Composite Structures 353 (2025) 118731

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Numerical investigation of composite metal foams using

substructure techniques
Márió Kovács a,* , Balázs Pere a , Imre Norbert Orbulov b
a
Department of Applied Mechanics, Faculty of Mechanical, Computer and Electrical Engineering, Széchenyi István University, Egyetem tér 1., Győr 9026, Hungary
b
Department of Materials Science and Engineering, Faculty of Mechanical Engineering, Budapest University of Technology and Economics, Műegyetem rakpart 3,
Budapest 1111, Hungary

A R T I C L E I N F O A B S T R A C T

Keywords: An accurate description of the mechanical behaviour of inhomogeneous, anisotropic composite materials, such as
Composite metal foams metal matrix syntactic foams (MMSFs), is a challenging task. This paper is based on the so-called substructure
Porous materials technique, which has never been used for the mechanical modelling of metal foams or MMSFs. Using this
Substructure technique
method, a finite element model with low computational requirements but still modelling inhomogeneous
FEM
properties can be constructed that reproduces the results of laboratory measurements with high accuracy. In all
cases, the tests presented in this paper are limited to the linearly elastic phase. As the most common investigation
of MMSFs is the compression test, this study considers only this loading case. A more detailed discussion of the
stress analysis of MMSFs is given at the end of this work. A potential cause of 45◦ crack planes during
compression tests and the evolution of stresses as a function of space-filling are also investigated.

1. Introduction the modification of material pairs or the geometric configuration of

hollow spheres [3,4]. Filling of hollow structures with metal foam is also
As materials have to meet more and more complex criteria, they often used to significantly improve the energy absorption capacity of the
often do not satisfy the requirements on their own. The consequence of structural elements, but only to increase their mass by small amounts
this is that the field of application of composite materials is constantly [5]. MMSFs have a wide range of practical applications. They are
expanding. This process has also resulted in the spread of metal matrix already increasingly used in buildings, vehicles and aircrafts. This
syntactic foams (MMSFs), which, due to their composite nature, offer spread is facilitated by the fact that they also have good sound-absorbing
better mechanical properties than conventional metal foams. The com- properties [6]. A novel application of composite metal foams is their use
posite structure of MMSFs is provided by the fact that the foam texture is as armour in sandwich structures [7,8]. The properties of metal foams
obtained by the filling of hollow spheres of a different material from the and MMSFs are usually investigated by static compression testing, which
matrix material instead of gas bubbles. These hollow spheres vary will be used numerically and for laboratory measurements in this work
greatly in diameter and wall thickness. Their diameter can range from [9].
10 mm to a few tenths of millimetres. The wall thickness is determined The process of degradation of metal foams and MMSFs is often
by the material and manufacturing technology. In ceramic spheres, for monitored by X-ray imaging to visualise the internal deformations. In
example, the wall thickness is only 2–3 % of the diameter, so these this way, the interior of composite foams can be examined step by step
hollow spheres can have extremely thin walls. Since the filling is always under different loading conditions [10–12]. Using an X-ray-based 3D
done with hollow spheres whose stiffness is higher than that of the modelling process, it is possible to make a digital representation of the
matrix material, this significantly increases the overall stiffness of the irregular internal structure. This 3D CAD model can be used for finite
metal foams [1,2]. As the stiffness of MMSFs can be increased, the aim is element analysis. The advantage of this method is that the properties of a
always to fill them with hollow spheres in the highest possible ratio in properly prepared finite element model are very similar to those of the
order to improve the mechanical properties. There are a growing num- test specimens used for laboratory measurements. However, it has the
ber of studies on the factors affecting the properties of MMSFs, such as major disadvantage that even for small foam models, a large investment

* Corresponding author at: Department of Applied Mechanics, Faculty of Mechanical, Computer and Electrical Engineering, Széchenyi István University, Egyetem
tér 1., Győr 9026, Hungary.
E-mail address: [email protected] (M. Kovács).

https://doi.org/10.1016/j.compstruct.2024.118731
Received 5 June 2024; Received in revised form 4 October 2024; Accepted 18 November 2024
Available online 20 November 2024
0263-8223/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
M. Kovács et al. Composite Structures 353 (2025) 118731

of time is required to prepare and correct errors. In addition, despite the shown in Fig. 1.
small size, this method results in a model with a high computational The figure above shows the random distribution of hollow spheres in
complexity due to the large number of degrees of freedom (DoF) the transparent matrix material. In order to understand this novel
[13–15]. methodology, the substructure technique is first introduced in general
Another approach to studying the mechanical properties of MMSFs is and then a specific adaptation for modelling composite metal foams is
homogenization techniques. There are several different approaches. The presented.
simplest is when generalised material properties are defined, as if the
internal structure were homogeneous. Another method is to build the 2.1. Basics of substructure technique
model by repeating representative volume elements (RVEs). This allows
faster and significantly simpler models to be built with less computa- The substructure technique is based on the discretized equilibrium
tional requirements, but in many cases the accuracy of these finite equation in case of small displacements:
element models is less than that of the X-ray-based method [16–19].
One of the main aims of this study is to combine the advantages of Ku = f (1)
the two methods described above and to tackle their disadvantages. To
achieve this goal, to the Authors’ knowledge, a never-before-used where K is the stiffness matrix of the structure, f is the load vector and u
technique for modelling metal foams and MMSFs, namely the sub- is the unknown nodal displacement vector to be determined. In addition,
structure technique, is used. This is a technique that was invented in the the method also requires the definition of boundary conditions. The
1970 s, but until now it has been used to solve completely different basic idea behind the substructure technique is to separate the finite
problems. Typically, it has been used for vibration testing of large element model into nodes of interest that are usually on the surface of
structures, including aircrafts [20,21]. Subsequently, it has also been the body (boundary points) and nodes that are inside the body (internal
increasingly used for structural analysis of bridges and large buildings points). After that, Eq. (1) can be partitioned according to internal and
[22]. This technique has been largely combined with the finite element boundary points:
method (FEM) and model updating methodology [23–25]. [
Kii Kib
][ ] [ ]
ui fi
Due to the time step used in the model updating technique often = (2)
Kbi Kbb ub fb
being small, the reduction of the DoF is especially important to reduce
the calculation time and therefore the substructure technique has come where Kii is the stiffness matrix concerning the interior points, Kib and
to the fore [26–28]. It is essential to note, however, that the substructure Kbi describe the connection between the boundary and the interior
technique also needs to be adapted slightly for each application [29]. points, while Kbb is the stiffness matrix concerning the boundary points.
As mentioned above, several studies have investigated how the In addition, ub is the boundary and ui is the interior column vector of the
properties of MMSFs are affected by the geometry of the hollow spheres displacements and f b and f i denotes the load vectors of the boundary and
and their filling percentage. For this reason, this study will include the interior points. Since this paper deals with a quasi-static compression
effects of hollow spheres on each other and on the matrix material. The test, no volumetric load is applied, so f i equal to zero. When the
mechanical properties and meshing criteria of hollow spheres by multiplication of the matrix equation (2) is done, it results in the
themselves have already been dealt with before, but here it is not following equations:
enough to examine one piece by itself [30–32]. In many cases, the mesh
is densified at the contact points of spheres, but this is not possible for a Kii ui +Kib ub = f i = 0 (3)
large foam model because the number of DoF would be increased
enormously due to the large number of spheres [33]. It is important to Kbi ui +Kbb ub = f b (4)
note that the matrix material plays an essential role, as it has a large
An important assumption in this method is that the displacements of the
influence on the variation of the stress, so considering only the hollow
interior points are excluded and only the displacements of the boundary
spheres would be inaccurate [34]. In many cases, the effect of changing
points are taken into account. If Eq. (3) and Eq. (4) are combined and the
a single geometric property is investigated, but the effect of several
displacements of the boundary points are derived from them, the result
different parameters, such as the thickness of the walls of the hollow
is:
spheres or the distance between them are less studied [35]. The effects of
parameters can best be traced for models where the DoF is not very high − 1
(Kbb − Kbi K−ii 1 Kib ) f b = ub (5)
due to the size of the whole structure. However, those models must still
contain more hollow spheres [36,37]. Eq. (5) allows to define only the displacements of the boundary points.
This work has two main aims. One main goal is to use the sub- This is particularly useful in the present case, because the deformations
structure technique to construct a model of tetrahedral elements with
low computational requirements that accurately models the properties
of MMSFs [38–40]. The tetrahedral elements were chosen for easy nu-
merical integration [41,42]. This method has been previously discussed
by the author in the two-dimensional case [43]. In this work, the linearly
elastic phase is investigated.
The other part of the paper examines the combined effect of geo-
metric size, spacing, and material of the hollow spheres on stress, which
cannot be investigated under laboratory conditions.

2. The substructure technique and its application to composite

metal foams

This paper is based on the methodology of the substructure tech-

nique, but in a different way from its common use. The application of
this technique to MMSFs has not existed before. Understanding the
complex internal structure of composite metal foams is essential for the
further parts of this paper. A sectional view of a model of an MMSF is Fig. 1. A sectional view of a model of an MMSF.

2
M. Kovács et al. Composite Structures 353 (2025) 118731

can only be detected on the outer surfaces in the real measurement. A

basic example of this simplification is shown in Fig. 2. Structures
simplified in this way are called super finite elements (SFEs). For better
visualisation the visibility of some elements has been turned off, so it is
possible to see the difference inside.
It is important to note that the properties of the SFE and the initial
model are 100 % identical since only equation ordering is done. It is not
very obvious in this example, but for a densely meshed model, this
simplification leads to a considerable reduction in memory requirements
[43].

2.2. Integration of composite metal foams and substructure technique

The substructure technique is usually used for the modal analysis of

large structural elements, but in this paper it is used differently. The
basic idea is that the SFEs described in the previous section are prepared
in such a way that each of them contains a carefully meshed foam
structure. An example of this is shown in Fig. 3, where a foam model is
shown in a sectional view. The 3D tetrahedral elements of the matrix
material are removed to better visualize the hollow spheres.
The method involves making a few of these models, which have
small enclosure sizes but different internal layouts. Then, each model is Fig. 3. MMSF cubes prepared for SFE.
converted into an SFE. Afterwards, the transformation of each finite
element model into an SFE is performed, so that the structure with
− 1
inhomogeneous properties can be simplified to mesh cubic surfaces − 1 − 1 − 1
(K1bb + K2bb + ⋯KNbb − K1bi K1ii K1ib − K2bi K2ii K2ib − ⋯KNbi KNii KNib )
only. A simple example of this method is shown in Fig. 2. Subsequently, (7)
(f 1b + f 2b + ⋯f Nb ) =ub
large-scale models with inhomogeneous properties can be constructed
from SFEs with a significantly reduced number of DoF compared to the
However, the number of loads does not necessarily have to be the same
original model. It follows that it is important that the SFEs must be
as the number of SFEs, as they may not all be loaded. Eq. (7) can be
coincident at the edges so that their nodes are common at the join. The
described more concisely as follows if the repeated additions are
inhomogeneity is increased by the fact that some of the SFEs have
replaced by sums:
different properties because their interiors were not initially identical. If
( )− 1 ( )
N is the number of SFEs and N≫1, then Eq. (2) can be written as: ∑N ( k )− 1 k ∑M r
⎡ ⎤⎡ ⎤ ⎡ ⎤ k=1
k k
(Kbb − Kbi Kii Kib ) f
r=1 b
= ub (8)
⎢ K1ii 0 ⋯ 0 K1ib ⎥⎢ u1i ⎥ ⎢ f 1i ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ 0
⎢ K2ii ⋱ ⋮ K2ib ⎥⎢ u2 ⎥ ⎢
⎥⎢ i ⎥ ⎢ f 2i ⎥
⎥ where M denotes the number of SFEs that are under load. If it is accepted
⎢
⎢ ⋮ ⋱ ⋱ 0 ⋮
⎥⎢ ⎥ ⎢
⎥⎢ ⋮ ⎥ = ⎢ ⋮
⎥
⎥ that there exists a solution to the original matrix equations (1) and only
⎢ ⎥ ⎢ ⎥
⎢ 0 ⋯ 0 KNii KNib
⎢ ⎥
⎥⎢ uN ⎥ ⎢ fiN ⎥ identical transformations are applied to this equation, then the solution
⎢ ⎥⎣ i ⎦ ⎢ ⎥
⎣ 1 ⎦ ⎣ N⎦ to the new system of equations (8) obtained must match the subset of
Kbi K2bi ⋯ KNbi K1bb + K2bb + ⋯KNbb ub 1 2
f b + f b + ⋯f b
solutions to the original system of equations. In Eq. (8) it may seem that
(6) the matrix operations behind the first sum must be performed on all
SFEs, which would be true if all the SFEs of all the foam cubes were
Eq. (6) above is the starting point for the description of this new different. However, in our method, only a few SFEs with different in-
methodology. Each step from here on describes this unique technique in ternal properties are needed and they are randomly assembled during
mathematical form, developed by the Authors. Performing the matrix the construction of the whole structure. For this reason, this matrix
multiplications and expressing the displacement vector of the boundary operation behind the first sum only needs to be performed as many times
points from the equation system results in a generalized form of Eq. (5): as many different SFEs are there. The results of these matrix operations
only need to be stored in memory and loaded only when the sum of the
SFE of the given type is about to occur. So, instead of the large number of
SFEs needed to build the whole model, only the few stored ones need to
be used randomly. The substructure technique uses the same system of
equations as in the conventional finite element methods, except that the
system of equations is rearranged. It follows that its accuracy is equiv-
alent to the conventional method. Furthermore, there is no need to
consider the converse, because this would have to be done in the same
way as in the conventional FEM. A simple two-dimensional example of
this methodology is shown in the schematic Fig. 4. It shows that a model
of 25 cells can be created by randomly assembling 9 different sub-
structures. Thus, only 9 substructures need to be stored in memory
instead of 25, and inhomogeneity can be ensured. In this case, each
different number represents a distinct type.
As a result of Eq. (8), the boundary displacement field of the entire
model assembled from the SFEs can be calculated. Thus, the internal
displacements of the SFEs are not known. A cardinal issue in using this
method is how many different substructures are needed and what their
Fig. 2. The meshed model and its SFE.

3
M. Kovács et al. Composite Structures 353 (2025) 118731

The second strongly influencing factor is the space-filling percentage

of the hollow spheres in the overall model. A space-filling rate of 50 %
was chosen, as this is a frequently used percentage.

3.1. Numerical example for a compression test with substructure

technique

The meshing software GMSH [44] was used to create the MMSF
models. In an experiment to verify the accuracy of the numerical
method, the MMSF specimen was compressed between two rigid plates.
An aluminium foil and lubricant were placed between the pressure
plates and the metal foam to avoid galling. The moving pressure plate
was modelled as a y − axial kinematic load. The plate at rest was
modelled as an elastic support where the y − axial displacement is
limited by a high stiffness value. The free displacement in the x and z
directions was allowed by the near-zero values of the stiffness matrix in
these directions. These boundary conditions are shown in the schematic
illustration in Fig. 5.
Since the kinematic load requires a specific load vector, the sum of
these vectors at each predetermined displacement value gives the force
that the pressure plates have to exert on the foam model. This force
value will be important for further comparison with laboratory results.
An example of a loading case on a finite element model is shown on the
left-hand side of Fig. 6. In the model assembled from the SFEs, one of the
Fig. 4. A schematic representation of the use of substructure techniques side plates was removed to show that the internal nodes no longer play a
for MMSFs. role in the displacement calculation. However, it can be seen that the
interpolated displacement field is different from the homogeneous ma-
inclusion sizes should be. But since inhomogeneity is not a quantifiable terial. Any arbitrarily selected substructure can be converted back into a
parameter, these cannot be precisely determined. This can also be seen conventional, solid structure. This is calculated by applying Eq. (9) to
from the fact that there is an extraordinary variety of MMSFs, since the the boundary displacement of the SFE. This gives the detailed
number of material pairs, hollow spheres, wall thicknesses and ratios in displacement field to the right-hand side of Fig. 6. However, this should
the matrix allow an infinite number of combinations. not be used too often, as it will significantly reduce the effectiveness of
Another approach could be to build the whole domain as one large the method. It is important to note that the substructure technique re-
model and include it in a single SFE. This would not be advantageous quires even more care in the preparation of the finite element mesh than
because then the meshing of a complex internal structure has to be re- a conventional model. This is because if an overly distorted element is
investigated and this leads back to the problem of finite element included in a substructure, it will cause inaccuracy as many times as that
model building based on X-ray imaging. substructure is built into the overall model.
Regardless of which method is chosen, only the displacements of the The efficiency of the method can be significantly increased if the
external points will be known. However, going back to Eq. (3), it can be point density at the boundaries of the SFEs is lower than inside the
seen that if the displacements of the boundary points are known, then substructures. At this point, it is useful to make a comparison between
after a rearrangement a formula can be obtained which can be used to three modelling aspects, which highlights the memory efficiency of the
determine the internal displacements of an arbitrary SFE: model on the left-hand side of Fig. 6. The first and most general method
( )− 1 has already been discussed, which involves a full meshing of the X-ray
− Krii Krib urb = uri (9)

where r is the number of the selected SFE and uri donates the internal
displacement field.

3. Numerical example and experimental investigation

This study shows the memory-efficient modelling of MMSFs. How-

ever, a detailed study has already been published by the authors of this
article, in which the memory efficiency of the method has been
demonstrated in detail [43]. It is necessary to determine the factors that
influence the mechanical properties. First, the materials used in the
simulations and experiments need to be defined. In this part of the
paper, a pair of pure aluminium matrix material and iron hollow spheres
material was selected because it is a common material pairing. Their
properties are summarised in Table 1.

Table 1
Elastic properties of the filler and the matrix.
iron aluminium
Fig. 5. Schematic representation of boundary conditions where the kinematic
Modulus on elasticity [GPa] 19.6 71.0
load is represented by the blue arrows and the elastic support is indicated by the
Poisson’s ratio 0.250 0.334
arrows in red.

4
M. Kovács et al. Composite Structures 353 (2025) 118731

Fig. 6. The complete model built from SFEs and the back-transformed displacement field of a selected SFE.

based model. In this case, there is not much possibility to increase the measuring 1000 hollow spheres on an Olympus SZX 16 type stereo
size of the finite elements to a large extent. For the case where the model microscope). The set of the MHSs was fixed by a stainless steel net at the
shown in Fig. 6 is not built using the substructure technique, but in the top of the assembly and the whole assembly was pre-heated to 300 ◦ C.
conventional way and the element size remains the same as shown on Subsequently and molten Al99.5 alloy (at 710 ◦ C) was poured on the top
the walls of the hollow spheres, the number of DoF is 6529008. When of the MHSs set. Ar gas was applied to force the molten matrix between
the substructure technique is used but the element size is not reduced the individual hollow spheres. After the casting and initial solidification,
towards the boundaries, this number is 1835181. This already repre- the whole structure was water cooled to room temperature. Afterwards,
sents a reduction of 71.9 %. However, if the element size increases to- the cast block was removed from the mould and the block was machined
wards the boundaries, as shown in Fig. 6, the number of DoF is only into 10.5 × 10.5 × 10.5 mm cubes for compressive tests. For further
24384. This represents a reduction of 99.63 % compared to the tradi- details about the production of MMSFs please refer to [45].
tional method, which already shows the inherent potential of the The compressive tests were conducted between flat plates in a
method. On the detailed displacement field of an SFE, it is already compression tool with four guide bars. The samples were lubricated by
possible to calculate a stress field on each pre-selected cubic region of Teflon foils to reduce the friction between the compressive plates. The
the entire model. An example of this is shown in Fig. 7, where Von Mises crosshead speed was set to 0.1 [mm/min] to ensure quasi-static condi-
stress is displayed. tion. During the measurement, the force and the displacement were
High-stress zones at 45◦ to the sides of the substructure can be registered by a 250 kN force cell and an extensometer. The force-
observed in the stress field. This phenomenon can be considered as one –displacement curves were transformed into engineering stress- engi-
of the reasons why the slip planes of the MMSFs appear at 45◦ . A more neering strain plots. The tests were conducted along the ISO13314
detailed understanding of stresses in MMSFs will be discussed in Section standard. During the finite element modelling, the material parameters
4. from Table 1 have been applied. Seven specimens were subjected to the
compression test, the curves of them are shown in Fig. 8. The complete
evaluation of the compressive results and the failure mechanisms of the
3.2. Experimental validation samples have been published in [45].
It can be seen that the curves differ from each other, which is a
For the validation of the numerical results, compression tests of the consequence of the inhomogeneous internal structure. Since the mea-
MMSF specimens with the same volume as the finite element model surement points do not coincide at all locations during the measure-
were performed. The MMSF samples were produced by low-pressure ment, the values of each curve must be determined by linear
infiltration. The metallic hollow spheres (MHSs, pure Fe in composi- interpolation at the same point on the x-axis. Afterwards, the data can be
tion, supplied by Hollomet GmbH, Dresden, Germany) were loosely used to construct the average pressure curve. This is shown in Fig. 9.
poured (ensuring ~ 50 vol% filling) into a steel mould coated by a However, this curve alone would not be enough to compare with the
graphite layer (N-77 graphite spray by DUE-CI ELECTRONIC s.n.c., results of the finite element model, so the standard deviation of the mean
Avesa, Italy) to prevent reactions between the matrix material and the values must be determined. This standard deviation becomes the one
mould. The nominal diameter of the hollow spheres was 1.92 ± 0.07 sigma value. Two times of this sigma has been selected for comparison.
mm, while their average wall thickness was 23 ± 0.6 µm (obtained by

Fig. 7. Stress field defined from the displacement field.

5
M. Kovács et al. Composite Structures 353 (2025) 118731

4. Analysis of the distance between hollow spheres on the

internal stress

For composite metal foams, there are only limited possibilities to

monitor the internal processes under load between the hollow spheres.
The stress changes in the matrix material and their influencing factors
can be observed only to a very limited extent or not at all under labo-
ratory conditions. However, the FEM offers the possibility to observe the
stress in the matrix material in detail. The model that was used was
made out of only four hollow spheres and loaded with compressive stress
in the direction of y-axis. Since the properties of MMSFs are most
influenced by the materials used and the percentage of space-filling,
these parameters were also varied during the investigation. The matrix
material was kept unchanged throughout, pure aluminium, because the
Fig. 8. Pressure curves of the test specimens. stress changes of the matrix material were monitored. The effect of the
wall thickness of hollow spheres has already been investigated in pre-
vious studies [34,35], so this parameter is not considered in this paper. A
cross-sectional view of the finite element model subjected to compres-
sive loading is shown in Fig. 11.
In this study, the dispersion of hollow spheres is not discussed
because, on the one hand, an ideal distribution is considered in this
section, so that the spheres are equally spaced. On the other hand, they
are so positioned in order to allow a more detailed study of the stress
zones created at 45◦ to the load direction, as shown in section 3.1. Since
this finite element analysis focuses on the matrix material, characteristic
parts have been selected that are also observed in the model presented in
section 3.1. An example is the region between the spherical shells at 45◦

Fig. 9. Average pressure curve.

This value is chosen because at one sigma confidence level there is a 33

% chance of excluding good data, but at two sigma it is only 5 %. This
means that there is a 95 % chance that values with a large error will be
excluded from the range. Thus, statistical considerations were decided
for the two sigma. Adding this value to the average curve in both the
positive and negative directions allows for the enveloping curves to be
fitted. These curves are shown highlighted in blue in Fig. 10. It is
important to note that the study deals with the linearly elastic phase, so
only this phase will be relevant.
The right-hand side of Fig. 10 shows a numerical approach between
the envelopes. This means that at the significance level of two sigma, the
measured data verify the numerical model at the linear interval. The use
of a Savitzky-Golay filter on the raw, initial curves for several different
parameters (polynomial degree, frame length) did not result in any
improvement in the accuracy of the linear section approximation, and
therefore no filtering was applied to the data.
Fig. 11. Cross-sectional view of a finite element model under load.

Fig. 10. Comparison between the numerical approximation and the measurements.

6
M. Kovács et al. Composite Structures 353 (2025) 118731

plane to the load. The centre of such a region is shown by point P1 (see
Fig. 11). These are the regions with the highest stress in the matrix
material, which may be one reason for the frequent 45◦ slip planes in
MMSFs. The other matrix area under consideration is the scenario where
a volume of matrix material is under investigation between two hollow
spheres, and these hollow spheres are aligned with the load. The centre
of this region is shown by point P2 (see Fig. 11). Such zones are at that
point where the stress is always lowest. By increasing the distances t1
and t2 under constant load, the change in the stresses at points P1 and P2
was observed. The change in these distances represents a variation in the
space-filling percentage. Since a stress value by itself is meaningless, it
will be compared to a point near the boundary of the model at which the
stress is no longer affected by the hollow spheres. This point is P3. In
each step, when the distance t1 and t2 was increased, the stress values for
both ceramic and iron hollow spherical shells were investigated. The Fig. 12. Stress difference between P1 and P3 as a function of increase in dis-
properties of iron and matrix material are summarised in Table 1. tance t1.
Commonly used materials were chosen as the ceramic shell, the
composition of which is summarised in Table 2:
The properties of the mixed ceramics were determined by a simple
mixing rule. The modelled hollow spheres have a radius of 0.96 mm and
a wall thickness of 0.023 mm. It would not be necessary to specify these
dimensions, because if they change proportionally, the diagrams would
not change. The reason for specifying these values is to be able to assign
a unit of measurement to the axes. In the simulation, the distance be-
tween the hollow spheres was increased until the stress at reference
point P3 was sufficiently close to the stress of the points P1 and P2.
Therefore, from this point onwards, the difference between the reference
point and the point P1 or P2 must be observed. Fig. 12 shows how the
stress difference between points 1 and 3 changes with increasing dis-
tance t1.
The distance t1 started from 0.02 mm and was increased until the
stress at point P1 was nearly equal to the reference point. It can be seen
Fig. 13. Stress difference between P3 and P2 as a function of increase in dis-
that iron and ceramic hollow spheres result in nearly equal stresses, so
tance t2.
this factor does not have a large effect on the stress of the matrix ma-
terial. At the beginning, even with a small increase in the distance t1, the
stress difference decreases steeply up to 0.2 mm. It can be concluded that on the rearrangement of elements of matrices already existing in the
reducing the high space-filling percentage causes a significant change in conventional case and on the rearrangement of equations, no additional
the properties of the composite foam, which is in agreement with the numerical error is added to the calculation. In Section 4 of the paper, an
general experience. After a distance of 0.2 mm for both types of hollow additional analysis of the evolution of internal stresses in relation to
spheres, the stress difference shows a degressive characteristic after a space-filling percentage is also presented. Other factors, such as the wall
slight increase. However, at point P2 the distance, t2 had to be signifi- thickness of hollow spheres or the dispersion of spheres, were either not
cantly greater to reach the stress at the reference point. Fig. 13 shows mentioned because they have already been discussed in previous
how the stress difference between points 3 and 2 changes with studies, or because the specific location of hollow spheres had to be fixed
increasing distance t2. in order to observe specific stress zones. The following conclusions can
In this case, there is also no significant difference between hollow be drawn from the literature review and the results presented in this
spheres of different materials. It can be seen that the stress reduction paper:
effect between hollow spheres oriented in the direction of the load is also
detectable at larger distances. Therefore, this phenomenon is not as 1. This specific application of the substructure technique has proven to
affected by the space-filling percentage as in the previous case. be a tool that, unlike X-ray-based techniques, can be used to easily
build inhomogeneous models of arbitrarily large dimensions. All that
5. Conclusions is needed is a few substructures with different internal structures and
their random assembly on the edges.
The aim of this study is to determine the linear elastic behaviour of 2. The experimental results have verified the numerical approximation
MMSFs under external loading using the substructure technique. This of the pressure diagram of MMSFs using substructure techniques at
special finite element technique has never been used before to model the quasi-linear phase at the two sigma significance level.
composite metal foams. It allows the creation of a model with low 3. From the numerical internal stress analysis, it was found that the
computational and memory requirements, which incorporates the variation of the space-filling has a large effect on the matrix material
properties of the complex internal structure. Since this method is based between the hollow spheres for which the line connecting the centres
of the shells makes an angle of 45◦ with the direction of loading. For
those hollow spheres where this is 0◦ , this effect is significantly
Table 2 smaller.
Amount and elastic properties of ceramic hollow spheres’ components.
Mass Percentage [%] Modulus on elasticity [GPa] Poisson’s ratio The substructure technique has been demonstrated for MMSFs in this
Al2O3 33 38.0 0.22
study, but could be extended to other inhomogeneous anisotropic ma-
Mullit 19 12.7 0.24 terials in the following studies. A further aim will be to extend the
SiO2 48 7.3 0.17 substructure technique to the characterisation of non-linear behaviour

7
M. Kovács et al. Composite Structures 353 (2025) 118731

(e.g. plastic deformations). In addition, it may be useful to further [15] Heitor D, Duarte I, Dias-de-Oliveira J. Aluminium alloy foam modelling and
prediction of elastic properties using X-ray microcomputed tomography. Metals
develop the method for dynamic simulations.
2021;11:925.
[16] Bardella L, Genna F. On the elastic behavior of syntactic foams. Int J Solids Struct
CRediT authorship contribution statement 2001;38:7235–60.
[17] Marur PR. Effective elastic moduli of syntactic foams. Mater Lett 2005;59:1954–7.
[18] Jhaver R, Tippur H. Processing, compression response and finite element modeling
Márió Kovács: Writing – original draft, Software, Methodology, of syntactic foam based interpenetrating phase composite (IPC). Mater Sci Eng A
Conceptualization. Balázs Pere: Writing – review & editing, Supervi- 2009;499:507–17.
sion, Formal analysis. Imre Norbert Orbulov: Visualization, Validation, [19] Koudelka P, Doktor T, Valach J, Kytýř D, Jiroušek O. Effective elastic moduli of
closed-cell aluminium foams–homogenization method. 11th IMEKO TC15 Youth
Investigation. Symposium on Experimental Solid Mechanics 20122012. p. 244-50.
[20] Przemieniecki JS. Matrix structural analysis of substructures. AIAA J 1963;1:
Declaration of competing interest 138–47.
[21] Craig Jr R. Substructure methods in vibration. 1995.
[22] Li J, Hao H. Numerical study of structural progressive collapse using substructure
The authors declare that they have no known competing financial technique. Eng Struct 2013;52:101–13.
interests or personal relationships that could have appeared to influence [23] Balmès E. Parametric families of reduced finite element models. Theory and
applications. Mech Syst Sig Process 1996;10:381–94.
the work reported in this paper. [24] Papadimitriou C, Papadioti D-C. Component mode synthesis techniques for finite
element model updating. Comput Struct 2013;126:15–28.
Acknowledgements [25] Dorosti M, Fey R, Heertjes M, van de Wal M, Nijmeijer H. Finite element model
reduction and model updating of structures for control. IFAC Proceedings Volumes
2014;47:4517–22.
This work was partially supported by the National Research, [26] Krysl P, Lall S, Marsden JE. Dimensional model reduction in non-linear finite
Development and Innovation Office (NKFIH), under grant agreement element dynamics of solids and structures. Int J Numer Meth Eng 2001;51:
OTKA-FK_21 138505. 479–504.
[27] Weng S, Xia Y, Xu Y-L, Zhu H-P. Substructure based approach to finite element
model updating. Comput Struct 2011;89:772–82.
Data availability [28] Chung E, Efendiev Y, Hou TY. Adaptive multiscale model reduction with
generalized multiscale finite element methods. J Comput Phys 2016;320:69–95.
[29] Ghannadi P, Kourehli SS. Investigation of the accuracy of different finite element
Data will be made available on request. model reduction techniques. Structural Monitoring and Maintenance 2018;5:
417–28.
References [30] Svensson SL. Finite elements on the sphere. J Approx Theory 1984;40:246–60.
[31] Lim T-J, Smith B, McDowell D. Behavior of a random hollow sphere metal foam.
Acta Mater 2002;50:2867–79.
[1] Movahedi N, Murch GE, Belova IV, Fiedler T. Functionally graded metal syntactic
[32] Shorter R, Smith J, Coveney V, Busfield J. Axial compression of hollow elastic
foam: Fabrication and mechanical properties. Mater Des 2019;168:107652.
spheres. J Mech Mater Struct 2010;5:693–705.
[2] Wouterson EM, Boey FY, Hu X, Wong S-C. Specific properties and fracture
[33] Taalebinezhaad M, Saka N. Elastoplastic Finite Element Analysis of Spheres in
toughness of syntactic foam: Effect of foam microstructures. Compos Sci Technol
Contact. Key Eng Mater 1989;33:213.
2005;65:1840–50.
[34] Demirtas A, Klinzing GR. Understanding die compaction of hollow spheres using
[3] Gladysz G, Perry B, McEachen G, Lula J. Three-phase syntactic foams: structure-
the multi-particle finite element method (MPFEM). Powder Technol 2021;391:
property relationships. J Mater Sci 2006;41:4085–92.
34–45.
[4] Gupta N, Woldesenbet E, Mensah P. Compression properties of syntactic foams:
[35] Fiedler T, Kim H, Belova I, Sloan S, Murch G, Öchsner A. Elastic finite element
effect of cenosphere radius ratio and specimen aspect ratio. Compos A Appl Sci
analysis on cross-sections of random hollow sphere structures. Mater Werkst 2010;
Manuf 2004;35:103–11.
41:250–6.
[5] Meriç D, Gedikli H. Investigation and optimization of energy absorption
[36] Sanders WS, Gibson L. Mechanics of hollow sphere foams. Mater Sci Eng A 2003;
performance of axially joined metal and composite foam-filled hybrid tubes. J Braz
347:70–85.
Soc Mech Sci Eng 2024;46:1–16.
[37] Vesenjak M, Ren Z, Fiedler T, Öchsner A. Impact behavior of composite hollow
[6] Bolat Ç, Akgün İC, Göksenli A. On the way to real applications: Aluminum matrix
sphere structures. J Compos Mater 2009;43:2491–505.
syntactic foams. European Mechanical Science 2020;4:131–41.
[38] Gellert M, Harbord R. Moderate degree cubature formulas for 3-D tetrahedral
[7] Garcia-Avila M, Portanova M, Rabiei A. Ballistic performance of composite metal
finite-element approximations. Commun Appl Numer Meth 1991;7:487–95.
foams. Compos Struct 2015;125:202–11.
[39] Shiakolas P, Lawrence K, Nambiar R. Closed-form expressions for the linear and
[8] Marx J, Portanova M, Rabiei A. Performance of composite metal foam armors
quadratic strain tetrahedral finite elements. Comput Struct 1994;50:743–7.
against various threat sizes. Journal of Composites Science 2020;4:176.
[40] McCaslin SE, Shiakolas PS, Dennis BH, Lawrence KL. Closed-form stiffness matrices
[9] Woldesenbet E, Gupta N, Jadhav A. Effects of density and strain rate on properties
for higher order tetrahedral finite elements. Adv Eng Softw 2012;44:75–9.
of syntactic foams. J Mater Sci 2005;40:4009–17.
[41] Felippa CA. A compendium of FEM integration formulas for symbolic work. Eng
[10] Adrien J, Maire E, Gimenez N, Sauvant-Moynot V. Experimental study of the
Comput 2004;21:867–90.
compression behaviour of syntactic foams by in situ X-ray tomography. Acta Mater
[42] Pedersen P. Analytical stiffness matrices for tetrahedral elements. Comput Methods
2007;55:1667–79.
Appl Mech Eng 2006;196:261–78.
[11] Croom BP, Jin H, Mills B, Carroll J, Long K, Brown J, et al. Damage mechanisms in
[43] Kovács M, Pere B, Orbulov IN. Application of Substructure Techniques to Syntactic
elastomeric foam composites: Multiscale X-ray computed tomography and finite
Metal Foams in a Finite Element Environment. Period Polytech, Mech Eng 2023;67:
element analyses. Compos Sci Technol 2019;169:195–202.
276–84.
[12] Kartal ME, Dugdale L, Harrigan J, Siddiq M, Pokrajac D, Mulvihill D. Three-
[44] Geuzaine C, Remacle JF. Gmsh: A 3-D finite element mesh generator with built-in
dimensional in situ observations of compressive damage mechanisms in syntactic
pre-and post-processing facilities. Int J Numer Meth Eng 2009;79:1309–31.
foam using X-ray microcomputed tomography. J Mater Sci 2017;52:10186–97.
[45] Szlancsik A, Katona B, Bobor K, Májlinger K, Orbulov IN. Compressive behaviour of
[13] Kozma I, Zsoldos I. CT-based tests and finite element simulation for failure analysis
aluminium matrix syntactic foams reinforced by iron hollow spheres. Mater Des
of syntactic foams. Eng Fail Anal 2019;104:371–8.
2015;83:230–7.
[14] Huang R, Li P, Liu T. X-ray microtomography and finite element modelling of
compressive failure mechanism in cenosphere epoxy syntactic foams. Compos
Struct 2016;140:157–65.