Available online at http://www.ijprt.org.

tw/

Chinese Society of Pavement Engineering

International Journal of Pavement Research and Technology 11 (2018) 846-860

http://www.ijprt.org.tw/

Horizontal deformation resistance of paving block superstructures

– Influence of paving block type, laying pattern, and joint behaviour
H.L. Hengl a, W. Kluger-Eigl b, M. Lukacevic a, R. Blab b, J. Füssl a,⇑
a
Institute for Mechanics of Materials and Structures, TU Wien, Austria
b
Institute of Transportation, TU Wien, Austria

Received 30 November 2016; received in revised form 21 June 2018; accepted 9 August 2018

Abstract

Paving block structures are experiencing an upswing in urban areas, mainly due to their ecological, economical and space forming
qualities. Unfortunately, this trend is being weakened by damages which occur even if all design standard were met. Very often unsightly
horizontal relative deformations between paving blocks are observed, a failure mechanism which is not taken into account in standards
sufficiently.
The identification and assessment of such horizontal deformation mechanisms of paving block superstructures represents the main
objective of the present work. This is realized by means of complex 3D finite element simulations, investigating six different laying pat-
terns with five different types of paving blocks, resulting in a strongly different joint behaviour of each configuration. The non-linear
interaction behaviour between paving blocks was identified experimentally, implemented into the numerical simulation tool, and subse-
quently allowed for the reproduction of very realistic horizontal deformation mechanisms.
Finally, the performance of several laying pattern and paving block type configurations were compared to each other, pointing out the
strength and weaknesses of each superstructure and revealing which combinations are performing best.
Ó 2018 Chinese Society of Pavement Engineering. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Horizontal loading; Block pavements; Finite elements analysis; Numerical simulation

1. Introduction and motivation do not take the special characteristics of this block-like
structures sufficiently into account. The success of a project
Paving block structures are experiencing an upswing in is therefore strongly dependent on the quality of planning
urban areas due to their ecological, economical and space and construction provided by experienced engineers and
forming qualities compared to common asphalt pavements. executing companies, rather than on reliable scientific
Even in areas with high traffic volumes paving block struc- knowledge implemented into design concepts and
tures are expected to represent a suitable alternative to standards.
common pavement building systems in future. This trend A brief overview of the rather modest research activities
is, however, continuously weakened by unexpected damage concerning paving block structures is given in the follow-
of newly built paving block pavements. The reason for this ing. One of the first investigations on paving block struc-
are immature design concepts and standards, which are tures with finite element simulations were done by
heavily based on the concepts for asphalt pavements and Nishizawa et al. [1] in 1984. He developed a numerical
model with the joints modelled through a set of springs
and 2D rectangular plate elements. In 1988 Jacobs and
⇑ Corresponding author.
Houben [2] undertook further investigations using 2D rigid
E-mail address: Josef.Fuessl@tuwien.ac.at (J. Füssl).

https://doi.org/10.1016/j.ijprt.2018.08.001
1996-6814/Ó 2018 Chinese Society of Pavement Engineering.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 847

block elements connected with linear springs. The first 3D define an optimal region of superelevation-to-road-width
finite element approach can be found in Huurman [3], ratios for superelevated profiles of paving block structures,
where the numerical model contains paving blocks inter- and in Hengl et al. [14] the temperature loading on paving
connected by four sets of three springs for each joint. block superstructures is investigated numerically.
The paving block superstructure is placed on four layers The main focus of all these research lied on the vertical
of bedding and subject to vertical loading. Displacement, structural response of paving block structures and the crit-
stress, and strain information for the paving blocks, as well ical stresses and strains in the underlying baselayers. In
as for the baselayers and subgrade could be obtained. addition to the vertical resistance, however, the structural
Higher bending stresses were determined in case of higher resistance to horizontal loads is assumed to represent an
joint stiffness. Hassani [4] used shell elements to model pav- equally important performance characteristic, especially
ing blocks and the joints between them. Lerch [5] and for paving block structures. According to experienced engi-
Ascher et al. [6] implemented a non-linear interaction neers, it is often observed in practice that due to braking
between paving blocks, represented by zero-thickness ele- and steering manoeuvre of heavily loaded trucks or buses
ments in their simulations, where elastic as well as shear damage is introduced into paving block structures. Such
moduli were adjusted iteratively during the analysis. Later, loadings are normally not part of design approaches and
this interactions were substituted by non-linear contact also hardly treated in scientific literature. An interesting
laws, allowing for elastic–plastic interactions between pav- experimental work is presented in Rachmat et al. [15],
ing blocks. Also Nejad and Shadravan [7] modelled inter- investigating the horizontal performance of paving block
actions with contact elements and found out that jointing pavement on sloped road sections. Neglecting these hori-
width, shape, size, and thickness of the paving blocks have zontal loadings can cause damage in form of permanent
a major impact on the vertical structural deformations. shiftings in the superstructure and chipped edges of paving
Mampearachchi and Gunarathna [8] performed a paramet- blocks, as can be seen in Fig. 1.
ric study under usage of a 3D finite element model to deter- For example, the Austrian regulation for paving block
mine necessary improvements for weaker support pavements RVS 03.08.63 defines the admissible paving
conditions of paving block structures. In the work of Oeser block structure as well as the thickness of the base courses
and Chandra [9] a 3D Cosserat theory is applied to a only based on the expected traffic volume. However, it is
sophisticated computational model using elastic as well as obvious that the paving block surfaces, the shape, as well
plastic interaction properties. A 3D numerical simulation as the laying pattern must have a significant impact on
tool for concrete paving slabs is presented in Füssl et al. the horizontal shifting resistance (responsible for fre-
[10–12], where non-linear and plastic behaviour between quently observed shortcomings) of the superstructure. This
paving slabs and between slabs and sandbed is taken into represents the main motivation for this work, which aims
account. In Hengl and Füssl [13] various parameter studies at a numerical simulation tool able to capture the effect
were carried out on comprehensive numerical models to of different laying patterns as well as different types of

Fig. 1. Damage in a paving block structure because of too little horizontal loading resistance.
848 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

paving blocks on the horizontal shifting resistance of the Thereby, each paving block was examined in every type
related paving block superstructure. Based on this motiva- of laying pattern. Finally, concluding remarks are given
tion, the following main objectives had been defined: in Section 5.

The definition of sufficiently large and suitable sub- 2. Identification experiments for the vertical joint behaviour
models of paving block superstructures allowing for
the determination of the horizontal shifting resistance The transmission of forces between paving blocks
of these structures without significant influence of the through the vertical joints of the paving block structure
boundary conditions. Furthermore, an automated gen- strongly affect, not to say define, the overall structural
eration of the whole superstructures’ geometry, to allow behaviour in horizontal direction. For this reason, special
for an efficient investigation of several paving block focus was laid on the identification of these properties.
geometries as well as laying patterns. Two different kinds of experiments were carried out at

An appropriate description of the non-linear interaction the TVFA Vienna to obtain shear and normal properties
behaviour, in tangential as well as normal direction of for all of the 6 paving block types aforementioned. The
the vertical joints, between paving blocks. Shear failure experiments and the results are presented in the following
in the joints and the opening of joints should be repro- two subsections. Additional information about such exper-
duced realistically. iments can be found in Füssl et al. [16], where similar tests

Finally, the finite-element-based determination of the were performed by the authors.
horizontal shifting resistance as well as the correspond-
ing deformation (failure) mechanisms of different paving 2.1. Normal joint behaviour experiment
block superstructures.
The experimental setup for the identification of the joint
The 5 types of paving blocks and the 6 different laying behaviour in normal direction is illustrated in Fig. 3.
patterns investigated in this work are shown in Fig. 2. To Thereby, two paving blocks are placed on a wooden multi-
get realistic interaction properties, describing the contact layer board and the 5 mm thick joint between them is filled
behaviour between paving blocks, two types of experiments with sand (0/2) and manually compacted. One paving
on paving blocks were carried out. The experimental setup block is completely clamped while to the other one a hori-
as well as the results are given in Section 2. In Section 3, the zontal force H is applied and the relative normal displace-
developed simulation tool is described in and the corre- ment un between these two blocks is measured with an
sponding numerical results are presented in Section 4. LVDT. The obtained relationships between the normal

Fig. 2. Paving block types and laying patterns investigated in this work: (a) Granite Cube, (b) Concrete Block, (c) Concrete Interlocking Block (CIB), (d)
Double-T Block, (e) Wave Block, and (1) Stretcher Bond, (2) Stretcher Bond 45°, (3) Herringbone, (4) Stacked Bond, (5) Stacked Bond 45°, (6)
Herringbone 45°.
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 849

Fig. 3. Experimental setup for the identification of the joint behaviour in normal direction.

Fig. 4. Relationships between the normal stress rn in the vertical joints and the relative displacement un between paving blocks (Concrete Blocks and
Concrete Interlocking Blocks) obtained from experiments.

stress rn and the displacement un for 5 performed tests are

plotted in Fig. 4. Since this information was assumed not to Table 1
Average normal joint behaviour extracted from the
be as decisive as the shear behaviour and less dependent on experimental results shown in Fig. 4, used as input to
the type of paving block, this experiment was only con- the numerical simulation tool.
ducted with Concrete Blocks and Concrete Interlocking Deformation un [mm] Pressure rn [MPa]
Blocks. Finally, an average response was evaluated (red
0.5 0
graph in Fig. 4), from which the input data for the numer- 0 0.0001
ical simulation tool were extracted (given in Table 1). 0.05 0.15143
0.1 0.35921
2.2. Tangential joint behaviour 0.15 0.62114
0.2 0.93722
0.25 1.30745
The experimental setup to identify the tangential joint 0.3 1.73183
behaviour is illustrated in Fig. 5. Thereby, three paving 0.35 2.21036
blocks are arranged in a row on a wooden multilayer board 0.4 2.74304
and the 5 mm thick joints in-between are filled with sand 0.45 3.32987
(0/2), which is mechanically compacted under dry condi- 0.5 3.97085
0.55 4.66598
tions. Then, in a first step, a constant load H is applied
 850 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

Fig. 5. Experimental setup to identify the tangential joint behaviour for (a,b) the Double-T Block, (c,d) the Concrete Block, and (c,f) the Wave Block.

in ‘‘row”-direction (horizontal direction in Fig. 5), allowing displacements in ‘‘row”-direction as well as the force F
frictional forces to be activated in the joints. In a second were measured continuously. This experiment was carried
step, an increasing load F is applied to the middle paving out for all 5 types of paving blocks, at each for up to 5 dif-
block transversely to the ‘‘row”-direction up to the point ferent constant loads H and with at least two tests per con-
of shear failure (see, e.g., Fig. 5(b, d, f)). The relative figuration to ensure repeatability. For all these tests the
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 851

Fig. 6. Shear stress resistance – Results of horizontal displacement experiments.

Table 2
Mohr–Coulomb friction parameters for all 5 paving block types obtained from
identification experiments.
Specimens Dimension [cm] Cohesion c [MPa] Friction angle [°]
Concrete Block 20/10/8 0 49.69
Granite Cube 18/18/18 0.0189 29.56
Wave Block 21.5/10.5/8 0.0422 71.15
CIB 20/10/8 0.4994 57.64
Double-T Block 19.5/16/7.6 0.7519 63.23

maximum (average) shear stress in the joint as a function of according to an industry standard, to 200/100/100 mm.
the average normal stress in the joint, defined through the Paving blocks with different geometry and dimensions are
load H, is plotted in Fig. 6. used as border stones to provide a straight boundary. Lin-
The data of each paving block type can be connected ear elastic material behaviour is assigned to the paving
through a Mohr–Coulomb friction law quite well (linear blocks, with an elastic modulus of 45 000 MPa, which
graphs in Fig. 6). The resulting Mohr–Coulomb friction was obtained by ultrasonic experiments on similar paving
parameters (cohesion and friction angle) for each paving blocks in Füssl et al. [11], and a poisson’s ratio of 0.15.
block type are given in Table 2, and were subsequently used Furthermore, a specific weight of 24 kN/m3 was assumed
as input to the numerical simulation tool. for the paving blocks, allowing to take their dead load
within the simulations into account.
3. Numerical simulation tool The interaction between paving blocks in normal direc-
tion is defined as a tabular pressure–overclosure relation-
Exemplarily for all models of different laying patterns, ship according to the normal joint behaviour experiments
the model for a stretcher bond 45° is displayed in Fig. 7. described in Section 2, whereas the exact values used can
It only consists of the paving blocks and the underlying be found in Table 1. The interaction behaviour in tangen-
sandbed. The other base layers were not modelled explic- tial direction between paving blocks is described by a
itly, since it can be assumed that they contribute only little Mohr–Coulomb friction law, with the cohesions and fric-
to the horizontal shifting resistance of the whole paving tion angles obtained from the tangential joint behaviour
block structure. At least they would not influence the per- experiments described in Section 2. For all investigated
formance comparison between different paving blocks and paving blocks these strength values can be found in Table 2.
laying patterns, which is the main focus of this work. For The default Mohr–Coulomb friction criterion in Abaqus is
all models the chosen modelling area is rectangular with not able to take a cohesion not equal to zero into account.
an approximate dimension of 5.8 m to 2.6 m. Small devia- For this reason, this criterion has been adapted and imple-
tions to this dimensions result from the different laying pat- mented as a user subroutine written in Fortran. More
terns. The dimensions of the paving blocks are set, details of the implementation of this tangential behaviour
852 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

Fig. 7. Geometry and boundary conditions of the numerical model illustrated by means of a stretcher bond 45° model.

will be given in a following subsection. At the lateral Exemplarily, the discretised version of the stretcher bond
boundary all paving blocks are supported by an elastic 45° model is shown in Fig. 8. A characteristic mesh size
foundation with a bedding modulus of 9 MPa/mm, which of 25 mm is used for the paving blocks, whereas the mesh
is an extrapolated value obtained from the normal pres- is refined (12.5 mm characteristic mesh length) in the area
sure–displacement relationship in Fig. 4. This bedding where the wheel-load is applied. The horizontal element
modulus approximately represents the stiffness of a vertical lengths of the mesh of the sandbed is 50 mm, finally result-
sand joint adjacent to a rigid border block. ing in a total number of elements for each model of around
To the underlying sandbed also a linear elastic material 100 000.
behaviour has been assigned, with an elastic modulus of The simulations were carried out as follows: In a first
350 MPa, a Poisson’s ration of 0.3 and a specific weight analysis step the dead load and in a second analysis step
of 18 kN/m3. Since the vertical structural response of the a fictitious braking performance of a single tire was applied
paving block structure was not investigated in this work to a single paving block. This fictitious loading consists of a
it seemed not to be necessary to model a more complex standard vertical tire load of 57.5 kN and of a horizontal
behaviour. For this reason, also the displacements of the loading of equal size, representing an absolute upper limit
lower surface of the sandbed were simply prevented in all of possible braking forces. The finite element analysis was
spacial directions. Between the sandbed and the paving carried out with the Abaqus/Standard solver on a HPC
blocks a ‘‘hard” contact was assumed in normal direction, computing cluster at TU Wien, using unsymmetrical
allowing the paving blocks to lift off unstressed while stres- matrix storage and including non-linear effects of large
ses are transmitted fully under pressure. In tangential direc- displacements.
tion classical Mohr–Coulomb friction is modelled with a To improve the stability of the numerical simulations an
frictional coefficient of 0.6 and no cohesion. elastic slip at all tangential contact interactions was
The whole model is discretised with 8-node hexahedron allowed. However, this elastic slip was restricted to
elements, except some of the border blocks with non- 0.0005 times the adjacent characteristic element size lel
rectangular geometry where wedge elements are used. and therefore has no significant influence on the numerical

Fig. 8. Discretisation of the stretcher bond 45° model with 8-node hexahedron elements.
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 853

Fig. 9. (a) Ideal frictional behaviour and (b) the approximation allowing a small amount of elastic slip used in this work.

result. The ideal frictional behaviour, where no relative means of two laying patterns, while the performance of
motion Dut is allowed for jsj < scrit , as displayed in Fig. 9 the 6 different laying patterns is compared in the second
(a) is thus approximated with the relationship shown in subsection by means of only the Concrete Block and the
Fig. 9(b). So, a small amount of elastic motion, related to Concrete Interlocking Block. Finally, a result overview of
the frictional shear stress through: all simulations conducted is given in the third subsection.
scrit 0:6  rn
Duel ¼  Dut ¼  Dut : fDut 2 R j jDut j 4.1. Performance of different paving blocks arranged in the
Duel;max 0:0005  le
same laying pattern
6 Duel;max g;
Fig. 10 shows the horizontal deformation fields for all 5
where rn represents the pressure between paving block and
different paving blocks laid in a stacked bond. For compar-
sandbed, is allowed. For more detailed information about
ison reasons all paving blocks were modelled with the same
the numerical implementation of Mohr–Coulomb friction
dimensions. In addition, it should be noted here that since
using the penalty method, which was used in this work,
the non-planar side surfaces of the Double-T Block as well
the reader is referred to Corp. [17].
as the Wave Block are already considered within the iden-
tified strength parameters obtained from experiments, the
3.1. Creation of different model geometries real geometry of these paving blocks doesn’t need to, or
even must not, be modelled at this structural scale.
Due to the large number of paving blocks modelled for For the visualisations in Fig. 10 the sandbed was
each laying pattern (as can be seen in Fig. 7) and to be able excluded and the deformations were scaled by a factor of
to flexibly adjust geometric parameters the models were 100. All deformation fields are related to the same fictitious
generated by an extensive python script consisting of more loading state as described before and, thus, show an
than 1000 lines of code. The script offers a GUI through impressive performance difference between these five differ-
which future users can define a variety of parameters, such ent paving blocks. For the Concrete Block, exhibiting zero
as the dimensions of the paving block, the size of the model cohesion between paving blocks, about six times the defor-
region, the material and contact properties, as well as the mation was obtained compared with the best-performing
type of laying pattern. Subsequently, the whole model is Double-T Block. Since there are no compressive stresses
generated automatically, including the geometry of the in transverse direction, the Concrete Block is not able to
border stones and all the contact interactions between the transfer shear stresses to adjacent rows and, thus, only
paving blocks and the paving blocks and the sandbed. Fur- the loaded row is shifted in load direction. The maximum
thermore, in combination with a bash script this automated displacement could be obtained by simply summing up
model generation provides a powerful tool for conducting the overlaps of the paving blocks in this row, which can
parameter studies. be interpreted as the compression of the sand joints plus
the displacement of the border block of the structure,
4. Numerical simulation results which is kept quite small by a relatively stiff elastic founda-
tion on the boundary. It can be assumed that the maximum
As already mentioned within the introduction, numeri- displacement would be even larger if more paving blocks
cal simulations have been performed for each configuration were modelled in load direction. A similar picture is drawn
of the 5 different paving blocks and the 6 different laying by the Granite Cube, also exhibiting only a very small
patterns proposed, with the intention to determine the cohesion of 0.019 MPa. For this reason, as expected, pav-
effect of those on the horizontal shifting resistance of the ing blocks with no interlocking effect (cohesion) are not
corresponding paving block superstructure. Trying to pre- suitable for this laying pattern. A completely different sys-
sent the results in a structured way, in the first subsection tem response was obtained for the three paving blocks hav-
the performance of all 5 paving blocks is compared by ing significant interlocking capabilities, the Wave Block,
 854 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

Fig. 10. Horizontal deformation fields of the 5 investigated paving blocks arranged in a stacked bond. A vertical and horizontal load of 57.5 kN each is
introduced on one paving block. Scaling of deformations: 100.

the Concrete Interlocking Block, and the Double-T Block. forces are introduced into the sandbed, an area with no
For all three, the obtained displacement field was proven to force transmission becomes visible. A deeper look at the
be independent on the boundary conditions, and without simulation results revealed that in these areas the paving
any compressive force in transverse direction they are able blocks are lifted and, thus, are having no contact with
to activate a large amount of paving blocks counteracting the sandbed.
the very concentrated introduced load. A far lesser impact of the type of paving block on the
This can also be seen very clearly in Fig. 11, in which the horizontal deformation was obtained for the stretcher
normal and shear contact forces between paving blocks are bond laying pattern, as can be seen in Fig. 12. Clearly, this
vividly illustrated, for the Concrete Blocks and the Con- laying pattern allows for a natural distribution of the intro-
crete Interlocking Blocks. duced load into the transverse direction for all types of
The great decrease in contact forces in the surrounding paving blocks. This is illustrated by means of the nodal
of the load introduction comes from the great shear forces normal contact forces plotted in Fig. 13 for the Concrete
which can be transferred to the sandbed by the paving Block superstructure.
blocks which experience a substantial vertical loading. At Nevertheless, still a significant influence of the joint
a certain distance from the load introduction a constant properties remains, leading to a 2.5 times higher deforma-
decrease in contact forces can be observed, reflecting the tion of the Concrete Block structure compared to the
shear transfer capability of the paving blocks to the Double-T Block structure. Anyways, the appropriate
sandbed only due to their dead weight. choice of the size of the modelling region, to avoid a strong
These large areas of constant shear transfer to the influence of the boundary conditions, can be seen well here.
sandbed (for the Concrete Interlocking Blocks) can be Only for the Double-T Block superstructure an even smal-
nicely seen in Fig. 11(f). Interestingly, between these areas ler maximum deformation would probably be obtained
and the load introduction zone where very high shear with a larger transverse modelling length.
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 855

Fig. 11. Visualisation of contact forces (in loading direction) between paving blocks, for the stacked bond laying pattern.
 856 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

Fig. 12. Horizontal deformation fields of the 5 investigated paving blocks arranged in a stretcher bond. A vertical and horizontal load of 57.5 kN each is
introduced on one paving block. Scaling of deformations: 100.

Fig. 13. Nodal normal contact forces in load direction for a stretcher bond with Concrete Blocks.

4.2. Performance of different laying patterns with the same The results for the stacked bond, the stretcher bond, and
paving blocks the herringbone are illustrated in Fig. 14. Fig. 14(a) to (d)
have already been shown before, but only in this compar-
In the following, horizontal deformation fields are ison it becomes obvious that the stretcher bond represents
shown for the 6 different laying patterns and for the Con- the ideal laying pattern for conventional concrete paving
crete Block as well as the Concrete Interlocking Block. blocks. The maximum horizontal deformation is only
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 857

Fig. 14. Horizontal deformation fields for three different laying patterns and two different paving blocks. A vertical and horizontal load of 57.5 kN each is
introduced on one paving block. Scaling of deformations: 100.

slightly higher than that of the concrete block with inter- is caused by a very concentrated, and thus bad, force distri-
locking effect in the stretcher bond, and it is smaller than bution into the superstructure. This is illustrated in Fig. 16,
the maximum deformation in all other laying patterns. showing the rather narrow bands of nodal normal contact
Conversely, CIBs are ideally suited for a stacked bond, forces in this laying pattern.
because their interlocking capabilities fully compensate Considering the Concrete Interlocking Blocks, the per-
the non-interlocking nature of this laying pattern. formance is better for all three laying patterns. This has
It is also interesting to see that the performance of the essentially two reasons: First, due to the 45° rotation more
Concrete Block is getting worse when laid in a herringbone joints are heavily loaded by shear forces and, thus, the
laying pattern instead of a stretcher bond while the perfor- interlocking capability of the joints get highly activated
mance of the CIB is getting even better. This is because the and, secondly, also paving blocks behind the load introduc-
interlocking capability of the herringbone can only be acti- tion are getting involved in the load transfer mechanism.
vated ideally if the vertical joints exhibit cohesive beha-
viour (as can be seen in Fig. 14(e) to (f)). 4.3. Overview of simulation results
The deformation fields of the three remaining laying
patterns, the 45° rotated ones, are shown in Fig. 15. Fig. 17 finally shows the maximum horizontal deforma-
As can be seen, even for this complex arrangements of tion in loading direction umax for all configurations of the 6
paving blocks the numerical simulation tool is able to deli- laying patterns with the 5 investigated paving blocks. In
ver plausible shifting mechanisms. The mechanisms them- general it can be said that both the type of laying pattern
selves are strongly characterized by the 45° orientation of and the type of paving block have a significant influence
the laying patterns. Looking at the conventional Concrete on the horizontal shifting resistance. However, disregard-
Block, only for the stacked bond a significant performance ing the stacked bond, the paving block type seems to play
improvement can be identified compared to its not rotated a more important role and cannot be compensated by the
counterpart, while for the other two laying patterns similar type of laying pattern easily. On average paving blocks
maximum deformations are obtained. For the stretcher with joints providing a decent interlocking effect perform
bond an even worse performance can be observed, which three times as good as paving blocks in the same laying pat-
 858 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

Fig. 15. Horizontal deformation fields for three different laying patterns and two different paving blocks. A vertical and horizontal load of 57.5 kN each is
introduced on one paving block. Scaling of deformations: 100.

Fig. 16. Rather bad distribution of nodal normal contact forces in a stretcher bond 45° with Concrete Blocks, leading to high horizontal deformations.

tern without interlocking capabilities. This may lead to (which could be reduced in case of poorly compaction),
huge performance differences also in practical applications and so on. The presented numerical simulation tool is cap-
and shouldn’t be ignored by design concepts and able of taking all this parameters into account. Two rela-
standards. tionships are exemplarily given in Fig. 18, showing the
Of course, there are still a lot of other parameters influ- significant influence of the paving block height as well as
encing this horizontal deformation resistance, such as the the normal contact stiffness in the vertical joints. A dou-
paving block height, the joint filling (which can partly be bling of this stiffness can reduce the maximum horizontal
described by the paving block height), the contact stiffness deformation by almost half.
 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860 859

Fig. 17. Maximum horizontal deformation in loading direction umax for the investigated 6 different laying patterns and 5 different paving blocks.

Fig. 18. Influence of the paving block height and contact stiffness in normal direction on the maximum horizontal deformation.

5. Conclusions
The fully automation of the model creation, enabled by

a complex Python script, allowed us to run a huge
In this paper, a numerical simulation tool has been pro- amount of simulations to finally identify an appropriate
posed allowing for a realistic reproduction of resistance model size for which the influence of boundary condi-
mechanisms of paving block superstructures under horizon- tions is not significant for the majority of the investi-
tal loading. To accomplish that, numerical models with gated configurations. Furthermore, much attention
more than 2000 non-linear interactions, with their beha- was paid to any kind of numerical regularisation mech-
viour defined through a user subroutine written in Fortran, anism, like artificial stiffness and viscous damping.
had to be run. The required strength properties to define the Many parameters studies were conducted to ensure that
non-linear joint behaviour of all five investigated paving such influences are negligible.
block types were obtained by special tangential shear exper-
The non-linear behaviour of the vertical joints, in nor-
iments. Based on these investigations, answers to the three mal as well as tangential direction, could be reproduced
questions raised in the introduction can now be formulated: quite accurately. Therefore, information from two kind
860 H.L. Hengl et al. / International Journal of Pavement Research and Technology 11 (2018) 846-860

of joint experiments were used and implemented into the [4] A. Hassani, Modelling and Structural Design of a Concrete Block
numerical simulation tool. Pavement System, International Conference on Concrete Block
Paving, 2006.

Mainly for this reason, very plausible 3D deformation [5] T. Lerch, Investigation of the Deformation Behaviour of Concrete
mechanisms could be obtained for several laying pat- Block Pavements under Simulated Traffic Loading (Ph.D. thesis), TU
terns with different types of paving blocks. Interesting Dresden, 2005.
insights into load transfer mechanisms could be gained, [6] D. Ascher, T. Lerch, M. Oeser, F. Wellner, 3d-fem Simulation of
showing a huge variety depending on the combination Concrete Block Pavement, Int. Conference on Concrete Block
Paving, 2006.
of laying pattern, type of paving block and the related [7] F.M. Nejad, M.R. Shadravan, A Study on Behavior of Block
joint behaviour. Finally, this allowed for a comprehen- Pavement using 3d Finite Element Method, International Conference
sive performance evaluation of several paving block on Concrete Block Paving, 2006.
pavements with respect to their horizontal deformation [8] W.K. Mampearachchi, W.P.H. Gunarathna, Finite-element model
resistance. approach to determine support conditions and effective layout for
concrete block paving, J. Mater. Civ. Eng. (2010)
[9] M. Oeser, H. Chandra, Segmented concrete block pavements:
In short, the simulation results have confirmed the high analysis with optimized numerical tool, 2010.
performance expectations of paving blocks with interlock- [10] J. Füssl, W. Kluger-Eigl, R. Blab, Mechanical performance of
ing effects. On average these paving blocks perform three pavement structures with paving slabs–Part I: full-scale accelerated
times as good as blocks without interlocking capabilities, tests as validation for a numerical simulation tool, Eng. Struct. 98
(2015) 212–220, https://doi.org/10.1016/j.engstruct.2014.10.054.
considering the same laying pattern. Even for the stacked [11] J. Füssl, W. Kluger-Eigl, L. Eberhardsteiner, R. Blab, Mechanical
bond laying pattern, where no structural load distribution performance of pavement structures with paving slabs–Part II:
occurs, paving blocks with interlocking effect can lead to numerical simulation tool validated by means of full-scale accelerated
reasonable horizontal deformation resistances. Further- tests, Eng. Struct. 98 (2015) 221–229, https://doi.org/10.1016/j.
more, as the horizontal deformation in those superstruc- engstruct.2014.10.055.
[12] J. Füssl, H. Hengl, L. Eberhardsteiner, W. Kluger-Eigl, R. Blab,
tures is highly affected by the stiffness of their sand-filled Numerical simulation tool for paving block structures assessed by
joints, it should be ensured that these joints are completely means of full-scale accelerated pavement tests, Int. J. Pavement Eng.
filled, well compacted, and their width and quantity (2016) 1–13, https://doiorg/10.1080/10298436.2016.1224410.
minimized. [13] H. Hengl, J. Füssl, The influence of superelevated profiles of paving
In summary, this work has the intention to demonstrate block structures on their load-bearing behavior, Eng. Struct. 117
(2016) 195–203, https://doi.org/10.1016/j.engstruct.2016.03.003.
the very different horizontal resistance of paving block [14] H. Hengl, W. Kluger-Eigl, R. Blab, J. Füssl, The performance of
pavements, to propose a method to identify this resistance, paving block structures with mortar filled joints under temperature
and to possibly contribute to future design concepts which loading, accessed by means of numerical simulations, Road Mater.
will hopefully cover this type of damage scenario. To Pavement Des. (2017) 1–20, https://doiorg/10.1080/
increase the computational efficiency of such kind of calcu- 14680629.2017.1330221.
[15] M. Rachmat, M.N. Hasanan, R.H. Mohd, L. Tung-Chai, Perfor-
lations, the application of finite-element-based limit analy- mance of concrete block pavement on sloped road section, Int. J.
sis formulations, as presented in Li et al. [18] for Pavement 6 (1) (2007) 136–145.
orthotropic strength behaviour and in Li et al. [19] includ- [16] J. Füssl, W. Kluger-Eigl, R. Blab, Experimental identification and
ing discontinuities, could be a possible future step. mechanical interpretation of the interaction behaviour between
concrete paving blocks, Int. J. Pavement Eng. 17 (2016) 478–488,
https://doi.org/10.1080/10298436.2014.993205.
References [17] D.S.S. Corp., Abaqus Analysis User’s Manual, Providence, 2016.
[18] M. Li, J. Füssl, M. Lukacevic, J. Eberhardsteiner, C. Martin,
[1] T. Nishizawa, S. Matsuno, M. Komura, Analysis of Interlocking Strength predictions of clear wood at multiple scales using numerical
Block Pavements by Finite Element Method, International Confer- limit analysis approaches, Comput. Struct. 196 (2018) 200–216,
ence on Concrete Block Paving, 1984. https://doi.org/10.1016/j.compstruc.2017.11.005.
[2] M. Jacobs, L. Houben, Wheel Testing and Finite Element Analysis of [19] M. Li, J. Füssl, M. Lukacevic, J. Eberhardsteiner, C. Martin, A
Concrete Block Pavement, Delft University of Technology, The numerical upper bound formulation with sensibly-arranged velocity
Netherlands, 1988. discontinuities and orthotropic material strength behaviour, J.
[3] M. Huurman, L.J.M. Houben, A.W.M. Kok, Development of a Theoret. Appl. Mech. 56 (2018) 417–433, https://doi.org/10.15632/
three-dimensional finite element model for concrete block pavements, jtam-pl.56.2.417.
in: Proceedings Fourth International Conference on Concrete Block
Paving, 1992.