Experimental and Numerical Analysis of Composite Folded Sandwich Core Structures Under Compression
- First Online:
- Cite this article as:
- Heimbs, S., Middendorf, P., Kilchert, S. et al. Appl Compos Mater (2007) 14: 363. doi:10.1007/s10443-008-9051-9
- 530 Views
The characterisation of the mechanical behaviour of folded core structures for advanced sandwich composites under flatwise compression load using a virtual testing approach is presented. In this context dynamic compression test simulations with the explicit solvers PAM-CRASH and LS-DYNA are compared to experimental data of two different folded core structures made of aramid paper and carbon fibre-reinforced plastic (CFRP). The focus of the investigations is the constitutive modelling of the cell wall material, the consideration of imperfections and the representation of cell wall buckling, folding or crushing phenomena. The consistency of the numerical results shows that this can be a promising and efficient approach for the determination of the effective mechanical properties and a cell geometry optimisation of folded core structures.
KeywordsFolded coreSandwich structuresCompression testingNumerical simulation
Sandwich structures with fibre-reinforced plastic faces and a cellular core have a successful history as lightweight structures in aircraft design due to their superior stiffness-to-weight ratio. However, to date their field of exterior applications is limited to fairings and control surfaces so that the primary structure of an airliner has not been realised with sandwich materials yet. The main drawbacks are the accumulation of humidity, the complexity in manufacturing and the vulnerability against impact loads. In this context the aircraft manufacturer Airbus has presented a promising sandwich fuselage concept named VeSCo (Ventable Shear Core) [1–4], which incorporates folded structures as a sandwich core material. Such folded cores not only solve the problem of humidity accumulation due to their open ventilation channels, they can also be produced efficiently in a continuous process. The issue of impact vulnerability is currently addressed by the transnational project CELPACT (Cellular Structures for Impact Performance), which is concentrating on advanced cellular core materials for potential applications in future primary aircraft structures with lower fabrication costs and improved impact performance. Besides other cellular materials, this project covers folded core structures with different geometries and different materials. In addition to experimental test series, investigations are focused on numerical simulation methods. Such numerical methods have proven their efficiency and reliability in the past in impact simulations on sandwich structures with honeycomb cores [5, 6] and folded cores [7–9]. In order to be able to evaluate the impact performance of a sandwich core material, knowledge of the nonlinear, effective mechanical properties especially for flatwise compression and transverse shear is required. Including a large number of different folded core geometries in such a study makes the experimental determination of these properties time and cost consuming. Much more desirable is a numerical determination method using meso-models and virtual testing simulations in order to get the effective stress–strain data. Recently, this virtual testing approach has successfully been applied to folded core structures . However, the development and validation of adequate meso-models poses different challenges. These are addressed in this paper that covers the investigation of the flatwise compression behaviour of folded core structures.
2 Folded Sandwich Core Structures
One is made of phenolic resin-impregnated aramid (Kevlar®) paper and produced in a continuous process by IFB Stuttgart, Germany. This process involves the embossing of the folding edges onto the flat sheet material and subsequent folding into the final geometry [11, 12]. The other type is made of a carbon fibre-reinforced plastic (CFRP) laminate in a [0°/90°/0°] lay-up. This folded core was produced at KTU Kazan, Russia in a discontinuous process by forming the flat sheet between two matrices . But in contrast to deep drawing, the matrices are transformable so that the prepreg material is folded without being elongated. The top and bottom folding pattern is a simple zigzag shape and therefore slightly different to the aramid paper foldcore's geometry.
The densities of both folded core structures are almost equal with 113 kg/m3 for the aramid paper structure and 102 kg/m3 for the CFRP foldcore.
3 Experimental Testing of Folded Core
In case of the CFRP folded core a slight cell wall buckling was also visible followed by a compression fracture at a remarkably higher stress level. The drop in the stress–strain-diagram is considerably steeper than for the aramid paper foldcore (Fig. 3b). There are three different phenomena which determine the post-damage behaviour (Fig. 4b): (1) sliding of the surfaces after failure in the cell wall middle, (2) formation of hinges through residual intact layers after failure in the cell wall middle and (3) continuous crushing after failure at the cell wall top and bottom. The higher the compressive strains, the more predominant is the third phenomenon and this cumulative crushing and fragmentation of the cell walls leads to an increasing stress level before densification.
4 Modelling and Simulation
The aim of the simulation activities was to investigate the approach of virtual testing, i.e. to perform dynamic compression test simulations with folded core meso-models. The benefit of this approach is obvious: If the effective mechanical properties of a cellular structure can be obtained reliably with numerical simulations, testing efforts can be minimised and a large number of different core geometries can be characterised with a minor amount of time and cost.
4.1 Model Development
For comparability reasons, two different commercial finite element software combinations were used including a general purpose pre-processing tool and an explicit solver: ANSYS + PAM-CRASH and MSC.PATRAN + LS-DYNA. In order to be able to easily generate folded core meso-models of different geometries with the pre-processing tools, parametric models were developed in the respective command languages APDL (Ansys Parametric Design Language) and PCL (Patran Command Language).
Folded core geometry (cell wall dimensions and angles, cell wall thickness)
Model size (length, width, height or number of unit cells)
Element size and element type
Additional simulation controls (termination time, etc.)
Random imperfections (node-shaking and geometry distortion).
So far the meso-scale models have a uniformly perfect geometry. In reality no cellular structure is neither uniform in geometry nor free of imperfections and irregularities. This affects the buckling load of the single cell walls and the whole structure's strength. Therefore, a meso-scale model without imperfections will always lead to an overestimation of stiffness and strength values. Neglect of imperfections can also cause discrepancies due to uniform unit cell behaviour.
In this work two different approaches to account for imperfections in the FE-models were investigated:
The other way is to keep the ideal mesh and reduce the cell wall properties in such a way that the effective structural behaviour matches experimental data. This inverse approach requires basic experimental data of the folded core. These data are the target values, while the cell wall's thickness as well as stiffness and strength are defined as parameters. Within a parameter identification loop with an optimisation software, compression and shear test simulations are performed in order to determine the set of parameters with the best correlation to the target values. Hereby, the lack of imperfections in the FE-model is compensated by the use of cell wall properties that are on purpose lower than in reality.
Both approaches were applied to the folded core models. Preliminary investigations on the CFRP cores showed that the imperfections in the model by geometry distortion or node-shaking reduce the global compressive strength, but not in a sufficient way that the simulation results of virtual compression tests agree with experimental data. On the other hand, the reduction of the mechanical properties and cell wall thickness led to a very good consistence, while still maintaining a realistic buckling pattern in the simulation. Reality is a combination of both: There are geometrical imperfections in the cellular structure and there are variations in the material properties and the paper thickness, especially in case of impregnated aramid paper. The weakest or thinnest areas of a cell wall determine the global properties, since damage will start from these areas, which justifies the approach of reducing the average values in the model.
4.3 Cell Wall Material Modelling
4.3.1 Aramid Paper Folded Core
The aramid paper exhibits different mechanical properties for tension and compression, analogue to other fibre reinforced composites. Generally, it is difficult to measure the compressive properties of the aramid paper. Single paper sheets will buckle before the maximum compressive stress is met. In the case of thicker layered aramid paper samples the interface effects are uncertain. In Fig. 10b the compressive stress–strain diagram of a small, thick-walled cylinder made of several layers of aramid paper under compression is shown. The cylinder is elastically compressed until maximum load. It is noted, that the maximum compressive stress is about half the level of the maximum tensile stress in Fig. 10a. Subsequently, the cylinder begins to crush around a ring-shaped initial zone at a constant stress level. The cylinder demonstrated only negligible buckling.
A rough estimate for the compressive paper behaviour was made based on the compressive cylinder test: The virtually constant crushing stress was assumed to indicate a “pseudo-plastic” material behaviour of the aramid paper for compressive loads, if buckling can be neglected. This was implemented into the material model, such that the paper deforms linearly elastic until the crushing stress is met and then behaves perfectly plastic.
The aramid paper was modelled with an isotropic material model in PAM-CRASH (material type 116). This is based on a linear elastic-perfectly plastic constitutive law with different behaviour for tensile and compressive loads. The material properties were used from the aforementioned experiments, the paper thickness was 0.35 mm. Local imperfections were incorporated by applying the node-shaking approach.
4.3.2 CFRP Folded Core
The CFRP folded core was modelled and simulated with LS-DYNA. Since the cell walls are a laminate with a [0°/90°/0°] lay-up of three unidirectional layers, a user-defined integration rule with three integration points across the thickness of each shell element was applied. The orthotropic composite material model MAT54 in LS-DYNA was used for the CFRP material. Each integration point represents one layer and is characterised by the respective fibre angle. Material model MAT54 is based on a linear elastic constitutive law with failure criteria by Chang-Chang. Additionally, failure strains are introduced, which control element layer erosion. The crush-front algorithm of MAT54 was also used, which reduces the strength of elements neighbouring eroded elements to facilitate the representation of a pre-damage and a progressing crushing [15, 16].
4.4 Influence of Modelling Parameters
Model size (number of unit cells): The number of unit cells in the FE-model has an influence on the effective stress–strain curve in the post-damage region. Even in an ideal model and especially in a model with imperfections not all cells fail at the same time, which results in a smoother effective stress curve in case of a larger number of cells.
Element size: Although coarser meshes reduce the computational cost, they are not able to accurately represent cell wall buckling phenomena, therefore the post-buckling stress levels are higher (Fig. 12). The element size also seems to slightly affect the Young's modulus of the folded core prior to buckling. The peak stress was comparable for different element sizes. It is worth noting that the element size effect is coupled to the shell thickness, since an even wider influence was observed for smaller thicknesses. Similar effects have been observed for the crushing of honeycomb cores . Finally, convergence studies were conducted in order to find an appropriate element size.
Element type: If shell elements with a uniformly reduced integration are used, which are state of the art in dynamic simulations, hourglassing effects can play an increasing role in case of cell wall folding and affect the simulation results. Therefore, shell elements with selective reduced integration and full integration were applied, which do not exhibit hourglassing effects. The effective stress–strain curve for the elements with selective reduced integration was similar, although the computational cost was more than twice as high. The fully integrated element was more expensive as well and led to variations in the post-buckling stress level, which may be ascribed to locking effects.
Loading rate: Due to the nature of explicit FE analysis with very small time steps, quasi-static loadings can not be simulated within an acceptable computing time. To reduce CPU time, the loading rate in the simulation has to be considerably higher than in the experiment. Therefore, velocities of 0.02–2 m/s were compared (Fig. 13). The maximum compression force at which buckling is initiated in the aramid paper folded core model differs for varying velocities. For higher velocities the samples tend to crinkle over a limited fraction of their height, whereas at smaller velocities the samples tend to buckle globally. Although the samples demonstrate different visible behaviour, the compression force levels are comparable during folding. For high velocities oscillation of the system is observable, which is caused by the rapid initial acceleration.
Global mass scaling: Mass scaling is a common technique to artificially increase the simulation time step and thereby reduce the computation time, when quasi-static tests are to be simulated with an explicit solver. This assumption is reasonable if the resulting kinetic energy remains negligible. The parameter study showed that the scale factor 1,000 is the upper limit, without having an effect on the effective stress–strain curves.
4.5 Virtual Testing Simulations
The experiences gained in the parameter studies were used to develop folded core models for virtual compression testing simulations with PAM-CRASH and LS-DYNA and a corresponding verification against experimental data. The primary aim was consistency of the respective compressive stress–strain relationships. However, a correct representation of the physical deformation behaviour of the cell walls is of the same importance, since only in that case reliable simulation results are obtained. In the following chapters simulations results are compared to experiments regarding cell wall deformation and stress–strain behaviour.
4.5.1 Aramid Paper Folded Core
The stress–strain curve indicates a good agreement within the collapse zone, which is dominated by folding and kinking. The cell wall collapse in experiment and simulation occurs at similar strain values. Subsequently, the edges buckle at likewise comparable strains. The folds observed within the experiment tended to form sharp kinks, whereas the folds in the simulation buckled in a sinusoidal shape. Nevertheless, the overall folding pattern in the simulation model agrees very well to the experiment. Following the stress–strain curve, the simulation showed a softer behaviour compared to the experiment for high compressive strains. This difference was attributed to the neglected friction in the simulation. For increasing contact forces friction is expected to exhibit a larger effect on the foldcore compressive behaviour.
4.5.2 CFRP Folded Core
The flatwise compression behaviour of folded structures made of aramid paper and CFRP was investigated experimentally and numerically. Two different failure mechanisms could be identified with folding and kinking cell walls for the aramid paper structure and crushing cell walls for the CFRP structure. Virtual testing simulations of these structures showed that not only numerous modelling parameters and an appropriate choice of material model for the cell wall paper influence the simulation results, also the consideration of imperfections is essential. Altogether, the results of virtual compression test simulations showed good consistency with experimental data. Only the post-failure behaviour of the CFRP core structure could not be represented satisfactorily with the respective material models used.
With the potential of parametric models and the ability to cover the stiffness and strength of the respective folded core structures with an acceptable degree of accuracy, a geometry optimisation can be performed in order to identify a folded core geometry with optimised mechanical properties with a minimum density. Using this numerical approach, this can be done much more efficiently than conducting experimental test series. This paper only covers the flatwise compression behaviour but this investigation can easily be extended to the transverse shear properties and the behaviour of foldcore sandwich panels under various impact conditions.
This work is part of the EU project CELPACT within the Sixth Framework Programme of the European Commission (contract AST5-CT-2006-031038, 2006–2009). The authors gratefully acknowledge the funding of the research activities. Sincere thanks are given to N. Voll (EADS) for his efforts in the CFRP folded core model as well as S. Fischer (IFB Stuttgart) for the aramid folded core testing.