Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Interlocked Composite Grids

Design and Manufacturing

DONGYUP HAN* AND STEPHEN W. TSAI


Department of Aeronautics and Astronautics
Stanford University
Stanford, CA 94305-4035, USA

(Received October 6, 2001)


(Revised May 24, 2002)

ABSTRACT: Composite grid structures made from pultruded unidirectional glass


or carbon ribs promise to provide unmatched performance/cost combination of any
composite panels. A new manufacturing method for a square grid using slotted joint
and adhesive bonding (Interlocked Composite Grid or ICG) with pultruded ribs has
been developed. Compared to previously proposed grids, it is mass producible at low
cost. It has large potential for many applications, particularly for large civil
structures. Equivalent stiffness models for the proposed ICG, in equivalent plate
stiffness matrix form and equivalent engineering constant form, have been
formulated and verified for the flexural behavior of ICGs. The effects of the slots
have been considered and integrated in the model. Design guidelines with equivalent
engineering constants, and approximate cost estimation rules have been established.
An example problem has been solved to demonstrate the simplicity of the design
rules. ICG beams and panels have been built and tested under static and dynamic
flexural loading. Valuable test data are provided and superior mechanical properties,
such as high damage tolerance, resilience, and durability have been demonstrated.

KEY WORDS: composite grids, unidirectional composites, pultrusion, equivalent


stiffness.

1. INTRODUCTION

RID STRUCTURES WITH unidirectional composite ribs are one of the most efficient
G forms of structures. Using unidirectional composite ribs, all the fibers can be aligned
along the rib directions and the stiffness and strength of the material can be utilized most
efficiently. The higher anisotropy means the higher translation of composite properties to
grid properties. The low transverse properties of the ribs are not critical to the
performance of the grids.

*Author to whom correspondence should be addressed. E-mail: dongyuphan@yahoo.com

Journal of COMPOSITE MATERIALS, Vol. 37, No. 4/2003 287


0021-9983/03/04 0287–30 $10.00/0 DOI: 10.1106/002199803028681
ß 2003 Sage Publications

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


288 D. HAN AND S. W. TSAI

Not only are composites the best match for the grids, grids are good for composites as
unidirectional composites have the most efficient fiber architecture. In unidirectional
composites, the fibers are kept straight and higher fiber volume fractions can be achieved.
Multidirectional laminated composites are susceptible to delamination and microcracking
due to the material mismatch. Woven fabrics degrade properties further because the fibers
are no longer straight as shown in Figure 1 [1] where the typical fabric properties are
compared with uni-plies. Uni-plies can achieve higher fiber fractions and the stiffness and
strength are higher by a factor of three or more.
Generally, composite grid panels have several advantages over the conventional solid or
sandwich panels as listed below.
. Unidirectional composite ribs have no material mismatch and they are unlikely to
delaminate and have higher impact and fatigue resistance than laminated structures. In
addition, by having separate ribs, cracks do not propagate to the next ribs and may
promote damage tolerance.
. For the same amount of material, grid panels are always thicker than the laminates and
have higher flexural rigidity.
. Being an open structure, the grids are not susceptible to moisture incursion as are
sandwich panels. Inspection and repair of grids are relatively easy also because they are
open and repeated structures.
. Other typical benefits of composite materials, such as light weight, high stiffness and
strength, high corrosion resistance, etc., are retained.
Grids can have ribs running in several directions for multidirectional loadings. According
to the number of rib directions, there are bi-grids, tri-grids and quadri-grids as shown in
Figure 2. Among these three different types, tri-grids and quadri-grids have balanced
longitudinal and shear properties, whereas bi-grids have practically zero shear and twisting

Figure 1. Stiffness and strength of uni-plies vs. woven fabrics.

(a) Bi-Grid (b) Tri-Grid (c) Quadri-Grid

Figure 2. Typical rib arrangements of grid structures.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 289

stiffness. However, for many applications, where the shear properties are not critical,
bi-grids perform better than the other two because they have higher longitudinal stiffness
and strength per unit mass. In addition, bi-grids are simpler to manufacture because they
have a lower number of ribs crossing at the intersections. The current study concentrates on
the design and manufacturing of bi-grids, especially flat square grid panels.
There has been significant previous research done in the area of analysis, design and
manufacturing of various types of grids. A few notable recent works most closely related
to the present work are reviewed.
Equivalent stiffness models and design optimization of various types of ‘‘ideal’’ grids
have been developed by Chen [2]. ‘‘Ideal’’ means that the grids have fibers aligned along
the ribs without any interference at the rib intersections. Using the equivalent stiffness
model, the relative performances among grids, conventional laminates, and sandwich
panels have been compared quantitatively. This approach is used in the current study
considering stiffness reduction at the rib joints.
Failure behavior and the effects of irregularities (inclusion of soft and hard points,
damaged ribs, repairs, and joined grids) of grids have been studied by Huybrechts [3].
Finite element formulation was done and the failure envelopes for the various patterns of
grids under different loadings were plotted.
Colwell [4] proposed several manufacturing methods to improve the quality of grids and
studied the application of rectangular grids as concrete reinforcements. Particularly,
Tooling Reinforced Interlaced Grid (TRIG) and Pin Enhanced Grid (PEG) were
proposed. In TRIG, Tooling becomes an integral part of the finished structure and
removes any design limitations incurred by the required separation of the part from the
tool. PEG uses the conventional mold and a number of pins to spread fibers at the rib
intersections in order to improve quality of the nodes and ribs. Current study proposes
entirely different manufacturing approach more suitable for large panels.
General overview of grids including various manufacturing methods, analysis, and design
were done by Tsai et al. [1]. Other notable work related to manufacturing has been done at
USAF research lab. A hybrid tooling concept [5] that uses expansion tooling inserts in a
thermally stable base tool has been developed specifically for launch vehicle application [6].
An optical boom structure was fabricated using TRIG core to study deployment accuracy
and stability [7]. The grid has been proven to be desirable alternative to honeycomb core in
ultra light stiffness controlled sandwich structures.

2. MANUFACTURING

The concept of a grid structure with unidirectional composite ribs is very attractive and
offers many advantages as pointed out earlier. However, manufacturing a high
performance composite grid is not an easy task. Rib intersections or joints are invariably
the weak links where failure initiates and stiffness reduction occurs. This is physically
unavoidable as fibers always interfere at the rib intersections. One of the main objectives
of the current study is to develop cost effective manufacturing processes for composite
grids with minimal property reduction at the rib intersections.
A new type of grid joint, a Cap Reinforced Slotted Joint and the resulting grid structure,
an Interlocked Composite Grid (ICG) are introduced. Fabrication of the flat square grid is
discussed in detail. Strengths and weaknesses of ICG panels and possible solutions are
pointed out.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


290 D. HAN AND S. W. TSAI

2.1 Interlocked Composite Grid (ICG)

The conventional slotted joint is weak and not suitable for high performance grid
structures. Our innovation is to bond rib-caps over the open slots to provide additional
load paths and to alleviate strength and stiffness reduction due to the slots. A schematic of
the assembly concept for an ICG is shown in Figure 3. The assembly starts with pultruded
unidirectional ribs and rib-caps. Slots half the rib height are cut into the ribs. An ortho-
grid is formed by inserting the ribs into matching slots of one another. Rib caps are then
bonded to the top and bottom sides of the ribs to bridge the open slots. The top and
bottom caps are orthogonal to each other and the whole structure is thus interlocked.
Pultrusion is a process that produces continuous lengths of constant cross-sectional
shapes. It is one of the most reliable, fastest, cheapest, and fully automated process for
manufacturing composite sections. For moderate section dimensions, the pulling speed is
around 1 m/min. Fiber volume fraction can be over 70%. The resulting stiffness and
strength is three or more times higher than most conventional laminated, woven or
braided fiber composites. The relative costs of various composite manufacturing processes
are shown in Figure 4 [8]. Costs of manual layup, automatic layup, and filament winding

upper cap

upper cap

slotted ribs rib


bonding

bonding
lower cap

lower cap

Figure 3. Interlocked composite grid.

6
5.4
cost relative to pultrusion

5
3.8
4

3
2.4
2
1.0
1

0
manual layup automatic layup filament winding pultrusion

Figure 4. Cost comparison of various composite manufacturing processes [6].

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 291

processes are compared to the cost of pultrusion, and show the dramatic cost saving of
pultrusion. The total cost includes the cost of tooling, material, labor, and equipments.
. For the first two layup processes in Figure 4, parts with simple geometry that is
comparable to pultruded and filament wound parts are considered, and the tooling
costs are almost same for all four processes.
. In terms of material costs, prepreg is required for the first two layup processes. Prepreg
cost is more than twice that of dry fibers used in pultrusion and filament winding.
. Pultrusion is a fully automated process while other processes include 5–20% of labor
costs.
. Equipment cost varies from simple pultruder, filament winder to manual and machine
layup. Autoclave is required for layup processes while it is not needed for the pultrusion
and may not be used for some filament wound parts. In addition, clean room, cold
storage, debulking, bagging, and cleanup are required for the layup processes.
. Other processes such as RTM (Resin Transfer Molding) and VARTM (Vacuum
Assisted Resin Transfer Molding) can produce complex parts. Their performance is
comparable to fabric composites having woven fibers and low fiber volume fractions.
These fiber architectures cannot compete with unidirectional composites when major
loading component is along the fiber direction as in grids.
Slots can be cut as the sections are pulled. Abrasive water-jet or synchronized saw blade
can be combined with pultrusion. Alternatively, many ribs can be stacked and cut in one
pass of a saw blade. Rib assembly is very easy and fast due to the slots and does not
require any special fixtures. Due to the simplicity of the steps, field assembly of large
panels is possible, if required. Also, extremely large panels are possible since pultrusion
can produce theoretically infinite lengths of sections.
In summary, the main features of an ICG are;
. Using pultruded section, the grid has very promising performance/cost ratio.
. Using slotted joints, the rib assembly is fast and easy.
. The rib caps are effective in stiffness recovery.
. Mass production of large panels is possible.

2.2 Improved ICG

Although the proposed ICG has unique features and superior properties, in some cases
it requires additional stiffness and strength. In this section, solutions to overcome some
intrinsic weaknesses of an ICG will be discussed.

Shear Reinforcement

For shear critical applications, there are several ways to improve the shear rigidity of an
ICG. For example, you can insert a solid block in each cell, bond skins with  45 layers, or
bond angle brackets at the joints. Considering the weight and manufacturing, the simplest
and most efficient way is to insert a diagonal rib (internal rib) to brace each cell. The
pictures in Figure 5 show standard and shear stiffened ICG made from carbon pultruded
sections. Internal ribs can be inserted before the cap bonding or after the caps on one side
are bonded. Internal ribs can be inserted without bonding, but require tight fits for better

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


292 D. HAN AND S. W. TSAI

Figure 5. Standard and shear stiffened ICG.

performance. Without bonding, the internal ribs are effective only when they are under
compression. In case the direction of shear is known for each cell, the internal ribs can be
inserted in a manner so that they are all under compression. Internal ribs will also give
better handling strength during and after assembly. The results from twisting tests on
standard and shear stiffened ICG are shown in Section 4.

Local Reinforcement and Edge Reinforcement

Low shear and transverse strength are the intrinsic weaknesses of unidirectional
composites and are the dominating failure mode of a grid. Combined with stress
concentration due to the slots, a crack always initiates at the roots of the slots and
propagates along the fiber direction. A thorough study of the material and processing
characteristics of pultrusion can yield some improvements in strength; however, usually the
ribs will need to be reinforced if strength rather than stiffness is a critical concern for the
design. Laminate construction of pultruded sections with woven layers and unidirectional
rovings can also improve the strengths, but this would increase the manufacturing time and
cost while decreasing the fiber content. Another solution is to use pultruded tubes as shown
in Figure 6 in light color. Pultruded tubes are cut into the length of the rib height and
bonded into critical cells where local stress concentration is expected. Fiber directions of the
tubes are perpendicular to the fiber directions of the ribs, which will improve load transfer
near the slots and increase the strength of the grid. This will also increase the buckling
strength of the ribs. The same pultruded tubes can be used to reinforce the edges of the grid
as well as any interior cells where concentrated load is applied.

Cap Alignments

As the size of the grid panel becomes large, cap alignment during the bonding becomes
more difficult and requires alignment guides. Proposed solutions are shown in Figure 7.
A simple solution is to put grooves in the caps. The groove can be made into the cap
during pultrusion using a die with the given shape. The groove in the cap is not only good

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 293

Figure 6. Local reinforcement and edge reinforcement using pultruded tubes.

Figure 7. Cap alignments.

for alignment but also improves bonding strength by providing more bonding surface.
Grooved ribs are another solution for alignment guidance. The grooves are cut after the
rib pultrusion when the slots are cut. One advantage of using grooved ribs is that the sides
of the caps are part of the surface of the grid, resulting in a flushed surface.
As a third option, grooved ribs and grooved caps can be used at the same time resulting
in higher bonding strength and flushed surface.

3. STIFFNESS MODELS

There are at least two different types of stiffness models that can be used to predict the
global behavior of grids. One is the Exact Stiffness Model where individual ribs, caps, and
joints of a grid are simulated exactly. Generally, it requires large computational memory,
time, and cost. Also, any changes during design process may require regeneration of all
meshes. Hence, the exact stiffness model is very inefficient and not suitable for the design
process. This has motivated the development of a simpler stiffness model, the Equivalent
Stiffness Model.
For a grid with sufficient number of cells, the equivalent stiffness model, where the rib
properties are smeared into a homogenized equivalent plate, can accurately predict the
global behavior of a grid. The Equivalent Stiffness Model is simpler and more efficient, so is

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


294 D. HAN AND S. W. TSAI

more suitable for design and optimization processes than the exact stiffness model.
Particularly, it is a very powerful tool for buckling and modal analyses where the solutions
require a number of iterations. Previous equivalent stiffness models formulated by Chen [2]
are for ‘‘idealized’’ grids where all the fibers are aligned along the ribs without any
interference at the rib intersections. The stiffness model for an ICG must include the effect
of the caps and the stiffness reduction due to the slots. In this section, the equivalent
stiffness model for a square ICG has been formulated in two different forms; (1) the
equivalent stiffness matrices (ABD and H matrices) and (2) the equivalent engineering
constants. The stiffness reduction due to the slots has been included in the formulation by
using the ‘‘cap efficiency factors’’ defined later. The equivalent stiffness matrices are more
general and accurate than the equivalent engineering constants. The equivalent engineering
constants are used in simple design equations for the preliminary design of an ICG. The two
stiffness models have been verified numerically and showed good agreement. It is easy to
extend the models for more general ortho-grids with rectangular cells [9].

3.1 Coordinate Systems and Sign Conventions

Local and global coordinate systems were used for the derivation of the equivalent
stiffness model (Figure 8). Local coordinate system (denoted by axes x, y, z) has the x-axis
aligned with the ribs with rib-caps on positive z. The y-axis is on the mid-surface of the
grid panel and perpendicular to the x and z axes. The orientation of the global coordinate
system (denoted by axes 1, 2, 3) is chosen for convenience where 1–2 plane is on the mid-
surface of the grid. Angle  is measured from global 1-axis to local x-axis.
The geometric dimensions for a square ICG and the rib material properties are shown in
Figure 9. Ribs and caps are made of the same material, and the longitudinal and shear
stiffness are denoted as Erib and Grib, respectively.

3.2 Cap Efficiency Factors

The stiffness of an ICG is reduced by the presence of the slots. To measure the
reduction, two cap efficiency factors, in-plane and flexural, have been defined. The
in-plane cap efficiency factor (Equation (3-1)), e0, indicates the stiffness reduction in
tension or compression and is defined as the ratio of axial rigidity of slotted T-section
beam to that of a solid T-section beam without slots. Similarly, the flexural cap efficiency

Local Coordinates
Y
X

2
Global Coordinates

Figure 8. Global and local coordinate systems.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 295

z h

x
Erib Young’s modulus of ribs and caps
L y
b Grib shear modulus of ribs and caps
L rib spacing
t
b rib thickness
h rib height
c

Figure 9. Geometric dimensions and notations for an ICG.

factor (Equation (3-2)), ef, indicates the stiffness reduction in bending and is defined as the
ratio of flexural rigidity of slotted to solid T-section beams. These factors range between 0
and 1 and are functions of L/h and R, where L is the distance between the slots, h is the rib
height, and R is defined as the area ratio of the cap to rib.

 
ðEAÞslot L
eo ¼ ¼f ,R ð3-1Þ
ðEAÞsolid h
 
ðEIÞslot L
ef ¼ ¼f ,R ð3-2Þ
ðEIÞsolid h

The axial rigidity of the slotted beam, (EA)slot in Equation (3-1), and the flexural rigidity
of the slotted beam, (EI)slot in Equation (3-2), are calculated using the commercial finite
element analysis package ANSYS. The rigidities of solid T-beams without slots are
calculated from simple beam theory.
For the ribs and caps in proper design range, the cap efficiency factors have consistent
values. The cap efficiency factors of glass sections are calculated and shown in Figures 10
and 11. The rib’s height to thickness ratios are in between 10 and 30, and the thickness of
the rib and cap are the same. L/h is varied from 0.33 to 5 and R from 0.1 to 1. Larger L/h
results in higher values for the cap efficiency factor because larger slot spacing results in
less slot interaction. Also, larger values for R results in higher values because the larger cap
area provides more load path to go around the slots. The increment of the in-plane cap
efficiency factor is gradual and the value ranges from 60 to 90%. Caps are more efficient
for the recovery of bending stiffness as shown in Figure 11. For R greater than 0.5 and L/h
greater than 1, the flexural cap efficiency factors are greater than 95%. In this range, the
bending stiffness reduction due to the slots can be neglected.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


296 D. HAN AND S. W. TSAI
In-plane Cap Efficiency Factor (Glass)
eo
1.0

0.9

R = 0.1
0.8
R = 0.25
R = 0.5
0.7
R = 1.0

0.6

0.5
0 1 2 3 4 5 6
L/h

Figure 10. In-plane cap efficiency factors for glass section.

Flexural Cap Efficiency Factor (Glass)


ef
1.0

0.9

R = 0.1
0.8
R = 0.25
R = 0.5
0.7
R = 1.0

0.6

0.5
0 1 2 3 4 5 6
L/h

Figure 11. Flexural cap efficiency factors for glass section.

The effects of slots on out-of-plane shear stiffness and torsional stiffness are neglected as
they are small compared to the longitudinal and bending stiffness. For different material
systems, the cap efficiency factors have similar values and trends [9].

3.3 Formulation of Equivalent Stiffness Matrices

Because uni-directional composite ribs have much lower transverse shear modulus than
longitudinal Young’s modulus, Mindlin’s Plate theory is adopted to account for the
transverse shear deformation. The stiffness reduction due to the slots is integrated into
the stiffness matrices (ABD and H matrices) using the cap efficiency factors defined in
Section 3.2. All the quantities are written with respect to the mid-plane where z ¼ 0 or
x3 ¼ 0, and the asymmetry due to the caps is included through a nonzero coupling matrix.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 297

c z′
t z z

neutral axis plane


d
h mid-plane
y d
neutral axis plane

Figure 12. Neutral axis planes and mid-planes of T-beams.

Stiffness matrices of an ICG are derived by considering the stiffness of a T-section beam
(combination of rib and cap). Axial and bending rigidities of a T-section beam are
calculated with respect to the neutral axis of the beam, and the cap efficiency factors are
applied to account for the stiffness reductions due to the slots. It is assumed that the slots
do not change the location of the neutral axis of the beam. The neutral axis planes of the
T-beams in x and y directions do not coincide when they are assembled into an ICG
(Figure 12). The mid-plane is chosen as a reference plane and the Parallel Axis Theorem
[10] is used to calculate the axial and bending rigidities with respect to the mid-plane.
The distance between the neutral axis plane and the mid-plane can be expressed in terms
of the rib and cap dimensions as in Equation (3-3).

ctðh þ tÞ
d¼ ð3-3Þ
2ðbh þ ctÞ

The corrected axial rigidity (EA)* and the bending rigidity (EI)* of the T-beam are shown
in Equation (3-4), where the origin of z0 axis is at the neutral axis plane. The in-plane and
flexural cap efficiency factors are applied to account for the stiffness reduction due to the
slots.
Z
ðEAÞ ¼ eo EdA ¼eo E rib ðbh þ ctÞ
Z  2 ! ð3-4Þ
 02 bh3 2 ct3 t h
ðEIÞ ¼ ef Ez dA ¼ ef E rib þ bhd þ þ ct þ  d
12 12 2 2

The total axial forces and moments with respect to the neutral axis of the T-beams in x and
y directions can be written as in Equations (3-5) and (3-6), where the prime superscript
denotes quantities with respect to the neutral axis.

0
N x ¼ ðEAÞ "0x
0 ð3-5Þ
N y ¼ ðEAÞ "0y
0
M x ¼ ðEIÞ kx
0 ð3-6Þ
M y ¼ ðEIÞ ky

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


298 D. HAN AND S. W. TSAI

Equations (3-5) and (3-6) are rewritten in Equations (3-7) and (3-8) with respect to the
mid-plane using the Parallel Axis Theorem, where the super script ‘‘o’’ denotes quantities
with respect to the mid-plane. Note that the direction of the offset from the mid-plane to
the neutral axis is different for x and y direction beams as shown in Figure 12.

N x ¼ ðEAÞ "ox þ dðEAÞ x


ð3-7Þ
N y ¼ ðEAÞ "oy  dðEAÞ y

M x ¼ dðEAÞ "ox þ ððEIÞ þ d 2 ðEAÞ Þx


ð3-8Þ
M y ¼ dðEAÞ "oy þ ððEIÞ þ d 2 ðEAÞ Þy

Dividing the Equations (3-7) and (3-8) by the unit cell length L, average axial forces and
moments per unit length along the boundary of the equivalent plate can be found.

N x ðEAÞ o ðEAÞ
Nx ¼ ¼ "x þ d x
L L L
ð3-9Þ
N y ðEAÞ o ðEAÞ
Ny ¼ ¼ "y  d y
L L L
 
Mx ðEAÞ o ðEIÞ ðEAÞ
Mx ¼ ¼d "x þ þ d2 x
L L L L
  ð3-10Þ
My ðEAÞ o ðEIÞ ðEAÞ
Mx ¼ ¼ d "y þ þ d2 y
L L L L

From Equations (3-9) and (3-10), the equivalent A, B, and D matrices with respect to the
local x, y, z coordinates are
2 3
ðEAÞ
0 0
6 L 7
6 7
½A on ¼6
6 0 ðEAÞ  7 ð3-11Þ
icg
4 07
5
L
0 0 0

2 3
ðEAÞ
d 0 0
6 L 7
6 7
½B on
¼6
6 ðEAÞ  7 ð3-12Þ
icg
4 0 d 07
5
L
0 0 0
2 3
ðEIÞ ðEAÞ
þ d2 0 0
6 L L 7
6 7
½D on
¼6
6 ðEIÞ 
ðEAÞ  7 ð3-13Þ
icg
4 0 þ d2 07
5
L L
0 0 0

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 299
z y

z
y Ty
y
x y

x x Myx

Tx Mxy
x

(a) (b)

Figure 13. (a) Positive torsion and rotation for beams; (b) positive twisting moment and rotation for a plate.

Note that there is no Poisson coupling since x and y direction beams are completely
separate. Also, note that the total stiffness ABD matrix is singular due to zero in-plane
shear stiffness, Ass, and twisting stiffness, Dss. The in-plane shear stiffness Ass is
zero because the joints are assumed to behave as hinges and have no resistance to the
rotation in the plane of the grid panel. For analysis purposes, Ass is set to a small number
on the order of 106 of Axx. The twisting stiffness Dss can be formulated by
considering the torsional rigidity of the T-section beam and the twisting rigidity of the
equivalent plate.
Figure 13 shows the positive torsion or twisting moment and rotations for beams and
plates. The beam torsion (T ) and the twisting moment (Ms) of the equivalent plate are
related as shown in Equation (3-14).

dx dy
Tx ¼ ðGJÞx Ty ¼ ðGJÞy
dx dy
ð3-14Þ
Tx ðGJÞx dx Ty ðGJÞy dy
Mxy ¼ ¼ Myx ¼ ¼
L L dx L L dy

where, the torsional rigidities of the beams are

1
ðGJÞx ¼ ðGJÞy ¼ ðGJÞ ¼ Grib ðhb3 þ ct3 Þ ð3-15Þ
3

From the force equilibrium, twisting moments Mxy and Myx are equal (Equation (3-16)).

Mxy ¼ Myx ¼ Ms ð3-16Þ

The twisting curvature ks of the equivalent plate and beam rotation angle  are related as

 
@y @x dx dy
ks ¼  þ ¼ þ ð3-17Þ
@x @y dx dy

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


300 D. HAN AND S. W. TSAI

From the Equations (3-14)–(3-17), Dss can be written in terms of rib material properties
and the dimensions of the rib and the cap:
 
@y @x dx dy Mxy Myx Ms L
ks ¼  þ ¼ þ ¼ þ ¼2
@x @y dx dy ðGJÞx =L ðGJÞy =L ðGJÞ

Ms ¼ Dss ks

Hence,
ðGJÞ Grib ðhb3 þ ct3 Þ
Dss ¼ ¼ ð3-18Þ
2L 6L
Usually, Dss is negligible compared to Dxx or Dyy since the torsional rigidity of a tall open
section beam is much smaller than the bending rigidity. For the preliminary design, the
twisting stiffness can be neglected.
The out-of-plane shear stiffness matrix H is derived from the transverse shear rigidity of
the T-beam. It is assumed that the shear is carried by the rib only and the shear rigidity
(GA) is shown in Equation (3-19).
ðGAÞ ¼ Grib bh ð3-19Þ
The average transverse shear forces per unit length along the boundary of the equivalent
plate are shown in Equation (3-20) where the shear correction factor, , of 5/7 is used.
ðGAÞ
Vx ¼  "xz
L
ð3-20Þ
ðGAÞ
Vy ¼  "yz
L
Hence the H matrix in local x-y-z coordinate is
" #
on  ðGAÞ
L 0
½H icg ¼ ð3-21Þ
0  ðGAÞ
L

The equivalent stiffness matrices of an ICG with respect to the local coordinate system
have been formulated by taking the averages of the beam rigidities. The stiffness matrices
with respect to an arbitrary global coordinate system can be obtained through coordinate
transformations.

3.4 Formulation of Equivalent Engineering Constants

A simpler equivalent stiffness in the form of engineering constants has been formulated.
Simple design guidelines for a square ICG have been established using the equivalent
engineering constants.
For a square ICG, there are five geometric design variables, b, h, c, t and L. Combining
some of these variables, two useful design parameters, total rib volume fraction f and cap
to rib area ratio R, are defined. The total rib volume fraction, f, is defined as the ratio of

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 301

the total volume of the ribs (without caps) to the volume of the plate where the thickness
of the plate is the same as the rib height h (Figure 14). Rib volume fraction is similar to
the fiber volume fraction of a laminate and directly related to the equivalent stiffness.
However, the fiber volume fractions for laminates are usually fixed by prepreg
manufacturers, whereas the rib volume fraction is an important design variable.
Another design parameter is the area ratio between the cap and the rib as defined in
Equation (3-22). The area ratio is useful in deciding the cap size relative to the rib size.
Cap Area ct
R¼ ¼ ð3-22Þ
Rib Area bh
The axial rigidity (EA) and flexural rigidity (EI) of the T-section (combination of
rib and cap) beam can be expressed in terms of the rib material properties, rib geometry,
and R. Some simplifying approximations have been made as shown in Figure 15.
These assumptions are reasonably accurate if the thickness of the cap is much smaller
than the height of the rib.
Under these assumptions, (EA) and (EI) of the section with slot effect can be written
in terms of the cap efficiency factors, rigidities of the rib, and some factors for the rigidities
of the cap:
ðEAÞ ¼ eo E rib Arib ð1 þ RÞ ð3-23Þ
 
 3R
ðEIÞ ¼ ef E rib Irib 1 þ ð3-24Þ
1þR

Equivalent in-plane stiffness, E ox and E oy , can be obtained by dividing Equation (3-23)


by the unit area Lh and can be written in terms of f and R as shown in Equation (3-25).
f
E ox ¼ Eyo ¼ eo E rib ð1 þ RÞ ð3-25Þ
2

2
f
unit cell Lb
b

Figure 14. Unit cell and rib volume fraction, f, of a square grid.

t
for t << h
Icap ~ 0
h
centroid of Acap at = h

Figure 15. Geometric assumptions for a T-section beam.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


302 D. HAN AND S. W. TSAI

Dividing Equation (3-24) with unit moment of inertia, flexural stiffness E fx and E fy are
written in Equation (3-26). These flexural stiffness will be used for the design of a grid
panel under flexure.
 
f 3R
E fx ¼ E fy ¼ ef E rib 1 þ ð3-26Þ
2 1þR

Poisson’s ratios are zero because the ribs in the x and y directions are completely separate.
In-plane shear stiffness is also zero because the joints are modeled as hinges. For out of
plane shear stiffness, it is assumed that the shear is carried by the ribs only and the
corresponding stiffness are expressed in terms of the total rib volume fraction f and the rib
material property. The Poisson’s ratio, in-plane and out-of-plane shear stiffness of the
equivalent plate are shown in Equation (3-27).

xy ¼ yx ¼ 0
Es ¼ 0 ð3-27Þ
f
Gxz ¼ Gyz ¼ Grib
2

4. EXPERIMENTS

Numerous ICG beams and panels with different cross section dimensions and cell sizes
were assembled and tested. Three different material systems, one glass and two carbon
systems, were used. The material properties are shown in Table 1.

4.1 ICG Beams

4.1.1 STATIC TEST


The stiffness reduction at the joints of an ICG was considered by the cap efficiency
factors. ICG beams were tested to validate the accuracy of the flexural cap efficiency
factors. For each specimen, two longitudinal ribs and number of cross ribs were combined
to avoid lateral buckling. The spacing between the cross ribs (slot spacing) and the cap size

Table 1. Material properties and section dimensions.


Glass Carbon 1 Carbon 2
Fiber E-Glass T300 Fortafil510
Resin Polyester Epoxy Vinylester
Specific gravity 2.125* 1.590* 1.505*
Fiber fraction (%) 70 62 49
Ex (GPa) 52* 143* 114*
Ey (GPa) 10 9 7
Gxy (GPa) 3.8 3.4 2.2
xy 0.268 0.268 0.292
Section dimension (cm) 0.48 4.98, 0.64 15.2 0.32 3.81 0.32 4.55

*measured values.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 303

were varied. The beams were tested under three point bending. The center displacement
and the applied load were measured to calculate the stiffness of the beam (P/ ).
The center displacement and the stiffness of the beam can be predicted from simple
beam analysis, including shear deformation (Equation (4-1)). In this equation, the bending
rigidity (EI) is modified by the cap efficiency factor ef to account for the stiffness reduction
at the joints. 5/7 is used for the shear correction factor .

Pl 3 Pl
¼ þ ð4-1Þ
48ef ðEIÞ 2 4ðGAÞ 2

The measured and predicted stiffness of the beams were compared. First, for the glass
section beams, the cross section dimensions and the slot spacing of each beam are shown
in Table 2. For each specimen type, three or four beams were tested.
The measured and predicted stiffness were compared in Figure 16 for the glass beams
listed in Table 2. The errors between the two values are less than 2% and the flexural cap
efficiency factors are very accurate.
Carbon section beams were also tested under three point bending. The geometric
dimensions and the material of the specimens are shown in Table 3. The measured and
predicted stiffness are compared in Figure 17 and showed errors less than 5%.
For all specimens, the initial failure occurred at the root of the slots and propagated
along the fiber direction. The commercial finite element analysis package ANSYS has
been used to solve the case for specimen type 4 in Table 3. The failure load of the specimen
was around 4900 N where the first load drop occurred. The finite element mesh and the
resulting stress distributions near the slot at the failure load are shown in Figure 18. Also
shown are the strength ratios (maximum stress/strength) for longitudinal tension,

Table 2. Geometric dimensions for glass section beams.


Material b (cm) h (cm) c (cm) t (cm) L (cm) L/h R Span l (cm)
Spec. 1 Glass 0.47 5 0.64 0.47 10.2 2.04 0.13 40
Spec. 2 Glass 0.47 5 2.41 0.47 10.2 2.04 0.49 40
Spec. 3 Glass 0.47 5 5 0.47 10.2 2.04 1.00 40

6.0E+06

5.0E+06
Stiffness (N/m)

4.0E+06
test
3.0E+06
model
2.0E+06

1.0E+06

0.0E+00
spec 1 spec 2 spec 3

Figure 16. Stiffness comparisons for glass section beams.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


304 D. HAN AND S. W. TSAI

Table 3. Geometric dimensions for carbon section beams.


Material b (cm) h (cm) c (cm) t (cm) L (cm) L/h R Span l (cm)
Spec. 1 Carbon 1 0.32 3.81 1.85 0.32 12.70 3.3 0.49 38
Spec. 2 Carbon 1 0.32 3.81 1.85 0.32 7.62 2.0 0.49 38
Spec. 3 Carbon 1 0.32 3.81 1.85 0.32 3.81 1.0 0.49 38
Spec. 4 Carbon 2 0.32 4.55 2.07 0.32 11.43 2.2 0.46 51

4.0E+06
3.5E+06
3.0E+06
Stiffness (N/m)

2.5E+06
test
2.0E+06
model
1.5E+06
1.0E+06
5.0E+05
0.0E+00
spec 1 spec 2 spec 3 spec 4

Figure 17. Stiffness comparisons for carbon section beams.

x
0.1
X

(a)
(b)

xy
0.7
S y
0.9
Y

(c) (d)

Figure 18. (a) Finite element mesh and the plots of stress distribution near the slot; (b) x; (c) xy; (d) y
distributions.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 305

transverse tension, and in-plane shear stresses at the failure load. For the calculation
of the strength ratio, typical strength values for unidirectional carbon composites
(T300/5208) were used where X ¼ 1500 MPa, X0 ¼ 1500 MPa, Y ¼ 40 MPa, Y0 ¼ 246 MPa,
and S ¼ 68 MPa [10]. As shown in (c) and (d) of Figure 18, the transverse tension and the
shear are the dominating failure modes. The transverse stress distribution shows
antisymmetry; the tension on one side and the compression on the other side. The
cracks are observed to propagate along the fiber direction to the side where transverse
tension is applied. Although analytical prediction is difficult, the failure load can be
estimated through numerical analysis. Failure modes other than material failure have not
been considered in the current study. Depending on the boundary conditions and
geometry of a grid, other failure modes have to be examined.

4.1.2 FATIGUE TEST


Tests on the fatigue strength of a grid have been performed. Three point bending tests
on glass section beams with 0.4 m spans have been performed. The specimens had the same
geometric dimensions as the specimen type 1 in Table 2. For these specimens, the static
failure load was around 7100 N. Four specimens were tested under different load level, 56,
50, 40 and 25% of the static strength. The load ratio (minimum load/maximum load) was
fixed at 0.1. The resulting numbers of cycles to failure at each load level are shown in
Table 4 where the final failure is defined when the specimen could no longer carry the
specified load. The resulting S–N curve is shown in Figure 19.
The failure modes in fatigue were similar to those under static loading. A crack initiates
near the root of a slot and propagates along the fiber direction. The rib cap bonding did

Table 4. Load level and the corresponding number


of cycles to failure.
Load Level Cycles to Failure
4000 N (56%) 11,700
3560 N (50%) 34,000
2850 N (40%) 690,000
1780 N (25%) Greater than 3,500,000

6000

5000
Load Level (N)

4000

3000

2000

1000

0
1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07
Cycles to Failure

Figure 19. S-N curve for fatigue test.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


306 D. HAN AND S. W. TSAI

not fail until much later in the cycling of the grid. Note that at 25% of the static strength,
fatigue failure did not occur up to 3.5 million cycles. Although the test data are very
encouraging, it requires more systematic measurements and further life tests before
long-term system reliability and design allowables can be established.

4.2 ICG Plates

The accuracy of the flexural cap efficiency factors has been validated through the set of
beam tests described in the previous section. In this section, various ICG panels were
tested and the results were compared to the predicted value from the equivalent ABD and
H matrices. Three large panel specimens were tested under central loading with four sides
simply supported. The applied load and the center deflection were measured to calculate
the stiffness of the panel. The picture of the test setup is shown in Figure 20. The glass
sections were used to assemble one 1.2 m 1.2 m panel and two 3 m 3 m panels. The
material, section dimensions and the rib spacing for each panel are listed in Table 5.
The measured and the predicted stiffness (P/ ) matched well and are shown in Figure 21.
The sample load versus displacement curve for the specimen type 2 in Table 5 is shown
in Figure 22 and reveals excellent mechanical performance. First, the grid is remarkably
resilient as the initial loading, unloading and reloading showed essentially no permanent
strain or hysteresis. Second, the ultimate failure strain was many times the initial failure
strain, yielding a highly ductile overall behavior. Cracks initiate in the middle ribs where
higher loads are carried. The crack propagates along the rib and next ribs start to fail one

four simply supported edges

P,

span l

Figure 20. Test setup for glass ICG panels.

Table 5. Geometric dimensions for glass section panels.


Material b (cm) h (cm) c (cm) t (cm) L (cm) L/h R Span l (cm)
Spec. 1 Glass 0.47 5 5 0.48 10 2 1 91
Spec. 2 Glass 0.64 15 7.5 0.64 30 2 0.5 274
Spec. 3 Glass 0.64 15 7.5 0.64 45 3 0.5 274

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 307
4.0E+06
3.5E+06
3.0E+06

Stiffness (N/m )
2.5E+06
test
2.0E+06
model
1.5E+06
1.0E+06
5.0E+05
0.0E+00
sp ec 1 spec 2 spec 3

Figure 21. Stiffness comparisons for glass section panels.

Figure 22. Load vs. displacement curve up to failure (specimen type 2 of glass panel).

Table 6. Geometric dimensions for carbon section panels.


Shear
Material Stiffening b (cm) h (cm) c (cm) t (cm) L (cm) L/h R Span l (cm)
Spec. 1 Carbon 1 No 0.32 3.81 1.84 0.32 7.6 2 0.48 46
Spec. 2 Carbon 1 Yes 0.32 3.81 1.84 0.32 7.6 2 0.48 46

by one. The grid is tough and damage tolerant and may well suit for bridge decks and
other critical applications.
Smaller carbon ICG panels shown in Figure 5 were tested for shear properties. The
material and the geometric dimensions of the specimens are listed in Table 6. The material
and cross-section dimensions of the internal ribs for the shear stiffened grid are the same as
the rib dimensions.
The panels were tested under central point loading with two opposite corners supported
(Figure 23). The loading and the center displacement were measured to calculate the
stiffness. The test data showed that the shear stiffened ICG has more than 80 times the

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


308 D. HAN AND S. W. TSAI

1.8E+06

1.5E+06

P,
1.2E+06

Stiffness (N/m)
test
9.0E+05
model

6.0E+05
ly
3.0E+05
lx
0.0E+00
spec 1 spec 2

Figure 23. Twisting test and stiffness of standard and braced ICG.

stiffness of the standard ICG. It is demonstrated that the internal ribs are very effective for
improving the shear properties.

5. DESIGN

Simple design rules for a square ICG have been developed and described in this section
along with sample design problems. The simplified design methodology is very useful for
the preliminary design to determine the geometric dimensions of an ICG. As studied
in literature [1] and [2], where the performances of idealized grids versus traditional
laminates and sandwich panels have been compared, grid structures have their
real benefits for flexural applications. Thus design for flexure of a flat ICG panel
is considered. The equivalent flexural engineering constants formulated in Section 3
were used to establish the design guidelines. Flexural stiffness of an ICG panel and a
sandwich panel were compared.

5.1 Design Variables

Given the loading and support conditions, the panel dimensions, material properties of
the ribs, and the deflection limit or the required stiffness of the panel, we would like to
determine the rib dimensions b and h, the cap dimensions c and t, and the cell size L of the
ICG panel. From these five independent design variables, the total rib volume fraction f
and the area ratio R have been defined earlier and repeated in Equation (5-1). The flexural
equivalent engineering constants are expressed in terms of f and R as repeated in Equation
(5-2). Stiffness reduction due to the slots is neglected and the cap efficiency factors are
not included.

2 Acap ct
f ¼ , R¼ ¼ ð5-1Þ
L=b Arib bh

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 309
 
f 3R
E fx ¼ E fy ¼ E rib 1 þ , Gxy ¼ xy ¼ yx ¼ 0 ð5-2Þ
2 1þR

5.2 Design Guidelines

5.2.1 OPTIMUM CAP SIZE


It can be shown that the T-section beam has its maximum bending rigidity per unit
weight when the cap size is half the rib size, i.e., when R equals one half [9]. The plot of
normalized I/A ((I/A)* ¼ (I/A)T-section/(I/A)rib ) is shown in Figure 24. At R equal to one
half, (I/A) of the T-section beam is maximum and has 33% higher value than that of the
rib only. The equivalent flexural stiffness for R equals one half is shown in Equation (5-3).

2
E fx ¼ E fy ¼ fE rib ¼ E rib ð5-3Þ
L=b

For manufacturing concern, the thickness of the cap (t) is fixed to the same thickness as
the rib thickness (b). Then the cap width (c) is fixed to half of the rib height (h). As c and t
are fixed relative to the rib dimensions, there are only three independent design variables,
rib dimensions b, h and the cell size L.

5.2.2 PROPER CELL SIZE AND OTHER DESIGN CONSTRAINTS


The minimum value of the cell size L can be determined by considering the flexural cap
efficiency factors. At area ratio R of one half, the flexural cap efficiency factors of the
three materials listed in Table 1 are plotted in Figure 25. It is shown that the flexural cap
efficiency factors drop significantly when L/h is less than one. When L/h is larger than one,
the cap efficiency factors are around 0.95 and the stiffness reduction due to the slots can be
neglected. The use of the equivalent engineering constants without considering the stiffness
reduction can be justified when designing an ICG with L/h larger than one and R larger
than one half. In addition, for the homogenization assumption to work and the equivalent

Figure 24. Moment of inertia per unit area of a T-section beam.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


310 D. HAN AND S. W. TSAI

Figure 25. Flexural cap efficiency factors for R ¼ 0.5.

stiffness model prediction to be accurate, it is necessary to have at least five ribs in each
span of an ICG panel. Hence, the cell size L has to be less than one fifth of the given span.
Other constraints on the rib section dimension b and h are assigned considering the
manufacturing practice of the pultrusion. The aspect ratio of the rib section h/b should be
larger than 10 to have a high moment of inertia, but < 30 from the manufacturing aspect
of the pultrusion process.
In summary, the design constraints are shown in Equation (5-4).

span h
h L , 10 30 ð5-4Þ
5 b

5.2.3 DESIGN STEPS


The first step is to make initial guesses for the independent design variables, L, b, and h.
The corresponding cap dimensions are fixed as discussed in Section 5.2.1. Then the
flexural equivalent engineering constants can be calculated for the initial ICG panel.
The second step is to calculate the stiffness of the initial panel under the given loading
and support conditions. The initial stiffness, usually applied load divided by the deflection,
can be calculated from numerical analysis or any existing approximate solutions.
Then, the stiffness ratio m is calculated where m is defined as the ratio between the
required stiffness and the initial stiffness. The stiffness ratio m indicates how the stiffness
of the initial design must be increased or decreased to meet the stiffness requirements.
The next step is to resize the rib and/or the cell according to Equation (5-5).

h3 b Li
m¼ ð5-5Þ
h3i bi L

The rib height h can be changed following the cubic relation, and the rib thickness b and
the cell size L can be changed following the linear relation. Two, or all three variables can
be changed at the same time. Changing the rib height is the most effective way, as it
follows the cubic relation, whereas others follow the linear relations.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 311

The last step is to check for the constraints in Equation (5-4). Other design constraints
given for the specific design problem should also be checked. If any of the constraints are
not satisfied, resize the ribs or cells until all the constraints are satisfied.
For the preliminary design described in this section, only bending deflection is
considered and the transverse shear deformation is neglected. The proposed design
guidelines do not address failure load requirements and one should check for material and
buckling failures during the final design.

5.3 Sample Design Problem

Design of an ICG floor panel is considered. Several design options using glass and
carbon systems have been proposed and the approximate cost estimation for each case has
been given.
The panel dimensions and the required stiffness are shown in Figure 26. The floor panel
has 6.1 m (20 ft) 6.1 m (20 ft) spans with four simply supported edges. It is loaded under
uniform pressure loading of 960 N/m2 (20 lb/ft2). The maximum deflection at the center of
the panel has to be less than 0.0064 m (0.25 in.). Thus the required stiffness in terms of the
total load divided by the maximum deflection, (P/ max)required, is 5.6 106 N/m. Other
design constraints include, first, that the rib height has to be less than 0.2 m (8 in.), which is
the reasonable thickness limit for a floor panel of a building. Second, the cell size has to be
< 0.6 m (24 in.) to better support the upper skin as flooring. The upper skin does not
contribute to the mechanical stiffness of the panel.

5.3.1 COST ANALYSIS


Before the design example is solved, an approximate cost analysis is proposed.
Estimation of the approximate cost of an ICG panel is shown in Table 7.
The total cost is the sum of the material cost and the assembly cost. The material cost
includes the cost of fibers, resin, and pultrusion process and is approximately 20 $/kg for

S.S.
2
p = 960 N/m (20 lb/ft )
max = 0.0064 m (0.25 in.) S.S.
h 0.2 m (8 in.) S.S. 6 m (20 ft)
L 0.6 m (24 in.)
S.S.
6 m (20 ft)

Figure 26. Panel dimensions and the required stiffness of the sample problem.

Table 7. Cost estimation for an ICG panel.


Material fibers, resin, and pultrusion 20 $/kg for carbon section
4 $/kg for glass section
Assembly slot cutting, surface preparation, 6.70 $/m of the rib length
and bonding

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


312 D. HAN AND S. W. TSAI

Table 8. Design options for the carbon section ICG and the total costs.
Total
h b L Ef rib Asse-
(m) (m) (m) f (MPa) Weight Material Length mbly Total
Carbon (in.) (in.) (in.) (%) (ksi) (kg) (lb) Cost($) (m) (ft) Cost ($) ($)
Initial 0.15 (6) 0.0064 (0.25) 0.6 (24) 2 2970 (431) 281 (620) 6200 122 (400) 800 7000
design
Option 1 0.19 (7.6) 0.0064 (0.25) 0.6 (24) 2 2970 (431) 356 (785) 7850 122 (400) 800 8650
increase h
Option 2 0.15 (6) 0.013 (0.50) 0.6 (24) 4 2970 (863) 562 (1240) 12400 122 (400) 800 13200
increase b
Option 3 0.15 (6) 0.0064 (0.25) 0.6 (12) 4 5950 (863) 562 (1240) 12400 244 (800) 1600 14000
decrease L

carbon sections using T300 or the equivalent carbon fibers and 4 $/lb for glass sections
using E-glass fibers. The assembly cost includes the cost of slot cutting, surface
preparation, and the cost of adhesives and bonding and is approximately 6.70 $/m of the
rib length. The releasing agent from the pultrusion has to be removed from the bonding
surface of the ribs and caps, which accounts for the cost of surface preparation. The cost
estimation of the assembly is from our experience of manufacturing the small and large
test specimens.

5.3.2 DESIGN WITH CARBON SECTIONS


The ICG panel shown in Figure 26 with carbon sections (E ¼ 141 GPa) has
been designed following the design steps described in Section 5.2. Initial values for the
rib height, rib thickness, and the cell size are 0.15 m (6 in.), 0.0064 m (0.25 in.), and 0.6 m
(24 in.), respectively. The total rib volume fraction for these initial guesses is 0.02. The
equivalent flexural engineering constants are calculated from the fixed R of 0.5 and
the total rib volume fraction of 0.02 ðE fx ¼ E fy ¼ 2970 MPa, Gxy ¼ 0, vx ¼ vy ¼ 0Þ:
For the given loading and support conditions, the equivalent plate with the initial
variables has been solved using finite element analysis. The initial stiffness (P/ )i
is 2.7 106 N/m and the stiffness ratio m is 2. Then the initial design variables were
changed to satisfy the Equation (5-6). Several options that satisfy the design constraints
are listed in Table 8.

h3 b 24
2¼ ð5-6Þ
63 0:25 L

For the three options listed in the table, each of the variables was changed to satisfy
Equation (5-6), and the weight and cost of each panel were calculated. For the first option,
the rib height was increased while others remain constant. Having higher ribs is more
efficient in terms of the weight and cost for the same flexural rigidity. The second and third
designs have the same total rib volume fractions and the same weight. However, the
number of ribs and the total rib length of the third design are twice those of the second
design. For the same rib volume fraction f, it is better to have thicker ribs than to have
more cells in terms of the assembly cost.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 313

Table 9. Design for the glass section ICG and the total costs.
h b L Ef Total rib Assem-
(m) (m) (m) f (MPa) Weight Matrial Length bly Total
Carbon (in.) (in.) (in.) (%) (ksi) (kg) (lb) Cost($) (m) (ft) Cost ($) ($)
Initial 0.15 (6) 0.0064 (0.25) 0.6 (24) 2 1075 (156) 372 (820) 1640 122 (400) 800 2440
design
Increase h 0.20 (8) 0.015 (0.60) 0.6 (24) 5 2585 (375) 1179 (2600) 5200 122 (400) 800 6100
increase b

Table 10. Comparison of carbon and glass ICG.


Ef (MPa) Weight Material Assembly
(ksi) (kg) (lb) Cost($) Cost ($) Total ($)
Carbon 2970 (431) 356 (785) 7850 800 8650
Glass 2585 (375) 1179 (2600) 5200 800 6100

5.3.3 DESIGN WITH GLASS SECTIONS


The same design problem shown in Figure 26 has been solved using glass sections
instead of the carbon sections, and the results are shown in Table 9. The initial variables
for the rib height, rib thickness, and the cell size are 0.15 m (6 in.), 0.0064 m (0.25 in.), and
0.6 m (24 in.), respectively. The equivalent flexural engineering constants for the glass
(E ¼ 51 GPa) are ðE fx ¼ E fy ¼ 1075 MPa, Gxy ¼ 0, vx ¼ vy ¼ 0Þ: The initial stiffness (P/ )i
of the initial design is 1.02 106 N/m. The stiffness ratio m is 5.5 and the initial design
variables have to be changed according to Equation (5-7).

h3 b 24
5:5 ¼ ð5-7Þ
63 0:25 L

As demonstrated in the previous design example with the carbon sections, the best
design option is to have taller ribs, and the next best design is to have thicker ribs rather
than smaller cells. For the current design with the glass sections, the rib height is
increased to its maximum limit (given in the problem specifications) of 0.2 m (8 in.) and the
rib thickness is increased to 0.015 m (0.6 in.) from the initial design. The cell size was
kept to its maximum limit of 0.6 m (24 in.) specified in the problem. Equation (5-7) and the
design constraints are satisfied. The weight and the cost of the final design are shown in
Table 10.
Comparing the best designs from carbon and glass sections, glass ICG is about 3.3 times
heavier and 30% cheaper than the carbon ICG. Considering the handling and installation
of the panel, the carbon ICG has greater benefits.

5.4 ICG Versus Sandwich Panel

An approximate estimation of the flexural stiffness of an ICG and a sandwich panel has
been made. A sandwich panel with [0/90]n skins and the same weight as the ICG was
considered. Figure 27 shows the geometric dimensions of the ICG and sandwich panel.
The rib height of the ICG and the core thickness of the sandwich panel are denoted as h.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


314 D. HAN AND S. W. TSAI
t [0/90] n skins s

(a) (b)

Figure 27. Geometric dimensions of (a) T-Beam in an ICG and (b) a sandwich panel.

The weight of the core is neglected. The total weight of top and bottom skins is same as
the weight of the ICG. Hence, the skin thickness of the sandwich panel is determined as
shown in Equation (5-8).

3
s ¼ fh ð5-8Þ
4

For optimum R value of one half, the flexural stiffness of the ICG is

E ficg ¼ fE rib ð5-9Þ

The effective stiffness of the skins are approximately one half the rib stiffness.

1
E skin ¼ E rib ð5-10Þ
2

Normalized flexural stiffness of the sandwich, E fsandwich , panel is shown in Equation (5-11).

 2  3 !
h h 9 9
E fsandwich ¼2 E skin s ¼ fE rib ¼ E ficg ð5-11Þ
2 12 4 4

From Equation (5-11), the specific flexural stiffness of the ICG is 44% that of sandwich
panel of the same height.
This comparison is only approximation, i.e., the uni-directional pultruded rib usually
has higher fiber volume fraction than cross-ply laminated skin with the same base
materials. Also the weight of the sandwich core is neglected. The specific stiffness of the
ICG will be close to 50% when these factors are considered. Each process results in
different properties and limitations. For example, overall thickness of a sandwich panel is
limited by that of the cores. Pultrusion has also its limits. Although cost comparison
cannot be easily made, the cost of ICG should be considerably lower as it does not require
bagging, curing that sandwich construction requires.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


Interlocked Composite Grids Design and Manufacturing 315

6. CONCLUSIONS

A simple and low cost manufacturing method for a flat ortho-grid panel has been
developed. This new type of grid, the ICG, takes advantage of carbon or glass pultruded
sections which are intrinsically low cost and possess high stiffness and strength in the fiber
direction. Utilizing the slotted joints, assembly of the pultruded sections into a grid pattern
is fast and simple. The property reduction due to the slots is recovered by the bonded rib-
caps which bridge the open slots. Although the assembly concept is very simple, the final
product has high performance and low cost.
The equivalent stiffness of an ICG has been formulated and shown accurate predictions
for the global flexural behavior. A simple design rule has been established using the
equivalent stiffness model as a tool. It has been demonstrated that the design of an ICG is
very simple and flexible. Also, an approximate cost estimation rule has been established.
Flexural stiffness of an ICG has been compared to that of a sandwich panel of the same
weight and height. Considering the cost of fabrication and the thickness limitation of
sandwich panels, the ICG can be proposed as a competitive high stiffness structures that
can substitute sandwich panels.
A large number of beam and panel specimens have been built and tested. For the
properties and failure modes tested, it has been demonstrated that the ICG has excellent
mechanical properties, such as high damage tolerance, resilience, and durability. The
initial failure occurred at the roots of the slots in combined shear and transverse tension
mode due to the stress concentrations. Stress distributions near the slots are very
complicated, and more elaborate stress analysis along with systematic life tests remain
as future tasks before design allowables and long term system reliability can be
established.
Finally, pultrusion of carbon composites must be studied. The carbon pultrusion lags
behind the glass pultrusion by lack of the experience. For many applications, carbon
pultrusion is required for its high stiffness, three times that of glass pultrusion, and alkali
resistance for concrete reinforcement.

REFERENCES

1. Tsai, S.W., Liu, K.S. and Manne, P.M. (1999). Manufacturing and Design of Composite
Grids. In: Miravete, Antonio, (Ed.), 3-D Textile Reinforcements in Composite Materials,
Chapter 5.
2. Hong-Ji Chen (1995). Analysis and Optimum Design of Composite Grid Structures, PhD Thesis,
Stanford University, Stanford, CA, USA.
3. Steven M. Huybrechts (1995). Analysis and Behavior of Grid Structures, PhD Thesis, Stanford
University, Stanford, CA, USA.
4. Tracy B. Colwell (1996). The Manufacturing and Application of Composite Grid Structures, PhD
Thesis, Stanford University, Stanford, CA, USA.
5. Meink, Troy E. et al. (1996). Hybrid tooling for Advanced Grid Stiffened (AGS) Structures.
In: Proceedings of the 1996 28th International SAMPE Technical Conference, Seattle, WA, USA,
Technology Transfer in a Global Community International SAMPE Technical Conference
Vol. 28, pp. 542–551, SAMPE, Covina, CA, USA.
6. Huybrechts, Steven et al. (1997). Advanced Grid Stiffened Structures for the Next Generation
of Launch Vehicles. In: Proceedings of the 1997 IEEE Aerospace Conference, Snowmass
Village, CO, USA, IEEE Aerospace Applications Conference Proceedings Vol. 1, 97CH36020,
pp. 263–269.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015


316 D. HAN AND S. W. TSAI

7. Huybrechts, Steven et al. (1998). Ultra Lightweight Structures for Deployed Optics.
In: Proceedings of the 1998 IEEE Aerospace Conference, Part 2 (of 5), Snowmass at Aspen,
CO, USA, IEEE Aerospace Applications Conference, Proceedings Vol. 2, IEEE Comp. Soc.,
Los Alamitos, CA, USA, 98TH8339, pp. 83–88.
8. Timothy G. Gutowski (1997). Cost, Automation, and Design, Advanced Composites
Manufacturing, p. 525, Wiley Inter-Science.
9. Dongyup Han (2001). Design and Manufacturing of Interlocked Composite Grids, PhD Thesis,
Stanford University, Stanford, CA, USA.
10. Tsai, S.W. (1992). Theory of Composite Design, Think Composites, Dayton, OH.

Downloaded from jcm.sagepub.com at CARLETON UNIV on May 4, 2015

You might also like