 Specific Energy Absorbed for Sandwich Structures with Syntactic Foam
the lower face sheet, likely due to a rupture in the SF core, with a crack propagating from
the tensile zone upwards towards the compression zone. Due to the dynamic nature of
failure initiation, it could not be confirmed if lower face sheet yielding preceded and
contributed to the core crack initiation or vice versa. Images of a failed SFS-b sandwich
are shown in Figure 5.15. Also, some magnified images of a failed SFS-b sandwich are
shown in Figure 5.16. In Figure 5.16 notice in (a) and (b) that the face sheets have
appeared to again fail in tension after yielding. Also notice the abrupt breaks in the
metallic ligaments in (c), as well as the SF still adhered to the metallic ligaments,
indicating good adhesion due to silane treatment. Though no images captured the actual
failure event, the abrupt face sheet separation seen in Figure 5.16 (a) and (b) indicates
86
that the failure initiated in the core, expanding downward to the face sheet and upward to
(a)
(b)
(c)
Figure 5.15 Failed SFS-b Specimen with (a) Side, (b) Top Face, and (c) Bottom Face
Views
87
(a) (b) (c)
Figure 5.16 Magnified Images (a), (b) of Failure Surfaces and (c) From Below
transition point, the behavior deviates from the initial linear elastic regime to the
compared to the theoretical mid-span deflection calculation from Section 1.5.1 [16]. The
load histories are used to calculate the theoretical deflections, along with geometric and
behavior, the quasi-static elastic properties used for the syntactic foam found by Jhaver
[17] were used once again. (Refer back to Figure 3.7 for the stress-strain responses from
which they were derived and Table 3.2 for the properties.) There is good agreement
between the theoretical and experimental values in the initial linear elastic regime for all
volume fractions. The predictions exhibit similar stiffness compared to those seen in the
measured data. Note that since the theoretical response is only for the linear elastic case,
the response after the transition point cannot be compared. The deviations after the
elastic region are possibly due to the fact that the elastic modulus for syntactic foam core
88
is assumed to be the same in tension as in compression. These comparisons between the
predicted and experimental behavior are shown below in Figure 5.17 to 5.19.
Material: SFSb20
9000 Temperature: 70°F
Specimen: FLEX-57
Cast: 23Jan2011
8000 Test: 10Feb2011
Rate: 0.025 in/min
7000
6000
Load (N)
5000
Measured
4000
Elastic
Prediction
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
Material: SFSb30
9000 Temperature: 70°F
Specimen: FLEX-59
Cast: 25Jan2011
8000 Test: 02Feb2011
Rate: 0.025 in/min
7000
6000
Load (N)
5000
Measured
4000
Elastic
Prediction
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
89
Material: SFSb40
8000 Temperature: 70°F
Specimen: FLEX-65
Cast: 01Feb2011
7000 Test: 10Feb2011
Rate: 0.025 in/min
6000
5000
Load (N)
Measured
4000
Elastic
3000 Prediction
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
core.
8000 SFSb20
SFSb30
7000 20%
6000 SFSb40
FLEX-57 (SFS20)
30%
Load (N)
5000
FLEX-59 (SFS30)
4000
40%
3000
FLEX-65 (SFS40)
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Crosshead Displacement (mm)
90
The transition deflection is virtually the same for all the three cases. After the
transition, nonlinear behavior is similar across the different cases with decreasing
stiffness as volume fraction increases. While SFSb30 has the greatest energy absorbing
capacity, the 30% and 40% Vf cases have the same specific energy absorbing capacity.
So if energy absorption is the driving factor, perhaps the 30% Vf case might be the best
core material. However, if both weight saving and energy absorbing capacity are both
important attributes, then the 30% and 40% Vf would both warrant consideration.
5.3 Summary
With the addition of the graded face sheets, the syntactic foam core sandwiches
(SFS-b) outperformed IPC foam core sandwiches in all metrics (comparisons are
discussed in Chapter 7). Not only did it outperform the IPC core sandwiches, but without
the metallic matrix in the core, the syntactic foam core sandwiches are lighter. It is easy
to see now that the IPC core sandwiches, though substantially better than conventional SF
core sandwiches, was an inferior material for flexural loading conditions due to its
weakness under the tensile conditions. As a single continuous phase the syntactic foam
does not experience the interphase separation that IPC foam does that essentially acts as
micro-defects, ultimately leading to tensile failure. The core integrity is better with the
syntactic foam. These results were not seen with the first type of SFS sandwiches. The
poor bond between the sanded face sheets and the core was not able to effectively
transfer load between the face and the core. Failure would occur prematurely when the
91
face sheet and core would debond, leaving the core to bear the entire load. The graded
face sheets in the SFS-b sandwiches remedied this problem, allowing for effective load
92
CHAPTER 6
CORRELATION METHOD
Digital image correlation (DIC) is a relatively new optical method and is based on
the principle of locating a point (or region of interest) in the deformed image of an object
relative to its location in the undeformed state [20,21,22,23]. The object surface is
decorated with a random pattern (in this case a random black/white speckle). The
decoration is illuminated using white light sources to produce diffusely reflected light
from the specimen surface. The light intensity (grayscales) of the surface is used for the
pattern matching process. The recording of the surface patterns is done using a digital
camera with a relatively high pixel count (typically on the order of one to ten megapixels)
converted into a discrete field D(x,y). The output from an image is typically a two
discrete intensity fields D and D’ are available for two load levels (or, reference and
deformed states), then a subset d from D (say, a 15 x 15 pixel subimage from a 1000 x
93
1000 pixel image) is chosen and its location in D’ (also a 1000 x 1000 pixel image) is
searched.
Figure 6.1. The highlighted box on the undeformed image indicates the area to track
during correlation. The box on the deformed image indicates the same area after
6.2.
(a) (b)
94
UNDEFORMED SUBIMAGE
•
P
∆x • u
O
v
•
• P’
O’
∆y
DEFORMED SUBIMAGE
can be expressed using affine coordinate transformations to relate O’(x’,y’) with O(x,y).
This requires an accurate interpolation method to express intensity variation within the
subimage.
D (O ) = D ( x, y )
D ( P ) = D ( x + dx, y + dy )
dx and dy. After deformation the intensity patterns recorded are expressed using
x ' = x + u ( x, y ); y ' = y + v ( x, y ) .
95
So,
∂u ∂u ∂v ∂v
nearly constant within the subimage, one can obtain u, v, , , , for each
∂x ∂y ∂x ∂y
subimage. This is generally done using one of the following measures: (1) magnitude of
intensity difference, (2) sum of the squares of intensity value difference, (3) normalized
is maximized to provide the optimal estimation of all the six displacements and
displacement gradients. A number of optimization methods can be used for this purpose
96
6.2 ARAMIS™ Image Analysis Software
Full field deformation and strain analysis was performed using the 2D image
correlation technique utilizing a spray-on speckle pattern. The software used for the
image post-processing was ARAMIS™ (GOM mbH, Germany). It is able to provide full
deformation during a test. A Nikon D100 digital camera was used during experiments to
record the grayscales (at 10-bit resolution) corresponding to the specimen deformation.
Although the camera had a 6 megapixel resolution, during time-lapse photography only
1500 x 1000 pixels were used for easy throughput of grayscale information into a laptop
using Nikon Capture software (Version 4.0). The entire image is regularly divided into a
the images track facets in each successive image. From each facet, a single measurement
results after computation. The facet size used (15 x 15 pixels) was a good compromise
between accuracy and computational time. The goal is to have a good representation of
the speckle pattern within the facet. This digital image processing provides full field
displacement and strain contours throughout the test. Furthermore, not only can this be
∂u ∂u ∂v ∂v
The deformation gradient tensor F (comprised of , , , ) creates a
∂x ∂y ∂x ∂y
functional connection of the coordinates of the deformed points Pv,i with the coordinates
of the undeformed points Pu,i [26]. They are connected through the relation
Pv ,i = u j + F gPu ,i ,
97
where,
The different coordinate systems defined through the calculation process are
shown in Figure 6.3. The coordinates of the points (e.g. pu and pv in the figure) are
calculated in the global x-y coordinate system. For 2D analysis, the local undeformed
coordinate system x’-y’ is parallel to x-y, but moved to coincide with the point pu. The
x”-y” coordinate system for the strain calculation is independent of rigid body movement
and rotation.
To calculate the deformation gradient tensor F, the coordinates of each point must
be known both in the undeformed and deformed state. This tensor can be interpreted as
98
an affine transformation which transforms a unit square into a generic quadrilateral. In
order to calculate F for a point, a number of neighboring points are needed. For this
calculation, a homogeneous state of strain is assumed for the set of adjacent points. The
width of the field is the reference length for evaluating strain. (For example in Figure 6.4,
the width of the field is three points or pixels. The reference length is then three pixels.)
The number of surrounding points can be changed, but a 3x3 neighborhood is shown in
Section 3.1 details the TPB test setup used in conjunction with DIC method and
time-lapse photography. The impact test counterparts were carried out using the speckle
99
6.4 Optical Measurement Results
computed and illustrated. With this capability, it is relatively simple to obtain local
displacement and strain information, rather that relying on the global load and
displacement measurements made using the testing machine. To ensure the ARAMIS™
results are reliable, local and global displacement measurements were compared for each
specimen just below the load roller (the machine displacement data given is for the
crosshead). An example of this is shown in Figure 6.5. This is one example, with every
specimen compared and shown in Appendix A. Since the camera recorded images in ~2
Material: IPC30
6000 Temperature: 70°F
Specimen: FLEX-16
Cast: 15Apr2010
Test: 09Jun2010
Rate: 0.025 in/min
5000
4000
Load (N)
3000 Measured
ARAMIS
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
100
Also, to check the strain calculations, a sample (129mm length x 5.7mm width x
21mm height) of PMMA (E = 3 GPa) was tested in three-point bending and the strains
calculated using ARAMIS™. These dimensions were chosen to match the planar
Figure 6.6. The measured load was used to calculate the maximum tensile stress using
My σx
σx = and theoretical elastic strain was calculated using ε x = . The optical
I x = 0; y = c E
measurement of strain and the theoretical prediction are plotted in Figure 6.6, with good
agreement between the two being evident. The error bars are determined from the plot in
Figure 6.7. Also plotted are strain (εx) and shear strain (εxy) variations, in Figures 6.7 and
6.8, respectively, along the beam height (y) at x=25mm (x=0 is the mid-span of the
VQ τ
beam). Theoretical shear strain is calculated using τ = and ε xy = . A rather good
Ib 2G
agreement is seen in the εx plot. Note the strange nonlinear behavior at the top and
bottom of the sample. Considering the known behavior of PMMA, it can be concluded
that any optical measurements around the top and bottom of the specimen have an
increased amount of error. This plot shows a maximum error of approximately 600
µstrains, which was determined to be the error associated with optical strain
measurements. The εxy plot trend was the same as the elastic prediction with an error of
101
Material: PMMA
70 Temperature: 70°F
Specimen: PMMA-1
Test: 04Mar2011
Rate: 0.025 in/min
60
50
Stress (MPa)
40
Measured
30 ARAMIS
20
10
0
0.0 0.5 1.0 1.5 2.0 2.5
Strain (%)
Material: PMMA
Temperature: 70°F
Rate: 0.025 in/min
9.2 Load: 1,650 N
6.7
4.2
1.7
y (mm)
Elastic
Predicition
-0.8
ARAMIS
-3.3
-5.8
-8.3
-10.8
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
ε x (%)
Figure 6.7 PMMA Strain (εx) Along the Height of the Beam
102
Material: PMMA
Temperature: 70°F
Rate: 0.025 in/min
9.2 Load: 1,650 N
6.7
4.2
1.7
y (mm)
Elastic
Prediction
-0.8
ARAMIS
-3.3
-5.8
-8.3
-10.8
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007
ε xy (rad)
Figure 6.8 PMMA Shear Strain (εxy) Along the Height of the Beam
The series of images shown in Figures 6.9 to 6.17 are typical ARAMIS™ results
for tests conducted on SFS sandwiches. Each of these experiments are summarized in
sets of three images: the first image corresponding to a point in the linear response
regime, the second image corresponding to just before failure, and the third image
corresponding to just after failure. Note the presence of strain concentration due to the
loading roller in particular. Though it is hard to conclusively determine given the low
amount of deflection in the beam, the u-displacement fields are similar to what is
expected from the beam theory. Note the units: u-displacement is measured in
millimeters, εx is measured in percent, and εxy is measured in strains. Also note that the
103
y
x
P=2346N
P=4946N
P=2869N
104
y
x
P=2346N
P=4946N
P=2869N
105
y
x P=2346N
P=4946N
P=2869N
106
y
x P=2236N
P=5468N
P=3618N
107
y
x P=2236N
P=5468N
P=3618N
108
y
x P=2236N
P=5468N
P=3618N
109
y
x P=2069N
P=4080N
P=2459N
110
y
x P=2069N
P=4080N
P=2459N
111
y
x P=2069N
P=4080N
P=2459N
112
6.4.2 IPC Results
The series of images shown in Figures 6.18 to 6.26 are typical optical results for
tests conducted on IPC sandwiches. In each of these sets of four images, the first image
corresponds to a point in the linear portion of the loading regime, the second image
corresponds to a point in the nonlinear regime, the third image corresponds to the point
just before failure, and the fourth image corresponds to the point just after failure. Notice
that with the IPC, the strain concentration effects due to the rollers are relatively muted,
as the interpenetrating core does a good job of evenly distributing the load. Again, u-
displacement fields qualitatively resemble the ones expected from the beam theory.
113
y
x P=824N
P=3536N
P=4922N
P=3905N
114
y
x P=824N
P=3536N
P=4922N
P=3905N
115
y
P=824N
x
P=3563N
P=4922N
P=3905N
116
y
x P=1097N
P=3997N
P=5680N
P=3741N
117
y
x
P=1097N
P=3997N
P=5680N
P=3741N
118
y
x P=1097N
P=3997N
P=5680N
P=3741N
119
y
x P=950N
P=2845N
P=4442N
P=3159N
120
y
x P=950N
P=2845N
P=4442N
P=3159N
121
y
x P=950N
P=2845N
P=4442N
P=3159N
122
6.4.3 Syntactic Foam (SFS-b) Results
The series of images shown in Figures 6.27 to 6.35 are typical ARAMIS™ results
for tests conducted on syntactic foam core sandwiches with graded face sheets (SFS-b).
In each of these sets of three images, the first image corresponds to a point in the linear
regime, the second image corresponds to a point in the nonlinear regime, and the third
image corresponds to the point just before failure. In the SFS-b sandwiches, the
concentration effects due to the rollers once again dominate. Also of note is that the
shear strains are higher compared to that of the SFS sandwiches due to the more efficient
load transfer throughout the structure enabled by the graded face sheets. Once again, the
u-displacement fields are what you would expect from the beam theory.
123
y
x P=3491N
P=7187N
P=8155N
124
y
x P=3491N
P=7187N
P=8155N
125
y
P=3491N
x
P=7187N
P=8155N
126
y
x P=2990N
P=6717N
P=7559N
127
y
x P=2990N
P=6717N
P=7559N
128
y
x P=2990N
P=6717N
P=7559N
129
y
x P=2207N
P=6054N
P=6735N
130
y
x P=2207N
P=6054N
P=6735N
131
y
x P=2207N
P=6054N
P=6735N
132
6.5 Impact Tests
The impact tests conducted in this work were only conducted on IPC and SFS-b
types using a single volume fraction (30%) of microballoons in the SF. The purpose of
these tests was to further elaborate on the failure mechanisms in the quasi-static case, as
Also, the feasibility of using DIC method in conjunction with a high-speed digital camera
Impact tests were conducted using an Instron Dynatup 9250-HV drop tower to
deliver low velocity impact to IPC foam core and SFS-b sandwich structures with 30%
volume fraction of microballoons in the SF. The nominal specimen size used for
dynamic testing was 127mm x 25mm x 18.5mm (5” length x 1” width x 0.73” height).
Each face sheet had a nominal thickness of 0.813mm, giving a core thickness of
~16.9mm. The specimen was supported over a 4.5 inch span and was impacted by a
Two different cameras were used to capture the dynamic event. The first was a
Phantom v710 digital high-speed camera with a single 1280 x 800 CMOS sensor. This
camera recorded a video of the event and was used to determine the overall impact
response on a large time scale. (Using a resolution of 800 x 600 pixels and a framing rate
of 13,000 frames per second, the total recording duration was ~0.75 seconds.) Images
from the Phantom camera were used to determine the time at which crack initiation was
likely to occur. (To view a video of the impact event created using the Phantom camera,
refer to the disk in the back cover of this thesis.) Using this estimated time, parameters
133
were selected for the second camera, a Cordin Model 550 high-speed digital camera,
which captured the speckle patterns during the dynamic event over a time period of ~600
higher spatial resolutions (1000 x 1000 pixels) and speeds and hence was preferred to the
Phantom camera.) The drop tower has an instrumented tup and a pair of anvils for
recording impact force and reaction force histories, respectively. Two high intensity
flash lamps, triggered by the camera, were used to illuminate the specimen surface. Two
separate computers were also used in the setup. One was used to record the force
histories, and the other was used to control the camera and store the digital images. A
schematic of the test setup with the Cordin camera is shown in Figure 6.36. The setup
consists of (1) high-speed digital camera, (2) drop tower impactor tup, (3) delay
generator, (4) specimen, (5) light sources, (6) DAQ – drop tower, (7) DAQ – camera, (8)
Region of Interest
(25mm x 25mm)
Figure 6.36 Schematic of Impact Test Setup: (1) high-speed digital camera, (2) drop
tower impactor tup, (3) delay generator, (4) specimen, (5) light sources, (6) DAQ – drop
tower, (7) DAQ – camera, (8) lamp control unit, and (9) drop tower controller
134
A picture of the impact test setup is shown in Figure 6.37 which shows (1) high-
speed camera, (2) drop tower impact tup, (3) specimen, and (4) light source.
4
1
Figure 6.37 Impact Test Setup: (1) high-speed camera, (2) drop tower impactor tup,
mirror optical system. The camera is capable of capturing images at a rate of 2,000,000
frames per second at a resolution of 1000 x 1000 pixels for each image. Thirty-two (32)
speed rotating mirror that reflects and sweeps light over the sensors. The 32 sensors
discrepancies between successive images. This is resolved by grouping the images into
pairs that consist of an undeformed and a deformed image recorded by the same sensor.
135
Before impact a set of 32 images was recorded in the area of interest with the desired
frame rate (50,000 frames per second for this work). After aligning the optics, the
rotating mirror was brought to the desired speed. The camera and lamps were triggered
by the operator. This set of 32 images (one per sensor) of the specimen surface was then
stored for use as undeformed, or reference, images. Maintaining all the same camera
settings, a second set of 32 images was recorded as the specimen experienced the impact
event. During the test, the camera and flash lamps were triggered by the tup as it
contacted a copper strip adhered to the top of the specimen. The first image was
identified, and then sequential pairs were made in the undeformed and deformed sets
(i.e., in the first set of 32 undeformed images, each image has a corresponding deformed
image in the second set). Each of the 32 pairs was analyzed separately. (To view a video
of the impact event created using the Cordin camera, refer to the disk in the back cover of
this thesis.) The impactor velocity used in this work was 5 m/s. The camera viewing
IPC foam core sandwiches with a 30% volume fraction of microballoons in the
SF core were impact tested. Figures 6.39 and 6.40 show a sample of correlated images
136
from the Cordin camera on an IPC30 specimen. Shown here are the u-displacement and
normal strain (εx) fields. Note that the correlated images contain twenty contours. The u-
t=510µs
t=530µs
t=550µs
Crack
initiation
137
t=510µs
t=530µs
t=550µs
Crack
initiation
The normal (dominant) strain, εx, image preceding visible crack initiation (t=530
µs) contains an anomaly not seen in the previous images in the area where the crack
forms. Also, no significant strain concentration at the impact edge is seen since the
impactor head is hemispherical and the front face of the specimen is far removed from
the specimen mid-plane. Since the crack first forms in the core, then propagates to the
lower face and upper compression zone, it can be concluded that the mechanism likely
138
responsible for both quasi-static and dynamic failure in the IPC foam core sandwiches is
also impact tested. Figures 6.41 and 6.42 show a sample of correlated images from the
Cordin camera on an SFSb30 specimen. Shown here are the u-displacement and normal
t=510µs
t=530µs
t=550µs
Crack
initiation
139
t=510µs
t=530µs
t=550µs
Crack
initiation
Similar to the IPC foam core sandwiches, the crack first forms in the tensile
region of the SFS-b sandwich core, only then propagating to the lower face and upper
compression zone. It can be concluded that the initial mechanism likely responsible for
both quasi-static and dynamic failure in the SFS-b foam core sandwiches is also tensile
Figures 6.43 and 6.44 show plots of the strain energy absorbed by the specimen as
a function of time for the IPC30 and SFSb30 sandwiches, respectively. The results are
140
summarized in Tables 6.1 and 6.2. The Impulse software (for the drop tower)
E (t ) = K (t ) + V (t ) + Ea (t ) = Const
where E(t) is the total energy in the system (which remains constant), K(t) is the kinetic
energy of the drop weight, V(t) is the potential energy of the drop weight, and Ea(t) is the
strain energy absorbed by the specimen up to time t. Graphically, Ea(t) is simply the area
under the load-deflection curve. It is calculated mathematically using the mass (m),
velocity (v(t)), and position (x(t)) of the drop weight (trapezoid integration used to find
1
Ea (t ) = m vimpact
2
− v 2 (t ) + mgx(t ) .
2
Material: IPC30
16 Temperature: 70°F
Core is Core is Overlay
intact cracked;
14
specimen
continues
12 to absorb
energy
Energy Absorbed (J)
10 Core crack
initiation
DYN-18
8 (IPC30)
DYN-20
(IPC30)
6 DYN-29
(IPC30)
0
0.0 0.5 1.0 1.5 2.0
Time (ms)
141
Material: SFSb30
16 Temperature: 70°F
Overlay
14
Core crack
12
Energy Absorbed (J) initiation
10
DYN-34
8 (SFSb30)
DYN-35
(SFSb30)
6 DYN-36
(SFSb30)
0
0.0 0.5 1.0 1.5 2.0
Time (ms)
142
Two major aspects of note are the rate of energy absorption and the total strain
energy absorbed by the specimen. While the SFSb30 sandwiches had a higher rate of
energy absorption, the IPC30 sandwiches actually had greater energy absorption capacity
under impact conditions. When compared to IPC30 sandwiches, the SFSb30 sandwiches
had 22.1% less strain energy absorption and a 14.2% drop in specific energy absorption.
Both architectures experienced visible core crack initiation at ~550 µs. Note the
difference in Figures 6.43 and 6.44 after that time. SFS-b sandwich energy absorption
only occurs from time zero up to the time of crack initiation. The IPC foam core
The impact tests in this work were only conducted using 30% volume fraction of
microballoons in the SF. The purpose of these tests was to further expound on the failure
mechanisms in the dynamic case. Also, the feasibility of using DIC method in
conjunction with a high speed digital camera to study impact behavior of sandwich
structures is illustrated. Contrary to the quasi-static case, IPC foam core sandwiches
appear to outperform the syntactic foam core sandwiches with graded face sheets. The
presence of the interpenetrating phases seems to have a positive effect under dynamic
conditions, though the crack still initiates in the tensile region of the core.
143
CHAPTER 7
CONCLUSIONS
In this thesis, flexural responses of sandwich structures with syntactic foam (20%-
40% volume fractions) core were studied. Three types of sandwich structures were
studied. The first type had a regular syntactic foam core (identified as SFS) with simply
adhered aluminum face sheets. The second one employed an aluminum-syntactic foam
seen in the SFS sandwiches. The third type consisted of a syntactic foam core with
graded face sheet (identified as SFS-b). Static three-point bend tests were carried out on
all three types of sandwich structures and global load-deflection responses were
peak load, mid-span deflection at failure, nonlinearity of the response, and strain energy
measurements, measured normal and shear strain fields, and optical microscopy were
144
In the case of the syntactic foam core sandwiches (SFS), the load-deflection
behavior was essentially linear elastic in nature. However, for the IPC foam core
sandwiches and syntactic foam core sandwiches with graded face sheets (SFS-b), there is
large scatter in measured peak load and crosshead deflection at failure. They also
showed very limited nonlinearity in load-deflection response. The IPC core sandwiches
reduction. The nonlinearity in this architecture is primarily due to face sheet yielding as
well as debonding between the phases of the interpenetrating composite foam core. The
SFS-b sandwiches, on the other hand, showed substantial improvement in both the peak
load and mid-span deflection at failure and exhibited nonlinearity in the response similar
to IPC foam core sandwiches. This is attributed to a gradual shear strain variation at the
graded face-core interface (unlike the SFS sandwiches) resulting in face sheet yielding
Considering that the SFS material possibly did not reach its full potential and
comparing the specific strain energies (which accounts for the weight difference in
adding the aluminum foam into the core), IPC20 had a 25% improvement, IPC30 had a
59% improvement, and IPC40 had an 88% improvement. While the peak load decreased
slightly, for the small penalty of added weight the IPC foam core significantly improved
the value of deflection-to-failure experienced and thus the strain energy absorbed.
145
With the IPC30 having the highest average values for all metrics in the IPC foam
core test group (most pronounced specific energy absorbed), an optimum IPC foam
Though IPC foam core sandwiches outperformed the SFS counterparts, the
syntactic foam core sandwiches with graded face sheets (SFS-b) outperformed the IPC
foam core sandwiches. This architecture was able to maintain linear load-deflection
behavior much longer than the IPC foam core architecture with an improvement in load
at the (linear to nonlinear) transition point of 244%, 230%, and 211% for SFSb20,
transition point were 222%, 234%, and 225% for SFSb20, SFSb30, and SFSb40,
respectively. Comparing the specific strain energies to those of IPC foam core
sandwiches (which accounts for the weight difference in the core), SFSb20 had a 71%
improvement, SFSb30 had a 76% improvement, and SFSb40 had a 118% improvement.
aluminum face sheets separated by an empty core. However, this being impractical, the
unfilled 20ppi aluminum foam core sandwich response was used to normalize the data.
146
25
SFS
IPC
SFSb
20
10
0
20% 30% 40%
Volume Fraction
Figure 7.1 Normalized Measured Peak Load for Sandwich Structures with SF and IPC
Foam Cores
14
SFS
12 IPC
SFSb
10
δFailure/δALS, nonlinear
0
20% 30% 40%
Volume Fraction
147
18
IPC
16
SFSb
12
10
0
20% 30% 40%
Volume Fraction
Figure 7.3 Normalized Measured Load at Transition Point for Sandwich Structures with
6
IPC
5 SFSb
4
δTransition/δALS, nonlinear
0
20% 30% 40%
Volume Fraction
148
14
SFS
IPC
12 SFSb
10
0
20% 30% 40%
Volume Fraction
Figure 7.5 Energy Absorbed for Sandwich Structures with SF and IPC Foam Cores
200
SFS
180 IPC
SFSb
160
Specific Energy Absorbed (J/kg)
140
120
100
80
60
40
20
0
20% 30% 40%
Volume Fraction
Figure 7.6 Specific Energy Absorbed for Sandwich Structures with SF and IPC Foam
Cores
149
The superior material based on all these metrics is the syntactic foam core
sandwiches with graded face sheets (SFS-b). The only instance it might not be
considered the optimal choice might be when considering the IPC foam core sandwiches
for their ability to use the metallic foam network to hold it together longer. Depending
Using the calculated strains from the image correlation process, plots of dominant
strain (εx) and shear strain (εxy) as a function of thickness were obtained. The strains
extracted were at x=25 millimeters (from the center of the beam). This distance from the
center was approximately half way between the loading and support rollers, minimizing
the stress concentration effects created by the rollers. The syntactic foam (SFS), IPC
foam, and syntactic foam (SFS-b) core material of the same volume fraction were
compared for a similar applied load. Figure 7.7 shows an example of the dominant
flexural normal strain εx results at a load of ~4,000 N for the 20% volume fraction of
150
Material: IPC/SFS
Temperature: 70°F
9.35 Rate: 0.025 in/min
Load: 4,000 N
6.85
4.35
1.85
y (mm)
FLEX-38
(SFS20)
-0.65 FLEX-10
(IPC20)
FLEX-57
-3.15 (SFSb20)
-5.65
-8.15
-10.65
-0.45 -0.35 -0.25 -0.15 -0.05 0.05 0.15 0.25 0.35 0.45
ε x (%)
Figure 7.7 20% Vf: Strain (εx) Variation Along the Height of the Sandwich Core
At 4,000N load, the SFS sandwich was still in the linear region, while the IPC
foam core and SFS-b sandwiches had entered the nonlinear regime. Notice that the SFS
and SFS-b sandwiches are generally exhibiting similar trends. This makes intuitive sense
considering the core material in each is the same syntactic foam. When comparing the
results with linear beam theory (where a linear strain variation through the thickness is
expected), one can observe that near the top and bottom of the beam, the behavior
deviates from linearity. This can be explained by the effects of face sheets restricting the
motion of the neighboring material points in the core, and by the increased amount of
image processing errors near the face sheets. Also, there is a shift in the neutral axis from
the center of the beam in the SF core cases. This can be explained by the severity of the
stress concentration effects of the loading point in the SF core cases. The expected
general trend of greater strain experienced in the IPC foam core compared to syntactic
foam of the same volume fraction is confirmed. Remember that the SFS sandwiches
151
failed because of the intense shear stresses seen at the faces, causing face-core
debonding. The main idea to take away from the shear strain plots is that both IPC core
and SFS-b sandwiches have done a better job of mitigating the intensity of the shear
strains (and thus stresses) at the faces. Figure 7.8 shows an example of the εxy results at
a load of ~4,000N for the 20% volume fraction of microballoons in the SF.
Material: IPC/SFS
Temperature: 70°F
9.35 Rate: 0.025 in/min
Load: 4,000 N
6.85
4.35
1.85
y (mm)
FLEX-38
(SFS20)
-0.65 FLEX-10
(IPC20)
FLEX-57
-3.15 (SFSb20)
-5.65
-8.15
-10.65
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030 0.0035 0.0040
ε xy (rad)
Figure 7.8 20% Vf: Shear Strain (εxy) Variation Along the Height of the Sandwich Core
In the impact tests, two major aspects of note are the rate of energy absorption and
the total strain energy absorbed by the specimen. While the SFSb30 sandwiches had a
higher rate of energy absorption, the IPC30 sandwiches actually had greater energy
absorption capacity under impact conditions. When compared to IPC30 sandwiches, the
SFSb30 sandwiches had 22% less strain energy absorption and a 14% drop in specific
energy absorption. The SFSb30 absorbed energy at a rate of ~20 J/ms, while the IPC30
152
absorbed at a rate of ~12 J/ms. Both architectures experienced visible core crack
initiation around the same time. SFS-b sandwich energy absorption only occurs from
time zero (impact) up to the time of crack initiation. The IPC foam core sandwiches,
however, continue to absorb energy even after the onset of a crack resulting in greater net
The primary focus of this work was to understand flexural failure behavior of IPC
and syntactic foam core sandwiches under quasi-static loading conditions. An attempt
was made to illustrate the feasibility of studying the same under impact loading
conditions. Experiments indicate an improved response of the IPC foam core sandwich
over the SFS sandwich, including energy absorption. However, in light of the failure
mechanism typically observed in the IPC foam core, an effort was made to create a
syntactic foam core sandwich with a graded face sheet to promote better load transfer
throughout the sandwich. The superior material that emerged and is recommended is the
characterize the core material in tension and bending. This study has not tested any of
the core material without face sheets, and therefore cannot speculate on the properties of
the core alone. One particular future objective of interest would be to obtain mechanical
properties (elastic modulus, Poisson’s ratio) of the core materials in bending and tension.
Jhaver [17] was able to obtain mechanical properties of the materials in question, but in
153
compression only. While the experimentally derived properties from compression should
experimentally.
The failures seen in the SFS sandwiches stemmed from debonding of the face-
core interface. The face sheet would typically separate from the core, leaving the core as
the only load bearing component. Very shortly after this debond, the core would
catastrophically fail. This is unlike the IPC sandwich where a more synergistic
relationship between the core and face was observed. However, with the IPC core
sandwich, interphase separation (between aluminum ligaments and SF) in the tensile
region was the source of failure. To better harness the face-core relationship for the
syntactic foam core, smoother transition from face sheet to core needs to exist. One
possible solution is to have face sheets with a thin layer of brazed aluminum foam similar
to that of the IPC. A better response was achieved by better linking the face and core
(SFS-b sandwiches), preventing the debond typically seen in the SFS sandwiches. The
useful extent of this thin layer of foam was not explored here. In this work all of the
graded face sheets had a layer of ~2 mm of aluminum foam brazed to the aluminum face
sheet. It would be beneficial to test varying thicknesses of this brazed layer of foam to
determine at what depth the best overall response occurs while still maintaining the
integrity of the syntactic foam core. Another closely related idea that could be explored
would be to change the method by which the core and face are fused. With such stiff
154
The impact tests in this work were only conducted using 30% volume fraction of
microballoons in the SF. The purpose of these tests was to further elaborate on the
failure mechanisms in the quasi-static case, as well as begin to understand the failure
mechanisms in the dynamic case. Future work should involve varying the microballoon
volume fraction. Also, the feasibility of using DIC method in conjunction with a high
Contrary to the quasi-static case, IPC foam core sandwiches appear to outperform the
syntactic foam core sandwiches with graded face sheets. The presence of the
interpenetrating phases seems to have a positive effect under dynamic conditions. These
tests serve as a first step on the path to dynamic characterization of sandwich structures
155
REFERENCES
1. Gibson, L.J., Ashby, M.F., 1988. Cellular Solids: Structure and Properties.
Pergamon Press, Oxford, England.
2. Wu, C.L., Weeks, C.A., Sun, C.T. Improving Honeycomb-Core Sandwich Structures
for Impact Resistance. Journal of Advanced Materials 26 (1995) 41-47.
6. Zhang, Q., Lee, P.D., Singh, R. Wu, G., Lindley, T.C. Micro-CT Characterization of
Structural Features and Deformation Behavior of Fly Ash/Aluminum Syntactic Foam.
Acta Materialia 57 (2009) 3003–3011.
7. Zhou, W., Hu, W., Zhang, D. Study on the Making of Metal-Matrix Interpenetrating
Phase Composites. Scripta Materialia, Vol. 39, No. 12 (1998) 1743–1748.
156
9. Styles, M., Compston, P., Kalyanasundaram, S. The Effect of Core Thickness on the
Flexural Behaviour of Aluminum Foam Sandwich Structures. Composite Structures
80 (2007) 532–538.
10. McKown, S., Mines, R.A.W. Static and Impact Behaviour of Metal Foam Cored
Sandwich Beams. High Performance Structures and Materials II (2004) 37-46.
11. McCormack, T.M., Miller, R., Kesler, O., Gibson, L.J. Failure of Sandwich Beams
with Metallic Foam Cores. International Journal of Solids and Structures 38 (2001)
4901-4920.
12. Chen, C., Harte, A-M., Fleck, N.A. The Plastic Collapse of Sandwich Beams with a
Metallic Foam Core. International Journal of Mechanical Sciences 43 (2001) 1483-
506.
14. Travitzky, N.A. Effect of Metal Volume Fraction on the Mechanical Properties of
Alumina/Aluminum Composites. Journal of Materials Science 36 (2001) 4459–4463.
15. Chang, H., Higginson, R., Binner, J. Microstructure and Property Characterisation of
3-3 Al(Mg)/Al2O3 Interpenetrating Composites Produced by a Pressureless Infiltration
Technique. J Mater Sci 45 (2010) 662–668.
16. Allen, H.G., 1969. Analysis and Design of Structural Sandwich Panels. Pergamon
Press, Oxford, England.
18. Sadler, E.J., Vecere, A.C. Silane Treatment of Mineral Fillers – Practical Aspects.
Plastics, Rubber, and Composites Processing and Applications 24 (1995) 271-275.
19. ASTM C393/C393M-06. Standard Test Method for Core Shear Properties of
Sandwich Constructions by Beam Flexure. 2006.
21. Chu, T.C., Ranson, W.E., Sutton, M.A., Peters, W.E. Application of Digital Image
Correlation Techniques to Experimental Mechanics. Experimental Mechanics 25 (3)
(1985) 232-244.
157
22. McNeill, S.R., Sutton, M.A., Miao, Z., Ma, J. Measurement of Surface Profile Using
Digital Image Correlation. Experimental Mechanics 37 (1) (1997) 13-20.
23. Sun, Z., Lyons, J.S., McNeill, S.R. Measuring Microscopic Deformations Using
Digital Image Correlation. Optics and Lasers in Engineering 27 (1997) 409-428.
24. Tyson, J., Schmidt, T., Galanulis, K. Advanced Photogrammetry for Robust
Deformation and Strain Measurement. Proceedings of SEM 2002 Annual
Conference, Milwaukee, WI, June 2002.
25. Tyson, J., Schmidt, T., Galanulis, K. Dynamic Strain Measurement Using Advanced
3D Photogrammetry. Proceedings of IMAC XXI, Kissimmee, FL, Feb 2003.
158
APPENDIX A
GLOBAL/LOCAL DISPLACEMENT COMPARISONS
The following are comparisons between the global load-displacement data from
the machine and the local displacement data from the image correlation for each of the
specimens used as examples throughout the thesis.
4000
Load (N)
3000 Crosshead
ARAMIS
2000
1000
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Displacement (mm)
159
6000 Material: SFS30
Temperature: 70°F
Specimen: FLEX-39
Cast: 04Oct2010
Test: 26Oct2010
5000 Rate: 0.025 in/min
3000 Crosshead
ARAMIS
2000
1000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Displacement (mm)
3500
3000
2500
Load (N)
Crosshead
2000
ARAMIS
1500
1000
500
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Displacement (mm)
160
6000 Material: IPC20
Temperature: 70°F
Specimen: FLEX-10
Cast: 30Sept2009
Test: 21Apr2010
5000 Rate: 0.025 in/min
3000 Crosshead
ARAMIS
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
4000
Load (N)
3000 Crosshead
ARAMIS
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
161
5000 Material: IPC40
Temperature: 70°F
Specimen: FLEX-14
Cast: 07Oct2009
Test: 21Apr2010
Rate: 0.025 in/min
4000
3000
Load (N)
Crosshead
2000 ARAMIS
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
7000
6000
5000
Load (N)
Crosshead
4000
ARAMIS
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
Figure 1.1-1 Flexural Response of SF Sandwich (20%) at Room Temperature
162
9000 Material: SFSb30
Temperature: 70°F
Specimen: FLEX-59
Cast: 25Jan2011
8000 Test: 02Feb2011
Rate: 0.025 in/min
7000
6000
5000
Load (N)
Crosshead
4000
ARAMIS
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (mm)
Figure 1.1-1 Flexural Response of SF Sandwich (30%) at Room Temperature
6000
5000
Load (N)
4000 Crosshead
ARAMIS
3000
2000
1000
0
0.0 0.5 1.0 1.5 2.0 2.5
Displacement (mm)
Figure 1.1-1 Flexural Response of SF Sandwich (40%) at Room Temperature
163
APPENDIX B
SELECT ARAMIS™ IMAGES
The series of images shown in Figure B.1 are typical ARAMIS™ εy results for
tests conducted on IPC sandwiches. This case (IPC30) is simply used as an example to
illustrate the εy trends. In the set of four images, the first image corresponds to a point in
the linear regime, the second image corresponds to a point in the nonlinear regime, the
third image corresponds to the point just before failure, and the fourth image corresponds
to the point just after failure. Note the units are measured in percent. Also note that the
correlated images for IPC30 contain twenty contours with an increment of 0.15%. The
The series of images shown in Figure B.2 are typical ARAMIS™ εy results for
tests conducted on SFS-b sandwiches. Again, this case (SFSb30) is used as an example.
In the set of three images, the first image corresponds to a point in the linear regime, the
second image corresponds to a point in the nonlinear regime, and the third image
corresponds to the point just before failure. As before, the units are in percent. The
correlated images for SFSb30 also contain twenty contours, but this time with an
164
y
x P=1097N
P=3997N
P=5680N
P=3741N
165
y
x P=2990N
P=6717N
P=7559N
166