Postbuckling Analysis of CFRP Stringer Stiffened Panels –
Benchmarking and Development of Fast Tools
Torsten Möcker1, H.-G. Reimerdes1, Richard Degenhardt2, Rolf Zimmermann2
1
Aachen University, Department of Aerospace and Lightweight Structures, Germany
2
DLR, Institute of Structural Mechanics, Braunschweig, Germany
Presently postbuckling calculations are extremely time consuming due to their high
nonlinearity and because they require a very deep understanding of the underlying theories.
The POSICOSS project develops new procedures ready for industrial application, which
reduce the design and analysis time by an order of magnitude. Before developing strategies
for these improved tools, the partners performed benchmarking in order to identify abilities
and deficiencies of existing tools. This presentation deals with a short summary of the
benchmarking investigation and gives an overview of the fast tools developed.
Three different benchmarks based on well-documented buckling tests were identified. Then,
tools to simulate the structural behaviour were selected, the benchmark computations were
performed and evaluated, and recommendations were derived.
The first benchmark was an axially compressed CFRP curved panel. The panel was stiffened
by stringers which were partially separated from the skin by impacting prior to the tests. The
two other benchmarks were CFRP plane panels with two different stringer types. The three
benchmark panels were axially compressed until collapse.
The partners applied different tools to simulate the behaviour of these panels. Linear and
nonlinear analyses as well as buckling analyses were performed in order to figure out the axial
stiffness, the buckling loads and the structural behaviour in the postbuckling region. A
considerable number of parameters was investigated.
The benchmarks will be presented, and selected results including an evaluation of abilities
and deficiencies of the tools as well as recommendations will be given. The nonlinear
simulations required long computing times, which are not acceptable for design purposes in
industrial applications and demonstrated the high need for fast tools.
In order to reduce the computational effort, several tools for the fast simulation of the
postbuckling behaviour were developed within the project. These tools can be divided into
two groups: on the one hand optimisation procedures that are used together with global
approximation strategies and on the other hand fast simulation procedures based on analytical
or on finite element solutions.
Regarding the optimisation tools, two different procedures were developed, both using global
approximations to estimate the structural behaviour. To build up the global approximations, it
is required to perform a limited number of finite element computations or physical experi-
ments for sample points within the design space. The first approach for the optimisation
procedure uses the finite element results in order to train and test the global approximations
that are obtained by neural networks as well as by radial basis functions. Within the second
approach, experimental design is employed for the definition of the sample points and the
information on the structural behaviour in the sample points is used to build up approximating
functions according to the response surface method. These approximation tools cannot only
be applied to perform optimisations, but also for the fast reanalysis of the structural behaviour
within the design space.
Next to the optimisation procedures, three tools for the fast simulation of the postbuckling
behaviour that are based on analytical or finite element solutions were developed. With these
tools three different levels of idealisation and accuracy are considered. The first approach
aims on predicting the total collapse load. For this purpose the model of a single beam is used.
The cross section of this beam consists of one stringer supplied by the skin of the panel. To
obtain the collapse load of the panel, the buckling load of the single beam is multiplied by the
number of stiffeners of the panel.
A more detailed idealisation of the structure is used for the second approach. The whole panel
is modelled by a discretisation of the structure in the circumferential direction only. No
discretisation in the longitudinal direction is required, as trigonometric functions are used to
describe the buckling and postbuckling behaviour in this direction. Accordingly, a significant
reduction of the degrees of freedom is achieved compared to a finite element model. Based on
this discretisation the total stiffness matrix of the structure is determined and buckling
analyses as well as nonlinear analyses are performed.
Finally, the third approach uses a hybrid reduced basis technique. First, the panel will be
analysed using a conventional finite element model. From the results of these analyses a small
number of shape functions (e.g. buckling modes) is extracted. Subsequently, these shape
functions are used to analyse the structural behaviour. In order to limit the error during the
nonlinear analysis, the shape functions are updated regularly. By the consideration of the
shape functions instead of a detailed finite element model, the degrees of freedom of the
problem can be reduced substantially.
An overview of the main concepts of these fast tools will be presented including a comparison
of the abilities and the expected computing times of the approaches. More information on the
procedures and examples for the application will be given in the detailed presentations on the
fast tools.
