Numerical modeling of aluminium foam on two scales

0451804 - ÚTAM 2016 RIV US eng J - Journal Article
Němeček, J. - Denk, F. - Zlámal, Petr
Numerical modeling of aluminium foam on two scales.
Applied Mathematics and Computation. Roč. 267, September (2015), s. 506-516. ISSN 0096-3003. E-ISSN 1873-5649
R&D Projects: GA ČR(CZ) GAP105/12/0824
Institutional support: RVO:68378297
Keywords : closed-cell aluminium foam * Alporas * multiscale modeling * homogenization * FFT * finite element modeling
Subject RIV: JI - Composite Materials
Impact factor: 1.345, year: 2015
http://www.sciencedirect.com/science/article/pii/S0096300315001162

The paper deals with computational modeling of aluminium foams on two distinct scales. The microscopically heterogeneous cell walls are modeled with continuum micromechanics models. Several analytical schemes and FFT-based homogenization are applied to predict elastic properties at the first level. Nanoindentation with sharp Berkovich tip is utilized to obtain input parameters for the homogenizations. Plastic properties are assessed directly from spherical nanoindentation at this level. Several geometrical simplifications are studied to model the upper foam level. At first, two dimensional models based on beam analogy and plane strain finite element (FE) models are studied for their ability to predict effective elastic and plastic foam properties. Finally, the behavior of the three dimensional voxel based FE model derived from micro-CT imaging is investigated. Models are compared in terms of their ability to predict experimental results and in terms of their computational demands. It is shown in the paper each model type has difficulties to quantitatively match experimental data in the whole tested range. Two dimensional beam models are capable to predict elastic properties but fail to predict plastic ones. Plane strain FE models are very compliant and lack three dimensional confinement. Three dimensional voxel model has the largest potential to predict experimental measurements but it is the most computationally demanding. It was found the performance of all models on the foam level is very much dependent on their porosity which is the main controlling parameter of the model behavior. Any deviations from experimentally assessed porosity leads to large deviations in the model prediction. Mutual model comparisons and possible solutions are provided in the paper along with computational aspects and requirements.
Permanent Link: http://hdl.handle.net/11104/0252894