You are on page 1of 12

Composites: Part B 27B (1996) 295-305

Copyright © 1996 Elsevier Science Limited


Printed in Great Britain. All rights reserved
ELSEVIER
1359-8368(95)00015-1 1359-8368/96/$15.00

Analysis and design of pultruded FRP shapes


under bending

J. F. Davalos*, H. A. Salim, P. Qiao, R. Lopez-Anido


Department of Civil and Environmental Engineering and Constructed Facilities Center
(CFC), West Virginia University, Morgantown, West VirgifJia 26506-6103, USA

and E. J. Barbero
Department of Mechanical and Aerospace Engineering and CFC, West Virginia University,
Morgantown, West Virginia 26506-6103, USA
(Received December 1994; accepted April 1995)

A comprehensive approach for the analysis and design of pultruded FRP beams in bending is presented. It is
shown that the material architecture of pultruded FRP shapes can be efficiently modeled as a layered system.
Based on the information provided by the material producers, a detailed procedure is preserited for the
computation of fiber volume fraction (Vr) of the constituents, including fiber bundles or rovings, continuous
strand mats, and cross-ply and angle-ply fabrics. Using the computed Vrs, the ply stiffnesses are evaluated
from selected micromechanics models. The wall or panel laminate engineering constants can be computed
from the ply stiffnesses and macromechanics, and it is shown that the predictions correlate well with coupon
test results. The bending response of various H and box sections is studied experimentally and analytically.
The mechanics of laminated beams (MLB) model used in this study can accurately predict displacements
and strains, and it can be used in engineering design and manufacturing optimization of cross-sectional
shapes and lay-up configurations. The experimental results agree closely. with the MLB predictions and
finite element verifications.

(Keywords: pultrusion; FRP shapes; modeling; micromechanics; laminated beams; experimental results; finite elements)

1. INTRODUCTION conventional materials can not always be readily applied


to FRP shapes. On the other hand, numerical methods,
Pultruded fiber-reinforced plastic (FRP) beams and such as finite elements, are often difficult to use, require
columns are increasingly used in Civil Engineering specialized training, and are not always accessible to
structures, due to their favorable properties like light design engineers. Therefore, to expand the structural
weight, corrosion resistance, and electromagnetic trans- applications of pultruded sections, a comprehensive
parency. In the design of fRP structural members, both engineering design method for FRP shapes should be
stiffness and strength properties are equally important developed. Such a design tool should allow practitioners
and depend on the material system and the geometry of to perform analyses of customized shapes as well as to
the cross-section. While changes in the geometry of FRP optimize innovative sections.
shapes can be easily related to changes iIi stiffness, Although it is generally accepted that pultruded
changes in the material constituents and fiber orienta- sections can be sitnulated as laminated composites2 ,
tionsdo not lead to such obvious results. In addition, there are no available guidelines for the computation of
shear deformations in pultruded FRP composite materi- the fiber. volume fractions (Vr) of the constituent
als are usually significant}, and therefore, the modeling materials in the laminate. Therefore, this paper presents
of FRP structural components should account for shear practical formulae and examples that permit the evalua-
effects. tion of Vrs of fiber-bundles or rovings, coninuous strand
Because of the complexity of composite materials, mats, and cross-ply and angle-ply fabrics. For pultruded
analytical and design tools developed for members of sections, it is not practical to evaluate the ply stiffnesses
experimentally, since the material is not produced
* Member ICCE by lamination lay-up. This paper presents selected
Analysis and design of pu/truded FRP shapes: J. F. Davalos et al.

micromechanics formulae for the computation of the laminated composites require accurate predictions of
laminae stiffnesses based on their respective Vfs. In ply stiffnesses, which can be evaluated throu.gh micro-
particular, the use of micromechanics formulae for mechanics by estimating the Vf of the constituents; and
compo·sites with periodic microstructure 3 is recom- (2) the modeling of member response requires a relatively
mended. The ply stiffnesses can then be used in classical simple, but accurate method that can be used in
lamination theory (CLT)4 to predict the laminate engineering analysis and design, which can be accom-
stiffnesses. In this study, predictions with CLT are plished by the proposed MLB model.
favorably compared to experimental results of FRP
coupon samples cut from full-size sections.
Based on a formal engineering approach of first-order 2.1. Prediction of material properties
shear deformation theory, the mechanics of thin-walled
Pultruded sections, such as H, box, and other shapes,
laminated composite beams (MLB) was developed 5 .
consist typically of arrangements of flat walls or panels.
Accounting for the cross-section geometry and material
Usually, the reinforcement used is E-glass, and the resin
anisotropy, this formulation permits the computation of
or matrix is either vinylester or polyester. Although these
beam stiffness coefficients and shear correction factor,
sections are not produced by lamination lay-up, the
that can be conveniently used in Timoshenko beam
arrangement of the constituent materials can be simu-
equations. Then, the· stress resultants are used to
lated as a layer system, and the stiffness properties of
compute ply stresses in a post-processing operation. In
each panel can be predicted by lamination theory in
this study, MLB is used as an efficient tool for the
terms of the ply stiffnesses computed through micro-
analysis and design of pultruded FRP sections. The
mechanics. A typical pultruded section may include the
predictions of MLB for H and box beams are v~lidated
following four types of layers (Figure 1): (1) A thin layer
with finite element analyses and an experImental
of randomly-oriented chopped fibers (Nexus) placed on
program.
the surface of the composite. This is a resin-rich layer
The experimental study presented in this paper
primarily used as a protective coating, and its contribu-
includes· two H and two box pultruded beams, which
tion to the laminate response can be neglected; (2)
are tested in bending to evaluate displacements and
continuous strand mats (CSM) of different weights
strains. The samples are subjected to 3-point and 4-point
consisting of continuous randon11y-oriented fibers; (3)
loadings, and the largest box section (10 x 10 x 3/4") is
stitched fabrics (SF) with different angle orientations,
tested at various span-to-depth ratios to evaluate shear
and (4) roving layers that contain continuousunidirec-
effects. The predictions of the MLB model agree well
tional fiber bundles, which contribute the most to the
with the experimental results, which are also verified with
stiffness and strength of a section. Each layer is modeled
finite element analyses using 8-node isoparametric shell
as a "homogeneous, linearly elastic, and generally
elements of ANSYS. Through the analytical and
orthotropic material, and to evaluate its properties, the
experimental program described in this paper, an
information provided by the material producer and
efficient and comprehensive analysis and design
pultrusion manufacturer are used to compute Vf of each
approach for pultruded FRP beams is illustrated. The
equivalent ply, as explained next.
proposed approach intends to bridge a gap between the
sophisticated modeling of composite materials and the
requirement for a simple but consistent tool for 2.1.1. Computation of fiber volume fraction. The Vf
engineering design. for each ply, either a roving or a CSM or a SF layer, is
Since there are no simplified and yet rigorous design
guidelines for pultruded FRP members, it is significant
that this paper presents a practical method that can be 62.5/ft - 61 yield roving
used for the analysis, design, and optimization of FRP 15.5 oz 90° SF
Nexus Veil
beams. The paper is organized in two major sections: (1) 1/2 oz CSM
Modeling considerations, which include prediction of
1 oz CSM
ply/laminate stiffnesses and member response; and (2)
analytical and experimental evaluations of FRP beams,
~=IIIIIIIII~~2: 12 oz ±45° SF
which include test results of coupon samples and full-size 62.5/ft - 61 yield
15.5 oz 90° SF
members.
1 oz CSM

2. MODELING OF PULTRUDED FRP SHAPES Total # of 61 yield rovings = 250


8"
FRP shapes are not laminated structures in a rigorous
sense. -However, they are produced with material
architectures that can be simulated as laminated config-
urations. In this section, the following two issues are
discussed: (1) The modeling of FRP sections as Figure 1 Lay-up of illustrative example
Analysis and design of pultruded FRP shapes: J. F. Davalos et al.

defined as the ratio of the volume of fibers present to the reasonably accurate stiffness predictions for roving and
total volume of the layer, and therefore, the relative SF layers. The formula for G I2 given by the composite
volumes of fiber and matrix are determined to evaluate cylinders model is
the ply stiffnesses. Similarly, the Vr of the whole section
G - G (1 + Vr)Gr + (1- Vr)G m
can be defined. For the CSM and SF layers, which are (5)
12 - m (1 - Vr)G f + (1 + Vr)G m
respectively specified in oz/ft 2 and oz/yd 2, the Vr can be
determined as follows: For comparative purposes and because of its popularity,
w the elastic properties of roving and SF layers are also
(Vr )cSMjSF = pt (1 ) computed by the mechanics of materials rule of mixtures
approach 4. For example, the expression for G12 is given
where, w is the weight per unit area in Ib/in 2, P is the by
unconsolidated density of the CSM or SF fibers in Ib/in 3,
and t is the 'as manufactured' thickness of the material _ GmGr
G 12 - (6)
(inches) as provided by the material producer. Thus, as (1 - Vr)Gr + VrG m
an approximation, it is assumed that the thicknesses of The stiffness of the CSM layers are estimated from the
these constituents remain constant during part manu- approximate expressions ll for randomly oriented com-
facturing, and that the voids are filled with resin. For the posites, based on the assumption that the material is
roving layers, the Vr is defined as: isotropic in the plane 12 . The following expressions given
in ref. 11 are used:
(Vr)r = nrA r (2)
tr E
II == - - 1 (7)
where, n r is the number of rovings per unit width (in-I) 2G
provided by the manufacturer, tr is the assumed in situ where, Eland E 2 are computed from any of the micro-
thickness of the roving-matrix layer, which is computed mechanics models described earlier, and E, G, and II are
by subtracting the thicknesses of the CSM and SF layers equivalent isotropic elastic constants. Recent experi-
from the pultruded laminate thickness; A r is the area of mental results for CSM strips have shown that the
one roving computed from: extensional stiffness, E, estimated with the above formula
1 is approximately 15% higher than the experimental
Ar == - (3) value 13 .
YPr
where, Y is the yield specified in yards/lb and converted 2.1.3. Prediction of panel stiffnesses. Once the ply
to in/lb, and Pr is the density of the fibers given by the stiffnesses for each flat panel or wall section of a pul-
producer. Once the Vrs for all the typical layers are truded shape are computed through micromechanics
computed, the ply stiffnesses are predicted using selected formulas, the stiffnesses of a panel can be computed
micromechanics formulas. from classical lamintion theory (CLF)4. In particular,
the compliance matrix corresponding to the extensional
2.1.2. Prediction ofply stiffness. Several formulae of stiffness matrix, A, can be used to evaluate the laminate
micromechanics of composites have been developed engineering properties in terms of the transformed stiff-
and used over time 6 . The degree of correlation between ness coefficients Qij. An example is given next to illus-
experimental data and theoretical predictions depends trate the computation of ply and laminate elastic
on the accuracy of the model used. In this study, the properties.
ply stiffnesses for the roving and stit,ched fabric layers
are computed using primarily a recent micromechanics
2.1.4. Example: elastic properties of a pultruded box
model for composites with periodic microstructure
section. Using the manufacturer's information and
developed by Luciano and Barbero 3. Detailed expres-
sions for the computation of the elastic constants E l , the micro/macromechanics approach described above,
E 2 , G I2 , and lI12 are given in the original paper along
the material properties of the box section shown in
Figure 1 are evaluated. This 4 x 8 x 1/4" pultruded
with experimental correlations. As an illustration, the
expression for the computation of the in-plane shear shape was manufactured using E-glass and vinylester
modulus G 12 is given by: resin by Creative Pultrusions Inc., and it consists of six
CSM layers, two cross-ply and two angle-ply SF layers,

G 12 == Gm - Vr [ -
S3
Gm + Gm -
1]
G
-1
(4)
and two roving layers. The Vrs' of the constituent layers
r are computed from equations (1) and (2) and the infor-
mation given in Table 1. As explained before, the thick-
where, the subscripts f and m refer to fiber and matrix ness of the roving layers is computed by subtracting the
respectively, and S3 is given by thicknesses of the CSM and SF layers from the total
83 == 0.49247 - 0.47603 Vr - 0.02748 Vf
thickness of the laminate (0.25").
To evaluate the ply stiffnesses through micromecha-
In addition, the composite cylinders model 7 which is nics, the properties of fiber and resin are given in Table 2.
based on the self-consistent theory8-IO, can provide To obtain accurate results, the elastic properties of the
Analysis and design of pultruded FRP shapes: J. F. Davalos at al.

Table 1 Ply properties

No. of plies and type Nominal weight* Thickness (in)


2 4
2, 1/2oz CSM 1/2oz/ft 2.17014 X 10- 0.0075* 0.315
4,1 oz CSM loz/fe 4.34028 x 10- 4 0.0200* 0.236
2, 15.50z 90° SF 15.50z/yd2 7.47492 X 10-4 0.0230* 0.353
2, 12 oz ± 45° SF 12.2oz/yd 2 5.88349 x 10-4 0.0220*t 0.291
2, 61 yield· roving 61 yards/lb 4.9497 x 10- 3 0.0325 0.793
* provided by material producer . .
t computed by subtracting the thicknesses of the CSM/SF layers from the lamInate thIckness

Table 2 Material properties of the constituents The compliance matrix [a], which is the inverse of A,
Material E (psi) G (psi) v
becomes:
6 6 6 7
E-glass fiber 10.5 X 10 4.1833 X 10 0.255 0.092 1.19 X 10- -3.831 X 10- 0 ]
Vinylester resin* 7.336 x 105 2.3714 X 105 0.30 0.041
[a] == -3.831 X 10- 7 1.536 x 10- 6 0
[
*obtained experimentally (Tomblin 1994)
0 o 4.902 X 10-6

From which, for example, Ex is simply given by


vinylester resin were obtained experimentally by testing
Ex == 1/( tall)' where t == 0.25 in. is the laminate thick-
all-resin rods in tension and torsion 14. The roving and SF
ness. Similarly, the remaining elastic properties are com-
ply properties are computed from the micromechanics 6
formulae described above (e.g. equations (4)-(6)), and Puted , and the results are: Ex ==6 3.361
.
X 10 psi, E y ==
6
2.604 X 10 psi, Gxy == 0.816 x 10 pSI, and v xy == 0.322.
the properties of the CSM layers are computed from
The accuracy of these values can be evaluated by testing
equation (7). The properties of the ±45° SF layer are
coupon samples, as illustrated later in this paper. The
computed by dividing the layer into two lamin~e of
-modeling of FRP shapes is discussed in the next section.
opposite fiber orientation angles. The ply stiffnesses of
the constituent layers are summarized in Table 3, and it
can be observed that the extensional stiffneses computed 2.2. Prediction of member response
by the rule of mixtures formulae agree with those given
In this study, the response ofFRP shapes in bending is
by the periodic microstucture and composite cylinders
evaluated using a formal engineering approach to the
models. However, for the in-plane shear stiffness, the G 12
mechanics of thin-walled laminted beams (MLB), based
value given by the rule of mixtures is approximately 34%
on kinematic assumptions consistent with Timoshenko's
lower for the roving layer and 500/0 lower for the CSM
beam theory. An overview of the MLB theory used is
layers in relation to the other two models.
presented next.
Using CLT, the engineering properties (Ex, Ey , vxy ,
The three different methodologies that are available in
and Gxy ) of the panel are computed by assembling the _
the literature to evaluate the stiffness of laminated
transformed stiffness coefficients into the extensional · dI5 .
composite beam elements under flexure are d Iscusse
stiffness matrix A. The expressions for Ai} are given as 4
The approach proposed by WItney h· et a.116 and T sal·17
N
considers that the effective moduli -of a laminated beam
Ai} == L(Qij)ktk i,}== 1,2,6 (8) are the corresponding engineering constants of the
k=I
laminate. Hence, the effective beam moduli are obtained
where, tk is the thickness of the k th ply, and for this from the reciprocals of the components of the laminate
example, the stiffness matrix A becomes compliance matrix, which is obtained by full inversion of
6 6 the laminate stiffness matrix. The basic assumption in
0.914 x 10 0.228 X 10 0 ]
A ==
[
0.228 X 106 0.708 X 106 ° this approach is that the lateral force and moment stress
resultants in a beam element are zero. Lopez-Anido et al. I5
o 0 0.204 X 106 showed experimentally that this approach represents the.

Table 3 Ply material properties computed by three micromechanics models

E I (10 6 psi) £2 (10 6 psi) Vl2 G I2 (106 psi)

Lamina PM CC ROM PM CC* ROM PM CC ROM PM CC ROM

1/2oz CSM 2.093 2.132 2.077 2.093 2.123 2.077 0.407 0.412 00412 0.744 0.755 0.337
1oz CSM 1.710 1.758 1.727 1.710 1.758 1.727 0.402 0.407 0.404 0.610 0.625 0.305
15.50z 90° fabric 4.118 4.276 4.181 1.183 1.141 1.092 0.389 0.281 0.284 0.457 00455 0.356
12 oz ± 45° fabric 3.505 3.674 3.576 1.056 1.022 1.006 0.396 0.284 0.287 0.405 0.404 0.327
61 yield roving 8.469 8.438 8.478 3.374 3.093 2.796 0.343 0.262 0.264 1.429 1.387 0.941
* E2 is an average of the computed upper and lowe: limi~s .
Note: PM = periodic microstructure, CC = compOSIte cylInders, ROM = rule of mIxtures
Analysis and design of pultruded FRP shapes: J. F. Davalos et al.

actual stiffness of laminated composite beam elements (Ns == M s == M sz == 0), and assuming that the off-axis
under flexure for span-to-width ratios 2: 6. Therefore, plies are balanced symmetric (Q16 == (316 == 0), the exten-
the· approach is adopted in this study to model pultruded sional, bending-extension coupling, bending, and shear
structural shapes by first-order shear deformation theory stiffnesses of the i th wall are expressed as: .
for thin- and moderately thick-walled laminated beams - -1 - -1
with open or closed cross-sections 5 • Ai == (811~ )i' Bi == (-{311 ~ )i'
- -1 - -1
In the present model, the stiffnesses of a beam are Di == (a 11 ~ ) i, Fi == (a66 ) i ( 10)
computed by adding the contribution of the stiffnesses of
the component panels, which in turn are obtained from
where ~ == a11 b11 - f3r 1
the effective beam moduli. The model accounts for where Ns ' Ms and Msz are laminate resultant force and
membrane stiffness and flexure stiffness of the walls, but moments and a16 and {316 are compliance elements.
shear lag warping effects are not considered. The position General expressions for the beam stiffness coefficients
of the neutral axis is defined in such a way that the are derived from the beam variational problem. Hence,
behavior of a beam-column with asymmetric material axial (A z ), bending (D x , D y ), and shear (Fx , F y ) stiffnesses
and/or cross-sectional·shape is completely described by that account for the contribution of all the walls can be
axial, bending, and shear stiffness coefficients (A z , D y , F y ) computed as
only. Furthermore, a shear correction factor is obtained n
from energy equivalence. An overview of the mechanics Az == LAibi
of laminated beams (MLB) theory is presented next. j=1

The basic kinematic assumptions in MLB are: (1) The n


contour does not deform in its own plane, and (2) a plane By == L[Ai(Yi - Yn) + B i cos ¢i]b i
section originally normal to the beam axis remains plane, i=1

but not necessarily normal to the beam axis due to shear


deformation. Straight FRP beams with at least one axis
of geometric and material symmetry are considered. The
Dy = 6n [ - ( -
Ai (Yi - Yn)
2
+ b2)
i· 2
12 sm <Pi (11 )

pultruded sections are modeled as assemblies of flat 2


walts. The compliance matrices [a]3x3' LB]3x3, [c5]3x3 are + 2B;(:Yi - Yn) cos <Pi + Di COs <Pi] hi
obtained from classical lamination theory (eLT)I? For n
each wall the position of the middle surface is defined by Fy == ~ - ·2
~ Fjb j sin ¢i
the function (Figure 2): i=1

The beam bending-extension coupling coefficient (Bx or


(9) By) can be eliminated by defining the location of the
neutral axis of bending (x n or Yn) as
where bi is the wall width, Yi is the position of the wall n
centroid, and ¢i is the orientation of the i th wall. L(YiAi + cos ¢iBi)bi
By defining stress resultant assumptions in each i=I
wall compatible with beam theory without torsion Yn == - ------- (12)
Az

_.__. . __. .._. . __._.._..__ _._._ __. .._ _. ... 1 . . . - _ -.-&0 ~

Figure 2 Cross-section geometry and reference coordinate systems


x
Analysis and design of pultruded F~P shapes: J. F. Davalos et al.

Introducing the coordinate y' == y - Yn, the extensional


and bending responses are decoupled. An explicit
expression for the static shear correction factor (Kx ; K y )
is derived from energy equivalence. As an approximation
in design, the shear correction factor for pultruded Hand
box sections can be taken as 1.0. General equations for
various FRP sections are presented in ref. 18.
Displacement and rotation functions can be obtained
by solving the Timoshenko beam theory equilibrium
equations. Deflections at discrete locations can be com-
puted by employing energy methods that incorporate the
beam bending and shear stiffnesses. General formulae for
maximum bending deflections (8b ) and shear deflections (b) H 6"x6"x1/4"
(8s ) for typical beam loadings and boundary conditions are
Figure 3 Layup configurations of pultruded H sections
available in manuals. For example, the maximum deflec-
tion for a four-point loading of a beam of span Land
concentrated loads (P) applied at L/3 is:

23PL 3 PL
8== 8b +8s == 648D +3KF (13)
[::Jp •••••••••••• 20zCSM
6.25/1n -61 yield

30zCSM

y y y (a) Box 4 I x4"x1/4"

For a beam loaded in the z-y plane, the axial, shear, 1/20z CSM -
30zCSM iijiiiijjiijijjjjir Veil
2 ot CSM Gl-4.2/in -28 yield
and bending stress resultants are N z , V y, and My. Then, 3 oz CSM • .;• • •l __ 2.B/in -28 yield
the strains and curvature at the middle surface of the i th
wall are given as
30zCSM
30zCSM :::llllllll1a~ 2.B/in -28 yield

........... r-2.8/in -28 yield


30zCSM - . ""%=ww,;-~. ;s' " ,,,',, ,
3 oz CSM ••• •• io--2.8/in -28 yield
2 oz CSM ~"·'i:> ,.,

3 ozCSM . 2.8/in -28 yield


10zCSM Veil

(b) Box 101 x101 x3/4"

(14) Figure 4 Layup configurations of pultruded box sections

and the stress resultants in the i th wall are


18"
Nz(Si' z) == AiEz + Bi'Xz, Nsz(Xi' z) == F/ysz, 14 -I
Mz(Si' z) == BiEz + D/Xz (15) 211 ~ I
th
Ply strains and stresses, at a location (Si' z) of the i Longitudinal
wall, can be obtained from the six stress resultants by a"


III( rsl

employing CLT 17 . The location of the shear center


S(x s,Ys) is defined in order to decouple bending and 2~1 I
Transverse

torsion. Evaluation of deflections and ply stresses for an


H section were satisfactorily correlated with a refined
layered ·shell finite element model 5 • The modeling con-
cepts discussed in this section are used in this paper to
evaluate the response of H and box FRP sections tested
in bending.

3. RESPONSE EVALUATION OF PULTRUDED


Box 10·x10·x3/4· D
Figure 5 Plan view of coupon samples cut from box section

FRP SHAPES
3.1. Experimental evaluation
The responses of two H and two. box FRP beams are
evaluated experimentally and analytically. The beams Four types of pultruded cross sections manufactured
are tested in bending, and deflections and strains are using E-glass fibers and vinylester resin are evaluated in
recorded. The test beams are analyzed using micro- this study. The sections include: two H (4 x 4 x 1/4" and
mechanics 3 and macromechanics 4 in combination with 6 x 6 x 1/4") beams, each consisting of 7 layers; one box
MLB 5 and shell finite elements 19 . The experimental and beam (4 x 4 x 1/4") consisting of 3 layers; and one box
analytical procedures and results are discussed in this beam (10 x 10 x 3/4") consisting of 171ayers. The stack-
section. ing sequence of the cross-sections of the test-samples are
Analysis and design of pultruded FRP shapes: J. F. Davaloset ai.

Table 4 Laminate properties comparisons of box lOx lOx 3/4" section

Exx Eyy v xy Gxy


6 6
Experimental 4.147 x 10 psi 2.044 x 10 psi 0.301 7.144 x 105 psi
(COV) (3.37%) (9.06%) (2.03%) (2.770/0)
Analytical 4.021 x 106 psi 1.991 x 106 psi 0.294 7.250 x 105 psi
%Difference -3.0% -2.6% -2.3% +1.5%

COY = coefficient of variance


Note: %difference is based on the experimental results

Table 5 Engineering constants of laminae for H and box sections

E} £2 G 12 tk
Beam section Layer* (10 6 psi) (10 6 psi) (10 6 psi) V12 (in)

H4x4x 1/4" 1.50z CSM 1.716 1.716 0.605 0.419 0.025


20z CSM 1.861 1.861 0.655 0.421 0.03
30 rovings (113 yield) 4.320 0.959 0.371 0.293 0.05
28 ravings (113 yield) 4.065 0.912 0.353 0.294 0.05
matrix 0.490 0.419 0.198 0.240 0.045
Box 4 x 4 x 1/4" 30z CSM 2.140 2.140 0.752 0.423 0.04
20z CSM 1.861 1.861 0.655 0.421 0.03
25 rovings (61 yield) 3.043 0.758 0.293 0.296 0.18
H6x6xl/4" 1.50z CSM 1.716 1.716 0.605 0.419 0.025
20z CSM 1.861 1.861 0.655 0.421 0.03
22 rovings (61 yield) 4.047 0.909 0.351 0.294 0.05
28 rovings (61 yield) 5.016 1.104 0.428 0.289 0.05
25 rovings (61 yield) 4.532 0.999 0.387 0.292 0.05
matrix 0.490 0.490 0.198 0.240 0.045
Box 10 x 10 x 3/4"t 1/20z CSM 2.090 2.090 0.743 0.297 0.0075
1oz CSM 1.716 1.716 0.610 0.423 0.02
20z CSM 2.090 2.090 0.743 0.297 0.03
30z CSM 2.290 2.290 0.810 0.295 0.04
28 rovings (28 yield) 5.370 1.520 0.595 0.280 0.06375
42 rovings (28 yield) 7.650 2.630 1.076 0.273 0.06375
*The rovings are reported as the total number per panel
t vinylester material properties were obtained experimentally

shown in Figures 3 and 4. The 10 x 10 x 3/4" box section bars. The results are reported in Table 4, along with the
was manufactured by Morison Molded Fiber Glass analytical predictions computed from micromechanics3
Company of Bristol, VA, and the other three sections and·CLT. The differences between the predicted and aver-
were produced by Creative Pultrusions Inc. of Alum Bank, age computed values are less than 3%. Therefore, for the
PA. The experimental program included testing of coupon other three sections tested in this study, the material prop-
samples, cut from one of the box beams, and full- erties were only evaluated analytically.
member samples.
3.1.2. Member testing. The two H and two box
3.1.1. Coupon testing. The laminate stiffnesses of the sections were tested in bending under 3-point and also
walls or panels of the FRP beams used in this study can 4-point loadings. For the 3-point test, the load was
be predicted using the micro/macromechanics approach applied at the midspan of the sample, and for the 4-point
described in Section 2.1, and to validate the accuracy of test of the 10 x 10 x 3/4" box section, the concentrated
the proposed models, coupon samples were cut from the loads were applied at third points, where for the other
10 x 10 x 3/4" box section and tested in tension and tor- three sections, the loads were applied at 20" from each
sion. Samples of 2-inch widths were cut along the longi- end, with a distance between loads of 32". Both H beams
tudinal and transverse directions of the box beam (see and the 4 x 4 x 1/4" box beam were tested for a span of
Figure 5). The longitudinal samples were 18" long, and 6ft; whereas the 10 x 10 x 3/4" box section was tested
the transverse samples were 8" long (limited by the 10" for five spans: 6, 8, 10, 12, and 14 feet. The samples
dimension of the box beam). Eight longitudinal and were tested in a load frame, and the load was applied
five transverse samples were tested in tension to deter- with a hydraulic jack and recorded using a load cell.
mine the extensional stiffnesses, Ex and Ey , and Poisson's For all the samples, the maximum deflections at
ratio v xy ' The shear stiffness Gxy was determined from midspans were recorded with linear variable differential
torsion tests of 10 longitudinal samples. Using the transducers (LVDTs), and the strains at midspan were
assumption of transverse isotropy20 and the measured recorded with strain gages. For the two Hsections and
torque and angle of twist, the shear stiffness was com- the 4 x 4 x 1/4" box section, which were tested 21 , only
puted from Saint Venant's torsion solution for rectangular the maximum strains at top and bottom flanges were
Analysis and design of pultruded FRP shapes: J. F.Davalos et al.

j------------------------------------------------------------------------------,
I I
displacements were recorded with a computerized data
I I
I I
acquisition system. The analytical evaluations are presented
i Manufacturer's Information ! first, and the results of the experimental/analytical response
~! !
~l
V'I
!
I
are presented and discussed at the end of this section.
g!
~
~odic Microstru~ !
- -!
!: : 3.2. Analytical evaluations
c.{ Ply Mechan.ical Properties i
I
I
IL
I
I
1I
Based on the modeling of pultruded FRP shapes
------------------------------ ---------------------------------------------. I
discussed in Section 2, the computer program FRPBEAM
was developed by Qiao et al. 22 to model and analyze
I
I
I
I
,
I
I FRP shapes in bending. The response of the four sections
I
I

Panel Mechanical Properties


I
I
I
I
tested in this study (Figures 3 and 4) are analyzed with
I
I
this program and also with the finite element program 19.
The computation of ply stiffnesses follows the procedure
.... presented in Section 2. A summary of the computed ply
o
en Beam Stiffness Coefficients properties are given in Table 5. The modeling of member
~l response are briefly discussed in this section.
o!
2! The displacements and strains of the four sections
Co il: Timoshenko evaluated in this paper are predicted using the theory
Beam Theory described in Section 2 and implemented in the computer
program FRPBEAM. A flowchart of the computational
i1
: Beam Displacements & procedure is given in Figure 6, and the three major
modules are: (1) A preprocessor to evaluate the ply
l
Panel Stress Resultants
properties, as discussed above; (2) a processor to
,1_______________________________ _ _
evaluate wall or panel stiffnesses with CLT (equation
~. r---------------------------- ----C@--------------------------' (10)), beam stiffness coefficients and shear correction
factor with MLB (equation (11)), beam displacements
~l and rotations with Timoshenko beam theory (e.g.
o!
~l Ply Strains & Stresses equation (13)), panel midsurface strai:'ls and curvatures
~! (equation (14)), and panel stress resultants (equation
enl
01
Q.t _ (15)); and (3) a post-processor to evaluate ply strains and
stresses with CLT. The computed pan~l stiffnesses of
Figure 6· Computational flowchart of program FRPBEAM
webs and flanges of the test beams are given in Table 6,
and the beam stiffness coefficients and shear correction
measured, while for the 10 x 10 x 3/4" box section, factors are given in Table 7, where the membrane and
strains were also measured at five locations on one web flexural components of the bending stiffness are listed
and at three locations on the opposite web. The tests separately. It is interesting to observe that the membrane
were conducted in the elastic range, and the strains and component, which is the first term of D y in equation (11),

Table 6 Panel stiffness coefficients of webs and flanges (given per unit width)

Flange Web

Extensional Bending Shear Extensional Bending Shear


A D P A D P
2 3 2 2 2 3 2
Beam Section (kip-injin ) (kip-in jin ) (kip-injin ) (kip-injin ) (kip-in jin ) (kip-injin 2 )
H 4 x 4 x 1/4" 816.0 3.72 118.0 611.0 3.61 108.0
Box 4 x 4 x 1/4" 690.0 3.13 102.0 690.0 3.13 102.0
H 6 x 6 x 1/4" 836.0 3.92 119.0 658.0 3.88 112.0
Box 10 x 10 x 3/4" 3010.0 134.00 527.0 3010.0 134.00 527.0

Table 7 Beam stiffnesses, shear correction factors

Shear stiffness Fy Shear correction


Beam section Membrane Flexural Total (1 061b-in 2 jin2 ) factor K y
H4x4x 1/4" 28.783 0.0298 28.844 0.406 0.9980
Box4x4xl/4" 24.271 0.0235 24.291 0.762 0.9959
H6x6xl/4" 93.325 0.0470 9.121 0.641 0.9977
Box 10 x 10 x 3/4" 1590.356 2.4676 1598.200 10.061 0.9959
Analysis and design of pultruded FRP shapes: J. F. Davalos et al.

Table 8 Deflectons and strains comparisons of FRP sections under 3-point loading

Maximum deflection (in/kip) Maximum strain (fLE/kip)

Beam section Span (ft) Exp. MLB 8MLB/8Exp. FE Exp. MLB EMLB/EExp. FE

H4x4x 1/4" 6 0.315 0.314 0.997 0.308 1232 1330 1.079 1400
Box 4 x 4 x 1/4" 6 0.327 0.344 1.052 0.331 1475 1480 1.003 1500
H6x6xl/4" 6 0.112 0.1116 0.996 0.1102 590 578 1.021 600
Box 10 x 10 x 3/4" 14 0.06258 0.06600 1.055 0.06206 129 130 1.008 135

Table 9 Deflections and strains comparisons of FRP sections under 4-point loading

Maximum deflection (in/kip) Maximum strain (fLE/kip)

Beam section Span (ft) Exp. MLB 8MLB/8Exp. FE Exp. MLB EMLB/EExP. FE

H4x4x 1/4" 6 0.2401 0.2263 0.943 0.210 810 738 0.911 750
Box 4 x 4 x 1/4" 6 0.2515 0.2525 1.004 0.246 910 823 0.904 830
H6x6xl/4" 6 0.0740 0.0781 1.055 0.0773 322 321 0.997 315
Box 10 x 10 x 3/4" 14 0.0504 0.0532 1.054 0.0510 77 78 1.013 75

is significantly larger than the flexural component (last ply stiffnesses proposed in this paper. For all four FRP
term of D y in equation (11 )). The total bending stiffnesses sections and a maximum load of 1000 lbs., the predicted
in Table 7 are computed by equation (11), accounting for maximum deflections and strains for the 3-point loading
the bending-extension effects (second term of D y in case are compared to the experimental values in Table 8.
equation (11)) in some of the sections which exhibited Similarly, the maximum deflections and strains for the
minor lay-up asymmetries. 4-point loading case are given in Table 9. The results
To verify the prediction accuracy with the FRPBEAM indicate that the predicted values with MLB agree very
program, the test beams are also analyzed with the well with. the· FE and experiinental v-alues, with
commercial finite element program ANSYS 19 using maximum differences of 5.5% for displacements and
Mindlin 8-node isoparametric layered shell elements 7.90/0 for strains. The MLB and FE results match closely
(SHELL 99), which include shear deformation. An the experimental values over the entire range of loading,
aspect ratio of one (length-to-width) is used to define as illustrated by Figure 7 for the midspan displacement of
the mesh for each member, and the input material the 6 x 6 x 1/4" H section under 3-point loading, and by
properties used are taken from Table 5. The analytical Figure 8 for the maximum strains of the 10 x 10 x 3/4"
and experimental results are correlated next. box section under 4-point loading.
To evaluate the shear effects, the 10 x 10 x 3/4" box
3.3. Comparison and discussion of results section was tested under five different spans (6, 8, 10, 12,
and 14 feet). The MLB explicit solution permits the
As shown in Table 4, the panel or laminate stiffnesses computation of the bending and shear components of
computed from the predicted laminae properties agree deflection, and as shown in Table 10, shear deflection for
closely with experimental results obtained for the coupon the 3-point loading accounts for approximately 27% of
samples of the 10 x 10 x 3/4" box section. These results
provide confidence in the computational procedure for
1000 ~------------------~1t-1
1000 P/2 ~ P/2
900

~o
900
800
800 I
r--.. 700 •I· r--..

g
700
600
;S
'-'" 600
~
~ 500
'"C 500 ~
~ ~ 400
0 400
...J 300
300
200 - - e - Experimental
200 - - e - Experimental -------+e---. :MLB
100
100 -------*---MLB - - . - FE (ANSYS)
---.- FE (ANSYS) O----.......-----.----r------,.--r----~-......----t
0 o 10 20 30 40 50 60 70 80
0 0.02 0.04 0.06 0.08 O. 1 0.12
Strain (JAS)
Center deflections (in)
Figure 8 Strains comparison at midspan for the 10 x 10 x 3/4" box
Figure 7 Deflections comparison for the H6 x 6 x 1/4" section section
Analysis and design of pultruded FRP shapes: J. F. Davalos et al.

Table 10 Midspan deflection comparisons for box 10 x 10 x 3/4" section

4-point loading 3-point loading

Exp. MLB . Exp. MLB

btotal btotal bshear btotal ratio btotal btotal bshear btotal ratio
L (ft) (in/kip) (in/kip) (%) (bMLB/ b Exp .) (in/kip) (in/kip) (%) (bMLB / b Exp .)

6 0.005621 0.005238 24.7 0.93 0.006320 0.006708 26.8 1.06


8 0.011235 0.011077 15.5 0.99 0.013002 0.014036 17.1 1.08
10 0.020172 0.020424 10.5 1.01 0.023672 0.025519 11.5 1.08
12 0.033899 0.034156 7.6 1.01 0.039677 0.042875 8.4 1.08
14 0.05041 0.053151 5.7 1.05 0.062582 0.066001 6.4 1.06

100
EXP p
0.05 ,.-.... Top
FE + = =
~ ..
'-'
MLB- oL oL

50
0.04 ~
..........
~ :lIW

~ 'i0
:g 0.03 4-point bending 0
(a)
:s
'-" :i~ * .....
~ 0.02 CJ
3-point bending tIJ -50 ... .. or ...
~
+
[J
e0 c 0 0 '-' 0
Bottom
0.01
~ -100
0.0 2.0 4.0 6.0 8.0 10.0
0.00 +---.......-----r----....------'T'"---r-----,.--~____i Width (in)
6 7 8 9 10 11 12 13 14
10.00 -r---~~-~----..,...----, ----.
Span (ft)
Figure 9 Deflection comparisons for various spans of the lOx lOx 8.00
3/4" box section

the total deflection for the 6-ft span, and it decreases to


:5 6.00
.........",

about 60/0 for the 14-ft span. As with the other FRP 'fJ (b)
~4.00
sections, the ratio of experimental to MLB predictions is
~
close to 1.0. The deflection comparisons with FE and
2.00
experimental results are shown graphically in Figure 9.
The longitudinal experimental and analytical strains for
three spans (6, 10, and 14ft) are compared in Figure 10, a .00 .......-r,.,..,.~r-1'"'1"T'T"T-M-rrr..,.,.,""I'-r'rrr~r-r-r-l~~r-r-r-r-4

-100.00 -50.00 0.00 50.00 100.00


showing close agreement. Figure lOa shows the strains
across the width of th.e top and bottom flanges, and Microstrains/load (tJSJkip)
Figure lOb shows the strains across the depth of the two Exp: c 6-ft + 10-ft * 14-ft MLB
apposite webs. To study the Poisson's effects, transverse
Figure 10 Longitudinal strains for the 10 x 10 x 3/4" box section: (a)
strains were also measured on top and bottom flanges, top and bottom flanges; (b) webs
and the experimental results were within 100/0 of the
MLB predictions.

The results of this study indicate that selected


4. CONCLUSIONS AND RECOMMENDATIONS micromechanics formulae can be used to accurately·
predict the ply stiffnesses of the constituent materials of
In this paper a comprehensive analytical and experi~ the laminate, including fiber-bundles or rovings, con-
mental program for the anaJysis and design of pultruded tinuous strand mats, and cross- and angle-ply fabrics.
FRP. shapes under bending is described. The interactive These relatively simple formulae can provide accurate
computer program FRPBEAM is developed to carry out results, provided that the constituent Vrs are correctly
the analysis of FRP beams, from the evaluation of ply estimated using the manufacturer's information and the
stiffnesses through micromechanics to the overall formulae given in this paper.
member response through the mechanics of laminated The experimental study discussed in this paper
beams (MLB) model. The proposed approach can be includes H and box FRP sections, and the MLB model
used as an efficient tool for engineering design by both predicts displacements and strains in excellent agreement
practicing engineers and pultrusion manufacturers. with the experimental results and finite element analyses
Analysis and design of pultruded FRP shapes: J. F. Davalos et ai.

with ANSYS19. The MLB model permits the computa- for strength, fracture toughness, and environmental effects.
NASA TM-83696, USA, 1984
tion ofshear and bending deflections separately, and has 7 Hashin, Z. and Rosen, B.W. The elastic modeling of fiber-rein-
the ability to analyze any cross-sectional shape with forced materials. J. Appl. Mecho 1964, (31), 223
some minor restrictions. Therefore, MLB can be used to 8 Whitney, J.M. and Riley, M.B. (September, 1966). Elastic prop-
erties of fiber reinforced composite materials. AIAA J. 1966,
conduct parametric studies, and it can be extended to 1537
optimization studies of existing and new shapes. 9 Budiansky, B. Micromechanics. Computers and Struct. 1983, 16,
In this study, distortional effects, such as shear-lag 3
10 Christensen, R.M. A critical evaluation for a class of micro-
and warping, which can be particularly significant for mechanics models. J. Mech. Phys. Solids 1990, 38, 379
thin-walled sections, were not considered. Moreover, the 11 Hull, D. 'An Introduction to Composite Materials', Cambridge
analysis is limited to linear-elastic response and does University Press
12 McCullough, R.L. 'Concept of Fiber-Resin Composites', Mar-
not include failure or buckling considerations. These cel Dekker, Inc., NY, 1971
additional concerns need to be addressed in the design of 13 Makkapati, S. Compressive strength of pultruded structural
FRP beams, and the present formulation can be used as FRP shapes, Master Thesis, West Virginia University, 1994
14 Tomblin, J.S. Compressive strength models for pultruded glass
the basis for further work in this area. fiber reinforced composites. Ph.D. Dissertation, Department of
Mechanical and Aerospace Engineering, West Virginia Univer-
sity, Morgantown, WV 26506, USA, 1994
ACKNOWLEDGEMENT 15 Lopez-Anido, R., Davalos, J.F. and Barbero E.J. Experimental
The FRP sections tested in this study' were partially evaluation of stiffness of laminated composite beam elements
under flexure. J. Reinforced Plastics and Composites 1995, (14),
contributed by Creative Pultrusions, Inc. and Morrison 349
Molded Fiber Glass Company. The authors thank 16 Whitney, J.M., Browning, C.E. and Mair, A. Analysis of the
Professor GangaRao and Mr Nagaraj for the use of flexure test for laminated composite materials. In 'Composite
Materials: Testing and Design (Third Conference)', ASTM
some of the experimental results presented in this paper. STP 546; American Society for Testing and Materials, 30, 1974
17 Tsai, S.W. 'Composites Design', Dayton, OH: Think Compo-
sites, 1988
18 Lopez-Anido, R. Analysis and design of orthotropic plates
REFERENCES stiffened by laminated beams for bridge superstructures. Ph.D.
Dissertation, Department of Civil and Environmental Engineering,
Mosallam, A.S. and Bank, L.C. Short-term behavior of pul- West Virginia University, Morgantown, WV 26506, USA, 1994
trtided fiber-reinforced plastic frame. J. of Struct. Engrg 1992, 19 'ANSYSUser;s Manual', Swanson Analysis System Inc., Hous-
118(7), 1937 ton, PA, 1992.
2 Barbero, E.J. Pultruded structural shapes: from the constituents 20 Davalos, J.F., Loferski, J.R., Holzer, S.M. and Yadama, V.
to the structural behavior. SAMPE J. 1991, 27(1), 25 Transverse isotropy modeling of 3-D glulam timber beams. J.
3 Luciano, R. and Barbero, E.J. Formulae for the stiffness of com- Mater. In Civil Engrg 1991, ASCE, 3(2), 125
posites with periodic microstructure. Int. J. Solids Struct. 1994, 21 Nagaraj, V. and GangaRao, H.V.S. Static and fatigue response
31(21), 2933 of pultruded FRP beams without and with splice connections,
4 Jones, R.M. 'Mechanics of composite materials', Hemisphere CFC Report, 94-183, Constructed Facilities Center, West Virgi~
Publishing Corporation, New York, NY, 1975 nia University, Morgantown, WV, 1994
5 Barbero, E.J., Lopez-Anido, R. and Davalos, J.F. On the 22 Qiao, P., Davalos, J.F. and Barbero, E.J. FRPBEAM: A com-
mechanics of thin-walled laminated composite beams. J. Com- puter program for analysis and design of FRP beams, CFC
pos. Mater. 1993, 27(8), 806-983 Report, 94-191, Constructed Facilities Center, West Virginia
6 Chamis, C.C. Simplified composites micromechanics equations University, Morgantown, WV, 1994

You might also like