International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Application of a new constitutive model to the analysis of plate load

tests in a pyroclastic rock
Angelo Amorosi a,1, Stefano Aversa b,2, Daniela Boldini c,n, Anita Laera d,3,
Marco Valerio Nicotera e,4
a
Politecnico di Bari, Department of Civil, Environmental, Construction and Chemical Engineering, via Orabona 4, 70125 Bari, Italy
b
Università di Napoli Parthenope, Department of Engineering, Centro Direzionale di Napoli, Isola C/4, 80143 Napoli, Italy
c
Università di Bologna, Department of Civil, Chemical Environmental, and Materials Engineering, viale Terracini 28, 40131 Bologna, Italy
d
Plaxis bv, Research, P.O. Box 57, 2600 AN Delft, The Netherlands
e
Università di Napoli Federico II, Department of Civil and Environmental Engineering, via Claudio 21, 80125 Napoli, Italy

art ic l e i nf o a b s t r a c t

Article history: This paper deals with a new critical state-based constitutive model for soft rocks and with its application
Received 2 February 2015 to the analysis of the response of a pyroclastic rock during in situ plate load tests. The model, formulated
Received in revised form in the single-surface plasticity framework, is characterised by the following main features: (i) a gen-
6 July 2015
eralised three-invariant yield surface capable of reproducing a wide set of well-known criteria, (ii) the
Accepted 13 July 2015
dependency of the elastic stiffness on the current stress state by means of a hyperelastic formulation and
(iii) the ability of simulating the plastic strain driven structure degradation processes by a set of ap-
Keywords: propriate isotropic hardening laws. The constitutive model was implemented in a commercial Finite
Pyroclastic rock Element code by means of an explicit modified Euler scheme with automatic sub-stepping and error
Constitutive modelling
control. The procedure does not require any form of stress correction to prevent drift from the yield
Numerical modelling
surface. The model was applied to simulate the response of a pyroclastic rock, the Neapolitan Yellow Tuff,
Plate load tests
to in-situ plate load tests conducted by 500 mm and 300 mm circular plates. In particular, in each lo-
cation a first test was carried out adopting the large plate, applying a loading and unloading cycle; this
was followed by a second loading stage performed on the same portion of rock by the smaller plate up to
larger stress levels. Test results pointed out some specific features of the rock response under such
loading conditions, including non-linear elastic behaviour and structure degradation, this latter high-
lighted by the overall reduction of the shear strength parameters. The numerical analyses showed a fairly
good agreement with the in-situ experimental data, substantiating the relevance of the selected con-
stitutive assumptions for the soft rock under investigation.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction behaviour or – for larger confining pressures – a ductile one.2 In

The mechanical behaviour of soft rocks is characterised by
features that fall in between those of soils and rocks.1 In fact, ir-
respectively of their depositional environment (e.g.: sedimentary,
volcanic), this class of materials shows non zero tensile strength
and related cohesion together with pressure dependant elastic
response and shear one, this latter leading to either fragile
n
Corresponding author. Fax: þ 39 051 2090247.
E-mail addresses: angelo.amorosi@poliba.it (A. Amorosi),
stefano.aversa@uniparthenope.it (S. Aversa), daniela.boldini@unibo.it (D. Boldini),
a.laera@plaxis.com (A. Laera), nicotera@unina.it (M. Valerio Nicotera).
1
Fax: þ39 080 5963675.
2
Fax: þ39 081 5476777.
3
Fax: þ31 15 2573 107.
4
Fax: þ39 081 5938936.
the last decades a number of experimental observations have
shown that the mechanics of soft rocks, like calcarenites and tuffs,
is significantly affected by the so called structure effects.3 Their
initial stiffness and strength are considerably affected by the initial
interparticle bonding, whose plastic strain-induced damage leads
to significant modification of the overall behaviour of the
material.4 The above experimental evidences have triggered the
interest towards the formulation of new constitutive models spe-
cifically devoted to soft rocks.5–7 Among others, it is worth men-
tioning the significant contributions by Nova and co-workers8–11,
aimed at casting into the critical state soil mechanics framework
the above soft rocks' mechanical features by means of hardening
plasticity theory.
The specific soft rock studied in this paper is the Neapolitan
Yellow Tuff (NYT). The city of Napoli (Italy) is located in the middle
of a wide volcanic region formed by the Vesuvius and the

http://dx.doi.org/10.1016/j.ijrmms.2015.07.003
1365-1609/& 2015 Elsevier Ltd. All rights reserved.
272 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282
Phlegrean Fields districts at the South-East and West of the town
respectively. The subsoil of the urban area mainly consists of
pyroclastic soils and rocks originated by the eruptions of the vol-
canic district of Campi Flegrei. Pyroclastic soils (pozzolana) gen-
erally overlay the lithic part of the Neapolitan Yellow Tuff and
other pyroclastic tuffs. It follows that shallow and deep founda-
tions, in the urban area of Napoli, are frequently seated on top of
an underlying tuff layer. The design of such foundations is typically
based on empirical approaches.
Pellegrino12 investigated on the mechanical response of small
diameter circular foundations on NYT by means of plate load tests
performed on the formation during the construction of a tunnel
excavated in the urban area of the city. In particular, loading and
unloading tests were first carried out along seven different ver-
ticals by means of a circular plate of 500 mm of diameter, up to a
maximum average vertical stress equal to 9 MPa. Those tests were
then followed by a coaxial reloading performed at the same lo-
cations adopting a smaller plate (d ¼300 mm), up to a maximum
average vertical stress of 25 MPa.
At that time no theoretical framework for describing the con-
stitutive relationship of a soft rock (i.e. the tuff) and related ad-
vanced numerical tools for simulating plate load tests were
available. In fact, the tests results were interpreted by means of the
finite element (FE) method assuming a non linear elastic con-
stitutive model.12 The numerical analysis performed by the Author
was not very satisfactory: for a single set of parameters he could
only manage to simulate the average behaviour observed during
the first series of plate load tests (d ¼500 mm), while failing to
reproduce the subsequent series of coaxial plate load tests per-
formed by the smaller plate (d¼ 300 mm).
More recently other researchers13 attempted to simulate the
same set of data performing FE analyses adopting a Cam-Clay based
model to account for some more recent findings on the mechanical
behaviour of Neapolitan tuff.14–16 The Authors adopted different
sets of parameters for the same tuff, depending on the presumed
degree of destructuration3,4,15 induced by the loading. This simpli-
fied approach was meant to account for the mechanical effects of
the first loading cycle on the subsequent plastic-strain-induced
damaged behaviour observed during the smaller plate’s tests.
In the following a new critical state-based hardening-plasticity
model for soft rock is proposed. The formulation is aimed at re-
producing some of the specific features of soft rock response, as the
non-linear elastic behaviour, the pressure dependency of the shear
response and the structure degradation, this latter implying a re-
duction in dimensions of the yield surface capable of modifying the
yield stress under isotropic compression and a modification of the
tensile and frictional components of the strength. The model relies
on a flexible formulation of the yield surface, originally proposed by
Bigoni and Piccolroaz17, which allows to adapt its shape under both
meridian and deviatoric sections to the available experimental data.
The model is then tentatively applied to analyse, by means of the FE
approach, the results of the in situ plate load tests performed by
Pellegrino.12
2. Physical properties and mechanical behaviour of Neapolitan
yellow tuff
Many pyroclastic rocks and soils outcrop in the Campania Re-
gion; their presence is primarily related to volcanic activity in the
Phlegrean Fields and only in a few cases of Vesuvius. Among the
rocks, the Neapolitan Yellow Tuff (NYT) is the most common in
the first hundred metres below the ground level, covering ap-
proximately 300 km2. The NYT forms the most important hills in
the Phlegrean areas and in the city of Napoli. This material origi-
nated approximately 11–13 kyears b.p. during one or more
volcanic eruptions in the Phlegrean Fields.18–20
NYT has been studied extensively at the University of Napoli
Federico II: the first systematic experimental investigation on its
mechanical behaviour dates back to more than forty years ago.2,12
A number of significant contributions were added starting from
the 80 s, once new experimental and theoretical tools were
available to study this peculiar soft-rock.14–16 Comprehensive re-
ports on the NYT physical and mechanical properties were pub-
lished by Evangelista and Aversa16 and Evangelista and co-
workers18
The structure of NYT is characterized by a fine matrix with
pumiceous inclusion of small sizes. Lithic inclusions are frequent
and their diameter is generally less than some millimetres. This
soft rock has a porosity ranging from 0.4 up to 0.6 and a dry unit
weight in the range from 10 kN/m3 to 14 kN/m3. The uniaxial
compressive strength of NYT varies from about 1 MPa up to
10 MPa and is strongly related to the dry unit weight but it is also
significantly affected by other structural features. Pellegrino2,
analysing the mechanical response of NYT and other tuffs as
emerged from isotropic and triaxial compression tests, observed
that they may show a rock-like or a soil-like behaviour as the
mean effective stress varies from low to medium and high values.
In the first case the tuff essentially behaves elastically up to plastic
yielding or failure while in the second one it shows a ductile be-
haviour characterised by accumulation of plastic strain from the
beginning of the test, similarly to a normally compressed and
unbounded clayey soil. In particular, the Author concluded that for
isotropically compressed triaxial tests the transition from the rock-
like to the soil-like behaviour occurred when the mean effective
stress achieved during the consolidation stage approximately
doubled the uniaxial compressive strength. Finally, the Author
sketched in the triaxial stress space the curve that bounded the
rock-like domain; this curve actually identifies the stress states
that produce the destructuration of the tuff, in accord with the
general framework proposed by Leroueil and Vaughan3 twenty
years later.
Nova8 carried out a first attempt to model the mechanical beha-
viour of the tuff by means of hardening plasticity theory. Subse-
quently, other researchers15,21 pointed out the influence of destruc-
turation phenomena on the mechanical behaviour of tuff. In parti-
cular, the experimental results obtained by Aversa and Evangelista15
demonstrated that during an isotropic compression test positive
hardening and degradation of the tuff developed at the same time.
These two competing phenomena resulted in both an expansion of
the yield surface and a contraction of its cross section in the meridian
plane (i.e. a reduction of the critical state soil mechanics parameter
M), leading to an overall reduction of the normalised shear strength
of the material. It results that both size and shape of the yield surface
of the tuff are modified by the strain history experienced by the
material.
The mechanical behaviour of the specific NYT at the test site
discussed in this paper was studied by means of laboratory tests
by Pellegrino.12 The material is characterised by a specific gravity
GS ¼ 2.47, a dry unit weight γd ¼12.0 kN/m3 and a void ratio
e¼1.20.
Uniaxial compression tests were extensively carried out, lead-
ing to a mean value of about 4.4 MPa (Fig. 1).
A series of drained triaxial tests were also performed, iso-
tropically consolidating the samples up to mean stress p ranging
from 1.1 MPa to 10.2 MPa. This set of data proves to be reliable in
terms of yield points identification, while providing less significant
information for large strain shearing, as most of the specimen
experienced significant strain localisation leading to shear band-
ing. The yield points as deduced by a careful analysis of the triaxial
data are shown in Fig. 2, together with a possible interpolating
yield surface, described in details in Section 6 of this paper.
 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282
Fig. 1. Uniaxial compression tests: comparison between some experimental results
and the numerical simulation.
Fig. 2. Yielding points and initial yielding surface in the meridian plane.
Fig. 3. Isotropic compression test: comparison between some experimental results
and the numerical simulation.
It is worth analysing the isotropic compression stages of all the
triaxial tests, reported in a compressibility plot in Fig. 3. Each test
shows a clear bend in the compression curve, indicating the oc-
currence of yielding under isotropic stress conditions, thus com-
pleting the picture sketched for deviatoric stresses in Fig. 2.
3. Plate load tests
The in situ plate load tests were carried out forty years ago by
Pellegrino.12 The tests were performed during the construction of
a tunnel excavated under one of the hill of the city of Napoli.
A massive steel reaction beam (5.40 m long and 3.30 m high),
273
shown in Fig. 4, was assembled inside the tunnel section (diameter
of about 12 m). The reaction beam was disposed on a rail track in
order to move it, thus allowing the tests to be carried out at dif-
ferent position. The load tests were performed at eight different
locations at the vertex of a rectangular grid formed by square
meshes 4 m  4 m. At each position the reaction beam was an-
chored by means of two drilled micropiles (∅ 220 mm, length
15.50 m) predisposed at the vertex of a regular grid analogous to
the one individuating the loading position; however the two grids
were disposed such that the minimum distance between the re-
action micropiles and the loading position was equal to 2 m. The
upper portion of each reaction micropile was cased by a metal pipe
to minimise the interaction between its shafts and the rock vo-
lume affected by the loading test. A hydraulic jack controlled by a
manual pump was used to apply the load (maximum total load
equal to 1.87 MN). The applied load was measured by a load cell
interposed between the jack piston and the reaction beam. The
displacements were measured by an array of dial gauges fixed on a
reference steel beam 7 m long disposed across the loading area.
Seven load tests were performed on a 500 mm diameter plate;
in these tests the mean load reached a value of 9 MPa. Seven more
tests were then performed at the same position considered in the
previous tests but on a 300 mm diameter plate; in these cases the
tests were driven up to failure. The load–displacement curves are
reported in Fig. 5.
A relative large scatter can be observed between the curves
obtained from the tests carried out by the 500 mm diameter plate;
this scatter is probably due to some bedding irregularities affecting
the contact between the loading plate and the tuff surface.
The results from the tests on 300 mm diameter plate seem to
be less scattered as a likely consequence of performing these tests
within the punch mark of the 500 mm diameter plate tests (see
Fig. 4b and c). The results of tests from C to H are qualitatively
similar, even if the observed stiffness and maximum settlements
were rather scattered probably due to the natural heterogeneity of
the material. On the contrary the load–displacement curve of tests
A and B are different from the others.
Test A was, indeed, the only test in which a clear collapse oc-
curred at relatively low mean vertical stress. It is likely that such
behaviour is related to the presence of some pre-existing dis-
continuities or inhomogeneity in the loaded tuff volume; however,
due the absence in the test report of any information other than
the tables indicating the applied loads and the measured settle-
ments, this hypothesis cannot be confirmed and hence it was
decided to not consider it in the following.
Test B was performed directly with the 300 mm diameter plate,
i.e. on a virgin volume of tuff. The load–displacement curve of test
B is substantially different from those observed in all the other
tests (see Fig. 5b). In particular the apparent yield stress is lower,
together with a larger slope of the following portion of the curve,
while the collapse load is not dissimilar from that observed in the
other tests. These differences can be related to the effects of two
competing phenomena that occur during the loading tests: volu-
metric hardening, which enhances the stiffness and strength of the
soft rock, and destructuration, which induces an opposite me-
chanical effects, together with a possible modification of the
shape of the yield surface, this latter observed by Aversa and
Evangelista15 and by Aversa and Lagioia.21 The above competing
phenomena affect the response of the tuff undergoing tests C to H
in a different way as compared to what occurs in test B, as in this
latter the absence of the preloading stage - by the 500 mm plate -
leads to an initially prevailing effect of the destructuration as
compared to that of positive hardening, thus modifying the overall
response.
The above qualitative interpretation of the results is subjected
to a quantitative validation in this paper, as the above features of
274 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282

Fig. 4. Views of the testing apparatus.13

Fig. 5. Load–settlement curves obtained during the plate loading tests performed with the 500 mm plate (a) and the 300 mm plate (b).

the mechanical behaviour of the soft rock are accounted for in the be made to tests C and D, as showing a consistent response for
proposed model and their role in the specific boundary value both 500 and 300 mm plate loading conditions, and to test B, due
problem under study is explored by means of FE analyses. to its specific – 300 mm only - testing conditions.
In the following, for the sake of simplicity, reference will only
 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282 275

4. Constitutive model: mathematical formulation ∂W ( εe) ∂ 2W ( εe)

σ ( εe) = C e ( εe) =
∂εe ∂εe ⊗ ∂εe (4a,b)
In this section the isotropic hardening model for weak rocks
(WR1) is described. The model, formulated in the single-surface A yield function F: S × R nint → R is defined such that the ad-
plasticity framework, is characterised by the following main fea- missible states in the stress space {σ , q} ∈ S × R nint are all those
tures: (i) a generalised three-invariant yield surface capable of which fall into the following convex set:
reproducing a wide set of well-known criteria17, (ii) the de- Eσ : = { (σ, q) ∈ S × R nint F (σ, q) ≤ 0} (5)
pendency of the elastic stiffness on the current stress state by
means of a hyperelastic formulation22 and (iii) the ability of si- The internal part and the boundary of this set represent the
mulating the plastic strain driven structure degradation processes elastic domain
by a set of appropriate isotropic hardening laws.23,24 It is for-
int Eσ : = { (σ, q) ∈ S × R nint F (σ, q) < 0} (6)
mulated in the framework of infinitesimal single-potential and
rate-independent elasto-plasticity, with reference to fully satu- and the yield surface
rated soils in isothermal conditions.
∂Eσ : = { (σ, q) ∈ S × R nint F (σ, q) = 0} (7)
In the following the soil mechanics sign convention is assumed
and all stresses are effective stresses. All tensor quantities are respectively. The aforementioned vector q ∈ R nint defines the set of
written in boldface form, while scalar quantities are not. Con- the nint internal variables (hardening variables) which can be both
sidering the following Cartesian basis ei , ej , ek , el ∈ R 3, such that scalar or tensorial quantities.
first-order tensors (vectors) and second-order tensors can be The evolution of plastic strains ε p is controlled by the flow rule
written as a = ai ei and α = αij ei ej , the following tensor calculus
∂G (σ, q)
notation is adopted: a⋅b = ai bi , ab = a ⊗ b = ai bj ei ej , ε̇p = γ ̇
∂σ (8)
α⋅β = αik βkl ei el , α: β = αij βij , α ⊗ β = αij βkl ei ej ek el . The L2 − norm of
a vector and a second-order tensor are indicated by ‖a‖ = a⋅a and that of the internal variables q by the hardening rule
and ‖α‖ = α: α , respectively. If ∇ = ∂/∂xj ej is the nabla operator, q̇ = γ h
̇ (σ, q) (9)
∇a = ∂ai /∂xj ei ej = ai, j ei ej is the gradient of vector a . The symbols
1
1 = δ ij ei ej and I = 2 (δ ik δjl + δ il δjk ) ei ej ek el indicate the unit second- where G: S × R nint
→ R is the plastic potential, h: S × → R nint R nint
a prescribed hardening function and γ ̇ the plastic multiplier sub-
order identity tensor and the fourth-order symmetric identity
ject to the so-called Kuhn–Tucker conditions
tensor, respectively. The symbol δ ij is the Kronecker delta.
γ ̇ ≥ 0, F (σ, q) ≤ 0, γḞ (σ, q) = 0 and to the consistency condi-
Within the context of the three-dimensional theory, where
tion γḞ ̇ (σ, q) = 0.
Ω ⊂ R 3 is the continuum body of reference and u: Ω → R 3 is the
displacement vector corresponding to the generic point x ∈ Ω , the
4.2. Proposed constitutive model
infinitesimal strain tensor will be:
1⎡ T⎤ The reversible behaviour is described by a hyperelastic for-
ε = ∇s u : = ⎣ ∇u + (∇u) ⎦
2 (1) mulation proposed by Houlsby and co-workers22 to include a non-
linear dependence of the elastic stiffness on effective stress. The
The strain tensor (such as the stress tensor) is a symmetric
adopted free energy potential has the following expressions:
second-order tensor, that is a linear transformation in
S: = {ξ: R 3 → R 3 ξ = ξT } where S is the vectorial space in which W ( εve , εse )
ξ: ξ = ξij ξij . pa
= k n / (2 − 2n) (1 − n)(2 − n) / (1 − n)
The adopted stress invariants are the mean pressure 2−n
p: =
1
(σ : 1), the deviatoric stress q : = 3
‖s‖ (where s : = σ − p1 is 2 3g 2
3 2 [k (εve ) + (εse ) ](2 − n) / (2 − 2n)
1−n (10)
the deviatoric part of the effective stress tensor) and the Lode
angle ϑ =
1
cos−1(
3 3 J3
)
1
(with J2 = 2 s : s and J3 =
1 3
(s : 1)). The where εve
and εse
are the elastic volumetric and deviatoric strain
3 2 J23/2 3
invariants. pa is a reference stress corresponding to the atmo-
volumetric and deviatoric strain invariants are εv : = ε: 1 and spheric pressure, while n, k and g are dimensionless experimen-
1
εs : = 2 ‖e‖, where e: = ε − 3 εv 1 is the deviatoric part of the strain tally determined parameters. The first and second derivative of
3
tensor. Eq.(10) with respect to elastic strain tensor εe provide the
corresponding stress state σ and elastic stiffness Ce , respectively,
as defined in Eq. (4).
4.1. Basic assumptions
The selected yield locus that delimits the elastic domain was
formulated by Bigoni and Piccolroaz17 for a wide range of fric-
In the context of the elasto-plasticity theory the total strain rate
tional-cohesive materials including soils and rocks. It is a function
ε̇ can be decomposed additively in an elastic (reversible) part ε̇e
of the three stress invariants p, q and ϑ according to the following
and a plastic (irreversible) part ε̇ p :
equation:
ε̇ = ε̇e + ε̇p (2) q
F (σ, q) = f (p) +
In this way the rate form of the stress-strain relationship can be g (ϑ) (11)
written as: where the so-called meridian function
e p
σ̇ = C : ( ε̇ − ε̇ ) (3) p + pt
f (p) = − Mpc ( Φ − Φm) [2 (1 − α ) Φ + α ] , with Φ = pc + pt
where Ce is the fourth-order elastic stiffness tensor. If a stored
energy function W (εe) : Ω × S → R exists (i.e. the material is hy- ∈ [0, 1] (12)
perelastic), the elastic constitutive relationship and the elastic accounts for the influence of the mean pressure and the deviatoric
stiffness tensor can be written as follows: function
276 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282

1 ⎡ 1 ∂F ∂F ∂F ⎤
g (ϑ) = h = pc ⎢ (
−ξv exp −ηv εvd ) −ξs exp ( −ηs εsp ) ⎥
(
cos π /3 − βπ /6 − (1/3) cos−1(γ cos 3ϑ) ) (13) ⎣ λ˜−κ˜ ∂p ∂p ∂q ⎦ (17)
describes the dependency on the Lode angle. Similar isotropic hardening laws were formulated for the other
In the original formulation17, f(p) and g(ϑ) were considered as two internal variables accounting only for the strain-induced de-
fixed, non-evolving, functions related to the constant values of the gradation by means of negative exponential functions:
following 7 material parameters:
pṫ = pt ⎡⎣ −ξvc exp −ηvc εvd
( ) ε ̇ −ξ
d
v sc exp ( −ηsc εsp ) εṡ p ⎤⎦ (18)
1. M40, which controls the pressure-sensitivity.
2. pc 40 and pt Z0, corresponding to the isotropic yield com-
Ṁ = M ⎡⎣ −ξvM exp −ηvM εvd
( ) ε ̇ −ξ
d
v sM exp ( −ηsM εsp ) εṡ p ⎤⎦ (19)
pression pressure and the isotropic yield tensile pressure,
respectively.
Finally, the loading/unloading criterion is expressed by the
3. 0 o α o1 and m 41, which define the distortion of the meridian
Kuhn–Tucker complementary conditions γ ̇ ≥ 0, f (σ, α , σK )≤
section (p-q plane).
0, γ ̇ f (σ, α , σK ) = 0 , leading to the standard consistency condi-
4. 0 r β r2 and 0 r γ o1, which model the shape of the devia-
tion for the plastic multiplier γ ̇ as expressed by γ ̇ f ̇ (σ, α , σK ) = 0.
toric section.
The constitutive model was implemented in the commercial
Finite Element code Plaxis by means of an explicit modified Euler
On the contrary, in the proposed model, M, pc and pt were as-
scheme with automatic sub-stepping and error control.25
sumed as internal variables capable to control the dimension and
shape of the yield surface thanks to specific isotropic hardening
laws described hereafter.
5. Description of the numerical model
Depending on the values assumed by materials parameters and
internal variables, the function F can be scaled down to different
A schematic sketch of the numerical model is shown in Fig. 6,
well-known yielding criteria, such as Modified Cam-Clay and
together with the adopted finite element discretisation. Due to the
Mohr–Coulomb among others. In general, it is capable to assume
geometry of the considered problem, axisymmetric conditions are
very different shapes according to meridian and deviatoric dis-
selected; in particular, the axis of symmetry corresponds to the
tortional parameters. plate axis and is made coincident with the left side of the model.
It is worth noting, in addition, that the limit α o 1 was speci- In addition, no displacements are allowed at the base of the model,
fically introduced in this study, since using the limit α o2, as while only vertical displacements are admissible at the vertical
proposed by Bigoni and Piccolroaz17, would lead in certain cases to sides.
negative values of F also for stress points located outside the yield The rock material extends for a dimension of 5 m in the hor-
surface (this issue being particular important during the model izontal direction and of 5 m in the vertical direction below the
implementation in relation to the definition of the trial state). ground surface, distance that was found sufficient to avoid any
The flow rule controlling the direction and intensity of the in- boundary effects.
crement of plastic strain is here assumed associated: The action exerted by the plate is simulated by a constant
∂F displacement vertical field applied at the upper surface, having a
ε̇p = γ ̇ length coincident with the plate radius (i.e. 250 or 150 mm). The
∂σ (14)
corresponding loading is obtained by dividing the vertical reaction
The evolution of the internal variables during plastic de- forces recorded at the nodes where the displacement field is ap-
formation is described by the hardening rules. The isotropic plied for the total plate area. A finer discretisation is adopted in an
hardening law for the internal variable pc is a modified version of
that originally proposed by Kavvadas and Amorosi23:
⎡ 1 p ⎤
pċ = pc ⎢
⎣ λ˜−κ˜ v v (
ε ̇ −ξ exp −ηv εvd ) ε ̇ −ξ
d
v s exp ( −ηs εsp ) εṡ p ⎥ = γ ̇ h
⎦ (15)

It is composed by two volumetric terms and a deviatoric one.

The first volumetric term is similar to the standard Modified Cam
Clay hardening law, here defined with reference to a bilogarithmic
compressibility plot characterised by the slopes λ̃ (compression
index) and κ̃ (swell-recompression index). The following terms
account for the volumetric and deviatoric strain-induced structure
degradation by means of two separate exponential damage-type
forms, in which the parameters ξv, ηv , ξs, ηs control the rate of
the phenomena. The negative exponential functions let the de-
gradation process be initially more intense and characterised by
decreasing efficacy as plastic strains cumulate. The major differ-
ence between this hardening rule and that formulated in Kavvadas
and Amorosi23 consists in the use of an increasing function of the
volumetric plastic strain as controlling variable for the volumetric
strain induced de-structuring, defined as:
t
εvd = ∫0 εv̇ p dt
(16)

Thus, the scalar hardening function h is: Fig. 6. Geometry of the problem and finite element discretisation.
 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282 277

area of 0.5 m by 0.5 m around the zone where the displacement is

applied.
No water is considered in the model.
A similar numerical model was also set up for the simulation of
laboratory tests, such as the isotropic compression test, the uni-
axial compression test and the triaxial drained compression test.
In all cases, displacement-controlled tests were performed.

6. Calibration of the constitutive model and discussion of the

numerical results

The material parameters and the initial value of the state

variables were calibrated using both laboratory and plate loading
test results. In particular, in the following reference is made to the
plate loading tests C and D (Fig. 5), which show a consistent re-
sponse for both plate loading test conditions (diameter d ¼500
and 300 mm).
The initial dimension and shape of the yield surface in the
meridian section (Fig. 2) was obtained considering the yielding
points detected during isotropic compression tests and triaxial
drained compression tests. The shape of the surface resembles that Fig. 7. Comparison between the settlements curves obtained in the first part of the
of the Cam-clay model (m ¼2, α ¼0); for this reason, and due to plate loading tests C and D (500 mm plate) and the numerical simulation PLT1.
the lack of data expressing the influence of the Lode angle, con-
forming values of the parameters were selected for the deviatoric Table 1
section (β ¼1, γ ¼0). In addition, the fit of the yield points makes Final set of material parameters and initial values of state variable of the model.
possible to fix the following initial values of the state variables:
N k g
M ¼1.2, pc ¼ 8 MPa and pt ¼ 5 MPa. 0.2 54,820 75,191
The hyperelastic parameters k, n and g were calibrated with
M pc (MPa) pt (MPa) α m β γ
reference to stress paths typical of uniaxial compression tests
1.2 8 5 1 2 1 0
(Fig. 1) and isotropic compression tests (Fig. 3). These latter tests
were also useful for estimating the hardening parameters λ̃ and κ̃ , λ̃ κ̃
0.04 0.0089
from the slopes of the virgin compression line and of the re-
compression line, respectively (Fig. 3). ξv ηv ξs ηs ξvc ηvc ξsc ηsc ξvM ηvM ξsM ηsM
For both stress paths the constitutive model is capable of re- 20 2 3 3 80 365 0 0 4 4 0 0
producing the observed non-linear response that characterises the
elastic behaviour, thanks to the adopted hyperelastic formulation.
This feature is commonly observed in many different rocks, in-
cluding the one investigated in this study. It is worth noting, on
the contrary, that with the adopted model it is not possible to
reproduce the extremely fragile behaviour occurring after the peak
during the uniaxial compression tests, associated to the develop-
ment of fractures inside the specimens whose modelling is not
feasible at the element level.
Finally, the strain-induced degradation parameters were cali-
brated thanks to a trial-and-error procedure, focusing on the si-
mulation of the first stage of the plate loading tests C and D
(Fig. 7), characterised by the use of a 500 mm plate. Those curves
are well reproduced by the numerical simulation (hereafter called
PLT1) obtained using the set of parameters summarised in Table 1,
which are thus adopted throughout all the following analyses.
With reference to simulation PLT1 it can be observed that, besides
the upward concavity displayed during the recompression part of
the curve, properly fitted by the numerical simulation, a good
match is achieved for the knee, where the response becomes
mainly controlled by the elasto-plastic response of the tuff. The
Fig. 8. Comparison between the settlements curves obtained during the plate
unloading response is non-linear as well, though in a less pro-
loading tests C and D (500 mm plate followed by the 300 mm plate) and the nu-
nounced fashion as compared to what observed experimentally. merical simulation PLT1.
Fig. 8 gathers together the overall response as obtained by the
simulation of the first stage of the plate loading test PLT1 (500 mm
the reloading and the unloading stages, as well as the horizontal
plate, relatively low pressures), followed by the simulation of the
translation of the curve knee, plotted in the w/d – p plane of Fig. 8,
second stage, when a 300 mm plate was used to further increase
the loading up to relatively high pressures (about 20 MPa) (ex- towards higher pressure values, as compared to what observed in
perimentally achieved only in test C). The calibration of the model, the first low pressure cycle.
performed with reference to the results of the first stage only, The numerical tool adopted for the simulation of the plate
proves to be effective in reproducing the overall response during loading tests allows to have an insight into the evolution of the
278 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282

0.095 0.220

0.085 0.196

0.075 0.172

0.065 0.148

0.055 0.124

B E F
0.045 0.100

0.035 0.076

0.025 0.052

0.015 0.028

0.005 0.004

-0.020
C E’ -0.005 F’
Fig. 9. Map of volumetric plastic strains at selected stages of analyses PLT1 and PLT2.

0.300 3.500

0.267 3.100

0.234 2.700

0.201 2.300

0.168 1.900

B E F
0.135 1.500

0.102 1.100

0.069 0.700

0.036 0.300

0.003 -0.100

-0.500
C E’ -0.030 F’
Fig. 10. Map of deviatoric plastic strains at selected stages of analyses PLT1 and PLT2.

state variables within the tuff. In particular, the distribution of the
volumetric plastic strains εvp , deviatoric plastic strains εsp , isotropic
yield compression pressure pc, variable M and isotropic yield
tensile pressure pt is considered, with reference to stages B and C
of Fig. 7, which correspond to the knee and to the maximum
loading pressure, respectively, of the settlement curve obtained
during the simulation of the first (low pressure) 500 mm plate
loading test. The contours of the internal variables are also
evaluated for stages E and F of Fig. 8, which, in turn, correspond to
the knee and to the maximum loading pressure, respectively, of
the settlement curve obtained during the simulation of the second
(high pressure) 300 mm plate loading test. The results are given in
Figs. 9–13 in terms of contours of the inspected variable as ob-
served within the 0.5 m  0.5 m square portion of rock adjacent to
the loading plates, where a denser discretisation was adopted
(Fig. 6). The following remarks can be made.
 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282 279

(MPa) (MPa)
13.0 46.0

11.9 41.4

10.8 36.8

9.7 32.2

8.6 27.6
B E F
7.5 23.0

6.4 18.4

5.3 13.8

4.2 9.2

3.1 4.6

C E’ 2.0 F’ 0.0

Fig. 11. Map of the isotropic yield compression pressure pc at selected stages of analyses PLT1 and PLT2.

1.30 1.30

1.24 1.23

1.18 1.16

1.12 1.09

1.06 1.02
B E F
1.00 0.95

0.94 0.88

0.88 0.81

0.82 0.74

0.76 0.67

C E’ 0.70 F’ 0.60

Fig. 12. Map of the parameter M at selected stages of analyses PLT1 and PLT2.

Initially, the increasing load pressure leads to the localisation of
positive volumetric plastic strain under the edge of the 500 mm
plate (stage B in Fig. 9). Further increments are associated to the
development of εvp in a wider and deeper area, with maximum
values of about 0.09 located around the edge of the plate (stage C
in Fig. 9). When the 500 mm plate is substituted by the 300 mm
one, followed by the reloading up to the previous maximum
pressure (stage E in Fig. 9), a similar pattern is observed, with
maximum values of εvp of about 0.09 near the tip of the smaller
plate. Additional pressure increments triggers the growth of the
area characterised by positive volumetric plastic strains, which at
the end of the test reaches a depth comparable to the plate dia-
meter (stage F in Fig. 9).
A similar trend can be depicted from the contours of the
280 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282
B E
C E’
Fig. 13. Map of the isotropic yield tensile pressure pt at selected stages of analyses PLT1 and PLT2.
deviatoric plastic strain during the different stages of the test, al-
though the overall magnitude of this invariant is larger in this case.
In fact, stage C (Fig. 10) is characterised by a maximum value of
about 0.3 at the edge of the plate. At the end of the numerical
simulation high values of εsp are concentrated along a narrow
vertical band, indicating the development of a punching failure
mechanism (stage F in Fig. 10).
No major modification of the isotropic yield compression
pressure is observed during the first loading stage, except for a
very narrow portion of tuff beneath the tip of the 500 mm plate
(stage C in Fig. 11). This pattern can be indicative of the occurrence
of a compensating effect between the destructuring process (ne-
gative hardening terms) and the compaction of the rock matrix
(positive hardening term). In the second loading stage, while at-
taining high pressures by means of the 300 mm plate, a sig-
nificantly larger volume of rock is characterised by positive hard-
ening with increasing values of pc, up to a depth comparable to the
plate diameter (stage F in Fig. 11).
During the simulation of the plate loading test a relevant por-
tion of rock material is subjected to a reduction of the state vari-
able M, which controls the shape of the yield surface and, related
to that, the frictional strength of the soft rock. The decay of M is
initially concentrated around the edge of the 500 mm plate (stage
B in Fig. 12) and then extending downward and leftward (stage C
in Fig. 12) reaching a minimum value of about 0.95 (for an initial
undisturbed value of 1.2). Further reduction is then observed
during the second loading stage carried out by the 300 mm plate
(stage E in Fig. 12), again initially concentrated around the edge of
the smaller plate and then extending at depth. At the end of the
test (stage F in Fig. 12) the whole material under the plate is in-
terested by a degradation phenomenon of the M parameter up to a
depth comparable with its diameter, leading to a decrease of M
down to 0.65 in the most destructured zones.
Isotropic yield tensile pressure too shows a reduction, initially
concentrated at the tip of the 500 mm plate (stage B in Fig. 13),
then extending under the whole plate dimension, reaching values
(MPa) (MPa)
5.20 5.20
5.05 5.05
4.90 4.90
4.75 4.75
4.60 4.60
F
4.45 4.45
4.30 4.30
4.15 4.15
4.00 4.00
3.85 3.85
3.70 F’ 3.70
as low as 4 MPa (stage C in Fig. 13). During the reloading with the
smaller 300 mm plate, the area involved in the decrease of pt does
not extend, while additional degradation is observed, as shown in
Fig. 13 for stage E. Upon further loading to high pressure (stage F in
Fig. 13) a substantially homogeneous value of pt is observed, equal
to about 3.7 MPa, involving a volume of rock down to a depth
comparable to the 300 mm plate diameter. It can be observed that
the evolution of the isotropic yield tensile pressure is more rapid
as compared to that of parameter M, this latter mainly occurring
between stages E and F (i.e. at high pressure and large values of
volumetric plastic strain), while the former more intensively af-
fecting the low pressure stages C to E, for correspondingly lower
volumetric plastic strain.
To better understand the above results, a new test was simu-
lated, named PLT2, assuming the 300 mm diameter plate loading
to be carried out directly on the virgin rock (i.e. without pre-
loading and unloading stages). The result of the above numerical
test is shown in Fig. 14, together with that of the second stage of
test PLT1 (that relative to the 300 mm plate) for which the nor-
malised settlements are rezeroed with reference to point D of
Fig. 7. The two curves do not coincide, except for their early elastic
portions, highlighting for PLT2 a higher rate of settlements accu-
mulation, leading to a lower knee point as compared to the other
test, followed by almost parallel responses at high pressures. This
behaviour, qualitatively resembling that observed by Pellegrino in
his test B (Fig. 5), carried out under similar conditions, can be at-
tributed to the more intense destructuring process occurring
during test PLT2 as compared to PLT1, this latter following a pre-
loading stage. In fact, the microstructural damage is initially more
intense and tends to reduce its role as further plastic strains are
cumulated, leaving space for the competing positive term of the
hardening to take place: not surprisingly, the virgin rock appears
more prone to experience destructuration as compared to a por-
tion of it that has already cumulated plastic strain during the
preloading stage. The relative position of the two curves of Fig. 14
can thus be justified by the following evidence: upon
 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282
Fig. 14. Comparison between the load-settlements curves obtained in the second
part of the analysis PLT1 and in the analysis PLT2.
destructuring the tuff stiffness is temporarily reduced while, as the
process saturates and volumetric strains keep cumulating, the rock
tends to show a stiffer behaviour; as such for medium pressures
the loading test PTL1 exhibits a more rigid response as compared
to the corresponding PTL2, while at high pressures the response is
nearly the same, leading to the approximately parallel curves.
Similarly to what already shown for test PLT1, the evolution of
the state variables during test PLT2 is analysed with reference to
the corresponding contours, plotted also in Figs. 9–13, as detected
for stages E′ and F′, equivalent in terms of loading pressure to
stages E and F of test PLT1. Positive plastic volumetric strains
concentrate in the vicinity of the plate tip (stage E′ in Fig. 9), for
values slightly larger than those observed in the in the numerical
simulation PLT1 at the same pressure level (stage E in Fig. 9). The
final high pressure distribution of volumetric plastic strains in
analyses PLT1 and PLT2 is similar, as illustrated in Fig. 9. An ana-
logous pattern characterises the contours of the deviatoric plastic
strains, which on average attain larger values than the volumetric
ones (Fig. 10). More in general, it is worth observing that no sig-
nificant increment of volumetric and deviatoric plastic strains
occurs between stages C and E of the analysis PLT1, while the
corresponding quantities detected at stage E′ in the PLT2 analysis
fully develop during this first loading.
With reference to the evolution of the isotropic yield com-
pression pressure (Fig. 11), it can be observed by comparing low
pressure stages E and E′, and high pressure ones F and F', that this
internal variable varies in a rather similar way during tests PLT1
and PLT2.
Values of parameter M associated to stage E′ (Fig. 12) are
characterised by a minimum of 0.85 (as compared to the minimum
value of 0.95 recorded for stage E in the same figure), indicating a
more intense destructuration process, consistently with the larger
volumetric plastic strains occurring in analysis PLT2. At stage F′
(Fig. 12) the final distribution of M is very similar to that obtained
in analysis PLT1 (stage F in the same figure).
The PLT2 analysis is characterised by a large decrease in isotropic
yield tensile pressure during the initial low pressure loading phase
(stage E′ in Fig. 13). The amount of destructuring-related reduction of
pt observed in PLT2 at the above stage is comparable to that occurred
during the initial loading stage of PLT1 (i.e. stage C of the 500 mm
loading test). In this latter case the further evolution of pt from stage
281
C to stage E, upon 300 mm reloading, is rather marginal as compared
to what observed during the corresponding direct loading on the
virgin rock simulated in the analysis PLT2 for the same level of
pressure (point E′): all the above confirms that the plastic strain-in-
duced damage on the variable pt occurs at such a high rate that most
of it already affects the early stages of the tests, irrespectively of the
dimension of the plate adopted. As for all the other state variables,
the final distribution of the isotropic yield tensile pressure in the
analyses PLT1 and PLT2 is rather similar (Fig. 13).
In general, from the previous observations, it is possible to
conclude that the destructuring process occurred during the initial
stage of analysis PLT2 is more intense and responsible for the
larger settlements recorded in this case (Fig. 14). The same process
was already developed during the first loading cycle of analysis
PLT1 and played somehow a more limited role during the re-
loading stages carried out by the smaller plate. However, the final
configuration of the state variables is very similar in the two cases
(Figs. 9–13) due to the large loading level reached at the end of
both tests.
This overall picture is also confirmed by the evolution of the
stress paths and yielding surfaces illustrated in Fig. 15 for a Gauss
point located at a depth of 0.5 m below the tip of the 300 mm
plate. In detail, Fig. 15a depicts the p-q plot as observed during test
PLT1: the first 500 mm plate loading phase leads to stage C, cor-
responding to the dotted yield surface characterised by a pc value
only slightly larger than the initial one, while a substantial re-
duction of pt is already observed at the same stage. The following
unloading phase brings the path within the yield surface, while
the subsequent high pressure reloading induces a dramatic ex-
pansion of the surface in terms of pc, for a barely constant value of
pt and a reduced one for M. Fig. 15b illustrates the corresponding
stress path as observed during test PLT2, directly carried out by the
300 mm on the virgin rock: during the low pressure loading
phase, up to point E′, the yield surface experiences an evolution
similar to what discussed above for the initial loading stage of test
PLT1, exhibiting a negligible modification of pc and a considerable
reduction of pt, attaining a high pressure configuration nearly co-
incident with that observed for test PLT1.
7. Conclusions
This paper proposes and discusses a numerical procedure to
back-analyse a set of in situ plate loading tests performed on a soft
rock (i.e. the Neapolitan Yellow Tuff). The tests were characterised
by two loading and unloading cycles, the first one performed by a
500 mm diameter plate up to relatively low pressures, followed by
a second phase carried out adopting a smaller plate (300 mm in
diameter) up to high pressures. The experimental programme also
included a test consisting in a unique 300 mm plate loading phase
carried out on the virgin rock up to high pressures, i.e. without
pre-inducing any damage to the tuff due to preloading phases.
The experimental observations allowed to highlight some
specific features of the tuff response, as the non-linear stress–
strain behaviour, which not only affects the virgin compression
stages but also influences its mechanics well before yielding, along
unloading and reloading stress-paths. Another key feature of the
material under study appears to be the marked destructuration
process occurring with different intensity and characteristics
during the stages of the tests characterised by predominant irre-
versible straining. This latter process is driven at the micro-scale
by the plastic strain-induced damage to the interparticle bonding
that characterises the intact rock. Its macro-scale effects consist of
a reduction of the stiffness and strength of the soft rock, the latter
leading to a decrease in the cohesion and friction of the material,
together with a shrink of the dimension of its yield surface.
282 A. Amorosi et al. / International Journal of Rock Mechanics & Mining Sciences 78 (2015) 271–282
Fig. 15. Comparison between the stress paths of a Gaussian point located below the tip e of the 300 mm plate as calculated in the analyses PLT1 and PLT2. The evolution of
the yielding surfaces at selected stages of the analyses is also shown.
Typically, destructuring processes tend to be more intense upon
first loading, showing a decreasing role as further straining is in-
duced in the geomaterial, down to a state where the debonding
process saturates and no further effects can be detected. All the
above compete with the opposite effects (positive hardening) in-
duced by the permanent accumulation of volumetric plastic
strains, related to the permanent reduction of the porosity of the
rock.
The above qualitative framework of interpretation is corrobo-
rated in the present paper by a set of quantitative FE analyses,
carried out to back analyse the in situ tests adopting a constitutive
model specifically formulated at the scope and capable of ac-
counting for the non-linearity and destructuring processes dis-
cussed above.
The model formulation and its calibration procedure are de-
scribed in the paper, together with a detailed analysis of the nu-
merical results, which appears to well reproduce the in situ test-
ing. Given this, the numerical analyses are used as a tool to vi-
sualise what is occurring to the tuff while experiencing the plate
loading tests, carefully looking at the evolution of key ingredients
as the volumetric and deviatoric plastic strains, the friction angle
(as a function of the variable M) and the dimension of the yield
surface by the variables pc and pt.
In conclusion, the validation of a numerical procedure capable
of simulating the response of this class of materials in geotechnical
problems of bearing capacity can be considered as a valuable tool
to improve our current design practise, while at the same time
representing a powerful instrument to detect and extend our
scientific knowledge on the mechanics of soft rocks.
References
1 Kavvadas MJ. Modelling the soil behaviour – selection of soil parameters. General
report. In: Proceedings of the 2nd International Conference on the Geotechnics of Hard
Soils-Soft Rocks . Athens; 12–14 October 1998, vol. 2: 1441–1481.
2 Pellegrino A. Mechanical behaviour of Soft Rocks under high stresses. In: Proceedings of
the 2nd ISRM Congress . Belgrade; 21–26 September 1970, vol. 2: 173–180.
3 Leroueil S, Vaughan PR. The general and congruent effects of the structure in natural
soils and weak rocks. Géotechnique. 1990;40(3)467–488.
4 Aversa S, Evangelista A, Leroueil S, Picarelli L. Some aspects of the mechanical be-
haviour of “structured” soils and soft rocks. In: Proceedings of the 2nd International
Conference on Geotechnical Engineering of Hard Soils-Soft Rocks. Athens; 20–23 Sep-
tember 1993, vol. 1: 359–366.
5 Desai CS, Salami MR. A constitutive model and associated testing for soft rock. Int J
Rock Mech Min Sci. 1987;24(5)299–307.
6 Adachi T, Oka F. An elasto-plastic constitutive model for soft rock with strain soft-
ening. Int J Numer Anal Meth Geomech. 1995;19(4)233–247.
7 Oka F, Kimoto S, Kobayashi H, Adachi T. Anisotropic behavior of soft sedimentary rock
and a constitutive model. Soils Found. 2002;42(5)59–70.
8 Nova R. An extended Cam-Clay model for soft anisotropic rocks. Comput Geotech.
1986;2:69–88.
9 Gens A, Nova R. Conceptual bases for a constitutive model for bonded soils and weak
rocks. In: Proceedings of an International Symposium on Geotechnical Engineering of
Hard Soils-Soft Rocks . Athens; 20–23 September 1993: 485–494.
10 Lagioia R, Nova R. An experimental and theoretical study of the behaviour of a cal-
carenite in triaxial compression. Géotechnique. 1995;45(4)633–648.
11 Nova R, Castellanza R, Tamagnini C. A constitutive model for bonded geomaterials
subject to mechanical and/or chemical degradation. Int J Numer Anal Meth Geomech.
2003;27(9)705–732.
12 Pellegrino A. Surface footings on Soft Rocks. In: Proceedings of the 3rd ISRM Congress.
Denver; 1–7 September 1974, vol. 2B: 733–738.
13 Aversa S, Evangelista A, Nicotera MV. Analisi a posteriori di prove di carico su piastra
eseguite su tufo napoletano. In: Proceedings of the XXIII Convegno Nazionale di Geo-
tecnica. Abano Terme (Padova); 16–18 May 2007: 111–122.
14 Aversa S. Mechanical behaviour of soft rocks: some remarks. In: Proceedings. of the
Workshop on Experimental Characterisation and Modelling of Soils and Soft Rocks. Na-
ples; 29–31 October 1991: 191–223.
15 Aversa S, Evangelista A. Mechanical behaviour of a volcanic tuff: yield, strength and
”destructuration” effects. Rock Mech Rock Eng. 1998;31(1)25–42.
16 Evangelista A, Aversa S. Experimental evidence of non-linear and creep behaviour of
pyroclastic rocks. In: Cristescu ND, Gioda G, editors. Visco-plastic Behaviour of Geo-
materials. Wien-New York: Springer-Verlag; 1994.
p. 55–102 CISM Courses and Lectures No. 350.
17 Bigoni D, Piccolroaz A. Yield criteria for quasibrittle and frictional materials. Int J Solids
Struct. 2004;41:2855–2878.
18 Evangelista A, Aversa S, Pescatore TS, Pinto F. Soft rocks in southern Italy and role of
volcanic tuffs in the urbanization of Naples. In: Proceeding of the 2nd International
Symposium on The Geotechnics of Hard Soils and Soft Rocks. Naples; 12–14 October
1998, vol. 3: 1243–1267.
19 Picarelli L, Evangelista A, Rolandi G, Paone A, Nicotera MV, Olivares L et al., 2007.
Mechanical Properties of Pyroclastic Soils in Campania Region. In: Proceedings of the
International Workshop on Characterisation and Engineering Properties of Natural Soils.
Singapore; 29 November–1 December 2006, vol. 4: 2331–2383.
20 Aversa S, Evangelista A, Scotto Di Santolo A. Influence of the subsoil on the urban
development of Napoli. In: Proceedings of the 2nd International Symposium on Geo-
technical Engineering for the Preservation of Monuments and Historic Sites. Naples; 30–
31 May 2013: 15–43.
21 Aversa S, Lagioia R. Model requirements and design criteria for soft rocks. In: Pro-
ceedings of the 2nd International Conference on The Geotechnics of Hard Soils-Soft Rocks.
Naples; 12–14 October 1998, vol. 3: 1483–1497.
22 Houlsby GT, Amorosi A, Rojas E. Elastic moduli of soils dependent on pressure: a
hyperelastic formulation. Géotechnique. 2005;55(5)383–392.
23 Kavvadas MJ, Amorosi A. A constitutive model for structured soils. Géotechnique.
2000;50(3)263–273.
24 Amorosi A, Boldini D, Germano V. Implicit integration of a mixed isotropic-kinematic
hardening plasticity model for structured clays. Int J Numer Anal Meth Geomech.
2008;32:1173–1203.
25 Sloan SW. Substepping schemes for the numerical integration of elastoplastic stress-
strain relations. Int J Numer Anal Meth Geomech. 1987;24:893–911.