Construction and Building Materials 24 (2010) 803810

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Experimental study of the size- and shape-effects of natural building stones

S.K. Kourkoulis *, E. Ganniari-Papageorgiou
Department of Mechanics, National Technical University of Athens, Zografou Campus, Theocaris Building, 157 73 Athens, Greece

a r t i c l e

i n f o

Article history:
Received 6 June 2008
Received in revised form 20 October 2009
Accepted 21 October 2009
Available online 24 November 2009
Dedicated to the memory of the late
Professor Ioannis Vardoulakis
Keywords:
Natural building stones
Kefalonia porous stone
Dionysos marble
Conchyliates shell-stone
Cyprus porous stone
Mechanical properties
Size effect
Shape effect

a b s t r a c t
The dependence of the mechanical properties of natural building stones on the size and the shape of the
specimens is studied experimentally. Attention is focused to the Kefalonia porous stone, a candidate substitute of the Kenchreae porous stone used by ancient Greeks for the erection of the Epidaurean Asklepieion. Series of uniaxial compression tests were carried out using both cubic and cylindrical specimens of
various sizes. A number of mechanical properties were determined including the peak stress, the modulus of elasticity, the stress drop after the peak stress, the peak strain and the strain energy density up to
the peak load. A strong dependence of the above properties on both the dimensions and the shape of the
specimens was concluded. In addition, it was indicated that the dependence of some of the above properties on the size of the specimens is not monotonous. The conclusions drawn are in good agreement with
similar ones obtained for Dionysos marble, the material used today for the restoration of the Parthenon
temple of Athens as well as for the Conchyliates shell-stone the material that has been used for the construction of the Zeus Temple at Olympia.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Restoring a monument is a complicated multidisciplinary scientic task. Series of problems, varying from elementary ones
(strength and deformability of the materials) to rather complex
ones (preservation of the structural system, determination of the
minimum interventions, their reversibility and of course their
durability) are to be considered and solved before decisions are
reached. Archaeologists, architects and civil engineers collaborate
in order to meet the nal target, the extension of the life of the
monument. The nal decisions are usually a compromise between
various, and often contradictory, points of view.
The structural stability of the monument is, of course, the most
important among the problems of the experts working for the restoration (although sometimes it is outweighed in public perception
by the aesthetics of the surface of the stone Fig. 1 [1]). The problem
is exacerbated in case large fragments of structural elements, or
whole elements of the structure, are missing and have to be replaced. The ideal solution is to have access to the source of the
authentic material. Unfortunately this is the exception rather than
the rule. In most cases there are no known functioning quarries of
the desired material or the areas of the ancient quarries have been
built over. In such a case, a new stone has to be chosen as a substi* Corresponding author. Fax: +30 210 77 2 1302.
E-mail address: stakkour@central.ntua.gr (S.K. Kourkoulis).
0950-0618/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2009.10.027

tution material. The substitution stone and the authentic one

should react similarly to environmental inuences, natural wear
and weathering and mechanical loads.
This paper focuses on some aspects of the mechanical behaviour
of such a stone, the porous stone of Kefalonia, which is considered
as a candidate substitute for the Kenchreae porous stone, the material used by ancient Greeks for building the Epidaurean Asklepieion, the most celebrated healing centre of the ancient world [2].
The authority and radiance of Asklepios, the healer god of antiquity, brought to the sanctuary nancial prosperity, which in the
4th and the 3rd centuries BC enabled the implementation of an
ambitious program aimed at housing the worship in monumental
buildings. The extensive remains of the site have been brought to
light during excavations conducted since the late 19th century to
the present day. The study of the Kefalonia stone is associated with
the restoration project in progress of three monuments of the
Asklepieion, namely the circular building of Tholos (the healing
gods assumed subterranean dwelling place), the Avaton or
Enkoimeterion (a large stoa used for the incubation and cure of
the sick) and the Propylon of the Gymnasium (a building complex
used for the sacred meals). All three follow the same construction
principles regarding the choice of material. Their foundations are
built of the locally available weak calcareous conglomerate, known
as foundation porous stone.
The comparative study of the original material of the monuments and the substitute one dictates the size of the specimens:

804

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

Fig. 1. An overview of a restored pillar from Avaton of the Epidauros Archaeological site, exhibiting the coexistence of ancient (Kenchreae porous stone) and new material [1].

They must be as small as possible since for obvious reasons it is not

permitted to prepare large specimens from the original stones.
Unfortunately it is common knowledge that the mechanical properties and the values of the mechanical constants of most building
materials depend on the size and the shape of the specimens and
this dependence is not yet fully understood. In this direction an
experimental study was undertaken to enlighten some aspects of
these phenomena for the Kefalonia porous stone and draw some
conclusions that could be of help for the engineers involved in
the restoration project of the Epidauros monuments, who are dealing with structural elements the dimensions of which are at least
one order of magnitude larger than those of the laboratory
specimens.
2. The size- and the shape-effects
The geometry and the shape of the specimens proposed by standards for the determination of the compressive strength of brittle
geomaterials differ concerning both their shape and size. The specimens most often used are either cylindrical or cubic. The dependence of the compressive strength on the shape of the specimens
has been widely studied and various empirical formulae have been
proposed for the relation between the strength obtained from
specimens of these geometries. In addition it is known that the
nominal strength of structures changes by scaling their size. This
phenomenon was already observed by Leonardo da Vinci who concluded that if two ropes have the same thickness the longest is the
weaker [3]. However, it was only at the beginning of the 20th century when quantitative results for the dependence of the nominal
strength of glass bers on their diameter were published by Grifth [4]. This dependence is caused by the dimensional inhomogeneity between the stress ([F][L]2) and the stress intensity factor
([F][L]3/2), which leads to a 1/2 constant slope of the graph of
the nominal strength versus the structural size in a bilogarithmic
diagram [5].
From this point on and for half a century the size effect was
attributed to the statistical nature of the distribution of aws within a structure and it was described by the weakest-link concept
introduced by Fisher and Tippett [6] and developed by Weibull
[7]. A non-statistical approach did not appear until early 1970s
when Walsh [8] published results that could not be described by
purely statistical approaches. This discrepancy and similar ones
observed experimentally for other materials placed limitations to

the use of the statistical approach. The main scepticism was that
the power law of the Weibull theory for the nominal strength implies the absence of any characteristic length. Such a conclusion is
unacceptable for the materials characterized as quasi-brittle which
exhibit a nite fracture process zone.
Today two approaches are identied: The rst one (deterministic-energetic) has been introduced by Bazant [9,10] and is based on
the observation that the failure of quasi-brittle materials is characterized by both energy and stress quantities, i.e. the fracture energy, Gf, and the tensile strength, ft. From dimensional analysis it
is concluded that such a material possess a characteristic length,
lo, depending on the size of the fracture process zone. The second
approach, introduced by Carpinteri [11,12], relates the size effect
to the fractal nature of the crack surfaces. Both theories were introduced for tensile loading of structures with preexisting cracks or
notches, for which the failure is caused by the localization of the
strain, which in turn results in a nite size fracture process zone.
However, it is believed today that the strain localization is the
cause of failure also for structures under uniaxial compression,
the difference being that the damage zone is larger. For the case
of concrete under uniaxial compression Kim and Yi [13] proposed
a modication of Bazants law for the dependence of the compressive strength, fo, on the dimensions of the specimen, assuming a ratio of height-to-diameter equal to 2:
0

Bf c
0 0
fo h
i1=2 afc fc
1 k fd0 dm
1 c

B, m, k1 are constants, da stands for the maximum aggregate size, fc0

is the compressive strength of the standard cylinder, d is the specimens diameter and a is the crack band length.
Understanding the size effect is of paramount importance since
the majority of laboratory tests are carried out using specimens
with dimensions of the order of 1040 cm. These results are then
extrapolated in order for conclusions to be drawn for structures
with dimensions of much larger size, of the order of a few meters.
Unfortunately up to now denite conclusions concerning the laws
governing the transition region from the size of the laboratory
specimens to the size of the real structures have not been reached
and as a result the design codes are still based on empirical or
semi-empirical formulas obtained from curve tting to the experimental results [10].

805

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

Fig. 2. SEM observations (a) and petrography (b) of the Kefalonia porous stone.

Table 2
Geometric characteristics, number of specimens per class of tests (after applying
Chauvenets criterion) and the results for the peak compressive stress and its standard
deviation for the cubic specimens.

Fig. 3. A close view of the surface of the material just after it was quarried.

For obvious reasons the phenomenon has been studied extensively (both theoretically and experimentally) for the case of concrete, the material most widely used in modern structures. The
experimental data available for other structural materials and natural building stones are rather limited. In this direction data concerning the size effect of a relatively soft natural building stone,
the porous stone of Kefalonia, are presented in the following paragraphs, together with some additional data for Dionysos marble
and for the Conchyliates shellstone, in an effort to establish the
dependence of some mechanical constants and parameters on
the size and the shape of the specimens used for the experiments.

3. Experimental procedure and results

3.1. The material and the specimens
The Kefalonia porous stone is a relatively soft stone with clearly layered structure quarried in the island of Kefalonia, in the Ionian Sea. The petrography and a
Scanning Electron Microscopy microphotograph of the material can be seen in
Fig. 2. It contains about 55% CaO and 0.3% MgO. Its bulk density is equal to about
1.47 g/cm3, its absolute density is 2.4 g/cm3 and the void ratio is 0.61. The texture
is relatively massive and its color is usually whitish-beige. It is characterized by

Edge
(mm)

Number
of tests

Peak compressive
stress, rp (MPa)

Standard
deviation
(MPa)

Load
rate (m/s)

50
60
73
93
120

15
14
19
13
14

14.7
21.2
27.9
24.7
18.9

1.2
2.1
4.4
2.7
1.8

2.5  106

high porosity (38%) and nets of internal pores and surface vents. This can be
clearly seen in the block of recently quarried material of Fig. 3, from which a number of specimens of the present study were cut.
Large series of uniaxial compression tests were executed with both cubic and
cylindrical specimens of different sizes in each case. For the cylindrical specimens
the diameter was kept constant equal to about 80 mm while the height ranged from
40 to 160 mm yielding apertures equal to 0.54, 0.75, 1.04, 1.25, 1.54, 1.75 and 2.00.
For the cubic specimens ve classes were tested with edges equal to about 50, 60,
73, 93 and 120 mm. Ten to twenty specimens were tested for each geometric class
depending on the scattering of the results for the peak compressive stress. Chauvenets criterion was used for the assessment of the tests and rejection of statistically
unaccepted results (maximum two per class). The geometric characteristics of the
specimens, the number of specimens tested, the peak compressive stress and its
standard deviation for each class of tests are summarized in Tables 1 and 2. The
specimens were cut in such a way that the compressive load acted normally to
the material layers. It is emphasized that all specimens without exception were
cut from adjacent volumes of freshly quarried material, without macroscopically
noticeable differences, in order to reduce as much as possible the inevitable inherent variability of the internal structure usually observed in porous geomaterials.
Special care was taken during the preparation of the specimens to ensure that
the bases of the specimens were parallel to each other and perpendicular to the load
direction. In addition a semi-spherical head interposed between the loading plate
and the moving piston ensured further the normality of the load on the bases of
the specimens (Fig. 4a).
For the measure of the axial displacements three LVDTs of sensitivity equal to
106 m were used, placed at 120o to each other in order to check the symmetry
of the loading (Fig. 4a). The rotation of the end platens detected was almost negligible. Additionally, in some of the tests a system of four strain-gauge rosettes was
used, antidiametrically glued on two mutually perpendicular diameters at the midheight of the specimens, for comparison reasons (Fig. 4b).
Concerning the interface between end-platens and specimen the majority of
tests in the present work were carried out using non-lubricated surfaces, in an effort to simulate the real conditions, since the material is to be used in direct contact

Table 1
Geometric characteristics, number of specimens per class of tests (after applying Chauvenets criterion) and the results for the peak compressive stress and its standard deviation
for the cylindrical specimens.
Diameter (mm)

Height (mm)

Height/diameter ()

Number of specimens

Peak compressive stress, rp (MPa)

Standard deviation (MPa)

Load rate (m/s)

80
80
80
80
80
80
80

43
60
80
100
120
140
160

0.54
0.75
1.00
1.25
1.50
1.75
2.00

10
10
14
10
12
12
10

25.0
25.5
24.9
23.3
21.5
20.2
19.6

2.3
2.7
3.1
2.8
2.4
1.9
1.8

2.5  106

806

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

Fig. 4. (a) The experimental set-up with only vertical and with vertical and horizontal LVDTs. The indications of the latter were not evaluated due to their scattering. (b) A
fractured specimen with strain gauge rosettes. Four rosettes were used at the ends of two mutually perpendicular diameters. (c) Typical cubic and (d) cylindrical specimens
after having been tested.

with existing structural elements of the monument. However, for comparison, additional tests were carried out with minimized friction between the specimens and
the loading platens. The minimization of friction was achieved by inserting two
sheets of Polytetrauoroethylene (PTFE), a thick one (thickness 2 mm) and a thinner
one (thickness 0.5 mm), between the platens and the bases of the specimens in the
series: Platens, thick PTFE sheet, thin PTFE sheet, specimens base. An amount of
stearic acid was placed both between the thick PTFE sheet and the loading platens
as well as between the two PTFE sheets. Stearic acid was used since it is very efcient in reducing friction under high loads. The role of the second PTFE sheet (the
thin one) was to prevent penetration of stearic acid particles in the pores of the
specimen, since in this way transverse strains are developed, which change locally
the stress eld. This technique was also used successfully by Exadaktylos et al. [14]
for Dionysos marble specimens and by Kourkoulis et al. [15] for Conchyliates
shellstone. The comparative study (lubricated versus non-lubricated specimens)
indicated that using lubrication not only causes quantitative changes of the
mechanical constants but also it changes the failure mode from the familiar one
of the Mohrs cones to an almost axial splitting one. This change is a serious problem for the restoration process: Although friction can be eliminated in laboratory
tests using lubrication, this possibility does not exist in real life situations where

any kind of lubricant becomes inactive with time. Therefore in case the conclusions
of an experimental programme are to be applied in the restoration praxis lubrication should be applied with scepticism.
The tests were carried out with the aid of a very stiff hydraulic Amsler loading
frame of capacity 1 MN, equipped with an electronic data acquisition and storage
system. The stiffness of the frame was considered innite since the maximum load
recorded did not exceed 1/5 of the capacity of the frame in any case. This is very
important in case the post-peak behaviour of the material is to be studied. The load
was applied quasi-statically and the displacement-control procedure was chosen.
The rate of displacement of the traverse was equal to 2.5  106 m/s and was kept
constant for all the experiments without exception. Typical fractured specimens are
shown in Fig. 4c and d.

3.2. Experimental results

Characteristic axial stress axial strain curves are plotted in Fig. 5. It can be
seen that for the majority of the experiments and for relatively small strains the
constitutive law is almost linearly elastic. This linearity is terminated by the

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

appearance of some axial surface cracks almost parallel to the longitudinal axis of
the specimens and soon afterwards a rather abrupt drop of the stress level is observed. In some cases, it is possible that the stress recovers and it increases again
exceeding, in some cases, even the level of the initial peak value. However, in general after the load drop the stressstrain curves become almost horizontal or
slightly inclined with small negative slope up to the nal disintegration of the specimen. A similar behaviour of the re graph has been observed for another porous
natural building stone, the Conchyliates shellstone [15,16]. It was concluded that
two competitive failure mechanisms are present: The rst one accounts for the failure due to cracking while the second one accounts for the failure due to the local-

(a)

40
Small Cylindrical specimens

(MPa)

30

20

10

0
0

0.02

0.04

0.06

(b) 20
Sm all Cu b ic sp ecim en s

(MPa)

15

10

807

ization of the deformation caused by the failure of successive material layers. This
second mechanism seems to be described accurately enough [17] by the honeycomb crushing model that was developed by Papka and Kyriakides [18].
By appropriate analysis of the raw experimental data the basic mechanical
properties of the material were determined. The quantities calculated are the peak
stress, rpeak, the modulus of elasticity, E, the strain at the peak stress, epeak, the
stress drop after the peak load, rd and the strain energy density absorbed by the
material up to the peak load or the rst visible crack (Eel). The above quantities
are shown schematically in Fig. 6. In order to study the size and the shape effects,
the mean values of the above quantities were plotted both versus the characteristic
size as well as versus the volume of the specimens.
In Fig. 7a, the variation of the peak stress is plotted for both cubic and cylindrical specimens versus either the height of the cylindrical specimens or the length of
the edge of the cubic ones. It can be seen from this gure that the dependence of the
peak stress on the dimensions of the specimens is different for cylinders and cubes:
For the cylindrical specimens the dependence is almost monotonous: Increasing the
height of the cylinders the peak stress decreases following a sigmoid curve. For the
smallest specimens (h = 40 mm) the value of the peak stress is equal to about
25 MPa while for the largest ones (h = 160 mm) the respective value is less than
20 MPa, i.e. there is a change of about 20%. For the cubic specimens, however the
situation is completely different: The peak stress increases with increasing dimensions, it reaches a maximum value equal to about 28 MPa for the cubes with edge
equal to 75 mm and then it decreases. The relative change for the range tested here
is more than 45%. The variation of the peak stress versus the volume of the specimens is plotted in Fig. 7b. The dependence is obviously the same and only the relative position of the two curves is slightly different.
In Fig. 8a, the variation of the modulus of elasticity is plotted versus the characteristic dimension for both types of specimens. The results are striking because
it is generally believed that this quantity is more or less insensitive to the changes
of the size of the specimens. However, the present series of experiments indicates
that at least for the soft porous stone studied here the modulus of elasticity depends
dramatically on the size of the specimens: For cylindrical specimens its values vary
from about 3 GPa to more than 8 MPa while for the cubic ones the respective values
vary between 3.5 GPa and 9 GPa. There is a change of more than 200% (!!) for both
cases. In addition it should be noted that the variation of the modulus of elasticity is
identical for both specimens types: Its values increase rapidly as the specimens
sizes increase and then they tend to become constant at a value around 9 GPa.
The variation of the elastic modulus versus the volume of the specimens is plotted
in Fig. 8b. Again the dependence is the same and only the stabilization tendency is
much more evident.
The variation of the remaining mechanical constants versus the size of the specimens is plotted in Figs. 911 for the strain energy density up to the peak load, the
stress drop after the peak strain and the strain at the peak stress, respectively. The
conclusions are more or less the same: The dependence of these quantities on the
size of the specimens is dramatic and in the case of the cubic specimens it is, also,
non-monotonous.

4. Discussion and conclusions

0
0

0.02

0.04

0.06

Fig. 5. Axial stress versus axial strain for some characteristic tests: (a) small cubes
(a = 50 mm) and (b) small cylinders (h = 40 mm).

The mechanical behaviour and failure of the Kefalonia porous

stone as well as the dependence of its mechanical constants on
the size and the shape of the specimens were studied. The most
striking conclusion of the study is that the above dependence appears to be irregular, at least for some of the parameters studied.
Indeed for the cubic specimens for which the similarity is pre-

E
peak

el

peak

Fig. 6. Denition of the quantities obtained experimentally.

808

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

Peak stress [MPa]

(a)

30

20

10
Cubic specimens
Cylindrical specimens
0
0

0.05

0.1

0.15

0.2

served (size effect) both the compressive strength (expressed in

terms of the peak stress recorded just before the rst stress drop
or the formation of the rst visible crack on the surface of the specimen) of this material as well as the strain energy density absorbed
up to the peak load depend on the size of the specimens according
to a clearly non-monotonic law.
A similar non-monotonic behaviour for the compressive
strength is also exhibited by Conchyliates shellstone, the material used for the erection of the Zeus Temple at Olympia [19]. In
this case, six classes of cylindrical specimens of aperture (ratio of
height to diameter) constantly equal to two were tested. In other
words, the similarity was preserved. The height of these specimens

Peak stress [MPa]

(b)

Strain energy density up to the peak load [MPa]

Characteristic dimension (Cylinders: height, cubes: edge) [m]

30

20

Cubic specimens
Cylindrical specimens
10

0
0

0.0005

0.001

0.0015

0.3
Cylindrical specimens
Cubic specimens

0.2

0.1

0
0

0.002

0.05

0.1

0.15

0.2

Characteristic dimension (Cylinders: height, cubes: edge) [m]

Volume of the specimen [m^3]

Fig. 7. The dependence of the peak stress on the size of the specimens. (a) Peak
stress versus the characteristic dimension. (b) Peak stress versus the volume of the
specimen.

Fig. 9. The dependence of the strain energy density up to the peak load on the size
of the specimens.

12
Cubic specimens
Cylindrical specimens

7.5

[MPa]

Modulus of elasticity [GPa]

(a) 10

5
Cylindrical specimens
Cubic specimens

2.5

0
0

0.05

0.1

0.15

0.2

0.1

0.15

0.2

Fig. 10. The dependence of the stress drop on the size of the specimens.
10

0.015

7.5

Cylindical specimens
Cubic specimens
5

Peak strain

Modulus of elasticity [GPa]

0.05

Characteristic dimension (Cylinders:height, cubes: edge) [m]

Characteristic dimension (Cylinders: height, cubes: edge) [m]

(b)

Cubic specimens
Cylindrical specimens
2.5

0.01

0.005

0.0005

0.001

0.0015

0.002

Volume of the specimens [m^3]

0
0

Fig. 8. The dependence of the modulus of elasticity on the size of the specimens. (a)
Modulus of elasticity versus the characteristic dimension. (b) Modulus of elasticity
versus the volume of the specimen.

0.05

0.1

0.15

0.2

Characteristic dimension (Cylinders: height, cubes: edge) [m]

Fig. 11. The dependence of the peak strain on the size of the specimens.

809

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

was equal to 100 mm, 150 mm, 200 mm, 250 mm and 300 mm.
The sixth class included much bigger specimens of height
600 mm and diameter 300 mm. The variation of the peak stress
versus the height of the specimens is plotted in Fig. 12. The similarity with the respective variation of the cubic specimens of the Kefalonia stone is obvious. In the embedded photograph a piece of
Conchyliates is shown exhibiting the strongly inhomogeneous
nature of the specic material.
For the sake of completeness it is to be mentioned here that the
results of other studies as for example the very recent experimental one by del Viso et al. [20] with cubic concrete specimens are not
in accordance with these for the cubic specimens of the present
study: The variation of the strength with increasing specimens
size appears to be monotonic. It could be therefore argued here
that the non-monotonic dependence observed here is inherent
only for relatively soft natural building stones. However, a study
by Vardoulakis et al. [21] with cylindrical specimens made from
Dionysos marble, the material used for the restoration of the Parthenon Temple of Athens, proved that the dependence of the peak
stress on the size of the specimens is again non-monotonic
(Fig. 13). Moreover, Vardoulakis and Kaklis [22] proved that the
same irregularity appears for the tensile strength of this marble,

Peak stress [MPa]

0
0

0.2

0.4

0.6

Height [m]
Fig. 12. The dependence of the peak stress on the height of the specimens for
Conchyliates shellstone.

as it is obtained from Brazilian tests (Fig. 14), although in this case

the increased scattering of the results masks the phenomenon.
Concerning the cylindrical specimens (non-preservation of the
similarity) the size effect appears to be monotonic. It is therefore
obvious that (at least in their present form) none of the theoretical
approaches mentioned in paragraph 2 can described in a satisfactory manner the present series of experiments. In fact the monotonous behaviour of the cylindrical specimens (shape-effect) seems
to be in accordance with the predictions of the energetic-statistical
theory [9,10] for the nominal strength at failure for specimens of
different height/diameter ratio. On the contrary the behaviour of
the cubic specimens (preservation of similarity) could be described
by the fractal theory of size effect [11,12].
Concluding it can be said that denite conclusions for the sizeand shape-effects for natural building stones cannot yet be drawn
and more experimental evidence is required. In this direction, a
study in progress attempts to answer the question related to the
tendency of the curves for even smaller height to diameter ratios.
Unfortunately the Kefalonia porous stone and the Conchyliates
shell-stone could not be used for the preparation of very small
specimens due to their remarkable inhomogeneity. Therefore the
Cyprus porous stone (another candidate substitute of the Kenchreae stone) was used since it is of extremely homogeneous nature.
Cylindrical specimens of constant diameter D = 10 cm and heightto-diameter ratios, r, ranging from 0.25 to almost 2.5 were subjected to dry uniaxial compression. Preliminary results, concerning
the dependence of the fracture stress on r are plotted in Fig. 15. It is
indicated that for very small specimens (r < 0.5) the fracture stress
increases dramatically with decreasing r as it is perhaps expected
taking into account the fact that no lubricant was used. In addition
it is observed that for intermediate size specimens the dependence
is again rather non-monotonous.
Recapitulating it can be said that all conventional mechanical
constants depend more or less on the size of the specimen in a
non-uniform manner (at least for the materials tested here).
Clearly their practical use becomes rather inefcient. Therefore,
the question arises whether a mechanical constant exists which
does not depend on the size of the specimen. In a recent study,
Kourkoulis et al. [15] studied the variation of the accumulated
strain energy density, Wel, versus the axial strain for the Conchyliates shell-stone, assuming that the working-load portion of the
re curve of such materials is restricted to only a part of the re
graph, well below the peak stress. The respective results are plotted in Fig. 16, for some classes of specimens. It is very interesting

120

10

Lubricated tests

Tensile strength [MPa]

Compressive strength [MPa]

Dry tests

80

40

150

300

450

Height [mm]
Fig. 13. The dependence of the compressive strength of Dionysos marble on the
size of the specimens.

40

80

120

Diameter [mm]
Fig. 14. The dependence of the tensile strength of Dionysos marble on the size of
the specimens.

810

S.K. Kourkoulis, E. Ganniari-Papageorgiou / Construction and Building Materials 24 (2010) 803810

size effect theories in their present form. A different approach is required which will take into account among others the internal
structure of the material, based on the data of additional larger series of experiments, especially in the case of strongly inhomogeneous materials (like the ones studied here).

100

Peak stress (MPa)

75

Acknowledgements
The authors express their gratitude to the late Professor Ioannis
Vardoulakis of the Department of Mechanics of the National Technical University of Athens for critical discussions during the preparation of the paper. His deep knowledge on the mechanical
behaviour of soft rock-like geomaterials was extremely valuable
for the authors. The assistance of Mr. Nikolaos L. Ninis of the Committee for the Preservation of the Epidauros Monuments in the execution of the experiments and the evaluation of the respective
results is also gratefully acknowledged.

50

25

0
0

0.5

1.5

2.5

r
Fig. 15. The inuence of the ratio r (height over diameter) on the fracture stress
under uniaxial compression for the Cyprus porous stone.

0.0045

Wel (MJ/m^3)

CS5:
y = 2.13x - 0.002

0.003

CS3:
y = 1.95x - 0.0006
CS1:
y = 1.84x - 0.0004
0.0015

0.001

0.002

0.003

Fig. 16. The variation of the strain energy density versus the axial strain for the
working-load portion of the re graph for Conchyliates shell stone.

to notice that the slopes of these lines vary in a rather restricted

range for all classes of specimens, indicating that this quantity is
not seriously affected by the size of the specimens. This observation can be partly explained by taking into account that the energy
density includes, albeit in an indirect manner, the effect of deformation in all three directions. In analyzing axial stress axial strain
data the inuence of specimen size, affecting the tendency of the
material to dilate and expand perpendicularly to the loading direction and determining its mode of failure, manifests itself in the variation of mechanical strength data. By using strain energy data
instead, this phenomenon is alleviated since the externally
measured energy input is the net sum of the energy input needed
to deform axially the specimen (assuming uniform deformation
and element material properties) and the energy released by internal deformation of the specimen (differentiated zones of contraction and dilation).
Coming to an end it can be said that both the size- and the
shape-effect of relatively soft porous building stones as well as
their failure mechanisms can not be described in a unied manner
either in terms of classical Strength of Materials or by using the

References
[1] Kourkoulis SK, Ninis NL. On selecting a compatible substitute for the
Kenhcreae porous stone used in the Epidaurean Asklepieion. In: Aifantis EC,
Kounadis AN, editors. Proc 6th national congress on mechanics. Thessaloniki,
Hellas: Giahoudi & Giapouli; 2001. p. 34856.
[2] Ninis NL, Kourkoulis SK. On determining appropriate parameters of
mechanical strength for numeric simulation of building stones. In: Hatzor
YH, editor. Stability of rock structures. Lisse, The Netherlands: A.A. Balkema
Publishers; 2002. p. 15362.
[3] Carpinteri A, Chiaia B. Embrittlement and decrease of apparent strength in
large-sized structures. Sadhana 2002;27(4):42548.
[4] Grifth AA. The phenomenon of rupture and ow in solids. Phil Trans R Soc
London 1921;A221:16398.
[5] Carpinteri A. Decrease of apparent tensile and bending strength with specimen
size: two different explanations based on fracture mechanics. Int J Solids Struct
1989;25:40729.
[6] Fisher RA, Tippett LHC. Limiting forms of the frequency distribution of the largest
and smallest member of a sample. Proc Cambridge Phil Soc 1928;24:18090.
[7] Weibull W. The phenomenon of rupture in solids. Proc R Swed Inst Eng Res
1939;153:155.
[8] Walsh PF. Fracture of plain concrete. Indian Concr J 1972;46:46976.
[9] Bazant ZP. Size effect in blunt fracture: concrete, rock, metal. J Eng Mech
1984;110:51835.
[10] Bazant ZP, Yavari A. Is the cause of size effect on structural strength fractal or
energetic-statistical? Eng Fract Mech 2005;72:131.
[11] Carpinteri A. Fractal nature of material microstructure and size effects on
apparent mechanical properties. Mech Mater 1994;18:89101.
[12] Carpinteri A, Chiaia B, Cornetti P. On the mechanics of quasi-brittle materials
with a fractal microstructure. Eng Fract Mech 2003;70:232149.
[13] Kim J-K, Yi S-T. Application of size effect to compressive strength of concrete
members. Sdhan 2002;27:46784.
[14] Exadaktylos GE, Vardoulakis I, Kourkoulis SK. Inuence of nonlinearity and
double elasticity on exure of rock beams II. Characterization of Dionysos
marble. Int J Solids Struct 2001;38:411945.
[15] Kourkoulis SK, Ninis NL, Bakolas V. The mechanical behaviour of Conchyliates
shellstone the size effect. In: Sih GC, Spyropoulos CP, editors. Proc int conf on
inuence of traditional mathematics and mechanics on modern science and
technology, Messini, Greece. Athens: Eptalofos (SA); 2004. p. 30715.
[16] Vardoulakis I, Kourkoulis SK, Zambas C. Modeling of the mechanical behaviour
of a conchyliates shellstone. In: Evangelista A, Picarelli L, editors. The
geotechnics of hard soils soft rocks. Rotterdam: Balkema; 1998. p. 91122.
[17] Vardoulakis I, Kourkoulis SK, Skajerstein A. Post-peak behaviour of rocks and
natural building stones in uniaxial compression. In: de Borst R, van der Giessen
E, editors. Material instabilities in solids. Chichester: J. Wiley & Sons; 1998. p.
20726.
[18] Papka DP, Kyriakides S. In-plane crushing of a polycarbonate honeycomb. Int J
Solids Struct 1998;62(3-4):23967.
[19] Kourkoulis SK. An experimental study of the mechanical behaviour of the
Conchyliates shell-stone: some irregularities of the size effect. Strain, in
press, doi:10.1111/j.1475-1305.2009.00618.x.
[20] del Viso JR, Carmona JR, Ruiz G. Shape and size effects on the compressive
strength of high strength concrete. Cem Concr Res 2008;38:38695.
[21] Vardoulakis I, Kourkoulis SK, Exadaktylos GE, Rosakis A. Mechanical properties
and compatibility of natural building stones of ancient monuments: Dionysos
marble (in Greek). In: Varti-Mataranga M, Katsikis Y, editors. Proc
interdisciplinary workshop The building stone in monuments. Athens
(Greece): IGME Publishing; 2002. p. 187210.
[22] Vardoulakis I, Kaklis J. An experimental investigation of the size effect in
indirect tensile test on Dionysos marble. In: Kounadis A, Providakis C,
Exadaktylos G, editors. Proc 7th national congress on mechanics. Greece:
Chania-Creta; 2004. p. 1517.