51st AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference<BR>18th AIAA 2010-3021

12 - 15 April 2010, Orlando, Florida

A Beam Formulation with Shell Capabilities

Erasmo Carrera1
Department of Aeronautic and Space Engineering,
Politecnico di Torino, Torino, 10129, Italy

Marco Petrolo2
Department of Aeronautic and Space Engineering, Politecnico di Torino,
10129, Italy, and Institut Jean Le Rond d'Alembert,
UMR7190 CNRS, Paris06, 75252 Paris, France

Rened beam theories suitable for aerospace structures are presented in this paper.
These theories have shell capabilities and permit the analysis of thin walled structures,
such as wings and helicopter blades. The formulation is given in the framework of the
Carrera Unied Formulation, CUF, which considers the order of the theory, N , as a
free parameter of the analysis. N is the order of the 1D displacement expansion. The
displacement components are, in fact, expanded in terms of the cross-section coordi-
nates, (x, z ), by using a set of 1-D generalized displacement variables. The rened
kinematic models are based on Taylor-type polynomials. The nite element formula-
tion is exploited in order to be able to face arbitrary cross-section geometries. FE's
matrices are obtained in terms of a few fundamental nuclei which are formally indepen-
dent of both N and the number of element nodes. A cubic (4 nodes) approximation
along the beam axis, (y ), is used. Structural analyses are conducted starting from
classical beam theories, rened models are then introduced to evaluate non-classical
eects. Dierent cross-section geometries, loading cases and boundary conditions are
considered. It is mainly concluded that rened models are mandatory to properly
detect shell-like mechanical behaviors. CUF hierarchical capabilities oer a powerful
systematic procedure to implement higher-order beam theories with no constraints on
N.

I. Introduction
The beam theory represents a reference technique to face the structural analysis of slender bodies
such as airplane wings, helicopter blades and bridges. In all these cases, a beam model oers signif-
icant advantages with respect to more cumbersome two-dimensional, 2-D, and three-dimensional,
3-D, elements, because it permits the proper evaluation of the mechanical behavior of a structure
via a 1-D analysis. Higher the order to which a beam theory is based on, higher its relevance. The
rening process of a model is needed to detect non-classical eects such as shell-like responses.
Two main methods are recognizable in the developing of a beam theory: axiomatic and asymptotic.
The former method establishes a kinematic model starting from the intuition of a scientist. The
latter investigates the role played by the various variables of the model in terms of a perturbation
parameter (usually a geometrical one such as the span-to-height ratio for beams). The 3-D problem
is then reduced to a 1-D model by exploiting an asymptotic series of a characteristic parameter, and
retaining those terms which exhibit the same order of magnitude when the perturbation parameter
vanishes.

1 Professor of Space Structures and Computational Aeroelasticity, Department of Aeronautic and Space Engineering,
Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Torino, Italy, erasmo.carrera@polito.it.
2 PhD Student, Department of Aeronautic and Space Engineering, Politecnico di Torino, Corso Duca degli Abruzzi,
24, 10129 Torino, Italy, marco.petrolo@polito.it., and Institut Jean Le Rond d'Alembert, UMR7190 CNRS, Paris06,
Case 162, Tour 55-65, 4, Place Jussieu, 75252 Paris, France.

Copyright © 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
The Euler-Bernoulli and Timoshenko beam models are classical examples of axiomatic approaches.
The former does not account for transverse shear eects on cross-section deformations. The lat-
ter provides a model that foresees a constant shear deformation distribution on the cross-section.
Both of them yield better results for slender than for short beams. Important contributions on the
development of rened theories by using an axiomatic approach have been furnished by Kapania
[1, 2] and Eisenberger [3]. Examples of asymptotic based rened theories can be found in works by
Volovoi et alii [4] and Yu et alii [5, 6].
Higher-order beam theories are herein developed in the framework of the Carrera Unied Formu-
lation, CUF, which was initially developed for plates and shells [7, 8], and it has recently been
extended to beams [913]. CUF oers a systematic procedure to obtain rened structural models
by considering the order of the theory, N , as a free parameter of the formulation. Cubic beam
elements (4 nodes) as well as dierent higher-order models for the cross-section displacement eld
are used. The Euler-Bernoulli and Timoshenko beam models are obtained as particular cases of the
linear formulation. The results are compared with benchmarks retrieved from the classical models
and with shell and solid models.

II. Preliminaries
The adopted coordinate frame is presented in Fig. 1. The beam boundaries over y are 0 ≤ y ≤
L. The displacements vector is:
© ªT
u(x, y, z) = ux uy uz (1)

Superscript "T " represents the transposition operator. The stress, σ , and the strain, ², are grouped
as follows:
© ªT © ªT
σ p = σzz σxx σzx , ²p = ²zz ²xx ²zx
© ªT © ªT (2)
σ n = σzy σxy σyy , ²n = ²zy ²xy ²yy

Subscript "n" stands for terms laying on the cross-section, while "p" stands for terms laying on
planes orthogonal to Ω. Linear strain-displacement relations are used:

²p = D p u
(3)
²n = D n u = (D nΩ + D nz )u

with:
     
∂
0 0 ∂
0 0 0 0 ∂y 0
∂z  ∂ 
   
Dp =  ∂
∂x 0 0  , D nΩ =  0 ∂
∂x 0  , D ny = 
 ∂y 0 0 
 (4)
∂ ∂ ∂ ∂
∂z 0 ∂x 0 ∂z 0 0 0 ∂y

The Hooke law is exploited:

σ = C² (5)

According to Eq.s 2, the previous equation becomes:

σ p = C̃ pp ²p + C̃ pn ²n
(6)
σ n = C̃ np ²p + C̃ nn ²n

In the case of isotropic material the matrices C̃ pp , C̃ nn , C̃ pn and C̃ np are:

     
C̃11 C̃12 0 C̃55 0 0 0 0 C̃13
T
C̃ pp =  C̃12 C̃22 0  , C̃ nn =  0 C̃44 0  , C̃ pn = C̃ np =  0 0 C̃23  (7)
0 0 C̃66 0 0 C̃33 0 0 0

2
For the sake of brevity, the dependence of the coecients [C̃ ]ij versus Young's moduli and Poisson's
ratio is not reported here. It can be found in Tsai [14] or Reddy [15].

III. Unied Formulation.

In the framework of the Carrera Unied Formulation (CUF), the displacement eld is assumed
as an expansion in terms of generic functions, Fτ :

u = Fτ u τ , τ = 1, 2, ...., M (8)

where Fτ are functions of the coordinates x and z on the cross-section. uτ is the displacement
vector and M stands for the number of terms of the expansion. According to the Einstein notation,
the repeated subscript τ indicates summation. Eq. (8) consists of a Maclaurin expansion that used
as base the 2D polynomials xi y j , where i and j are positive integers. The maximum expansion
order, N , is supposed to be 4. Table 1 presents M and Fτ as functions of N . For example, the
second-order displacement eld is:

ux = ux1 + x ux2 + z ux3 + x2 ux4 + xz ux5 + z 2 ux6

uz = uy1 + x uy2 + z uz3 + x2 uz4 + xz uz5 + z 2 uz6 (9)
uz = uz1 + x uz2 + z uz3 + x2 uz4 + xz uz5 + z 2 uz6

The Timoshenko beam model (TBM) can be obtained by acting on the Fτ expansion. Two conditions
have to be imposed. 1) a rst-order approximation kinematic eld:

ux = ux1 + x ux2 + z ux3

uy = uy1 + x uy2 + z uy3 (10)
uz = uz1 + x uz2 + z uz3

2) the displacement components ux and uz have to be constant above the cross-section:

ux2 = uz2 = ux3 = uz3 = 0 (11)

The Euler-Bernoulli beam (EBBM) can be obtained through the penalization of ²xy and ²zy . This
condition can be imposed by using a penalty value χ in the following constitutive equations:

σxy = χC̃55 ²xy + χC̃45 ²zy

(12)
σzy = χC̃45 ²xy + χC̃44 ²zy

The classical theories and the rst-order models require the assumption of opportunely reduced
material stiness coecients to correct Poisson's locking (see Carrera and Brischetto [16, 17]). Unless
dierently specied, for classical and rst-order models Poisson's locking is corrected according to
Carrera and Giunta [9].
Introducing the shape functions, Ni , and the nodal displacement vector, qτ i :
© ªT
q= quxτ i quyτ i quzτ i (13)

The displacement vector becomes:

uτ = Ni Fτ qτ i (14)

For the sake of brevity, the shape functions are not reported here. They can be found in many books,
for instance in [18]. Elements with 4 nodes (B4) are formulated, that is, a cubic approximation
along the y axis is adopted. It has to be highlighted that, while the order of the beam model
is related to the expansion on the cross-section, the number of nodes per each element is related
to the approximation along the longitudinal axis. These two parameters are totally free and not
related to each others. An N -order beam model is therefore a theory which exploits an N -order
polynomial to describe the kinematics of the cross-section. The stiness matrix of the elements and
the external loadings, which are consistent with the model, are obtained via the Principle of Virtual

3
Displacements:
Z
δLint = (δ²Tp σ p + δ²Tn σ n )dV = δLext (15)
V

Where Lint stands for the strain energy, and Lext is the work of the external loadings. δ stands for
the virtual variation. The virtual variation of the strain energy is rewritten using Eq.s (3), (6) and
(14), in a compact format it becomes:

δLint = δ qTτi Kijτ s qsj (16)

Where Kijτ s is the stiness matrix in the form of the fundamental nucleus. Its components are:
Z Z Z Z
ijτ s
Kxx = C̃22 Fτ,x Fs,x dΩ Ni Nj dy + C̃66 Fτ,z Fs,z dΩ Ni Nj dy+
ZΩ Z l Ω l
C̃44 Fτ Fs dΩ Ni,y Nj,y dy
Ω l

Z Z Z Z
ijτ s
Kxy = C̃23 Fτ,x Fs dΩ Ni Nj,y dy + C̃44 Fτ Fs,x dΩ Ni,y Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kxz = C̃12 Fτ,x Fs,z dΩ Ni Nj dy + C̃66 Fτ,z Fs,x dΩ Ni Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kyx = C̃44 Fτ,x Fs dΩ Ni Nj,y dy + C̃23 Fτ Fs,x dΩ Ni,y Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kyy = C̃55 Fτ,z Fs,z dΩ Ni Nj dy + C̃44 Fτ,x Fs,x dΩ Ni Nj dy+
ZΩ Z l Ω l (17)
C̃33 Fτ Fs dΩ Ni,y Nj,y dy
Ω l

Z Z Z Z
ijτ s
Kyz = C̃55 Fτ,z Fs dΩ Ni Nj,y dy + C̃13 Fτ Fs,z dΩ Ni,y Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kzx = C̃12 Fτ,z Fs,x dΩ Ni Nj dy + C̃66 Fτ,x Fs,z dΩ Ni Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kzy = C̃13 Fτ,z Fs dΩ Ni Nj,y dy + C̃55 Fτ Fs,z dΩ Ni,y Nj dy
Ω l Ω l

Z Z Z Z
ijτ s
Kzz = C̃11 Fτ,z Fs,z dΩ Ni Nj dy + C̃66 Fτ,x Fs,x dΩ Ni Nj dy+
ZΩ Z l Ω l
C̃55 Fτ Fs dΩ Ni,y Nj,y dy
Ω l

The virtual variation of the work of the inertial loadings is:

Z
δLine = ρüδuT dV (18)
V

4
where ρ stands for the density of the material, and ü is the acceleration vector. Eq. 18 is rewritten
using Eq.s 3, and 14:
Z ·Z ¸
δLine = δ qτ i Ni
T
ρ(Fτ I)(Fs I)dΩ Nj q̈ sj dy (19)
l Ω

where q̈ is the nodal acceleration vector. The last equation can be rewritten in the following compact
manner:

δLine = δ qTτi Mijτ s q̈ sj (20)

where Mijτ s is the mass matrix in the form of the fundamental nucleus. Its components are:
Z Z
ijτ s ijτ s ijτ s
Mxx = Myy = Mzz = ρ Fτ Fs dΩ Ni Nj dy
Ω l (21)
ijτ s ijτ s ijτ s ijτ s ijτ s ijτ s
Mxy = Mxz = Myx = Myz = Mzx = Mzy =0

It should be noted that no assumptions on the approximation order have been done. It is therefore
possible to obtain rened beam models without changing the formal expression of the nucleus
components. This is the key-point of CUF which permits, with only nine FORTRAN statements, to
implement any-order beam theories. The shear locking is corrected through the selective integration
(see [18]). The undamped dynamic problem can be written as it follows:

M ä + Ka = p (22)

where a is the vector of the nodal unknowns and p is the loadings vector. Introducing harmonic
solutions, it is possible to compute the natural frequencies, ωi , for the homogenous case, by solving
an eigenvalues problem:

(−ωi2 M + K)ai = 0 (23)

where ai is the i-th eigenvector.

IV. Numerical results

Dierent static and dynamic assessments are considered hereafter. Previous works have dealt
with the structural investigation of airfoil-shaped beams and unconventional wing congurations.
The deformed free-tip cross-section of a cantilevered three-cell wing which undergoes a torsional
loading is shown in Fig. 2 [11]. This result has been obtained via a fourth-order model, the out-of-
plane warping is clearly detected. Fig. 3 [19] shows a natural mode of a joined-wing which has been
obtained with the CUF beam and an MSC Nastran shell model. The comparison of the natural
frequencies highlights an almost perfect agreement.
A simply supported beam with a thin-walled annular cross-section is considered as a rst assessment
of the present work. The geometric characteristics of the model are shown in Fig. 4. The outer
diameter, d, is equal to 2 [m]. The thickness is supposed to be as high as 0.02 [m]. The length-to-
diameter ratio, L/d, is equal to 10. An isotropic material is used. The Young modulus, E , is equal
to 75 [GPa]. The Poisson ratio, ν , is equal to 0.33. A vertical force, Puz , is applied in [0, L/2, 1]
and it is equal to 12.5 [MN]. The mechanical behavior of the structure is investigated through the
vertical displacement of two points. The rst one, T, coincides with the loading point, the latter,
B, is placed at [0, L/2, −1]. A reference value is obtained via the following formula:

Puz L3
uzb = (24)
48EI
Where I is the cross-section moment of inertia. uzB and uzT are reported for increasing order beam
models in Table 2. EBBM and TBM predict the same vertical displacement for both points. In
particular, the Timoshenko beam shows a smaller rigidity. Linear and parabolic models furnish

5
slightly dierent results. It has to be highlighted that the second-order model gives a smaller
displacement value than the rst-order one because the Poisson locking correction is activated only
for linear expansions. Remarkable dierences are present in the solutions obtained through cubic
and fourth-order models. uzB is considerably smaller than in case of lower-order expansions. This
eect is related to the increase of uzT which is the vertical displacement of the point where the load
is applied. uzT exploits a larger quote of the external work making the local eects more relevant
than the global ones. This mechanical response is due to the presence of the concentrated load and
is not detected at all by lower than cubic theories. This aspect is further claried by Fig. 5. Linear
and fourth-order models are compared by considering the deformed cross-sections at y = L/2. The
deformability of the cross-section is not detectable by the linear model.
A clamped-clamped beam with a thin-walled annular cross-section is considered as the second
assessment of the proposed model. The geometrical and material features are the same of the rst
case. The static analysis is conducted by applying a vertical force, Puz , equal to 5.0 [MN] at [0,
L/2, 1]. Table 3 presents the vertical displacements of the loading point for dierent beam models.
The last row is related to a shell model solution which has been made in MSC Nastran. The second
column shows the total number of degrees of freedom of each model, that is, the computational
cost. A good convergent behavior is observed with a considerable smaller computational eort
than that required by the shell model. Fig. 6 shows the deformed congurations of the beam for
dierent theories. EBBM is not able to detect the local eects of the concentrated load. Higher the
order, more detailed the result becomes. A further investigation is furnished in Fig. 7 where the
deformed loading point cross-sections are reported for dierent beam theories and compared with
that obtained by MSC Nastran. It can be observed that models higher then the fourth are able to
detect the displacement eld of all the cross-section points but the loading one. It occurs models
higher than the tenth-order to obtain a detailed description of the loading point displacement eld.
The free vibration analysis of the clamped-clamped model has been made as the last assessment
of this work. Table 4 reports the rst two bending frequencies for dierent beam models. Shell
and solid models have been used for comparison purposes. The position within the eigenvalue
vector of each frequency is shown in the superscripts. Each frequency appears twice because of the
symmetry of the cross-section. The rst bending frequency is properly detected by the third-order
model although a fourth-order model is needed to predict the exact position of the frequency in the
eigenvalue vector. A fth-order model is needed to detect the second bending frequency whereas a
sixth-order theory is able to exactly locate the position of the frequency in the eigenvalue vector.
Fig.s 8, 9 shows two natural circumferential modes of the beam which are characterized by the
presence of two and three lobes along the cross-section. Table 5 presents the natural frequencies
of the rst two-lobe mode for dierent theories. At least a third-order model is needed to detect
this mode. The proper evaluation of the value of the natural frequency is furnished by the seventh-
order model. It is signicant to observe that the shell and the seventh-order beam model are able
to compute a lower frequency than the solid one, that is, these two models are able to predict a
lower stiness of the model than that of the 3D elements. The rst three-lobe mode frequencies are
reported in Table 6. A fourth-order model is needed to detect this mode. The eighth-order model is
able to furnish the same natural frequency of the shell model. The dierences with the 3D elements
are more relevant than in the two-lobe case.

V. Conclusion
The static and free vibration analyses of thin-walled structures based on higher-order beam
theories have been presented in this paper. The Carrera Unied Formulation, CUF, has been used
for the systemic implementation of rened models. According to CUF, the element stiness and
mass matrices are obtained in compact forms, named fundamental nuclei, that do not depend on
the theory approximation order, that is, the order of the model is assumed as a free parameter of the
modelling. Elements based on classical theories have been derived as particular cases. The numeri-
cal analyses have been conducted for the investigation of the static and dynamic behaviors in terms
of deformed congurations, natural frequencies, and vibration modes. Comparisons with shell and
solid models of commercial FE codes have been made. The following main conclusions can be drawn.

1. The builded beam formulation permits to deal in an unied manner with arbitrary cross-

6
 section geometries and thin-walled structures.
2. The use of higher-order theories has permitted the overcome classical beam model limitations.
3. The comparison with shell and solid models has shown the shell capabilities of the rened
beam theories in detecting the localized eects of concentrated loads and shell-like natural
modes.
4. The computational eort requested by the present beam model is strongly smaller than those
needed by shell and solid elements.
The use of the proposed beam models appears suitable for aeroelastic applications that include airfoil
in-plane and out-of-plane deformations and for extensions to wings made of advanced composite
materials. These topics could be subject of future works.

Acknowledgments
The nancial support from the Regione Piemonte project MICROCOST is gratefully acknowl-
edged.

References
[1] Kapania, K. and Raciti, S., Recent Advances in Analysis of Laminated Beams and Plates, Part I:
Shear Eects and Buckling, AIAA JOURNAL, Vol. 27, No. 7, 1989, pp. 923935.
[2] Kapania, K. and Raciti, S., Recent Advances in Analysis of Laminated Beams and Plates, Part II:
Vibrations and Wave propagation, AIAA JOURNAL, Vol. 27, No. 7, 1989, pp. 935946.
[3] Eisenberger, M., An exact high order beam element, Computers and Structures , Vol. 81, 2003, pp. 147
152.
[4] Volovoi, V. V., Hodges, D. H., Berdichevsky, V. L., and Sutyrin, V. G., Asymptotic theory for static be-
havior of elastic anisotropic I-beams, International Journal of Solid Structures , Vol. 36, 1999, pp. 1017
1043.
[5] Yu, W. and Hodges, D. H., Elasticity Solutions Versus Asymptotic Sectional Analysis of Homogeneous,
Isotropic, Prismatic Beams, Journal of Applied Mechanics , Vol. 71, 2004, pp. 1523.
[6] Yu, W., Volovoi, V. V., Hodges, D. H., and Hong, X., Validation of the variational asymptotic beam
sectional analysis (VABS), AIAA Journal , Vol. 40, 2002, pp. 21052113.
[7] Carrera, E., Theories and nite elements for multilayered plates and shells, Archives of Computational
Methods in Engineering , Vol. 9, No. 2, 2002, pp. 87140.
[8] Carrera, E., Theories and nite elements for multilayered plates and shells: a unied compact formula-
tion with numerical assessment and benchmarking, Archives of Computational Methods in Engineering ,
Vol. 10, No. 3, 2003, pp. 216296.
[9] Carrera, E. and Giunta, G., Rened beam theories based on Carrera's Unied Formulation, Interna-
tional Journal of Applied Mechanics , In Press.
[10] Giunta, G., Carrera, E., and Belouettar, S., A Rened Beam Theory with only Displacement Variables
and Deformable Cross-Section, 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynam-
ics, and Materials Conference, Palm Springs, California 4 - 7 May 2009 , 2009.
[11] Carrera, E., Giunta, G., Nali, P., and Petrolo, M., Rened beam elements with arbitrary
cross-section geometries, Computers and Structures , Vol. 88, No. 56, 2010, pp. 283293, DOI:
10.1016/j.compstruc.2009.11.002.
[12] Carrera, E., Petrolo, M., and Nali, P., Unied formulation applied to free vibrations nite element anal-
ysis of beams with arbitrary section, Shock and Vibrations , Vol. 17, 2010, pp. 118, DOI: 10.3233/SAV-
2010-0528.
[13] Carrera, E. and Petrolo, M., On the Eectiveness of Higher-Order Terms in Rened Beam Theories,
Journal of Applied Mechanics , In Press.
[14] Tsai, S. W., Composites Design , Dayton, Think Composites, 4th ed., 1988.
[15] Reddy, J. N., Mechanics of laminated composite plates and shells. Theory and Analysis , CRC Press,
2nd ed., 2004.
[16] Carrera, E. and Brischetto, S., Analysis of thickness locking in classical, rened and mixed multilayered
plate theories, Composite Structures , Vol. 82, No. 4, 2008, pp. 549562.
[17] Carrera, E. and Brischetto, S., Analysis of thickness locking in classical, rened and mixed theories for
layered shells, Composite Structures , Vol. 85, No. 1, 2008, pp. 8390.
[18] Bathe, K., Finite element procedure , Prentice hall, 1996.
[19] Carrera, E., Petrolo, M., and Varello, A., Advanced Beam Formulations for Free Vibration Analysis
of Conventional and Joined Wings, Submitted.

7
 Tables

Table 1 Mac Laurin's polynomials.

N M Fτ
0 1 F1 = 1
1 3 F2 = x F 3 = z
2 6 F4 = x2 F5 = xz F6 = z 2
3 10 F7 = x3 F8 = x2 z F9 = xz 2 F10 = z 3
... ... ...
(N +1)(N +2)
N 2 F (N 2 +N +2) = xN F (N 2 +N +4) = xN −1 z . . . F N (N +3) = xz N −1 F (N +1)(N +2) = z N
2 2 2 2

Table 2 Vertical displacements of dierent cross-section points computed via increasing order
beam elements, simply-supported beam.
uzT [m] uzB [m]
Ref. Sol. Eq. 24 −0.456 −0.456
EBBM −0.456 −0.456
TBM −0.474 −0.474
N=1 −0.475 −0.472
N=2 −0.474 −0.467
N=3 −0.644 −0.339
N=4 −0.928 −0.159

Table 3 uz displacements at the loading point for dierent beam theories and comparison with
a shell model, clamped-clamped beam.
Theory DOF's uz [m]
EBBM 155 −0.046
TBM 155 −0.053
N=1 279 −0.053
N=2 558 −0.052
N=3 930 −0.114
N=4 1395 −0.229
N=5 1953 −0.335
N=6 2604 −0.386
N=7 3348 −0.486
N=8 4185 −0.535
N=9 5115 −0.564
N = 10 6138 −0.584
N = 11 7254 −0.597
Shell 49500 −0.670

8
Table 4 First two bending mode frequencies of the hollow cylinder and their position in the
eigenvalue vector for dierent beam theories and comparison with shell and solid models,
clamped-clamped beam.
Theory DOF's f1 [Hz] f2 [Hz]
1,2∗
EBBM 155 32.598 88.0723,4
TBM 155 30.3041,2 76.4473,4
N=1 279 30.3041,2 76.4473,4
N=2 558 30.7301,2 77.3383,4
N=3 930 28.7541,2 69.4485,6
N = 4 1395 28.7473,4 69.4029,10
N = 5 1953 28.7453,4 69.39713,14
N = 6 2604 28.7453,4 69.39717,18
Shell 49500 28.4893,4 68.94017,18
Solid 174000 28.3693,4 68.68717,18
(∗): positions of the frequencies in the
eigenvalue vector.

Table 5 First two-lobe mode frequencies of the hollow cylinder and their position in the
eigenvalue vector for dierent beam theories and comparison with shell and solid models,
clamped-clamped beam.
Theory DOF's f [Hz]
EBBM 155 −
TBM 155 −
N=1 279 −
N=2 558 −
∗
N=3 930 38.7553,4
N = 4 1395 25.1561,2
N = 5 1953 20.5011,2
N = 6 2604 20.4501,2
N = 7 3348 17.3631,2
Shell 49500 17.4061,2
Solid 174000 18.9321,2
(∗): positions of the frequencies
in the eigenvalue vector.

9
Table 6 First three-lobe mode frequencies of the hollow cylinder and their position in the
eigenvalue vector for dierent beam theories and comparison with shell and solid models,
clamped-clamped beam.
Theory DOF's f [Hz]
EBBM 155 −
TBM 155 −
N=1 279 −
N=2 558 −
N=3 930 −
N = 4 1395 75.69011,12
N = 5 1953 65.1869,10
N = 6 2604 52.3869,10
N = 7 3348 50.3727,8
N = 8 4185 40.1027,8
Shell 49500 40.4277,8
Solid 174000 46.4447,8
(∗): positions of the frequencies
in the eigenvalue vector.

10
 Figures

z
y

Fig. 1 Coordinate frame of the beam model.

Fig. 2 Torsion analysis of a wing model via a fourth-order beam model.

11
Fig. 3 Natural mode of a joined-wing via CUF, f = 47.512 [Hz], and MSC Nastran Shell,
f = 47.118 [Hz].
z

y x
O

Fig. 4 Reference coordinate frame for the annular cross-section.

(a) N = 1 (b) N = 4

Fig. 5 Deformed loading point cross-sections in case of linear and fourth-order models, simply-
supported beam.

12
 (a) EBBM (b) N = 4

(c) N = 8 (d) N = 11

Fig. 6 Deformed beams in case of various theories, clamped-clamped beam.

Undeformed N=9 N=5 N=1

Shell N=8 N=4 TBM
N = 11 N=7 N=3 EBBM
N = 10 N=6 N=2

Fig. 7 Deformed congurations of the loading point cross-section for dierent beam theories
and comparison with a shell model, clamped-clamped beam.

13
 (a) Two lobes. (b) Three lobes.

Fig. 8 Two- and Three-lobe natural modes, clamped-clamped beam.

(a) Two lobes. (b) Three lobes.

Fig. 9 Two- and Three-lobe natural modes, 2-D sketches.

14