Available online at www.sciencedirect.

com

Computers and Structures 86 (2008) 919929

www.elsevier.com/locate/compstruc

Design and analysis of stiened composite panels including

post-buckling and collapse
R. Degenhardt

a,*

, A. Kling a, K. Rohwer a, A.C. Orici

b,c

, R.S. Thomson

a
DLR, Institute of Composite Structures and Adaptive Systems, Lilienthalplatz 7, 38108 Braunschweig, Germany
School of Aerospace, Mechanical and Manufacturing Engineering, Royal Melbourne Institute of Technology, GPO Box 2476V,
Melbourne, Vic. 3001, Australia
Cooperative Research Center for Advanced Composite Structures Limited, 506 Lorimer Street, Fishermans Bend, Vic. 3207, Australia
b

Received 8 March 2007; accepted 30 April 2007

Available online 3 July 2007

Abstract
The European aircraft industry demands reduced development and operating costs, by 20% and 50% in the short and long term,
respectively. Contributions to this aim are provided by the completed project POSICOSS (5th FP) and the running follow-up project
COCOMAT (6th FP), both supported by the European Commission. As an important contribution to cost reduction a decrease in structural weight can be reached by exploiting considerable reserves in primary bre composite fuselage structures through an accurate and
reliable simulation of post-buckling up to collapse. The POSICOSS team developed fast procedures for the post-buckling analysis of
stiened bre composite panels, created comprehensive experimental data bases and derived suitable design guidelines. COCOMAT
builds up on the POSICOSS results and considers in addition the simulation of collapse by taking degradation into account. The results
comprise an extended experimental data base, degradation models, and improved certication and design tools as well as extended design
guidelines.
One major task of POSICOSS and COCOMAT is the development of improved analysis tools that are validated by experiments performed within the framework of the projects. Because the new tools must comprise a wide range of various aspects a considerable number of dierent structures had to be tested. These structures were designed under dierent objectives (e.g. large post-buckling region). For
the design process, the consortiums applied state-of-the-art simulation tools and brought in their own design experience. This paper deals
with the design process as performed within both projects and with the applied analysis procedures. It is focused on the DLR experience
in the design and analysis of stringer-stiened CFRP panels gained within the scope of these two projects.
2007 Elsevier Ltd. All rights reserved.
Keywords: Collapse; Post-buckling; Composites; Simulation tools; Experiments; Degradation

1. Introduction
The European aircraft industry demands reduced development and operating costs, by 20% and 50% in the short
and long term, respectively. The European Commission
(EC) project POSICOSS, which lasted from January 2000
to September 2004, and the 4-year follow-up project
COCOMAT, which started in January 2004 (cf. Fig. 1),

Corresponding author.
E-mail address: richard.degenhardt@dlr.de (R. Degenhardt).

0045-7949/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2007.04.022

contribute to this aim [14]. Both projects are under the

co-ordination of DLR, Institute of Composite Structures
and Adaptive Systems. The main goal is the exploitation
of considerable reserves in primary bre composite fuselage
structures through an accurate and reliable simulation of
post-buckling up to collapse. Collapse is specied by that
point of the loaddisplacement curve where a sharp
decrease occurs thus limiting the load-carrying capacity.
The POSICOSS team developed fast and reliable procedures for post-buckling analysis of bre composite stiened
panels, created experimental data bases and derived design
guidelines [1,2,517]. Alternative fast methods were

920

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

GARTEUR SM-AG-25

Improved MATerial Exploitation

at Safe Design of COmposite
Airframe Structures by Accurate
Simulation of COllapse

Postbuckling and
Collapse Analysis

COCOMAT (EU,6th FP, STREP)

2000

2001

2002

2003

2004

2005

2006

2007

POSICOSS (EU, 5th FP)

Improved POstbuckling SImulation
for Design of Fibre COmposite
Stiffened Fuselage Structures
Fig. 1. Timetable of the EU projects GARTEUR SM-AG-25, POSICOSS and COCOMAT.

developed also in parallel to the POSICOSS project [1720].

The COCOMAT project builds up on the POSICOSS
results and goes beyond by simulation of collapse. It
improves existing tools for design and analysis, sets up
design guidelines suitable for stiened panels taking skin
stringer separation and material degradation into account,
and it creates a comprehensive experimental data base
[3,4] concerning such structural components. Another project was carried out by the Action Group 25 of the Group
for Aeronautical Research and Technology in Europe
(GARTEUR) Structures and Materials panel. It investigated three dierent benchmarks with well-documented
buckling tests in order to identify abilities and deciencies
of available analysis tools as well as to establish recommendations for buckling, post-buckling and collapse analysis of
thin-walled aerospace structures [21,22]. One of the benchmarks investigated in this GARTEUR activity was taken as
a start design for the design process in POSICOSS.
The improved tools, developed within the POSICOSS
[814] and COCOMAT project, have to be validated by
test results. Since appropriate test data was not available,
both projects were forced to create new experimental data
bases for curved stringer-stiened carbon bre reinforced
plastic (CFRP) panels as well as for complete cylindrical
shells. To that end suitable panels and cylindrical shells
were designed under own project objectives. Some of the
test structures are already manufactured, inspected and
tested. Both projects dierentiate between shell structures
for validation and for industrial application. The validation structures are designed as to specic limiting application aspects of the software to be validated, e.g. small or
large stiness reduction in the post-buckling regime. The
industrial structures were designed with regard to industrial applications, mainly by existing procedures used in
day-to-day industrial design practice.
For the design and analysis the partners brought in their
own experience and they utilised dierent available software tools. Two dierent kinds of tools were applied: fast
tools suitable for an economic design process and very
accurate but necessarily slow tools required for the nal

certication. Geometrical nonlinear computations up to

collapse were performed while the material was assumed
linear elastic. The onset of degradation of the structure
and of the skinstringer separation was determined using
dierent failure criteria.
This paper is focused on the experience in design and
analysis of stringer-stiened CFRP panels DLR gained
within the scope of the POSICOSS and COCOMAT projects. In [3,6,21,23], the authors have already published
partly some results presented here. However, this paper
adds new data and gives a useful comparative study
between the outcomes obtained within the dierent projects
GARTEUR SM-AG-25, POSICOSS and COCOMAT.
2. Design of composite panels
2.1. Introduction
Designing a structure always involves some kind of optimisation. Regardless of the material, kind of structure or
application, the objective function in this optimisation process depends on the purpose of the structure. In general,
one can distinguish between industrial structures and validation structures. The validation structures are designed
as to specic limiting aspects of application of the software
to be validated, e.g. type of shell theory (design going into
the limits of the theory), type of buckling before post-buckling (local or global), mild or strong stiness reduction in
the post-buckling regime, or multiple or single modes of
buckling before post-buckling. Industrial structures are
designed with regard to industrial applications, mainly by
existing procedures and requirements used in day-to-day
industrial design practice. For these structures, there exist
usually multi-objective requirements concerning weight,
load-carrying capacity and costs.
Fig. 2 illustrates a realistic (experimentally measured)
load-shortening curve of an axially compressed stiened
CFRP panel representing a stringer dominant design. It
explains the terminology of three marked load levels. The
lowest one usually provokes the rst local buckling where

Collapse load

921

Load

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

1st global (stringerbased) buckling

1st local buckling

Realistic curve
Simplified curve

Shortening

Fig. 2. Denition of rst local and global buckling load and collapse load.

the buckling mode is restricted to local skin buckles

between the stringers. The second level causes the rst global buckling which is stringer based-buckling. The highest
load level is reached at collapse. The thicker red curve is a
simplied representation of the real load-shortening curve
with knees at these characteristic load levels. COCOMAT
aims at improving design capabilities by accounting for
the complete load-shortening behaviour up to collapse.
This paper concentrates on the description of the design
process as performed within the EC projects POSICOSS
and COCOMAT. The structures considered are curved
stringer-stiened panels and cylinders made of CFRP
material. Within each project a large number of structures
were designed. Because the maximum number of tests was
limited only designs appropriate under the manufacturing
and testing conditions were selected. The new test results
build a large experimental data base, which is necessary
for the validation of the new tools developed to simulate
the buckling and post-buckling behaviour up to collapse.
Two kinds of tools are considered: reliable fast tools reducing design and analysis time by an order of magnitude,
which allow an economic design process, whereas very
accurate but in most cases necessarily slow tools are
required for the nal certication. For the industrial applicability, these tools must be on the one hand validated by
appropriate experiments and on the other hand their applicability must be proven on real industrial panels.
Sections 2.3, 2.4 and 2.5 describe the design process of
DLR as partner within the projects POSICOSS and COCOMAT. It starts with a description of a benchmark from
the GARTEUR Action Group 25 which was taken as a
start design for the POSICOSS structures. Then the design
process within the POSICOSS project is described. The
panel design within COCOMAT builds up on the experience on POSICOSS and is described nally. Section 2.2
gives a comparison of the material properties and geometrical data used for all panel designs.

2.2. Geometrical and material data of the panel designs

This section comprises the material properties and
geometrical data used for all panel designs described in
Section 2. Table 1 contains material properties for the prepreg material IM7/8552 UD, which was used throughout,
Table 2 gives material properties for the adhesive, which
connects the skin with the stringers, Table 3 compares
nominal geometrical data, Fig. 3 shows the dierent stringer types used in the test and their nite element (FE) modelling with the connection to the skin and Fig. 4 gives the
assumed boundary conditions.
2.3. Start design
The design of the POSICOSS structures started out
from the results of a pre-damaged benchmark, which was
formerly tested at DLR. This benchmark was intensively
investigated within the GARTEUR SM Action Group 25
Postbuckling and Collapse Analysis [21,22] and at the

Table 1
Material properties for CFRP prepreg IM7/8552 UD
Stiness

Unit

0 tensile
modulus
90 tensile
modulus
0 compression
modulus
90
compression
modulus
In plane shear
modulus
Poisson ratio

GARTEUR AG25
(start design)

POSICOSS

COCOMAT

GPa

192.3

164.1

GPa

10.6

8.7

GPa

141

146.5

146.5

GPa

11

9.7

9.7

GPa

6.3

6.1

5.1

0.3

0.31

0.28

922

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

Table 2
Material properties of the adhesive Redux 312 [24]
Stiness/strength

Unit

Value

E1
m12
Max. compressive stress
Max. shear stress
Max. normal stress

MPa

MPa
MPa
MPa

3000
0.4
48
38
8.3

beginning of the POSICOSS project and will be subsequently detailed.

As depicted in Fig. 5, the benchmark represents an axially compressed CFRP panel consisting of a skin with
nominally cylindrical shape and stiened by T-shaped
stringers. The stringers were partially separated from the

skin by impacting prior to the tests. The damaged areas

were measured by ultrasonic inspection to introduce them
into the FE model for an accurate numerical simulation.
Finally, the panel was axially compressed until collapse.
Dierent commercial and self-developed nite element
tools were applied to simulate the behaviour of this panel
during loading up to collapse. Linear and nonlinear analyses as well as buckling analyses were performed in order to
observe the axial stiness in the pre-buckling region, the
buckling loads of the panel and the structural behaviour
in the post-buckling region. A major challenge of this
benchmark was the simulation of the damaged region
through contact elements. Furthermore, a considerable
number of parameters like skinstringer connection, stringer ange modelling, number of nite elements, damping,

Table 3
Nominal geometrical data and lay-up for the panel designs
Nominal geometry/lay-up

GARTEUR AG25 (start design)

Panel length (mm)

Free length (buckling length) (mm)
Radius (mm)
Arc length (mm)
Number of stringers
Distance stringer to stringer
Distance stringer to longitudinal edge
Laminate set-up of skin
Laminate set-up of stringers (cf. Fig. 3)
Blade
Flange
Ply thickness (mm)
Stringer height (mm)
Stringer width (mm)

l = 800
lf = 620
r = 400
a = 419
n=6
d = a/6
e = d/2
[90, +45, 45, 0]s

POSICOSS

COCOMAT

Panel 1/3

Panel 2/4

l = 780
lf = 740
r = 400 (1000)
a = 420
n=3
d = a/6
e = d/2
[+45, 45, 0]s

n=4
d = a/8
[90, +45, 45, 0]s

[(+45, 45)3, 06]s

cf. Fig. 3
t = 0.125
h = 14
f = 37.9

Test

l = 780
lf = 660
r = 1000
a = 560
n=5
d = 132 mm
e = f/2 = 16 mm

[(45, 45)3, 06]

h = 14
f = 32

3 mm

12.5 mm

32 mm

FE-model
with skinstringer
connection

GARTEUR AG 25 / POSICOSS
Fig. 3. Stringer types of the panel designs.

COCOMAT

14 mm

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

923

General
Blue:

Rigid body (loaded edge)

Green:

Fully fixed

Red:

Party fixed (free in axial direction)

Yellow: (see row below)

Lateral
edges

Detail

Edge support

free

Filler

clamped

Gliding plane
Test panel

25 mm
GARTEUR AG 25 / POSICOSS

COCOMAT

Fig. 4. Boundary conditions. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

180
160
140

Load [kN]

120
100
80

DLR,

ABAQUS/Explicit

Karlsruhe, FEAP, 3D elements

60

Karlsruhe, FEAP, 2D elements

40

QinetiQ,

LUSAS

Samtech, SAMCEF

20

Test 67

0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Shortening [mm]

Fig. 5. Finite element analyses of the undamaged DLR benchmark taken as start design [21].

imperfections, loading velocity, boundary conditions,

numerical method or kind of nite elements were investigated. Fig. 5 illustrates some load-shortening curves
obtained by numerical simulations of the undamaged
panel. As a main result it turned out that all FE software
tools considered were suitable in general for the simulation
of buckling, post-buckling or collapse behaviour of such
panels. Specic abilities and deciencies of the nite element tools were evaluated. Recommendations with respect
to the inuence of parameters, the initial buckling load, the
convergence behaviour, the simulation of load introduction
and boundary conditions as well as the imperfection sensitivity were derived.
In order to check the inuence of the pre-damage this
benchmark was also analysed with damage ignored. It
turned out that this undamaged panel has almost no
reserve capacity in the post-buckling region because its
local skin buckling load is very close to the global stringer

buckling load. This is due to a rather skin-dominant

design. Since under the design objective large post-buckling region such a design represents a highly undesirable
case this undamaged panel was taken as a start design
for the POSICOSS project.
2.4. Design process within POSICOSS
The main design objective within POSICOSS was to
obtain a signicant post-buckling area before collapse.
DLR aimed to design four panels and two complete cylindrical shells that should be suitable for software validation
purposes. The undamaged benchmark described in the previous section has almost no post-buckling region and in
that sense represents a highly undesirable design. It was
therefore taken as a start design, which was then modied
in the following way in order to increase the load-carrying
capacity in the post-buckling range (cf. Fig. 2):

924

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

(1) For the rst cylinder design, the number of stringers

was reduced to trigger the local buckling at a lower
level and to increase the load-carrying capacity in
the post-buckling region.
(2) In the second cylinder design, in addition to the
reduced number of stringers, the 90-layers of the
skin were removed in order to increase the sensitivities to torsion loading. The radius of both cylinders
was xed to 400 mm due to testing constraints.
(3) For the purpose of comparison, the panels should be
as similar as possible to the cylinders. Therefore, the
rst two panel designs were taken as 60 sections
from the cylinder designs.
(4) Two additional panel designs were dened that dier
from the rst two, only by an increase in the radius
from 400 mm to 1000 mm in order to examine the
inuence of the radius and to get closer to the real aircraft fuselage structures.

This design process resulted in four dierent stiened

panels and two dierent complete cylindrical shells as illustrated in Fig. 6. Except for one panel, which has the smallest post-buckling region, all designs were manufactured,
some of them two or three times each in order to increase
the reliability, and were tested at the DLR buckling test
facility until collapse. Fig. 7 illustrates the comparison of
the test and simulation of one tested panel. It can be seen
that the panel design has a large post-buckling region as
it was planned. There is a good agreement between simulation and test up to the rst global buckling. From that
point on the dierences become larger. However, because
degradation is not considered within that simulation a better agreement in the deep post-buckling region is not
expected. In addition, the modelling of the boundary conditions for the clamping of the lateral edges of the panel
showed a large inuence on the axial stiness in the postbuckling region after the rst global stringer buckling.

Fig. 6. POSICOSS designs (DLR, cf. Fig. 3).

120
First ply failure by Tsai-Wu

Axial load [kN]

100

80

60

40
Experiment P12
ABAQUS/Standard, Stabilize 2e-6, 20mm boundary

20

ABAQUS/Standard, Stabilize 2e-6, 20mm boundary, with

buckling-mode-imperfection (20%)

0
0

0.5

1.5

2.5

3.5

Shortening [mm]

Fig. 7. (Color online) POSICOSS design panel P12 comparison test and simulation.

4.5

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

More details are given in [5]. Considering degradations as

well as additional investigations to applied longitudinal
edge conditions are topics of the project COCOMAT.
2.5. Design process within COCOMAT
To simulate accurately the collapse load of stringer-stiened CFRP panels the COCOMAT group improves slow
certication tools and fast design tools that are capable
of taking degradation into account. The group considers
the following degradation modes: skinstringer separation,
delamination in the stringer blade and degradation on the
composite structure itself. For the validation of the tools
an appropriate experimental data base is not available
(the POSICOSS test data base could only be taken as a
baseline because degradation was not considered during
the project). Therefore new curved stringer-stiened CFRP
panels, which shall be manufactured and tested, were
designed. As in POSICOSS, the group designed two kinds
of panels: validation panels and industrial panels. Being a
research establishment DLR concentrated to design one
validation panel which is called Design 1 in the following.
The objective was to accomplish a large post-buckling
region and an early onset of skinstringer separation.
The design process for Design 1 started with a panel
conguration with a radius of 1000 mm tested within POSICOSS. The objective was to increase the post-buckling
region further, especially to have a certain load capacity
after the rst global buckling. The reason is that the inuence of skinstringer separation on the collapse load
should be investigated and this kind of degradation usually
occurs after the rst global stringer buckling. Several parametric studies for the variation of the lay-up of the skin and
stringer, number of stringers, stringer geometry and position of the stringers were performed. During the design

925

process the onset of dierent kinds of degradation, such

as skinstringer separation, delamination in the stringer
blade and failure in the composite laminate structure have
been estimated by simple extension of the available software tool. In order to check the inuence of degradation
on collapse the panels with a large post-buckling region
and the indication of skinstringer separation (failure in
the adhesive layer) as early failure mode were favoured.
There was another important change of Design 1 in
comparison to the POSICOSS design. For Design 1, the
clamping boundary conditions of the lateral edges of the
panel, which were applied on all POSICOSS experiments,
were released because the modelling of these boundary
conditions showed a signicant inuence on the axial stiness in the post-buckling region after the rst global stringer buckling (cf. Fig. 7). However, in order to avoid skin
buckling starting in that laterally free area the stringers
were moved in circumferential direction to the lateral
edges. In addition, dierent designs were analysed in order
to ensure that the onset of skinstringer separation starts in
the middle stringers and not in those at the edges.
On the basis of structural and fracture mechanics analyses one design (Design 1) was selected as being the most
suitable for the experimental investigation into degradation
and collapse of stiened composite panels. Fig. 8 illustrates
the load-shortening curve of this design in comparison to a
POSICOSS design. Design 1 exhibits a large post-buckling
region, even after the rst global stringer buckling which
starts in the center of the panel.
3. Analysis of composite panels
For the design of the panels, described in the previous
section, the FE software ABAQUS/Standard (Abaqus)
was applied. Geometrical nonlinear computations with an

140

POSICOSS
panel
Start
desi gn

120

Design 1

Load [kN]

100

80

60

40

1st failure
First estimated
in the adhesive
failure in the adhesive

20

Design 1
0
0

0.5

1.5

2.5

3.5

Shortening [mm]

Fig. 8. (Color online) Load-shortening curve of the COCOMAT panel design in comparison to the start design from POSICOSS.

926

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

incremental iterative NewtonRaphson method with articial damping (stabilize-method) up to collapse were performed. The material is linear elastic. In order to model
degradation DLR developed Abaqus user subroutines,
which consider the skinstringer debonding using stressbased failure criteria.
3.1. Nonlinear nite element analysis without degradation
To analyse the pre- and post-buckling behaviour of the
panels the four-node shell element S4R of Abaqus has been
used. Fig. 9a depicts some details of the FE model (e.g.
spring elements, which have been applied to introduce
the stiness of the longitudinal edge supports in the computer model of the POSICOSS project).
The approach to conduct the FE analysis consists basically of four stages (Fig. 9b): The preprocessing, a linear
eigenvalue analysis to extract buckling modes, which are
subsequently used as initial imperfections in the nonlinear
analysis utilising the built-in NewtonRaphson technique
with adaptive/articial damping, and nally the postprocessing. This nonlinear solution method has been proven
to be relatively stable for the considered stringer-stiened
panels. Figs. 7 and 8 depict the loaddisplacement curves,
which have been obtained by utilising the analysis procedure described in Fig. 9b with and without initial geometric
imperfections.

Spring-elements

The validation of the numerical simulation is performed

by a comparison with experimental results on the so-called
global and local level. On the global level of validation the overall load-shortening as well as the full scale
deformation patterns are compared. Fig. 10 shows such a
comparison of buckling patterns obtained by experiment
and simulation. The experimental data was obtained using
ARAMIS a 3D-optical measurement system based on
photogrammetry. On the local level measurements from
strain gauges are considered and compared to numerically
calculated strains. Details to this concept can be found in
[6]. Fig. 7 shows a comparison of the load-shortening
curves of simulation and experiment of the POSICOSS
panel P12 [5]. Up to the rst global buckling load a very
good agreement can be observed. From that point the simulation and experiment begin to dier. There are two explanations for that. Firstly, no degradation is taken into
account, so in the deep post-buckling region a good agreement cannot be expected. Secondly, the high sensitivity of
the modelling of the lateral clamping boundary conditions
has most probably caused the starting of divergence shortly
before global buckling.
The objective for the design of the panels within COCOMAT was a large post-buckling region and an early onset
of skinstringer debonding. At that stage of the project no
tools were available that could take this kind of degradation into account. However, for the design process it is sufcient to know when degradation starts. As a work-around
the available tool Abaqus was utilised in following way. To
determine the onset of degradation of the composite lamina structure itself failure criteria that are available in Abaqus (e.g. Tsai-Wu and Tsai-Hill) were applied. The
adhesive, which connects the skin with the stringers, was
modelled with 3D solid elements (cf. Fig. 3). The occurrence of the maximum allowable stress in the adhesive
was taken as indication for the onset of degradation.
Detailed results can be found in [23]. The load-shortening
and failure predictions in Fig. 8 were calculated using this
approach. One can see that the onset of skinstringer deb-

Fig. 9a. Details of the FE model (POSICOSS).

Real Structure
CFRP-Panel

FE-Model

Linear Eigenvalue Analysis

Buckling Modes

Buckling Load

scaled imperfektions

Measured
Imperfections

rough
estimate

Nonlinear Analysis
Newton-Raphson-Method + automatic / adaptive
damping to stabilize the analysis (*STATIC, STABILIZE)

Postprocessing
(Load-Shortening-Curve, deformation of the structure, ...)

Fig. 9b. Analysis procedure in Abaqus.

Fig. 10. Out-of-plane deformations of one tested POSICOSS panel at

identical load levels.

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

onding almost coincided with the beginning of global stringer buckling. This behaviour is plausible and was expected
because the onset of stringer buckling causes also signicantly higher stresses in the adhesive layer.
3.2. Nonlinear nite element analysis with degradation
One main task of the COCOMAT project is to improve
slow certication tools. Within this task DLR concentrates
on the improvement of Abaqus in order to allow for skin
stringer separation. To solve this problem the adhesive
layer between skin and stringer is modelled using 3D nite
elements. The mechanical behaviour of these elements is
described by means of new self-developed Abaqus user
subroutines. Three user subroutines are developed, which
dier in their numerical approach. At this stage they use
simple stress-based failure criteria. However, it is possible
to implement more enhanced and probably more accurate
degradation models in two of the three subroutines. Within
COCOMAT new degradation models, which are based on
experimental investigations, are currently developed and it
is planned to implement them into the subroutines. The
three subroutines can be described as follows:
(1) User Dened Field (USDFLD): Allows dening only
simple failure criteria, which reduce selected material
properties.
(2) UMAT explicit: The stresses are calculated from the
previous increment results explicitly. It has the
advantage to control the failure propagation and
the degradation of the adhesive layer.
(3) UMAT implicit: The stresses are calculated from the
current stiness matrix implicitly. This version was
also extended for nding the rst element failing in
each increment. This increases the analysis time
dramatically.
The position of the user subroutines within the Abaqus
calculation process is given in Fig. 11. The last two user
subroutines allow for monitoring the propagation of the
failure in the adhesive and the implementation of complicated user-dened degradation models. As a rst approach,
a simple stress-based failure criterion for the adhesive was
implemented into all three user subroutines. The degradation of the adhesive is simulated by decreasing the Youngs
modulus to a small fraction of the initial value for those
nite elements for which the maximum allowable stress
(cf. Table 2) is reached.
All three user subroutines were tested and compared on
small and large models and showed a good agreement
between each other. The application of one user subroutine
on the COCOMAT panel Design 1 and the comparison
with the experiment and other software tools without degradation is shown in Fig. 12. This gure illustrates the
load-shortening curve of Design 1. Up to the rst global
buckling at about 1 mm shortening, it shows an excellent
agreement between all curves. From that point on only

927

Start of increment
Calculate integration point field variable
from nodal values
USDFLD

Start of iteration
Calculate
Calculate ,

UMAT
e xplicit / implicit

Define loads:

Fig. 11. Position of the user subroutines within the Abaqus calculation
process.

Fig. 12. (Color online) Comparison of experiment and dierent simulation tools [23].

the Abaqus analysis with the user subroutines shows a

good agreement with the experiment. However, it must
be noted that it is not sucient to compare only the
load-shortening curve, because the global buckling pattern
of the simulation and experiment are dierent. In addition,
the subroutines predicted more damaged adhesive areas
than observed in the experiment. Fig. 14 illustrates the
damaged areas for the connection between skin and stringers after the collapse test. It was obtained by ultrasonic
inspection (left) and the optical lock-in thermography
which show a good agreement between each other.
Fig. 13 shows the numerical simulation of failure propagation of the adhesive layer at four load levels of Fig. 12 [25].
The comparison with the experimental result of Fig. 14
with the simulation in Fig. 13 shows that within the simulation too many adhesive elements failed. This demonstrates that further improved degradation models as
under current development within COCOMAT are
needed.

928

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

Fig. 13. (Color online) Numerical simulation of failure propagation of the adhesive layer at four load level of Fig. 12 [25].

Fig. 14. (Color online) Ultrasonic aw echo (left) and thermographic (right) investigation visualising damages mainly between skin and stringer after the
test.

4. Summary and conclusions

This paper illustrates the design process and the experience gained at DLR on stringer-stiened panels and cylinders through the work done in the nished EU project
POSICOSS and the running EU project COCOMAT.
The FE tool Abaqus was applied for the design process
and is extended by means of self-developed user subrou-

tines that simulate the order of degradation of the skin

stringer separation. The numerical calculations have been
successfully validated with experimental data up to the rst
global buckling. For the simulation of the deep post-buckling region degradation must be taken into account. Here
Abaqus user subroutines were developed, which consider
stringer debonding using simple stress-based failure criteria. First application of the Abaqus user subroutines

R. Degenhardt et al. / Computers and Structures 86 (2008) 919929

yielded promising results. However, improved degradation

models which are currently under development within the
COCOMAT project are needed. It can be expected that
this design and analysis experience will be of advantage
for the design of future composite fuselage structures.
Acknowledgements
The nished project POSICOSS was supported by the
European Commission, Competitive and Sustainable
Growth Programme, Contract G4RD-CT-1999-00103.
The running project COCOMAT is supported by the European Commission, Priority Aeronautics and Space, Contract AST3-CT-2003-502723. The information in this
paper is provided as is and no warranty is given that the
information is t for any particular purpose. The reader
thereof uses the information at its sole risk and liability.
References
[1] Zimmermann R, Rolfes R. POSICOSS improved postbuckling
simulation for design of bre composite stiened fuselage structures.
Compos Struct 2006;73:1714.
[2] www.posicoss.de.
[3] Degenhardt R, Rolfes R, Zimmermann R, Rohwer K. COCOMAT
improved material exploitation at safe design of composite airframe
structures by accurate simulation of collapse. Compos Struct
2006;73:1758.
[4] www.cocomat.de.
[5] Klein H, Zimmermann R, Kling A. Buckling and postbuckling of
stringer stiened bre composite curved panels tests and computations. Compos Struct 2006;73:15061.
[6] Kling A, Degenhardt R, Klein H, Tessmer J, Zimmermann R. Novel
stability design scenario for aircraft structures simulation and
experimental validation. In: Proceedings of the 5th international
conference on computation of shell and spatial structures, Salzburg,
Austria, 14 June; 2005.
[7] Hansen L, Degenhardt R, Kling A. Auslegung von axial gedruckten,
versteiften Paneelen. Internal DLR Report, IB 131-2004/37; November, 2004.
[8] Kling A, Degenhardt R, Zimmermann R. A hybrid subspace analysis
procedure for non-linear postbuckling calculation. Compos Struct
2006;73(2):16270.
[9] Bisagni C, Lanzi L. Post-buckling optimisation of composite stiened
panels using neural networks. Compos Struct 2002;58(2):23747.

929

[10] Mocker T, Reimerdes HG. Postbuckling simulation of curved

stiened composite panels by the use of strip elements. Compos
Struct 2006;73(2):23743.
[11] Rikards R, Abramovich H, Auzins J, Korjakins A, Ozolinsh O,
Kalnins K, et al. Compos Struct 2004;63(2):24351.
[12] Gal E, Levy R, Abramovich H, Pavsner P. Buckling analysis of
composite panels. Compos Struct 2006;73(2):17985.
[13] Lanzi L, Giavotto V. Postbuckling optimisation of composite
stiened panels: computations and experiments. Compos Struct
2006;73(2):20820.
[14] Rikards R, Abramovich H, Kalnins K, Auzins J. Surrogate modelling in design optimisation of stiened composite shells. Compos
Struct 2006;73(2):24451.
[15] Rikards R, Abramovich H, Green T, Auzins J, Chate A. Identication of elastic properties of composite laminates. Mech Adv Mater
Struct 2003;10(4):33552.
[16] Bisagni C, Cordisco P. Testing of stiened composite cylindrical
shells in the postbuckling range until failure. AIAA J
2004;42(9):180617.
[17] Bisagni C, Cordisco P. Postbuckling and collapse experiments of
stiened composite cylindrical shells subjected to axial loading and
torque. Compos Struct 2006;73(2):13849.
[18] Se-Hee Oh, Kwang-Soo Kim, Chun-Gon Kim. An ecient postbuckling analysis technique for composite stiened curved panels.
Compos Struct 2006;74(3):3619.
[19] Linde P, Schulz A, Rust W. Inuence of modelling and solution
methods on the FE-simulation of the postbuckling behaviour of
stiened aircraft fuselage panels. Compos Struct 2006;73(2):229
36.
[20] Burmann P, Rolfes R, Tessmer J, Schagerl M. A semi-analytical
model for local post-buckling analysis of stringer- and frame-stiened
cylindrical panels. Thin Wall Struct 2006;44:10214.
[21] Degenhardt R, Rohwer K, Wagner W, Delsemme JP. Postbuckling
and collapse analysis of stringer stiened panels a GARTEUR
activity. In: Proceedings of the 4th international conference on thinwalled structures, Loughborough, England; 2224 June, 2004.
[22] van Houten R, Zdunek A. GARTEUR (SM) AG-25 post-buckling
and collapse analysis. Final Technical Report. GARTEUR TP-149
NLR-TR-2004-463, Amsterdam; 2004.
[23] Orici AC, Thomson RS, Degenhardt R, Kling A, Rohwer K,
Bayandor J. Degradation investigation in a postbuckling composite
stiened panel. Compos Struct, in press. doi:10.1016/
j.compstruct.2007.01.012.
[24] Hexcel Composites. Product data. www.hexcel.com; 2005.
[25] Reinsch A, Degenhardt R. Numerical nite-element-analysis of
degradation in stiened CFRP panels. DLR Internal Report 1B
131-2005/34. Braunschweig: Institute of Composite Structures and
Adaptive Systems, DLR; 2005.