Nonclassical Behavior of Thin-Walled Composite Beams

with Closed Cross Sections

Lawrence W . Rehfield Ali I<. Atilgan Dewey l I . Hodgcs

Professor, Aer-o. Science arid NATO Scllolor Professor
Engineering Post Doctoral Fellow
Univ. of Calif., Davis

School of Aerospace Engirleerirlg

Georgia Institute of Technology,
Atlanta, Go.

This paper focuses on two nonclassical effects in the behavior of thin-walled composite heams: elastic hending-
shear coupling and restrained torsional warping. These nonclassical effects are clarified and analyzed in some
simple examples involving cantilevered beams. First, elastic bending-transverseshear coupling is shown to he
important in the analysis of beams designed for extension-twist coupling. It is found that the lateral dellections
ran be off by more than a factor of two if this coupling is ignored. This eaupling stems from plies with off.axis
fihers in the beam. The presence of these plies affects significantly the modeling approach (i.c., determination
of the constitutive equations) in that transverse shcar must appear in the kinematics so that its coupling with
bending will he exhihitcd in the elastic constants. This linding is in accord with "exact" beam theories which
develop the beam displacement and crass sectional orientation in terms of six kinematical variables instead of
the three or four found in some previously published works on composite blade modeling. A second nonclassical
effect, torsional warping rigidity, is shown to be important far certain box beams having a thin-walled, closed
crass section. The importance of including these nonclassical phenomena in a complete theory is discussed in
l i ~ h of
t the magnitude of their effects for various values of configuration parameters.

Introduction walled closed cross sections and their physical behavior. Wc

intend furthermore to determine the cxtent to which bcnding-
A erospace vehicle structures are largely composed of thin-
walled elements stiffened by beam-like members and are
increasingly being madc of composite materials. There are ccr-
shear coupling and torsional warping rigidity arc essential ele-
ments of such an understanding.
tain modeling assumptions that are typically associated with The subjcct of compositc rotor blade modeling was reviewed
so-called classical analyses of isotropic beams which will not in Ref. I . There exist quite general approaches to the dcter-
suffice for beams made of compositc matcrials. The usual clas- mination of sectional constants ranging from powerful finite
sical analyses must be reviscd to include certain nonclassical element methods such as Rcf. 2 to simple analytical methods
effects. Two of these nonclassical effects, bending-shcar cou- such as Rcf. 3. Referencc 2 shows that thcre are two classcs
pling and torsional warping rigidity, are the subjects of thc of warping involved in the calculation of sectional elastic con-
present paper. Here "shear" refers to transverse shcar in the stants. .The particular solution (also called the St. Venant so-
sense of Timoshenko theory. Whilc shcaring strains and tor- lution) ignores all end effects that arise from rcstraining the
. - - .
sional wamine rieiditv are treated in some classical analvses. . .
the influence of such effects is usually small for isotropic beams.
warping. This solution allows the determination of a 6 X 6
matrix of elastic constants for the blade cross section. Thus,
lExceotions to this include isotrooic beams with oocn cross shear deformation must be included in the blade deformation
1 sections, which arc not considered herein, for which warping model in order for these constants to contributc to the strain
rigidity is known to be important.) Forcomposites, on the other energy.
hand, these effects may not be small. The analysis which fol- The homogeneous (or boundary layer) solutions, however,
lows is intended as a contribution towards undcrstdnding the- allow the end effects to be treated to a varying degree of ac-
oretical foundations for analysis of compositc beams with thin- curacy depending on how many of thc restrained warping
"modes" are retained. Each of these modes has a characteristic
- length which determines how rapidly its effects decay with thc
Presented at the American Helicopter Society National Technical Spe- distance from the ends. In order to make usc of these solutions
cialists' Meeting an Advanced Rotorcran Struclures, Oct 25-27. 1988. in the determination of sectional elastic constants, additional
Wiiliamsburg, Va. kinematical variables, which serve as amplitudes for thcir modes,
42
MAY 1990 COMPOSITE BEAMS 43
,
must be incorporated into the deformation model. These ad-
-
ditional variables can be the derivatives of existine ones. Sub-
sequent work in Ref. 4 shows that, among the out-of-plane
restrained modes, the torsional warping mode is the most sig-
nificant.
For an arbitrary composite beam the in-plane and out-of-
plane St. Venant solutions can be quite significant. When we
restrict the discussion to thin-walled beams, however, the St.
Venant warping solutions do not significantly affect the stiff-
ness constants (Ref. 5). Thus, a useful contribution to the
understanding of composite blade modeling would be to cx-
amine a simple thin-walled blade theow includine at least the
full 6 X 6 matrix of elastic constants wide also examining the
effects of additional constants associated with the out-of-olane
torsional warping.
The simplest theory required to examine the importance of
bending-shear coupling and restrained out-of-planc torsional
warping is that of Ref. 3, a linear composite beam theory,
which scrves as the starting point for this study. Results from
this theory were shown to agree well with NASTRAN finite
element results for the static deformation of a model rotor blade Fig. 1 Schematic of thin-walled beam configuration.
(Ref. 6). Very good correlation between the theory and cx-
oeriments was also obtained for both box beams (Ref. 7) and In order to obtain an expression for the axial deflection a,
circular tubes (Ref. 8). In Ref. 7 Rehfield's theory if, 3)
was able to oredict strain distrtbution in the beam cross section.
-
some assumotions must be made concernine the tranverse shear
strains. As in the usual theory of torsion for thin-walled bcams
Also, the correlation between Rchfield's theory and the cx- made of isotrouic materials. thc shear strain is assumed to be
oeriment in Ref. 8 was verv eood. independent of s. Therefore, let y,, = y,.(x) and y,, = yrl(x)
In this paper, we proceed-by first summarizing the basic be the transverse shear strains of any cross section. They are
equations of Rehfield's theory, in which distortion in the plane assumed to be uniform for each cross section so that there is
of the cross section, local shell bending and twisting moments, no warp due to transverse shear; that is, a pure transverse shear
the hoop stress resultant, and initial twist and curvature arc not strain results in a plane cross scction. Furthermore, let y =
considered. The significance of the nonclassical effects is eval- y(x) be the shear strain due to twisting. Thcrcfore, from the
uated by means of simple examdes involving cantilevered beams. strain transformation law and elementary geometrical consid-
The importance of bending-shear coupling k assessed for beams erations, the membrane shear strain in the beam wall is given
designed for extension-twist coupling. Finally, the importance by
of restrained torsional warping in composite beams is asscssed
for a family of thin-walled box beams. The differences relativc
to isotropic cases are highlighted.

Synopsis of the General Theory Introducing the position vector r from the reference axis of the
The starting point of our considerations is the linear theory beam to an arbitrary point in the wall of thc beam and a unit
for thin-walled, composite beams developed in Ref. 3. After vector n normal to the wall at the arbitrary point and directed
the kinematics of thc theow are summarized. we will then toward the interior of the cross section, one can express the
outline development of the equilibrium equations and the force- shear strain in terms of the deformation as
deformation relationships from the principle of virtual work.

Kinematics
A thin-walled beam with closed, single-cell cross section is where u, is the tangential component of displacement given by
shown in Fig. I . Thc coordinate direction x is along a straight,
but as vet unsoecified, rcference axis while v and z are the
transveise cootdinales of the cross scction measured from thc
reference axis. The circumferential coordinates is taken alone
the middle surface of the wall. The beam undergoes stretching
Following Ref. 3, one can find the form of the axial displace-
bending, twisting, and transverse shearing. Introducing a frame
which c~n~nc~dcvwith the c r o s s s e c t ~ o n ~undcli~rn;cd
~i~~e ~ C ~ I I I ,ment component, u, by ignoring any effects of tapcr along the
one can dcc(~n~p~lsc the displnccmcnt ftclJ 111' thc bean1 into a spanwise direction and by enforcing the continuity condition
around thc circumference of the cross section. The result is
ngi(l-hudy tran\l~tionan11r6t;ltion of thc frarnc, and a warping
of the crocs suction relat~vcto th;~ttranslntcd and rotated (i;imc.
Considering only small displacements and rotations and ig-
noring distortion of the cross section in its own plane, one can
immediately represent the transverse displacement components where U = U(x) is the axial component of the displacement
in the form of the point where the reference axis passes through a given
cross section, p, = P,,(x) and P, = P.(x) are the cross section
rotations, positive in a right-handed sense about the axes y and
z, respectively,*

*In other words, U. V, and Ware components of the rigid-body dis-

Here V = V(x) and W = W(x) arc transverse components of placement of the point in lhc cross sectional frame where lhc rcference
the displacement at the point where the reference axis passes axis passer through it; +, P,, and PI are components of ihc rigid-body
+
through a given cross section, and = +(x) is the twist angle. rotation of the cross sectional frame.
44 L.W. REHFIELD JOURNAL OF THE AMERICAN HELLCOFTER SOCIETY
*
a = rrz- w.. where N is the axial force, Q, and QI are the shear forces, M,
is the torsional moment, M, and M, are the bending moments,
and Q,,.is the generalized warping related force (or bimoment).
Here q,, q,, and y, are applied, distributed forces, rn,, !nY,and
and Jl is the torsional warping function given by tn, are applied, distributed moments, and q,,, is an applied hi-
moment. The generalized internal forces and the resulting cqui-
librium equations are identical to those in Ref. 3.
Composite thin-walled construction, herein, is characterized
by the membrane stiffness matrix K which relates the non-zero
stress resultants to the mcmbranc strains. The constitutive re-
which satisfies the condition that $,.Jlds = 0. Here A is the lations are (see also Ref. I)
enclosed area of the cross section, c = $ l . d ~is the circumfer-
ence, and r represents the cross section shape. The axial strain
is then obtained as

The stiffness K l l corresponds to uniaxial extension, K2, cor-

and the shear strain is responds to shear, and K12 is a coupling modulus. They are
related to the usual laminate stiffness matrix A (Ref. 9) as
follows:

A?2
where U,,is the axial strain, and y,, and y , are the shear strains
K , , = A,, - -
A22
at the beam reference axis. while 6. ... .
.. ,0........ and 0.,,... are the
twlst and bcndingcurvaturcs, rcspcct~vcly.6 ,,is thc additional
kinematical varixhle associated with torsion~lwarnine. With
both shear deformation and torsional warping pre'seG in the
theory it is possible to examine the roles of these nonclassical
effects. K2, = A,, - A:
-
,
A22

For N plies, the laminate stiffnesses-are determined by simply

Farce-DeformationAnalysis by Principle of Virtual Work adding the plane stress stiffncsses, Q,, for each ply. Thus,
For thin-walled beams, local shell bending and twisting mo-
ment resultants can be ignored, and thus, the beam reacts ex-
ternal forces by membrane action in the wall. Introducing axial
and shear stress resultants, N,, and N,,, respectively, and as-
suming that there is no internal pressure so that the hoop stress where h, is the thickness of the kth ply. The ply stiffnesses
resultant, N,, can be ignored, one can write the principle of depend upon the material and fiber orientation.
virtual work as The deformational variables or generalized strains are easily
identified from the strain expressions. Arrayed in a column
matrix u they are

where SW is the virtual work of the external forces. Application

of the calculus of variations with the usual assumptions re- Similarly the generalized internal forces can be put in a column
garding continuity results in the following equations of equi- matrix form as
librium:

N,." + 4.. = 0
l'hc relationship hctwce~lthc bcam and its reicrcncc axis (the
Q,., + Y, = 0 ~(x~rdlnatc dimction x) has not yct bccn spzcillcd: howcvrr. it
is convenient to choose it in such a way ihat
Q:., + 4, = 0

Mr..- Q,,..,, + m.. - Y,,.,.. = 0

4,. ~,,yd=
s 0

M,,, - Q , + m , = O

M,,, + Q,. + m, = 0 (11) This choice defines the reference axis as the tension axis found
in Ref. 3. This is the axis for which the application of aresultant
where the generalized internal forces are defined as tensile force will not produce any bending. It is also possible
to define the y and z axes as principal flexural axes which
uncouple bending about these orthogonal axes in cross section.
The necessary condition for this is that
 .
Since the force and the deformation are linearly relatcd, a
COMPOSITE BEAMS

circular cross section and a choice of material and fiber ori-

45

symmetric 7 x 7 stiffness matrix, C, can then be defined such entation so that the extension-twist coupling C,, is non-zero.
that A simple case is that of a slender cantilevered beam with a
-
circular cross section as shown in Fia. 2. For examole. consider
\uch a hcan~wit11 diamctcr of 2 in and wllh a circ;~nfcrentlall~

By virtue of the procedure and choice of axes defined above,

. . made of lMhlK6376 Crsohlte/
uniform stiffnes\ (CUS). la,ur,
Epoxy. The material properties used in determining the ilastic
the elements of C consist of 25 independent stiffness constants constants are E,, = 23.1 x lo6 psi, E2, = 1.4 x lo6 psi,
v12 = 0.338, andG12= 0.73 X 106psi. From theseproperties

{,,K22 (2) d . ~
2 and Eq. (21), the 6 X 6 matrix of elastic constants can be
CII = jr KI ds;
I C22 = (lclcrm~ncd:the rcsults are presented i n Tahle I. A hlade with
these pn>pcnies 1s ondcr dcvclop~~~cnt - . Kc-
at NASA I.anrlcv
search Center (Ref. 6).
CI2 = jr K12dyds ds; C2, =
r
dy dr
KZ2- - ds
ds ds
As can be seen in Table I, for this type of design there are
other nonzero coupling terms (i.e., off-diagonal terms) in ad-
dition to C,,, which are C2, and C3,. These terms couple the
dz displacements in the two orthogonal directions by coupling the
transverse shear strain along each axis with the bending strain
about that axis. The extension, twist, and warping terms are
2A
C14= -
C
jr K12ds; C, = -

CZS= 4,. ds
K12
dY
zds; CC = - j,.K12 ddsY yds IM6 / R6376 and T300 / 5208
Graphite / Epoxy
[elT; t = 0.0055" (single ply)
C34 = F ir ;i; K22
dz dz
ds; Cs = r K12-
ds
zds [0,8,-8,90]~;t = 0.022" (balanced layup)

c
36 = - 4,. dz
K12 ZY~S.' c# = (F)' jr K22d~
[e,e-90,e,(e-9o)~,ei~; t = 0.033" (CUS)

2A
C,, = -
C
jr K12zds; C,6 = -
2A
7 jr K12yds

Cii = jr Kl1JI2ds; Cs6 = - j,.Kllyzds = 0

Fig. 2 Schematic of circular tube cross section,

Now that we have the stiffness matrix, it is possible to ex-
amine special cases that illustrate some nonclassical effects. In
Table 1 Stinnesses for a composite circular crass section
order to apply forces and calculate beam deformations, how- Material IM6lR6376 GraphitelEpoxy [20, -70, 20,(-70),, 201,
ever, it is necessary to invert Eq. (20) to obtain the flexibility Plv thickness 0.0055 in.: D = 2 in.
relationship
Stiffmesses Calculated Values
...
C... Ib
C,4, Ib-in
0.1972
0.6680
X 10'
x 106
where S = C-I. This inversion is only carried out for certain Czz, lh 0.2317 x lo6
simplified cases below. C,., Ib-in -0.3340 x lo6
c,,, lb 0.2317 X 106
Shear Deformation with Bending-Shear Coupling C,,, Ib-in -0.3340 x lo6
C,, Ih-in' 0.4634 X 106
The first nonclassical effect cxamined is that of shear de- C.., Ib-in2 0.9862 x 10'
formation and its coupling with bending. To illustrate this cou-
pling, the terms in thc stiffness matrix arc cvaluated for a
46 L.W. REHFIELD JOURNAL OF THE AMERICAN HELICORER SOCIETY
*
decoupled from these effects; thus, one can consider just the Therefore, from Eqs. (24) the curvature about the y-axis can
remaining 4 x 4 matrix of stiffness constants so that be written easily in terms of beam flexibility terms and applied
loads as

1
P,,,x = S 5 ~ M S= - S55 -2 q:(L -x ) ~ (28)

Integration of Eq. 28 and application of the boundary condition

Clearly, now, if one does not consider transverse shear defor- p, = 0 at the root yields the section rotation about the y-axis
mation in the model development, as in Refs. 10 and 11 for
example, there is no possibility of correctly accounting for the
bending-shear coupling terms in Eq. (23). These terms will
soften the model, and the question naturally arises whether
these coupling effects can he important.
To see the effect more explicitly, consider the inverse of Eq. Now, from the first of Eq. (6)
(23)

In light of Eqs. (27) and (241, y.,, can be written in terms of

beam flexibility terms and applied loads as

where
Substitution of Eq. (31) into Eq. (30) results in

sss q:
w,.<= S,iqf(L -x) - -
6
[(L - x ) ) -L3] (32)

Finally, with the boundary condition that W vanishes at the

root, integration of Eq. (32) gives an expression for W

Clearly, if one ignores the coupling effect, the transverse shear

and bending flexihility coefficients are simply the reciprocals where 5 = and where S33and S5, are the correct (including
of the transverse shear and bending stiffnesses, respectively, coupling) shear and bending flexibilities, respectively. It can
i.e., be recognized that the flexibility terms correspond to the en-
gineering flexihility constants found in Ref. 6; s,, corresponds
to the flapwise bending flexibility, and S3, corresponds to the
transverse shear flexihility (due to Timoshenko).
To examine a simpler expression, consider only the tip de-
(coupling ignored) flection

The fact that the correct flexibility coefficients are larger than The second term in parenthesis corresponds to the direct trans-
the ones in which coupling is ignored is now plain. verse shear flexihility effect. This term has relative importance
In order to see the magnitude of the effect we can calculate only when the ratio % becomes significant compared to
SssL
the deflection of a beam under uniform distributed load in the unity; for a beam of given cross sectional geometry and ma-
z direction so that q, = qf where q: is a constant. From the terial, this ratio becomes larger as the beam becomes shorter.
equilibrium equations, Eqs. (I I), and the zero shear force and It mav or mav not be imoortant for a oarticular value of slen-
bending moment boundary conditions at the tip, the shear force derness, depending on the ratio of extension and shear moduli.
and bending moment become
.
..
Holvever. rhe irnoortarlce o f the elastic couoli~m-derertnir1i11~
the correct values of S,] and S5s-l~as notl~ingto do with sle11-
derrress of the beam! Rather, it depends on the magnitude of
the coupling C:, relative to C2,C,,. In order to assess this
effect, clearly one must determine a complete set of elastic
constants (C,, i, j = 1 , 2. . . . 6 at least). The approach of
Refs. 10 and I I will not suffice when the beam is designed
for extension-twist coupling.
For a beam whose elastic constants are given in Table 1,
Fig. 3 shows the tip deflection determined with two approxi-
MAY 1990 COMPOSITE BEAMS 47

0 = 23 deg. Any model which ignores that amount of bending-

C
0 shear coupling will be off by at least a factor of 2 in predicting

-g8 W(no shear flexibility)/W(correct) the deflections. As 0 increases beyond 23 deg the amount of
coupling decreases. The CUS construction gives a symmetric
distribution about 0 = 45 deg at which no coupling exists.
The maxima are reached around 23 deg and 67 deg. After its
maxima, the amount of coupling decays more rapidly than in
the single ply case.
Torsional Warping Rigidity
We now turn to another nonclassical effect, the influence of
torsional warping rigidity. In order to proceed, we first need
to calculate the solution of the coupled warping-torsion-exten-
sion equation. Then, the effect of the warping stiffness on the
behavior of a cantilevered box beam will be examined.
Fig. 3 Efict of beam slenderness on the relative importance of bend- Determination of Twist Distribution
ing-shear coupling and transverse shear flexibility for tip deflection. Consider a beam subjected to a discrete twisting moment,
M:, at the free end with no axial force, implying that N = 0.
Set m, = q, = q , , = 0. Taking the twisting moment equilib-
mations: (1) without transverse shear flexibility and (2) without riumequation, the fourth of Eqs. (1 I), and writing themoments
bending-shear coupling. Both are normalized by the correct tip in terms of kinematical quantities by using the stiffness matrix,
deflection from Eq. (34). Neglecting only the direct transverse one obtains
shear flexibility is seen to be inconsequential for slender beams
since, as the length to diameter ratio increases, the normalized
displacement tends toward unity. On the other hand, if only M.: = C T ~-, c77+,.ztr
~ (35)
the bending-shear coupling is neglected, we see about a 50
percent reduction in the displacement which is independent of Here the effective torsional stiffness (for zero axial force) is
the slenderness! given by
Also shown in Fig. 3 is the lateral displacement Vnormalized
by W. Unless the beam is extremely slender, the presence of
bending-shear coupling is seen to induce non-neglible lateral
displacements. In light of the importance of the lead-lag de-
flection and flap-lag elastic coupling in rotor blade stability
problems (e.g., see Ref. 12). this would appear to be another where p = A%-
KI Kzz'
I
reason to include bending-shear coupling. The boundary conditions arise naturally from the principle
Fig. 4 shows the magnitude of coupling C$5normalized with of virtual work. Atx = 0 the rotation and warping displacement
C2,C5,. (For circular cross sections with constant stiffness around are restrained so that = + +,, = 0, and at x = L the warping
the cross section, this becomes
CnCx
= A = A%- =
CnCe K,,Kn
is free so that +,,
= 0. The classical solution is the particular
solution of Eq. (35) or
p.) Here we use the material T30015208 GraphiteIEpoxy with
the following properties: E , , = 21.3 X lo6 psi, E,, = 1.6
x lo6 psi, v,, = 0.28, and G , , = 0.9 x lo6 psi. It is seen
that the balanced construction does not exhibit any coupling.
On the other hand, a single ply gives maximum coupling around
where 5 = f . With restrained warping, the general solution
has the form

where the homogeneous solution +,,,can be expressed in terms

of exponentials
+,,, = cOIe-At + ~ i 2 ) ~ "+
F COI
(39)

Here h is a decay length parameter given by

A large value of A indicates rapid decay of the solution as the

distance from the end increases. Evaluation of the constants
from the boundary conditions yields
0 30 60 90
Orientation Angle

Fig. 4 Variation of normalized bending-shear coupling parameter

with respect to ply angle.
cz,cs
48 L.W. REHFIELD JOURNAL OF THE AMERICAN HELICOFTER SOCIETY

Assuming that e k < I (which is tru; for practical situations), It should be observed that A2, can be conveniently regarded
Eq. (41) reduces to as a product of "material" and "geometric" parts as long as
stiffnesses are uniform around the cross section. Thus, Eq. (47)
can be written as

It can easily be seen that the tip rotation is

The geometric part, A,, i s the same for both orthotropic and
isotropic beams. However, the material part, A,,,, is different.
Thus, the classical tip rotation is reduced by a factor related Figure 6 shows how A,,, changes with fiber orientation. The
to the decay length. If A >> 1, then the effect is insignificant; material used in Fig. 6 is T30015208 GraphitelEpoxy. In Fig.
but if A is, for instance, less than 25, the tip rotation can be 7 the variation of A, is shown with respect to slenderness pa-
significantly reduced. rameter and the breadth of the cross section. For a given box
beam, the boundary layer parameter can be found by multi-
Influence of Warping Stiffness for Box Beams plication of the numbers coming from Figs. 6 and 7.
Now that the solution is known in terms of A, we shall Because A is relatively large for slender, thin-walled box
determine the value of A for the cross section under consid- beams made of isotropic materials, it is well known that the
eration. For the sake of simplicity, suppose that material prop- effects of warping are not very important in such beams. For
erties do not change over the cross section. To obtain the example, a thin-walled beam with the geometry depicted in
effective torsional stiffness, C,, given in Eq. (36), we use C l l , F i g . 5 , w i t hEG- = 2 . 5 , a = 0 . 2 5 , a n d u = 10,hasX=58.42.
Ct4, and Cq4whichresults in
Here, warping makes a difference of only 1.7 percent in the
tip rotation due to twist.
On the other hand, A can be much smaller for certain com-
posite beams, giving the "boundary layer" effect more sig-
nificance. Indeed, for a thin-walled box beam with the same
For the rectangular cross section (Fig. 5) being used geometry as depicted in Fig. 5 , with a = 0.25, u = 10, and
0 = 15 deg, except made of T30015208 GraphiteIEpoxy under
"normal" conditions, we obtain A = 22.35. In this case,
warping makes a difference in the tip rotation due to twist of
approximately 4.5 percent. Consider another box beam section
where 2b is the height of the cross section, 2a is the width, with a = 0.1, u = 10, and 0 = 0 deg, made of AS 3501-6
and a = bla. For the rectangular cross section the warping GraphitelEpoxy with hygrothermal effects. The material prop-
stiffness becomes erties used in the calculation are E l l = 19.3 X lo6 psi,
E2, = 0.33 X lo6 psi, v12 = 0.41, and GI2 = 0.25 X lo6
psi. Thus, A = 8.76, and warping makes a difference in the
tip rotation due to twist of approximately 11.4 percent. The
classical and nonclassical twist angle predictions and the bound-
Thus, AZ is then found for the rectangular cross section as ary layer effects for these cases can be seen in Fig. 8; here,
thenormalized twist angle is defined as $$,. The boundary layer
zone is determined as the distance where the amplitude of the
twist rate + , f is within 5 percent of the classical twist rate
(unity).
where u = -is Za
a slenderness parameter. The solution for the
twist, given in Eq. (42), is identical to that obtained in classical
theories for isotropic beams; only the value of the parameter T 300 / 5208 and AS 3501-6
h i s different. Indeed, for isotropic materials one finds that Eq. Graphite / Epoxy
47 reduces to [BIT; t = 0.0055"
G 48u2
A2 = - - (isotropic case) (48)
E (1 -

which agrees with the result obtained by Von Karman and

Christensen (Ref. 13).**

**A slightly different result was determined by Benscoter (Ref. 14).

The grealest difference between the theories of Refs. 7 and 8 occurs when
the cross section is square (a = I), which is the value of a for which the
warping displacement and stress vanish at every paint in the cross section.
A limited numerical study in Ref. 15 suggests that the differences between
these two theories are not very great. Fig. 5 Schematic of box beam cross section.
 COMPOSITE BEAMS 49

1.01 No Warping Rigidily 1

0.0 I
0 30 60 90
Orientation Angle
Normalized Axial Coordinate
Fig. 6 Variation of material part of the boundary layer parameter
A, with respect to ply angle.
Fig. 8 Variation of normalized twist angle with respect to the axial
coordinate for various values of A.

The percentage reduction in twist angle at the tip increases

this that restrained torsional warning

, -
for beams for smaller values of a. W e hereby conclude from
-
can affect certain elobal
deformation results. Extremely small values of such parameters
ent in any general-purpose analysis. It is further observed that,
without the presence of shear deformation in the kinematics,
as the slenderness LI2a. the thickness ratio a.the material ratio the proper form of the coupling terms in the flexibility matrix
KZ2IK,,,and ply angle'8 being chosen to gi"e a small h,, can cannot be obtained. The influence of this coupling is far more
result in a value for A that influences the tip rotation in a significant in the case analyzed than the direct (Timoshenko)
significant way. Thus, this effcct should be wcighcd carefully cffcct of transverse shear flexibility and is independent of the
before being excluded from composite rotor blade analyses. slenderness of the beam. This coupling also induces an elastic
"flap-lag" type coupling thc influence of which on rotor blade
stability is well known. Finally, even the direct shear flexibility
Concluding Remarks term may not be negligible for the composite case in general,
because there are materials for which the shear modulus may
Two main conclusions have been drawn in the present work: be much.smaller than the extension modulus (i.e.. . . K,, <<a<

1. In structural models designed for extension-twist coupling K3t).

(the circumferentially uniform stiffness case), an important de- 2. Torsional warping is found to be significant enough to
gree of bending-shear coupling is present which causes the warrant its inclusion in composite beam analyses in certain
structure to he significantly more flexible in bending than it circumstances. A boundary laycr parameter caused by the re-
would be if the coupling were ignored. In light of possible uses strained warping at the ends is identified. Although this pa-
of extension-twist coupling in future designs, effects such as rameter is relatively large for slender, thin-walled beams made
coupling between bending and shear deformation must be pres- of isotropic materials, it can be much smaller for composite
beams. A smaller boundary layer parameter yields longer dccay
length, along which the end effects prevail and stiffen the
structure. Thus, the smaller this parameter the larger the e m r
in the twist angle predictions. In some rather unusual cases,
the error in thc twist angle predictions at the tip can rcacb more
than 10 percent. Therefore, inclusion in the cross sectional
stiffness matrix of the torsional warping rigidity, which stems
from the inclusion of an additional variable to the kinematical
field, would be important for certain laminated structures.

Acknowledgments
This work was supported by the U.S. Army Research Office
under contracts DAAG29-82-K-0094 and DAAL03-88-C-0003.

References
'Hodgcs, D. H., "Rcview of Composite Rotor Blade Modeling." AlAA
Joan~nl,Vol. 28, (3). 1990.
'Giavotto. V., er a / . , 1983. "Anisotropic Beam Theory and Applica-
PI - tions." Comprrrers otrd Srntcrsres, Vol. 16.
0 10 20 30 40 SO 'Rehfield, L. W.. "Design Analysis Methodology far Composite Rotor
Blades." Proceedings of the 7rh DoDlNASA Corfere,lce on Fibroes Co,?,-
Slenderness Ratio ~osiresin Srrrrcrrrrol Desie,,~,(V(a)-1)-(V(a1-15). 1985.
'Bauchau. 0. A., "ABeamTheory for Anisotropic Materials," Joerrcal
Fig. 7 Variation of geometric part of the boundary layer parameter of Applied Mrcl~onics.Vol. 52. (2). 1985.
As with respect to ply angle. 'Hedges. D. H., Atilgan, A. R., Fulton. M. V., and Rehfield, L. W.,
50 L.W. REHFIELD JOURNAL OF T H E AMERICAN HELICOPTER SOCIETY
*
"Dynamic Characteristics of Composite Bcam Slructures." Proceedings 'oMansfield, E. H., and Sobey, A. I . , "TheFiberComposite Helicopter
of the Amsricon Helicopter Society Nntiorrnl Speciolisrs' M r ~ r O r gon Ro- Blade-Pan I: Stiffness Properties-Part 2 Prospecls for Acraelastie Tai-
torcrqfr Dy,rantics, Arlington, Tex., Nav 13-14, 1989. loring," Aero,rnaticnl Qaorrerly, Vol. 30, (2). 1979.
"Hodger, R. V.. Nixon, M. W., and Rehfield, L. W., "Comparison of "Hong, C.-H., and Chopra, I., "Acroelaatic Stability of a Composite
Composite Rotar Blade Models: A Coupled Beam Analysis and an MSCl Blade." Jo~#ntalof tl8c Atnerican Helicopter Sociery, Val. 30, (2), 1985.
NASTRAN Finite Element Model," NASA Technical Memorandum 89024, "Ormistan, R. A,. "Concepts forlmproving HingelessRotorStability."
1987. Proceedings of rlre A~nericartHelicopter Sociep Midem1 Region Syrnpos-
'Bauchau, O., Coffenbeny. B. S., and Rehfield, L. W., "Composite iww of Rotor T e d ~ , ~ o l o g Essington,
y, Pennsylvania, Aug 1976.
Box Bcam Analysis: Theory and Experiments." Jos,?,nl of Reinforc~d "Von Karma", T., and Christcnsen, N. B.. "Methods of Analysis for
Plosricr and Co,,~posites.Vol. 6. (I). 1987. Tonion with Variable Twist," Jorrrrlnl ofAero,?#oericnl Scierrcr, Vol. I I ,
8Nixon, M. W., "Extension-Twist Coupling of Cornpasite Circular Tubes ({I), 1944.
with Applicalion to Tilt Rotar Blade Design," Proceedirtgr of rlre 28th "Benscoter, A. U.,"A Theory ofTomion Bending far Multi-Cell Beams."
Strucrsres. Strscrr,rol Dyrmnzics, and Morerials Co,fcret~c. Apr 6-8, Jonr,zal of Applied Meclta,tics, Vol. 21, (I), 1954.
1987, Monterey, Calif., AlAA Paper No. 87-0772. "Murray. N. M., i~ztroductio,~ to I l ~ eTlrcory of Tlzbn-WalledStrrrcners,
910nes, R. M., Mechafrics of Co,ntposite Mnterials, McCraw Hill, 1975. Oxford Science Publishers, 1985.