Professional Documents
Culture Documents
Technical Report
a r t i c l e i n f o a b s t r a c t
Article history: Metallic foams are new materials mainly produced by expansion in a proper chamber and mainly char-
Received 30 April 2008 acterized by internal voids: a material characterized by a very low density is obtained in this way. A lot of
Accepted 21 June 2008 foamed components are commonly produced, directly by injecting a gas or foaming agent into molten
Available online 2 July 2008
metal inside a closed die. However, secondary operations on these materials can play an important role
in order to enhance the foam production flexibility. From the above considerations, the deformation
behaviour of an aluminium foam was investigated by compression tests. The study compares three dif-
ferent numerical analyses highlighting their points of strength and weakness in order to verify their
applicability in process design. More in detail, two models based on the implicit formulation were inves-
tigated; in one case, the billet material was set as porous object with the material density which was cal-
culated and updated as part of the simulation. The second implicit analysis, instead, was built using the
plastic material formulation; the porosity, in this case, was physically created introducing voids within
the workpiece.
The latter simulation class was carried out through an explicit investigation; an efficient model con-
struction was proposed introducing spherical surfaces connected each other with plans. Experimental
data were used to validate the calculated results and a discussion concerning the three different numer-
ical analyses was finally reported.
Ó 2008 Elsevier Ltd. All rights reserved.
0261-3069/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2008.06.051
F. Gagliardi et al. / Materials and Design 30 (2009) 1282–1287 1283
method to efficiently construct both 2D and 3D models that cap- 3. Numerical analyses
ture the main features of the actual experimental material. A
foamed titanium was investigated since applications are growing Three different numerical analyses were carried out in order to
in different industrial fields [20]. However, at the early stage of point out, first of all, the differences and their discrepancies from
foaming, when porosity is less than 25%, pores are mostly rounded, experimental results. In fact, each of them starts from different ba-
generally equiaxed, and un-merged [21]. Then, three-dimensional sics and, up to now, the best methodology is not yet described, also
(3D) pore size and pore distribution information were derived from because it is impossible to take into account only one performance
several 2D sections using the function of a microstructural distance index. However, these techniques are the most utilized in forming
distribution to characterize the spatial location of pores. Then, the process simulations. The simulations were set up taking into ac-
microstructure of the foam at high porosity was obtained by a geo- count the same input variables. The friction modelling was based
metric simulation of pore growth and movement during foaming on the constant shear hypothesis, fs ¼ mk [12]. In the above model
process. The built model was used to develop a numerical cam- fs is the frictional stress, k the shear yield stress and m is the shear
paign simulating uniaxial stress–strain tests. factor. The latter was set equal to 0.75 according to Manisekar et al.
In this work, experimental and numerical upsetting analyses [24], who investigated a cold compression process without any
were carried out on AlulightÒ aluminium foams; the numerical lubrication.
campaign was developed using both explicit and implicit formula- The three analyses used the following material flow rule,
tion. The first one was run through Ls-Dyna; a novel model con- r ¼ 210e0:2 MPa; such data were obtained through tensile tests
struction [22] was proposed introducing spherical surfaces that on specimens cut from Al6061 solid material. Finally, the mesh
represent the internal voids existing in a metallic foam. The code was modelled using approximately 200,000 elements; one fourth
Deform 3D, instead, was used for two different implicit studies. of the whole geometry was used due to symmetry conditions. On
In fact, beyond the workpiece definition in which the porous mate- the contrary, the upper and lower dies were modelled as rigid
rial model was used, a further analysis was carried out ‘‘subtract- tools. The peculiarities of different simulations are highlighted in
ing” the voids from the homogeneous material by Boolean the following.
operations. These voids, similarly to the explicit model, had differ-
ent dimensions according to the real spatial density distribution. 3.1. Implicit porous formulation
The latter one was rebuilt using an image processing analysis
(IPA) technique, proposed by some of the authors [13,23]. All the The yield condition proposed by Shima and Oyane [25] was uti-
numerical analyses were finally compared highlighting their points lized. This yield function depends on the first invariant of the stress
of strength and weakness related to predicting ability and CPU re- tensor, on the second invariant of the deviatoric stress tensor and
quired time. on a scalar parameter, which is the relative density, R; its formula-
tion is reported in Eq. (1):
2. Experimental campaign
1
A ½ðrx ry Þ2 þ ðry rz Þ2 þ ðrz rx Þ2 þ ðs2xy þ s2yz þ s2zx Þ
Compressive tests were performed to obtain experimental data 6
to be used for the validation of the proposed numerical analyses. A
þ 1 ðrx þ ry þ rx Þ2 ¼ dr20 ð1Þ
As above mentioned, an AlulightÒ foam was selected, based on 3
the AA6061 Aluminum alloy; the material presents an average
density equal to about 20% of the solid material. The foam was where A = 2 + R2, d = 2R21 and the strain hardening is given by the
compressed between two steel platens on an electronically con- aforementioned flow rule. It must be highlighted that the formula-
trolled hydraulic testing machine with a load cell able to provide tion of the porous material takes into account the possibility of
a maximum ram load equal to 100 kN. Circular billets were used growth or closure of the existing porosity. In other words, at the
for the upsetting tests; their initial thickness and diameter were, end of each deforming process step, the value of the relative den-
respectively 10 mm and 50 mm (Fig. 1a). The workpieces were ob- sity, calculated at the integration points within each element, is up-
_
dated taking into account that the relative density variation rate, R,
tained cutting along the billet axis cylindrical parts (Fig. 1b).
The upper die velocity was set equal to 1.0 mm/s to avoid any and the volumetric strain rate, e_ c , are linked by
dynamic effect during the forming sequence. Moreover, the test R_
was repeated five times in order to consider the effects due to e_ c ¼ ð2Þ
R
the material in-homogeneity. The crushing foam sequence was ob-
By integration of the equation, the updated relative density is calcu-
served and the ram load was acquired.
lated as
Z
R ¼ R0 exp e_ V dt ffi R0 ð1 DeV Þ; ð3Þ
where R0 is the relative density at the previous step and DeV is the
volumetric strain increment during the analyzed step.
Following the above considerations, it is easy to understand
that the porous materials formulation can be applied introducing
an initial relative density less than 100%.
For this analysis, Deform 3D was utilised; in this software, the
porous objects are treated as plastic objects (compressible rigid
viscoplastic materials) except that material density is calculated
and updated during the simulation.
The limiting strain rate and the flow stress must be specified at
the fully dense state. The material density is specified at each ele-
ment. Objects with changing material densities such as materials
Fig. 1. (a) The billet used for the upsetting test and (b) the initial cylindrical part. used in powder forming, should be modelled as porous objects.
1284 F. Gagliardi et al. / Materials and Design 30 (2009) 1282–1287
Fig. 4. (a) The single triplet of spheres and (b) their union to built the model.
Fig. 6. Upper die load for the different analyses for upsetting process.
Table 1
‘‘Time consuming” for the three investigated numerical analyses
However, for the case here analyzed, the porous model needs less
elements to correctly mesh the billet due to its simplicity. Never-
theless, the same number of tetrahedra was used for both implicit
analyses carried out through Deform 3D; this choice was suggested
with the aim of comparing the effective time difference between
the two numerical studies using the same initial variables. From
this point of view, even remeshing action has relevant influence;
in fact, the plastic model is characterized by irregular mesh that,
during forming simulation, needs more remeshing sequences to
complete the analyses. This aspect determines a consistent CPU
Fig. 8. The compressed billet: (a) experimental and (b) explicit model.
time increment with a substantial reduction of velocity in simula-
tion analyses.
It has to be highlighted, obviously, that the quality of the model
moved through Boolean operations, leading the initial billet totally building time and the CPU time is not the same: the latter, in fact,
holed. For the explicit simulation, instead, the model was defined is a continuous operation that can be run with few human dedi-
directly by the beads surfaces introducing another variable due cated resources. On the contrary, model building requires an active
to the definition of suitable planes; in fact, these ones have to guar- human work and, for this reason, sometimes a longer CPU time can
antee the material continuity and, obviously, their contribution on be preferred.
the total mass cannot be neglected.
However, moving the attention on the required simulation time 5. Conclusions
the scenario totally changes; the explicit simulation resulted to be
the faster as compared to both the implicit analyses. The behaviour of metallic foams under deformation is very dif-
This conclusion can be easily explained taking into account the ficult to simulate with the available numerical codes. This is due to
potentiality of the explicit formulation concerning the simulation the obtainable accuracy and the long CPU time depending on the
time reduction properly using mass scaling and time-step vs. chosen methodology. In the paper an interesting benchmark was
velocity control. carried out, not with the aim to select the best solution but to high-
Most onerous analysis was carried out using the porous formu- light all the aspects involved in the different options.
lation; this is in agreement with the above considerations concern- At the end, it is possible to state that a general purpose model is
ing the direct iteration method and its number of iterations trying not yet available because a good prediction of the process mechan-
to approach the converged step solution. In Table 1, the CPU time ics is possible but a proper model setup is required. Unfortunately,
for each numerical investigation is reported. In the case of porous the last issue can be expensive in terms of involved human
materials, the direct method is the only method currently available resources.
in Deform 3D [12]. This method does not have fast convergence The use of explicit codes currently is the right way but more ef-
capabilities, and consequently, using the same number of ele- fort have to be spent performing automated procedures able to
ments, a porous simulation is longer than a simple plastic analysis. build reliable models in a reasonable time.
Fig. 9. Implicit plastic simulation before and after upsetting process: (a) internal and (b) external side of the billet.
F. Gagliardi et al. / Materials and Design 30 (2009) 1282–1287 1287
Acknowledgements [11] Chen D-C. Rigid-plastic finite element analysis of plastic deformation of
porous metal sheets containing internal void defects. J Mater Process Technol
2006;180:193–200.
This research is funded by Italian Ministry of University and Re- [12] Deform-TM 3D V5.0.3-user manual.. Columbus, OH, USA; 2006.USA: DeformTM,
search (MUR). Scientific Forming Technologies Corporation; 2006.
[13] Filice L, Gagliardi F, Shivpuri R, Umbrello D. Numerical simulation of foam
The authors would like to thank Mr. F. Pulice and M. Risoli for
forging to reproduce biomechanical components. In: Proceedings of ninth
their contribution to the development of this work and Prof. L. Numiform, Porto, Portugal 2007; p. 1225–31.
Fratini from University of Palermo for his kind support in FEA. [14] Contorno D, Filice L, Fratini L, Micari F. Forming of aluminum foam sandwich
panels: numerical simulations and experimental tests. J Mater Process Technol
2006;177:364–7.
References [15] Filice L, Fratini L, Umbrello D. On the finite element simulation of secondary
operations on metallic foams. Mater Sci Forum 2007:539–43.
[1] Simone AE, Gibson LJ. Aluminum foam produced by liquid-state process. Acta [16] Meguid SA, Cheon SS, El-Abbasi N. FE modelling of deformation localization in
Mater 1998;46:3109–23. metallic foams. Finite Elem Anal Des 2002;38:631–43.
[2] Sugimura Y, Meyer J, He MY, Bart-Smith H, Grenstedt J, Evans AG. On the [17] Czekanski A, Elbestawi MA, Meguid SA. On the FE modeling of closed-cell
mechanical performance of closed cell AL alloy foams. Acta Mater aluminum foam. Int J Mech Mater Des 2005;2:23–34.
1997;45:5245–59. [18] Shen H, Brinson LC. Finite element modeling of porous titanium. Int J Solids
[3] Banhart J. Manufacture, characterisation and application of cellular metals and Struct 2007;44:320–35.
metal foams. Prog Mater Sci 2001;46:559–632. [19] Shen H, Oppenheimer SM, Dunand DC, Brinson LC. Numerical modeling of pore
[4] Banhart J, Baumeister J. Deformation characteristics of metal foams. J Mater Sci size and distribution in foamed titanium. Mech Mater 2006;38:933–44.
1998;33:1431–40. [20] Salimon A, Brechet Y. Potential applications for steel and titanium metal
[5] Kenesei P, Kàdàr CS, Rajkovits ZS, Lendvai J. The influence of cell-size foams. J Mater Sci 2005;40:5793–9.
distribution on the plastic deformation in metal foams. Scripta Mater [21] Murray NGD, Dunand DC. Microstructure evolution during solid-state foaming
2004;50:295–300. of titanium. Compos Sci Technol 2003;63:2311–6.
[6] Benouali AH, Froyen L, Delerue JF, Wevers M. Mechanical analysis and [22] Gagliardi F, Filice L, Umbrello D. Simulation of aluminium foam behavior in
microstructural characterisation of metal foams. Mater Sci Technol compression tests. Submitted to AMPT 2008.
2002;18:489–94. [23] Gagliardi F, Filice L, Umbrello D, Shivpuri R. Forging of metallic foams to
[7] Körner C, Singer RF. Processing of metal foams – challenges and opportunities. reproduce biomechanical components. Mater Sci Eng A 2008;480(1–2):510–6.
Adv Eng Mater 2000;2:159–65. [24] Manisekar K, Narayanasamy R, Malayappan S. Effect of friction on barrelling in
[8] Reyes A, Hopperstad OS, Berstad T, Hanssen AG, Langseth M. Constitutive square billets of aluminium during cold upset forging. Mater Des
modeling of aluminum foam including fracture and statistical variation of 2006;27:147–55.
density. Eur J Mech A/Solids 2003;22:815–35. [25] Shima S, Oyane O. Plasticity theory for porous metals. Int J Mech Sci
[9] Lopatnikov SL, Gama BA, Haque MJ, Krauthauser C, Gillespie Jr JW, Guden M, 1976;18:285–91.
et al. Dynamics of metal foam deformation during Taylor cylinder–Hopkinson [26] LS-DYNA keyword user’s manual. Livermore, California, USA: Livermore
bar impact experiment. Compos Struct 2003;61:61–71. Software Technology Corporation; 2003.
[10] Hanssen AG, Hopperstad OS, Langseth M, Ilstad H. Validation of constitutive
models applicable to aluminium foams. Int J Mech Sci 2002;44:359–406.