Professional Documents
Culture Documents
a r t i c l e
i n f o
Article history:
Received 26 July 2012
Accepted 11 June 2013
Keywords:
Analytical formulation
Stiffened panels
Composite materials
Post-buckling
a b s t r a c t
The paper presents an analytical formulation for the post-buckling analysis of composite aeronautical
panels with omega stiffeners loaded in compression and shear. The formulation relies on an energy principle and the method of Ritz. In the rst step, the panel is an assembly of plate elements, and the buckling
analysis is performed. In the second step, the panel is an elastically restrained skin, and the post-buckling
behaviour is studied. The comparisons with nite element analyses and experimental results from the
literature reveal the ability of the formulation to assess the post-buckling response.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
The European consortium MAAXIMUS (More Affordable Aircraft
structure through eXtended, Integrated, & Mature nUmerical Sizing) [1], consisting of 58 industries and universities, works to demonstrate the fast development of a highly-optimized composite
airframe. The project is divided into a physical platform, focused
on the development of composite technologies for low weight aircraft, and a virtual platform aiming to reduce the time required for
the identication of the best structural solutions.
This paper presents part of the activity performed by Politecnico di Milano during MAAXIMUS in the context of the virtual platform. The work focuses on the development of a fast
computational method, which can be used to assess the post-buckling response of composite stiffened panels during the early design
stages. The fast tool aims to improve the computational efciency
of the current design procedure, which mainly relies on expensive
nite element analyses. Not only the tool can be used for a fast
optimization of composite fuselages, but also to account for more
degrees of freedom compared to today standard practice. Indeed,
the amount of design solutions to be explored can be increased,
and consequently the efciency of the nal design can be improved. For this reason, the availability of a fast method represents
a crucial aspect to move from a more costly and empirical optimization to a faster structural design optimization loop.
Among the fast design methods in the literature, analytical and
semi-analytical formulations represent an attractive strategy
thanks to their computational efciency.
Corresponding author. Tel.: +39 0223998390.
E-mail address: chiara.bisagni@polimi.it (C. Bisagni).
0045-7949/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruc.2013.06.002
39
8
9 2
Nx >
A11
>
>
>
>
>
6
>
>
>
>
N
A12
y >
>
>
> 6
>
<N >
= 6
6
A16
xy
6
6B
>
M
x >
>
>
6
>
> 6 11
>
>
>
> My >
> 4 B12
>
>
>
>
:
;
M xy
B16
A12
A16
B11
B12
A22
A26
B12
B22
A26
A66
B16
B26
B12
B16
D11
D12
B22
B26
D12
D22
B26
B66
D16
D26
9
38
B16 > x >
>
>
>
> y >
>
>
B26 7
>
>
7>
>
>
>
7>
<
=
7
c
B66
xy
7
7
>
w;xx >
D16 7>
>
>
>
>
7>
>
> w;yy >
>
D26 5>
>
>
>
:
;
2w;xy
D66
where Nx,Ny,Nxy are the membrane forces per unit length, Mx,My,Mxy
the bending and twisting moments per unit length, x,y,cxy the
membrane strains, w the out of plane displacement and the derivatives of w represent the curvatures. The comma followed by the
coordinate denotes the differentiation with respect to that coordinate. The coefcients Aik and Dik dene the in plane and out of plane
stiffness of the laminate, respectively, and the coefcients Bik determine the coupling between in plane and out-of-plane behaviour.
Each laminate composing the panel can be made of an arbitrary
number of layers under the assumptions that:
the laminate is symmetric with respect to the midplane, i.e.,
Bik = 0
the coupling between extension and shear is null, i.e.,
A16 = A26 = 0
The formulation accounts for the terms D16 and D26 as typical
aeronautical panels can exhibit coupling between bending and
twisting.
3. Linear buckling analysis
In the rst step, the buckling analysis is performed referring to a
plate representation of the panel cross section. The approach analyzes a multistiffened panel by studying a representative unit composed of two stiffeners, one bay and two half-bays. An example is
reported in Fig. 2(a), where the plate representation of an omega
stiffened panel is shown with eleven plate elements. Five elements
model the skin, including the portion of skin under the stiffeners,
and six elements model the stiffeners. A generic plate element
Fig. 2. Plate assembly representation of the stiffened panel: (a) subdivision in plate
elements, (b) generic element dimensions and reference system.
40
Np
X
Pi U p
where Np is the number of elements that compose the representative unit, and Up is a penalty term which is added to the functional
to enforce compatibility of the rotations along the edges of adjacent
plates. The potential energy of each plate element is obtained as the
sum of three contributions:
Pi U b U m V load
where Ub is the strain energy for bending, Um is the membrane energy and Vload is the potential of the external loads. The membrane
energy term is introduced only for curved elements due to the coupling between in plane and out of plane behaviour. It is neglected in
case of at plates, as the in plane and out of plane response are
uncoupled.
Referring to Kirchhoff thin plate theory and classical lamination
theory, the bending energy of the i-th element is expressed as:
Z
1
D11 w2;xx 2D12 w;xx w;yy D22 w2;yy 4D66 w2;xy
2 A
4D16 w;xx w;xy 4D26 w;yy w;xy dA
Ub
where A is the area of the plate element, w and Dik are the out of
plane displacement and the bending stiffness of the plate element.
The membrane energy is written in terms of Airy stress function
F, and is dened as:
Um
1
2
Z
a11 F 2;yy a22 F 2;xx 2a12 F ;xx F ;yy a66 F 2;xy dA
V load
1
kNx
2
Z
A
w2;x dA kN xy
w;x w;y dA
8
on x 0; a
>
<w 0
w0
on y 0; b
>
: Du 0 along the edges of adjacent elements
ij
M;N
X
mpx
npy
qmn sin
sin
a
b
m;n
where qmn are the unknown amplitudes, while M and N are the
number of shape functions along the longitudinal and transverse
direction. It can be observed that the rst two essential conditions
of Eq. (7) are identically satised, whereas the third one is imposed
with a penalty approach as described in the next paragraph.
The unknown amplitudes of the generic i-th element are
collected into the column vector qi, and the minimum potential
energy principle is applied:
@ Pi @U b U m @V load
0
@qi
@qi
@qi
@U b U m ~
Ki qi
@qi
@V load
kGi qi
@qi
10
qT1
qT2
qTNp
oT
11
The stiffness and the loading stiffness matrices of the single plate
~ i and Gi are assembled according to the denition of
elements K
~ and G.
Eq. (11) to obtain K
The last step of the procedure regards the enforcement of the
compatibility condition of the rotations along the longitudinal
edges of adjacent plate elements. The condition is imposed by
introducing penalty terms, which are torsion springs along the
common edges of adjacent plates. Their expression is:
U p;ij
1
kt
2
Z
0
Du2ij dx
12
~ K
KK
13
K kGq 0
14
where K and G are the stiffness and the loading stiffness matrices,
while q is the vector collecting the unknown amplitudes describing
the out of plane deections of the plate elements.
The eigenvalue problem of Eq. (14) is solved numerically and, to
improve the efciency of the procedure, it is decomposed into a set
of smaller subproblems accounting for the symmetry and antisymmetry of the buckling modes.
4. Post-buckling analysis
The structural model used in the second step is reported in
Fig. 3, and is representative of the skin of the stiffened panel. The
length and the width are denoted with a and b, respectively. The
transverse edges are simply supported, while the stiffeners are accounted for by introducing a torsion spring of stiffness kt along the
longitudinal edges. The stiffeners are also characterized by an axial
stiffness, which is denoted as EA. It is observed that, for a closedsection stiffener, the axial stiffness EA accounts not only for the
contribution of the stiffener itself, but also for the portion of skin
under the stiffener.
To represent the presence of the surrounding structure, whose
effect is to support the in plane motion of the panel, the longitudinal edges are free to move along the transverse direction, but are
forced to remain straight.
A Cartesian coordinate system is taken over the skin midsurface
with the x-axis directed along the longitudinal direction, and the yaxis along the transverse direction.
It is observed that the elastically restrained model, compared to
the plate assembly one, offers the advantage of limiting the computational time. Indeed, it allows a reduction of the number of
the degrees of freedom, and does not require to iterate at each step
of the incremental procedure to determine the fraction of load carried by each plate element [3,17]. A further complication in the use
of the plate assembly model would be represented by the need for
nding a different solution of the compatibility equation for each
different expression of the out of plane displacement.
4.1. Evaluation of the elastic restraint
The stiffness of the elastic restraint kt is determined by imposing that the buckling load of the elastically restrained panel is
equal to the buckling load of the plate assembly model. The approach assumes that the stiffeners restraint the rotation of the skin
edges, therefore representing an intermediate condition between
the simply-supported edges and the clamped edges [3], and carry
axial loads according to their axial stiffness EA.
The buckling eigenvalue problem is derived referring to the
minimum potential energy principle. In this case, the total potential energy is expressed as:
P U b U m V load U k
15
1
U k kt
2
Z
0
w2;y dx
y0
1
kt
2
a
0
w2;y
dx
yb
41
After applying the minimum potential energy principle, the problem is written in the form of Eq. (14), where the matrix K is now
function of the unknown stiffness kt.
The panel buckling load is obtained as the sum of two contributions. The rst one is the load acting on the skin, which is obtained
by integrating the skin buckling force per unit length along the panel width. The second term is the load acting on the stiffeners, and
can be derived assuming that both the skin and the stiffeners undergo the same displacement. The total load Ptot is:
Ptot
Z
0
Nx
dy
x0
EA
DU
a
17
16
w w0 ;xx
R
18
where w0 are the initial imperfections. They are introduced to account for small deviations from the nominal conguration. Furthermore, the presence of small imperfections facilitate the numerical
solution of the problem, avoiding problems related to the
convergence.
The fulllment of the compatibility requirement of Eq. (18) allows to impose the equilibrium of the solution referring to the
minimum potential energy principle, which is written in terms of
out of plane displacement and Airy stress function. In particular,
the total potential energy is obtained as:
P Um Ub Uk V c V s
19
where the terms Um and Ub are the membrane and the bending energy, whose expression is reported in Eqs. (4) and (5). It is worth
observing that the strain energy expression of Eq. (19) is simplied,
as the contribution due to the stiffeners axial stiffness is neglected.
42
1
V c Nx b
a11 F ;yy a12 F ;xx w2;x w;x w0;x dx
2
0
Z
V s Nxy
w;x w;y w;x w0;y w;y w0;x dA
Z
w0
m;n
r;s;t;u1 brstu
4
a11 nb 2a12
qrs qtu qrs q0tu qtu q0rs
2 2
4 i
a66 ma nb a22 ma
24
brstu rstu r 2 u2
brstu rstu r 2 u2
if r t m and s u n
if jr tj m and js uj n
if jr tj m and s u n
25
if r t m and js uj n
and:
21
~f mn b
~mn q q0
mn
mn
26
with:
mpx
npy
sin
sin
a
b
22
1
F Nx y2 Nxy xy
2
4a2 b
20
q0mn
PM;N;M;N
M;N
X
fmn
2M;2N
X
fmn cos
m;n
M;N
X
~f mn sin mpx sin npy
a
b
m;n
mpx
npy
cos
a
b
23
where the rst two terms describe a state of uniform axial and
shear force. The third term is responsible for the force redistribution
due to the large deections, and the fourth term describes the distribution of the in plane forces related to the coupling between in
plane and out of plane behaviour due to the panel curvature.
The Airy stress function amplitudes fmn and ~f mn do not introduce
further degrees of freedom in the problem, as they are expressed as
function of the amplitudes qmn from the solution of the compatibility equation. In particular, after substitution of Eqs. 8, 22 and 23
into Eq. (18), it is obtained:
2
1
~mn m
h
i
b
a Rp2 n4 a11 mn2 2a12 a66 m4 a22
b
ab
27
@P
0
@q
28
43
@2P
@ 2 P @q
@2P @K
0
2
@ g@q @q @ g @q@ K @ g
29
Kq;g GK;g 0
30
Kq;g GK;g 0
K2;g qT q 1
31
The nonlinear Eqs. (31) are nally solved with a perturbation approach [22], where the unknowns q and K are represented with a
rst order Taylor expansion.
5. Implementation of the fast tool
The fast tool is implemented in a computer code working in
Matlab [23] environment. It results suited to perform quick analyses, to study different congurations and to realize parametric
studies. Thanks to the user-friendly graphical interface, the program allows for a simple introduction of the input data.
The length and the width of the panel, as well as the dimensions
of the stiffener, are rstly introduced as inputs. Then, the program
requires the material properties in the form of elastic moduli and
ply strengths. The values can be set manually or can be loaded
from a database. Furthermore, the stacking sequences of the skin
and the stiffener are introduced.
The fast tool offers the possibility of choosing between two
solution procedures. The rst one is the linear buckling analysis,
and relies on the plate assembly model, which is the rst step of
the procedure. The second possibility is the post-buckling analysis,
and refers to the complete two-step procedure.
The input is completed by setting up the analysis parameters.
The program requires to dene the pre-buckling load and the number of shape functions used in the plate assembly model. In case of
post-buckling analysis, further inputs are expected. They include
the number of shape functions for the post-buckling procedure,
the total applied load, the number of buckling eigenvalues used
as initial imperfections and their amplitudes. Different outputs
can be requested on the basis of the chosen type of analysis. In case
of buckling analysis, the user can require a summary of the results,
including the buckling load, the corresponding mode and the linear
stiffness. In case of post-buckling analysis, the available outputs
are the forcedisplacement curve, the deformed shape, and the
contour of three different failure criteria (maximum stress, maximum strain, Tsai-Hill).
6. Analysis of an omega stiffened panel
The fast tool is applied to study a curved omega stiffened panel
subjected to loading conditions of combined compression and
shear.
The load history is divided in two phases:
1. pure shear load per unit length of 62 N/mm
2. compression load with an axial displacement of 2.10 mm, and
shear load per unit length kept constant at 62 N/mm
one bay, and two half bays. The height of the stiffener is 24 mm,
the crown top is 27 mm long and the web angle measured from
the normal to the skin is equal to 10. The skin consists of quasiisotropic laminate of 12 plies with a stacking sequence of [45/
45/0/90/ 45/45]s, for a total thickness of 1.5 mm. The
stringers are composed of a laminate of 9 plies with stacking sequence of [45/0/ 45/0/90/0/ 45/0/45] and a total thickness of 1.125 mm.
The results computed with the fast tool are compared with
analyses performed using the commercial nite element code Abaqus [24].
The nite element model of the panel is reported in Fig. 5. The
skin and the stiffeners are modeled with four-noded S4R shell elements, with a mesh of about 5 5 mm chosen on the basis of a
preliminary convergence analysis. The loaded edges are simply
supported, whereas a constraint equation is applied to the longitudinal edges to enforce periodic conditions in terms of out of plane
displacements and rotations around the longitudinal axis. Two
analyses are performed, eigenvalue and nonlinear static.
The input data are introduced by means of the graphical user
interface. It is interesting to highlight the choice of the analysis
parameters, which are the number of shape functions and the initial imperfections. They can be dened case by case on the basis of
a convergence analysis. In the rst step, the out of plane deections
are represented using 10 10 functions for the skin, and 10 7
for the stiffener elements, which are a good compromise between
computational efciency and accuracy of the results. Considering
that the plate assembly model exploits the symmetries and the
anti-symmetries of the buckling modes, the total number of degrees of freedom for the buckling analysis is equal to 235.
In the second step, the skin deection is represented with a
number of 12 12 functions. The resulting problem is given by
145 equations, consisting in 144 equilibrium equations and 1 constraint equation requested by the arc-length procedure. Compared
to the rst step, the number of shape functions is increased. This
choice is motivated by the need to calculate the stresses, which
are related to the second derivatives of the displacements, and consequently characterized by a slower convergence rate.
Regarding the introduction of the initial imperfections, it is
important to avoid the convergence of the solution to the unbuckled conguration. The initial imperfections can be determined
through a Fourier analysis of experimental data, if available. Otherwise it is possible to use a linear superposition of buckling modes,
which is the strategy here adopted. In the present study, the initial
imperfections are taken equal to the rst buckling mode, with an
amplitude to thickness ratio of 5%.
The outputs are then dened. In particular, the results of the
rst step analysis are required in terms of buckling load and buckling mode, while the results of the post-buckling analysis are dened as force displacement curve, deformed surface at different
load levels and contour of the maximum stress index.
The buckling load obtained from the rst step of the procedure
is characterized by a compressive force per unit length equal to
59.2 N/mm, and a shear force per unit length of 62 N/mm. The
buckling mode shape reveals 6 skewed halfwaves on the skin,
whereas the plate elements composing the omega stiffener do
Table 1
IM7/8825 ply mechanical properties.
Material properties
150000
9000
5300
0.32
0.125
44
Fig. 5. Curved omega stiffened panel: (a) nite element model, (b) stiffener cross section.
2.5
2
1.5
1
0.5
0
0
Analytical
ABAQUS
0.5
1.5
2.5
Fig. 7. Contour of the out of plane displacement: (a) analytical, (b) Abaqus.
45
Fig. 8. Contour of the maximum stress failure index: (a) analytical, (b) Abaqus.
Table 2
T300/934 ply mechanical properties.
Material properties
Youngs modulus E11 [MPa]
Youngs modulus E22 [MPa]
Shear modulus G12 [MPa]
Poissons ratio m12
Ply thickness [mm]
124000
11400
5500
0.3
0.145
The stress eld is plotted at the last step of the solution. In this
example, the maximum stress criterion is computed for each ply of
the laminate. The contour is reported for the ply undergoing the
highest value of failure index, which is automatically identied
by the fast tool. In this case, the procedure identies the second
ply of the skin, which is oriented at 45, as the most stressed
one. The contour is reported in Fig. 8, with a maximum value equal
to 0.50 in proximity of the longitudinal edges. The numerical results of Fig. 8 are characterized by a maximum stress failure index
of 0.65. The results obtained with the fast tool can be used as a preliminary estimate of the stress level exhibited by the structure. The
46
Fig. 10. Experimental [25] and analytical congurations in the post-buckling range:
(a) and (b) three halfwave conguration, (c) and (d) mode jumping, (e) and (f) ve
halfwave conguration.
47
[3] Byklum E, Amdahl J. A simplied method for elastic large deection analysis of
plates and stiffened panels due to local buckling. Thin-Walled Struct
2002;40(11):92553.
[4] Harris GZ. Buckling and post-buckling of orthotropic laminated plate. In: 16th
AIAA/ASME/SAE Structures, Structural Dynamics, and Materials Conference,
AIAA-1975-813, May 2729 1975, Denver, CO.
[5] Romeo G, Frulla G. Analytical/experimental behavior of anisotropic rectangular
panels under linearly varying combined loads. AIAA J 2001;39(5):93241.
[6] Diaconu CG, Weaver PM. Postbuckling of long unsymmetrically laminated
composite plates under axial compression. Int J Solids Struct 2006;43(22
23):697897.
[7] Rhodes J, Harvey JM. Plates in uniaxial compression with various support
conditions at the unloaded boundaries. Int J Mech Sci 1971;13(9):787802.
[8] van der Neut A. The stiffness in compression of imperfect elastically restrained
plate strips at various in-plane boundary conditions. Report LR-245, Delft,
1977.
[9] Mittelstedt C, Beerhorst M. Closed-form buckling analysis of compressively
loaded composite plates braced by omega-stringers. Compos Struct
2009;88(3):42435.
[10] Brubak L, Hellesland J. Semi-analytical postbuckling and strength analysis of
arbitrarily stiffened plates in local and global bending. Thin-Walled Struct
2007;45(6):62033.
[11] Brubak L, Hellesland J. Semi-analytical postbuckling analysis of stiffened
imperfect plates with a free or stiffened edge. Comput Struct 2011;89(17
18):157485.
[12] Mijukovic O, Coric B, Pavlovic MN. Transverse-stiffener requirements for the
post-buckling behaviour of a plate in shear. Thin-Walled Struct
1999;34(1):4363.
[13] Chai GB, Banks WM, Rhodes J. The instability behaviour of laminated panels
with
elastically
rotationally
restrained
edges.
Compos
Struct
1991;19(1):4165.
[14] Boay CG, Wah KP. Postbuckling formulation for symmetric laminated panels
with various edge support conditions. Mech Adv Mater Struct
2001;8(1):1528.
[15] Bisagni C, Vescovini R. Analytical formulation for local buckling and postbuckling analysis of stiffened laminated panels. Thin-Walled Struct
2009;47(3):31834.
[16] Vescovini R, Bisagni C. Single-mode solution for post-buckling analysis of
composite panels with elastic restraints loaded in compression. Compos: Part
B 2012;43(3):124758.
[17] Buermann P, Rolfes R, Tessmer J, Schagerl M. A semi-analytical model for local
post-buckling analysis of stringer- and frame-stiffened cylindrical panels.
Thin-Walled Struct 2006;44(1):10214.
[18] Hyer MW. Stress analysis of ber-reinforced composite materials. New
York: McGraw-Hill; 1998.
[19] Reddy JN. Mechanics of laminated composite plates and shells: theory and
analysis. Boca Raton: CRC Press; 2004.
[20] Loughlan J. The buckling performance of composite stiffened panel structures
subjected to combined in-plane compression and shear loading. Compos
Struct 1994;29(2):197212.
[21] Paik JK, Lee MS. A semi-analytical method for the elasticplastic large
deection analysis of stiffened panels under combined biaxial compression/
tension, biaxial in-plane bending, edge shear, and lateral pressure loads. ThinWalled Struct 2005;43(3):375410.
[22] Steen E. Application of the perturbation method to plate buckling problems.
Technical Report 98-1, University of Oslo, Department of Mathematics,
Mechanics Division, 1998.
[23] . Matlab. Users Guide 2011. Mathworks, Inc.; 2011.
[24] . ABAQUS, version 6.11. Users Manual. Providence, RI, USA: SIMULIA World
Headquarters; 2011.
[25] Falzon BG. The behaviour of damage tolerant hat-stiffened composite panels
loaded in uniaxial compression. Compos Part A: Appl Sci Manuf
2001;32(9):125562.
[26] Falzon BG, Steven GP. Buckling mode transition in hat-stiffened composite
panels loaded in uniaxial compression. Compos Struct 1997;37(2):25367.