Structure and Infrastructure Engineering

Maintenance, Management, Life-Cycle Design and Performance

ISSN: 1573-2479 (Print) 1744-8980 (Online) Journal homepage: http://www.tandfonline.com/loi/nsie20

Limit state design approach for the safety

evaluation of the foundations of concrete gravity
dams

Maria Luísa Braga Farinha, Laura Caldeira & Emanuel Maranha das Neves

To cite this article: Maria Luísa Braga Farinha, Laura Caldeira & Emanuel Maranha das
Neves (2015) Limit state design approach for the safety evaluation of the foundations of
concrete gravity dams, Structure and Infrastructure Engineering, 11:10, 1306-1322, DOI:
10.1080/15732479.2014.964265

To link to this article: http://dx.doi.org/10.1080/15732479.2014.964265

Published online: 03 Oct 2014.

Submit your article to this journal

Article views: 125

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=nsie20

Download by: [University of Nebraska, Lincoln] Date: 28 October 2015, At: 20:03
 Structure and Infrastructure Engineering, 2015
Vol. 11, No. 10, 1306–1322, http://dx.doi.org/10.1080/15732479.2014.964265

Limit state design approach for the safety evaluation of the foundations of concrete gravity dams
Maria Luı́sa Braga Farinhaa*, Laura Caldeirab1 and Emanuel Maranha das Nevesc2
a
Concrete Dams Department, National Laboratory for Civil Engineering, Av. do Brasil 101, 1700-066 Lisbon, Portugal; bGeotechnical
Department, National Laboratory for Civil Engineering, Av. do Brasil 101, 1700-066 Lisbon, Portugal; cInstituto Superior Técnico,
Technical University of Lisbon, Av. Rovisco Pais, 1049-001 Lisbon, Portugal
(Received 29 January 2014; final version received 26 June 2014; accepted 12 July 2014; published online 3 October 2014)

The application of the limit state design (LSD) in the geotechnical area has increased over the last two decades, but this
approach is not yet widely used in dam safety evaluation. This study aims to widen the use of the LSD application for large
dams, in particular concrete gravity dam foundations. This paper starts with a brief reference to the LSD approach in recently
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

published guidelines for dam design, followed by a detailed description of the LSD formulation when applied to the
foundation of concrete dams. The relevance of the joint application of the concepts of ultimate limit states and of numerical
methods is highlighted. Comments are made regarding the criteria adopted in order to determine the characteristic values
of the material mechanical properties, with an emphasis on discontinuities, taking into account the spatial variability.
The sliding safety assessment of the foundation of a concrete gravity dam using the LSD and a discrete element model, both
in persistent and in an accidental design situation, is presented. Results led to the conclusion that the LSD methodology may
be followed for dam foundation design with the partial factor values prescribed in Eurocode 7.
Keywords: civil and structural engineering; geotechnical engineering; codes of practice and standards; dam safety;
foundations; failure modes; mathematical modelling

1. Introduction ways of simulating the distribution of uplift pressures

The majority of concrete gravity dam failures are due to along the base of the dam, the possibility of crack
problems in the foundation rock mass (Deroo & Boris, formation and propagation along the dam –foundation
2011; International Commission on Large Dams [ICOLD], interface and stability criteria, expressed in terms of
1995), thus great care has to be taken at the design stage. minimum values for the overall safety factor.
The design of the rock foundation requires stress – strain, Concerning the overall safety factor, several defi-
seepage and sliding stability analyses. The stresses on the nitions have been proposed (Alonso, Carol, Delahaye,
foundations of gravity dams are relatively low, so bearing Gens, & Prat, 1996; Asadollahi & Tonon, 2010; Kovari &
capacity failure of the rock mass is unlikely and its safety Fritz, 1989, 1993). In all cases, it provides a measure of the
is controlled by sliding either along the dam –concrete distance from the limit equilibrium. Nevertheless,
interface or along rock mass discontinuities or sub- generally, uncertainty about water pressure distribution
horizontal weak layers in the foundation, close to the base and shear strength parameters (cohesion and friction angle
of the dam. The conventional approach to assess the in effective stresses) along the slip surfaces requires
sliding stability of concrete gravity dams is based on parametric studies to be carried out. The above-mentioned
principles of limit equilibrium (Londe, 1973), and on an conventional deterministic approach is followed by many
overall safety factor. Both the dam and its foundation are design codes, among which is the Portuguese regulation
assumed to be rigid bodies and the method consists in for dam design (NPB, 1993). The ICOLD is aware of the
assuming a failure mode and calculating the driving and shortcomings of the overall safety factor concept and is
resisting effective forces acting on the sliding surface, with moving towards probability approaches (ICOLD, 1988,
the ratio of these two forces being the overall safety factor. 1993).
Regarding the stability evaluation of gravity dams, Actually, two alternative approaches may be employed:
different criteria have been adopted by different bodies, semi-probabilistic and probabilistic procedures. The
and great effort has been made in order to analyse the former allows the consideration of the different levels of
differences in the results (CFBR, 2006; European Club of uncertainty regarding the various actions and material
ICOLD, 2004a, 2004b; FERC, 2002; USACE, 1983, 2000, properties, through the use of characteristic values of
2005). These differences are mainly related with different actions, material properties and geometry data (called basic

*Corresponding author. Email: [email protected]

q 2014 Taylor & Francis
 Structure and Infrastructure Engineering 1307

variables) and partial factors. This is the method prescribed Farinha, 2010; Gimenes & Fernández, 2006; Lemos, 1999;
in Eurocode 7 (EC7) for geotechnical design (CEN, 2004). Mostyn, Helgstedt, & Douglas, 1997). However, the
The latter would allow a probability of failure to be uncertainties in these flow models still need to be
estimated, taking explicitly into account the uncertainties in evaluated.
the basic variables (ICOLD, 1993). However, the This paper presents a study about the joint application
calculation of probabilities of failure in rock engineering of the concept of ULS with partial factors and of numerical
presents serious difficulties, mainly due to the usual lack of methods in the safety assessment of the foundation of a
knowledge concerning the probability distribution func- concrete gravity dam, in static conditions, for two different
tions and the spatial variability of the material properties, design situations: the persistent situation and the
and to the current inability to reproduce statistical accidental situation related to the total clogging of the
descriptions of the joint patterns in numerical models. drainage curtain. According to EN 1991-1-7 (CEN, 2006),
Despite these difficulties, a few studies were presented in a risk analysis which includes an estimate of the likelihood
the last decade using probabilistic approaches for the of the clogging of the drainage system is required for a
seismic analysis of dams (e.g. Lupoi & Callari, 2012). CC3 type of structure, like a large dam. In the present case,
Approximate solutions of the probability of failure were as high levels of quality control are required not only
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

obtained by Bernstone, Westberg, and Jeppsson (2009) to during dam construction but also during operation, this
assess the safety of an operating concrete dam, taking into event is not likely to occur, thus, it is considered as an
account recorded uplift data. accidental situation.
The European standard EC7 (CEN, 2004) is based on Based on prescribed values of EC7 for the partial
the limit state design (LSD), namely ultimate limit state factors, a safety verification of the foundation design of
(ULS) and serviceability limit state (SLS) design (CEN, Pedrógão concrete gravity dam, in Portugal, will be
2004). Its implementation has resulted in many changes to presented. This example only envisages the ULS
geotechnical design (Orr, 2012). Nowadays, an LSD corresponding to the overall stability concerning sliding.
approach is not routinely used in dam safety evaluation, Comments are made regarding the criteria adopted in order
contrary to current practice in both geotechnical and to determine the characteristic values of the rock mass
structural design. However, EC7 may be extended to strength properties, with an emphasis on discontinuities,
special structures, such as dam foundations and tunnels, or taking into account the spatial variability. The potential
to the design of foundations of nuclear power plants and failure modes regarding sliding which are analysed
offshore structures. In these cases, though, there may be involve the dam foundation rock mass, taking into account
additional requirements. the orientation of the main sets of discontinuities within
This study aims to widen the use of the LSD the dam foundation. Analysis was carried out for the
application for large dams, in particular concrete gravity critical failure mechanism, which was adequately
dam foundations. Gravity dams resist the thrust of the identified. Dam safety assessment of sliding follows EC7
reservoir water with their own weight. The seepage flow (CEN, 2004).
through the foundation, in the upstream – downstream
direction, gives rise to both seepage and uplift forces,
which, in turn, reduce the stabilising effect of the
structure’s weight. Due to the great influence that uplift 2. LSD approach in dam design
forces have on the overall stability of gravity dams, the Taking into account the dominant role of the Eurocodes
distribution of water pressures along the base of the dam concerning matters of structural safety all around Europe,
should be as accurately predicted as possible at the design it is not surprising that the subject of LSD has been raised
stage using numerical models. The hydraulic gradients by the European Club of the ICOLD. However, until now,
also deserve particular attention, due to their role in the the official publications of this organisation have made
internal erosion processes. little mention of it. Nevertheless, the final report on sliding
However, these particular aspects only started to be safety of existing gravity dams of the European Club
tackled in a reliable way in the 1980s, with the (European Club of ICOLD, 2004a) refers, in its Appendix
development of numerical models which simulate the dedicated to Regulatory Rules, Guidelines and Normal
hydromechanical (HM) interaction, which is particularly Practice in different countries, to the Chinese Technical
important in this type of structure, using the finite element Standards related to the design of hydraulic engineering
method (e.g. Noorishad, Ayatollahi, & Witherspoon, structures, namely the Design Specification for concrete
1982). In more recent years, several studies have been gravity dams, which clearly states that hydraulic structures
carried out using models of flow in discontinuous media, must be designed for the ULS and the SLS, the latter called
with discrete element models, mainly for gravity dams, normal operation limit states. In addition to these Chinese
taking into account the water pressures resulting from the standards, guidelines on design criteria for concrete
seepage conditions (Barla, Bonini, & Cammarata, 2004; gravity dams based on the LSD were published in
 1308 M.L.B. Farinha et al.

Australia, in 1991, and some documents and recommen- projects, safety reviews and design of remedial works
dations were published in France, concerning the safety (Peyras et al., 2008; Royet & Peyras, 2013).
and serviceability verification for both gravity dams
(CFBR, 2006, 2012) and embankment dams (CFBR,
2010). 3. Formulation
It is relevant to make a brief reference to the above- 3.1. General aspects
mentioned Chinese standard, according to which the safety In this paper, the use of LSD for dam safety assessment is
evaluation against sliding involving the dam foundation is based on EC7 (CEN, 2004), as mentioned in the introduction.
identified as an ULS. The different partial factors are This European standard describes the general principles and
combined for the following design situations: sustained requirements for geotechnical design, primarily in order to
(corresponding to persistent situations according to ensure safety (resistance and stability), serviceability and
Eurocodes), transient and occasional (corresponding to durability of geotechnical structures. As such, it should be
accidental situations) status. In the Chinese standard, ULS used in conjunction with EN 1990 Eurocode: basis of
verification is based on the following expression: structural design (CEN, 2002b), EN 1991 Eurocode 1:
actions on structures (CEN, 2002a) and EN 1998: design of

 1

Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

g0 C E gG Gk ; gQ Qk ; Ak # R X k =gm ; ð1Þ structures for earthquake resistance (CEN, 2005). EC7

gd consists of two parts: Part 1, General Rules (CEN, 2004),
which presents the general rules of geotechnical design; and
where g0 is the importance factor of the structure Part 2, Ground Investigation and Testing (CEN, 2007),
(g0 ¼ 0.9, 1.0, 1.1), C is the factor of design situation which discusses the use of field investigations and laboratory
(C ¼ 1.0, 0.95, 0.85) and gd is the structure coefficient, testing for geotechnical design.
equal to 1.2 for the sliding limit state. EC7 serves as a reference document for the geotechni-
The remaining symbols are those used in Eurocodes: cal design of foundations for special foundation works
E() is the calculation model of the effects of the actions; [clause 2.1(21) of EC7 (CEN, 2004) and clause 1.1(2) of EN
R() is the calculation model of the resistances; gG , gQ and 1990 (CEN, 2002b)], such as the foundations of concrete
gm are the partial factors for permanent actions, variable dams. This study focuses on large dams, particularly on
actions and ground parameters (material properties), concrete gravity dams and the safety analysis of their
respectively; and Gk , Qk and Ak are the characteristic foundations.
values of permanent, variable and accidental actions,
respectively. For the assessment of the ULS of sliding, a
rigid body limit equilibrium approach is used. The overall 3.2. DAs and design situations
shear resistance is evaluated by means of the Mohr – EN 1997-1 (CEN, 2004) and EN 1990 (CEN, 2002b)
Coulomb criterion based on cohesion and internal friction include, as options, the following three different DAs for
angle in terms of effective stresses. The procedure for ULS verifications in persistent and transient design
calculating the characteristic values is not included in the situations (Frank et al., 2004). DA1 requires, in principle,
above-mentioned Appendix of the European Club report, two calculations involving two sets of partial factors
but will be given particular attention in this paper. (Combinations 1 and 2). Where it is obvious that one of
It is also interesting to mention that the LSD approach these sets governs the design it will not be necessary to
proposed in the previously mentioned guidelines on design carry out calculations for the other (Annex B.1 (2), CEN,
criteria for concrete gravity dams published by the 2004). This DA may be termed an action and material
Australian National Committee on Large Dams factor approach with partial factors applied at the source,
(ANCOLD) in 1991 was recently replaced by a factor of that is to actions, rather than to the effects of actions, and to
safety or working stress approach, in revised guidelines shear strength parameters, rather than to resistances.
published in September 2013 (ANCOLD, 2013). This DA2 requires a single calculation to either actions or
move away from the LSD approach was justified by the effects of actions and to resistances. It may be termed an
need to reflect the preferred design approach (DA) action effect and resistance factor approach. It must be
amongst Australian dam engineers. noted that, in this case, the application of partial factors to
Regarding French publications in this area, the the effects of actions does not deviate significantly from
documents adopt the limit state format similar to the conventional overall factor safety approach. DA3
Eurocodes. The recommendations mark an important requires a single calculation where partial factors are
departure from conventional deterministic practices, applied to actions or effects of actions from the structure
requiring the designer to set at source the safety factors and to ground strength parameters and may be termed an
on the various basic variables involved in the compu- action effect and material factor approach.
tations. Concerning gravity dams, this standardised A National Annex allows each European country to
method is now being used routinely in France for new set, within certain limitations, the safety levels for civil
 Structure and Infrastructure Engineering 1309

engineering works through the partial factors values, actions. Only one strength parameter is used in the
called Nationally Determined Parameters. The National following – the friction angle in effective stresses at the
Annex may also specify the procedure to be used when rock mass discontinuities. The corresponding partial
alternative procedures are given in the Eurocode, namely factors are 1 and 1.25, respectively, for Combinations
the DAs. In this study, the Portuguese National Annex will 1 and 2 (C2) of the persistent design situation and 1.1 for
be followed, which requires DA1 to be adopted for the the accidental design situation (A). Figure 2 shows the
ULS design in both the persistent and transient situations. load cases and corresponding partial factors. An additional
For the accidental situations, the partial factors included in load case considering both permanent actions as
this Annex are also assumed. unfavourable (Combination 1 with load case 2 – C1.2)
The actions as well as the shear strength of the is also shown, for comparison with conventional
materials will be represented by their characteristic values. procedure in dam design. However, in gravity dams it is
The design values result from the application of partial obvious that the self-weight always has a stabilising effect
factors to the characteristic values. Obviously if the partial in the ULS verification of sliding.
factor is 1, the characteristic and design values are equal.
A particular and important aspect of the actions acting on a
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

concrete gravity dam, when compared with the structures 3.3. Application of LSD in a numerical modelling
covered by EC7, is the large values of the water pressure framework
on the upstream face of the dam, as well as the significant The current practice of ULS verification assumes a rigid
weight of the dam body itself. Two design situations are perfectly plastic behaviour associated with a previously
considered, as shown in Figure 1: the persistent situation selected failure mechanism, usually requiring the study of
considering the retention water level (RWL) at the a large number of mechanisms in order to identify the
reservoir (Figure 1(a)), and the accidental situation critical one. The constitutive models based on stress –
involving the complete clogging of the drainage system strain relations are only used in SLS verifications.
of the dam foundation (Figure 1(b)). Nevertheless, models of the latter type can and shall be
According to the Eurocode system, for the persistent used, with the adequate design values, in the ULS
design situation, the following partial factors must be verifications, since they enable the prompt and unequi-
used: (i) Combination 1 – partial factors for permanent vocal identification of the critical failure mechanism, and
actions greater or equal to 1 (gG $ 1, namely 1.0 or 1.35) allow consideration of the interaction between the effects
and for the shear strength equal to 1 (gM ¼ 1); (ii) of different actions. Therefore, they provide more
Combination 2 – partial factors for permanent actions adequate knowledge of the forces in place.
equal to 1 (gG ¼ 1) and for the material properties greater The above-mentioned procedures imply the use of
than 1 (gM . 1, namely 1.25). For the accidental design numerical analysis to determine the effects of actions.
situation, the partial factors for permanent actions are However, it is very important to stress the singular
equal to 1 (gG ¼ 1) and for the shear strength properties nature of the geotechnical actions when compared with
greater than 1 (gM . 1, namely 1.1). the structural ones. The latter are independent of the
In the present case, for Combination 1, two permanent mechanical characteristics of the materials, while the
actions (water pressure on the upstream face of the dam former, if originated at the ground, either soil or rock mass,
and dam weight) must be combined with different partial are dependent on the mechanical characteristics of the
factors, as one is unfavourable and the other favourable ground. However, more importantly, the stress transmitted
(Combination 1 with load case 1 – C1.1). The to the ground by the external action can generate, in an
recommended value of gG is 1.35, for unfavourable elemental area of the sliding surface and simultaneously,
permanent actions, and 1.0, for favourable permanent both a shear stress (having a destabilising effect) and a

(a) (b)

WP WP
D

G.C. ϕ’k ϕ’k

G.C.

NOTE: G.C., grout curtain; D., drainage curtain; WP, water pressurewith the reservoir at the retention water level;
ϕ’k, characteristic friction anglein effective stresses.

Figure 1. Design situations: (a) persistent situation considering the reservoir at the RWL and (b) accidental situation involving the
complete clogging of the drainage system of the dam foundation.
 1310 M.L.B. Farinha et al.

Persistent design situation Accidental design situation

(a) (d)
C1.1 A
RWL RWL
1.0 Gk 1.0 Gk
1.35 Wk 1.0 Wk

tan ϕ’k /1.0 tan ϕ’k /1.1

G.C. G.C.

(b)
C1.2
RWL
1.35 Gk
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

1.35 Wk

D
G.C.- grout curtain
tan ϕ’k /1.0
G.C.
D - drainage curtain
RWL - retention water level
(c) Wk - characteristic value of
C2 the hydrostatic pressure
RWL
1.0 Gk
1.0 Wk Gk - characteristic value of the
dead weight

ϕ'k - characteristic value of

D the friction angle in
effective stresses
tan ϕ’k /1.25
G.C.

Figure 2. Combinations, load cases and corresponding partial factors for the different design situations: (a) C1.1; (b) C1.2; (c) C2; (d) A.

normal stress (having a stabilising effect), the latter is due to actions; and the partial factors of other actions, F k;i , must
the frictional nature of the mobilised resistance on the be referenced to this value, so the following expression
assumed surface. must be used (Gulvanessian, Calgaro, & Holichy, 2002):
Therefore, due to the difficulties in distinguishing the  
effects induced by the different actions, these actions, gf ;i
Ed ¼ gF;1 E F k;1 ; F rep;i ; i . 1; ð2Þ
when introduced in the model, should not be affected by gf ;1
any partial factor, the failure mechanism and the results of
all numerical analyses always being induced or calculated where Ed is the design effect of the actions, Ef } is the
by a strength reduction procedure. The critical failure result of the calculation model, F k;1 is the dominant action
mechanism may be determined through the analysis of the for the considered limit state associated with gF;1 , F rep;i are
foundation displacements, considering design values of the remaining actions, gf ;i are the corresponding partial
strength affected by an increasing partial factor. This is factors to F rep;i and i is the number of remaining actions in
done until a failure surface is determined, which is addition to the dominant one. The numerical analysis is
assumed as the critical failure mechanism for the overall performed with the characteristic values of the dominant
sliding. To ensure that such ULS will not occur, or, rather, action and the characteristic values of the remaining
to check its sufficiently low probability of occurrence, actions affected by the relation gf ;i =gf ;1 .
analysis is carried out for the identified mechanism. The critical failure mechanism is reached by a
In numerical analysis, when there is more than one progressive reduction of the strength from its design
partial factor for actions, a reference value, gF;1 , must be value, by performing a sequence of analysis, until failure
selected (in general associated with the dominant action) occurs. The final value of the reduction factor correspond-
and applied to the whole effect of the combination of ing to the last stable situation is designated by strength
 Structure and Infrastructure Engineering 1311

reduction factor (SRF), which is the well-known overall that the procedure described above for C1.1 has been
safety factor. An auxiliary factor, the over-design factor developed for the analysis of a combination of favourable
(ODF), is then introduced as and unfavourable loads. In this case, however, the dam
weight has an important role in terms of resistance, and
SRF thus the procedure presented will result most probably on
ODF ¼ : ð3Þ
gF;1 the conservative side.
For C1.2, the numerical analysis was carried out with
This ODF gives an additional margin of safety beyond that the characteristic value of the water pressures at the
required by Eurocode for the studied limit state. The safety upstream face of the dam, the characteristic value of
verification implies that this factor be larger than or equal the weight of the dam body and the design value (equal to
to 1. the characteristic one) of the shear strength of the rock
In the application of this procedure to the example mass discontinuities. The value of the ODF is also
shown in Figures 1 and 2, the dominant action is the water obtained from the final value of the SRF divided by 1.35
pressure on the upstream face of the dam, therefore (gF;1 ). For Combination 2 (C2), the calculation was
according to Equation (2), its partial factor was not performed using the design value of the shear strength,
initially applied in the numerical model. Figure 3 presents
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

applying the partial factor, gM , to its characteristic value.

a schematic representation of the use of actions’ partial In this case, only gtan w0 is required and, according to EN
factors when numerical analysis is carried out. 1997-1, its value is 1.25. In a procedure similar to the one
For C1.1, the numerical analysis was performed with described for Combination 1, an SRF value corresponding
the characteristic value of the water pressures at the to the last stable situation was obtained. For this
upstream face of the dam, 0.74 (1/1.35) of the combination, the ODF coincides with the SRF.
characteristic value of the weight of the dam body and For the accidental design situation (A), the calculation
the design value (equal to the characteristic one) of the was made using the design values of the friction angle in
shear strength of the rock mass discontinuities. The design effective stresses at the rock mass discontinuities, applying
value of w0 (tan w0d ¼ tan w0k =gM ¼ tan w0k ) is reduced by the partial factor gtan w0 with a value of 1.1. In a similar
applying a strength reduction procedure, SRF . procedure, an SRF value was obtained. For this
1 ðtan w0 ¼ tan w0d =SRFÞ and the resulting horizontal combination, the ODF also coincides with the SRF. A
displacement at the crest of the dam (failure indicator) is value of ODF equal to 1 indicates that both sliding safety
registered. The value of the ODF is obtained from the final and the available margin of safety are exactly what is
value of the SRF divided by 1.35 (gF;1 ). It should be noted required by EC7 (CEN, 2004). If ODF . 1, the margin of
safety is more than adequate and ODF , 1 means
(a) C1.1 inadequate safety (though not necessarily meaning failure).
RWL
1/1.35 Gk
Wk
3.4. Characteristic values
The basic variables involved in the overall sliding stability
of a gravity dam are the reservoir water level, the unit
1.35 × D weights of both the dam body and the foundation, the shear
strength parameters along the sliding surface and the
G.C.
geometry of the dam, and its foundation discontinuities.
For the persistent and accidental design situations, the
reservoir water level was assumed to be at the RWL. Thus,
for this variable, as well as for the unit weights of the
(b) C1.2
structure and of the foundation, whose variability is small
RWL
[coefficient of variation (COV) less than 5%], the
1.0 Gk
corresponding characteristic values are taken as the
Wk
mean values. Thus, the uncertainties are associated with
the shear strength parameters at the failure surface and the
1.35 ×
geometry of the foundation discontinuities.
D
Due to the lack of data, the uncertainties associated
with the geometry of the rock mass discontinuities were
G.C. not explicitly taken into consideration. Instead, an
unfavourable configuration was assumed in the numerical
Figure 3. Use of the actions’ partial factors of Combination 1 modelling (Rocha, 1978). It is important to highlight that,
when numerical analysis is carried out: (a) C1.1; (b) C1.2. being an ULS verification, the peak values of the shear
 1312 M.L.B. Farinha et al.

Equation (6) can be used when this variance is unknown,

being replaced by the variance of the test results.
Regarding failure mechanism dimension, no criterion
is indicated in EC7. For the establishment of this criterion,
it is necessary to take into account the spatial variability
and averaging of the parameter value along the failure
surface, as explained below. Neglecting the other
components of the uncertainty associated with the
determination of the geotechnical parameters, namely,
measurement and interpretation errors and statistical
uncertainty, and considering only the spatial variability,
for a normal distribution, the characteristic value f0k;var can
be estimated based on the following equation:

tan f0k;var ¼ tan f0 ð1 2 k5% COVinherent GÞ; ð7Þ

Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Figure 4. Evaluation of the characteristic values with a where COVinherent represents the coefficient of variation
statistical basis using a normally distributed variable X.
induced by the spatial variability and G is the variance
reduction function resulting from the averaging process of
the parameter along the spatial extent (average length L,
strength or any contribution of the dilatancy must be
area A or volume V) of the governing failure mechanism.
replaced by the critical state values, that is the strength for
Several functions are available for this variance
large relative displacement between discontinuities’ reduction function, the linear, the exponential and the
surfaces, with a constant shear stress and no volumetric quadratic exponential being the most common. However,
variation (the distance between the discontinuities’ in the current cases, these all give very similar results.
surfaces is constant, not dependent on the distortion). In the following, the simplest function, the linear, is
The evaluation of the characteristic value of the shear adopted:
strength with a statistical basis is performed (Figure 4), in
8

accordance with EC7 (CEN, 2004), calculating the lower < Ld 1 2 3L d
if Ld . 1;
value of the critical state friction angle in effective stresses G ¼
2
ð8Þ
associated with the 0.05 fractile, f0k;inf , if the extent of the : 1 2 3Ld ; if Ld # 1;
failure surface involved in the sliding mechanism is small,
using the following Equation (a normal distribution for the where d represents the fluctuation scale (related to the
shear strength variable being assumed): effective autocorrelation distance) and L is the length of
the sliding surface in the considered direction. The square
tan f0k;inf ¼ tan f0 2 k5% stan f0 ð4Þ of total variance reduction function can be obtained by
multiplying the square of the variance function in each
or of the 95% lower confidence limit for the population direction.
mean of the same parameter, f0k;mean , if a larger rock mass The comparison of the values obtained with Equations
volume is involved in the failure mechanism, as in (4) to (7) allows the assessment of the extent of the failure
  stan f0 surface, that is whether the extent of the failure surface is
tan f0k;mean ¼ mf0 95% ¼ tan f0 2 k5% pffiffiffi ; ð5Þ considered either small or large. Thus, if the characteristic
n
value calculated with Equation (7) is larger than that
  stan f0
tan f0k;mean ¼ mf0 95% ¼ tan f0 2 t5% pffiffiffi ; ð6Þ obtained with Equations (5) or (6), the critical sliding
n surface can be considered large, otherwise, Equation (4)
must be used. Despite not being foreseen in EC7, the spatial
where stan f0 is a known standard deviation of the variable
variability is explicitly considered in the study presented
tan f0 , given in the bibliography or derived from test
here, based on an assumed scale of fluctuation, in order to
results (then represented by stan f0 ), k5% is the value of the
analyse its influence on ULS sliding verification.
standard normal variable associated with a probability of
5%, kmf0 Þ95% is the 95% lower confidence limit for the
mean, mf0 , and t5% is the variable value of the t-student
4. Application to Pedrógão dam
distribution, with a number of degrees of freedom equal to
the number of available tests (n) minus 1, associated with a 4.1. General characteristics
probability of 5%. Equation (5) can be used when the Pedrógão dam (Figure 5) is a straight gravity dam located
variance of the distribution is known a priori and on the River Guadiana, in the south-east of Portugal, with a
 Structure and Infrastructure Engineering 1313

(a) (b) (m) Block 6A-7

95
90
RWL = 84.8
85
80
75
70 33.8
65
60
55
50
45 44.4
40
RWL – retention water level

Figure 5. Pedrógão dam: (a) downstream view from the left side of the uncontrolled spillway and (b) cross-section of the central area
of the dam.
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

maximum height of 43 m and a total length of 448 m, of in the area of the river bed below the central area of the
which 125 m is of conventional concrete and 323 m of dam, which is going to be analysed, are described in detail
roller-compacted concrete. The dam is part of a multi- in Caldeira, Farinha, Maranha das Neves, & Lemos
purpose development designed for irrigation, energy (2013).
production and water supply (Miranda & Maia, 2004) During excavation it was possible to identify the main
and creates a reservoir which allows the turbines of joint sets in the different foundation areas (EDP, 2004).
Alqueva dam, the main structure of the development, Figure 6 depicts the average position of the main sets of
located about 23 km upstream from Pedrógão, to pump rock joints, in the bottom of the valley, in relation to the
water from the downstream to the upstream reservoir in a dam. The sub-horizontal joint set may be relevant
reverse motion. The dam has an uncontrolled spillway regarding strength and/or deformability of the dam
with a length of 301 m with the crest at an elevation of foundation and must be pointed out due to its substantial
84.8 m, which is the RWL. Figure 5(b) shows a cross- extension at the valley bottom, near the riverbed and
section of the central area of the dam, in which the base downstream from the dam. The hydro-geological studies
length in the upstream – downstream direction is 44.4 m led to the conclusion that the ground-water table was
and the dam height is 33.8 m. It should be noted that the controlled by the River Guadiana. The results of the
cross section of this dam is unusual, as in the majority of Lugeon tests revealed a low permeability rock mass at
gravity dams the base length is around 20% lower than depth, except in localised zones or along discontinuities.
their height. Figure 7 illustrates the areas of poor geomechanical
Since the beginning of the first filling of the reservoir, properties and of highest permeability in the foundation of
in November 2005, Pedrógão dam has shown a peculiar Pedrógão dam.
seepage pattern, with some local high flow rates and From samples collected in boreholes PD5, PD7 and
uplift pressures. A discontinuous HM model of the dam PD12 at given depths (see Figure 7 and Table 1), five
foundation was developed, in which the main seepage
paths, identified with in situ tests, were represented, which
allowed recorded discharges and water pressures during
normal operation to be accurately interpreted (Farinha,
2010). In the study presented here, however, a design stage
is assumed, and thus the current performance of the dam is
not taken into account.

4.2. Characteristics of the foundation rock mass

The foundation of Pedrógão dam consists of granite with
small to medium-sized grains and is of good quality with
the exception of the areas located near two faults in the
main river channel and on the right bank, where the
geomechanical properties at depth are poor. The most Figure 6. Average position of the main sets of rock joints
relevant geotechnical aspects of the foundation rock mass (a, b, c, d and g) in relation to the dam.
 1314 M.L.B. Farinha et al.

analysed area
(block 6A-7)

(m) R.B. L.B.

100 1 2 3 3A 4 4A 4B 5 5A 5B 6 6A 6B 7 7A 8 9 10 11 12 13
90
80
70
60
50
40

PD5
PD7
PD6 PD8
PD12

poor geomechanical properties area of both poor geomechanical

Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

properties and high permeability

areas of highest permeability

Figure 7. Areas of poor geomechanical properties and of highest permeability in the foundation of Pedrógão dam. Location of the
boreholes drilled for geological and geotechnical investigation in the central area of the dam.

compression and shear tests of the discontinuities’ planes Figure 8 shows the influence of spatial variability on the
were carried out at different normal effective stresses at the characteristic values, as a function of the scale of fluctuation
Portuguese National Laboratory for Civil Engineering for L ¼ 94 m (Equations (7) and (8)). Conservatively, the
(LNEC, 2004). The results presented in Table 1 are the length in the direction perpendicular to the cross section of
normal stiffness at a normal stress of 1 MPa, shear stiffness the dam was considered small and the corresponding
corresponding to a shear displacement of 0.2 mm at the variance reduction function was assumed to be equal to 1.
same normal stress, ultimate friction and ultimate dilatancy Figure analysis leads to the conclusion that the character-
angles. The friction angle at the critical state (without istic value considering the spatial variability, f0k;var ,
dilatancy) is presented in the same table and used in the decreases gradually with an increase in the scale of
calculation of the characteristic values. This variable fluctuation, and that it is higher than f0k;mean for values of the
presents a low variation when compared with the other scale of fluctuation lower than 20 m. So, according to EC7,
strength parameters, so its use constitutes a great advantage if the scale of fluctuation is lower than 20 m, the failure
in the probabilistic context. surface must be considered large and the characteristic
The mean value and the COV of the friction angle value equal to f0k;mean . Otherwise, the failure surface extent
sample are equal to 34.38 and 7.4%, respectively. Following is small and the characteristic value is equal to f0k;inf .
the recommendation of Schneider (1999), a value of 10% In order to assess the influence of spatial variability in
was adopted for COV. Two different characteristic values terms of dam sliding stability, a scale of fluctuation of
were calculated using Equations (4) and (5) proposed in 0.60 m (equal to the discontinuities’ average spacing) is
EC7: f0k;inf , corresponding to the lower limit of the assumed in this paper. Given that the critical sliding surface
population, is 29.78, and f0k;mean , corresponding to the lower has a dimension of 94 m, a characteristic value considering
limit of the mean value, is 32.38. the spatial variability, f0k;var , of 33.98 is obtained.

Table 1. Compression and shear test results in effective stresses of the discontinuities’ planes.

Borehole Sample Normal stiffnessa Shear stiffnessb Ultimate friction Ultimate dilatancy Friction angle at the
identification depth (m) (MPa/mm) (MPa/mm) angle (8) angle (8) critical state (8)
PD5 20.15 15.2 2.9 41.3 5.5 35.5
22.70 25.4 1.9 37.5 6.5 32.9
PD7 14.20 39.4 2.4 38.7 4.2 36.2
PD12 37.55 23.6 2.5 41.8 6.3 35.3
38.20 13.2 3.0 39.8 7.8 31.4
a
Normal stiffness at normal stress of 1 MPa.
b
Shear stiffness corresponding to a shear displacement of 0.2 mm at a normal stress of 1 MPa.
 Structure and Infrastructure Engineering 1315

35 spacing; a coarse mesh with an average zone size of 5.0 m

μφ' £ 5.0 m proved to be sufficient).
34 Although there is no evidence from site investigations,
Characteristic value φk (º)

an additional rock mass joint was assumed, dipping

33
φ' k,mean
towards upstream (see a angle in Figure 9), in order to
EC7
allow a failure mechanism of sliding along foundation
32 discontinuities to develop. This hypothetical situation may
simulate a combined mode of failure, where the failure
31 φ' k,var path occurs both along the dam –foundation interface and
30 EC7 φ' k,inf through the rock mass, in geology where the rock is
horizontally or near horizontally bedded and the intact
29 rock is weak (USACE, 1994).
0 20 40 60 80 The foundation model is 200.0 m wide and 80.0 m
Scale of fluctuation δ (m) deep. The dam has the crest of the uncontrolled spillway
33.8 m above ground level and the base is 44.4 m long in
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Figure 8. Influence of spatial variability on the characteristic the upstream – downstream direction, as shown in Figure 5
values proposed by EC7, as a function of the scale of fluctuation,
for L ¼ 94 m. (b). The analysis was carried out utilising software UDEC
(Itasca, 2004), which allows the interaction between the
hydraulic and the mechanical behaviour to be studied in a
4.3. Two-dimensional HM discontinuum model fully coupled way. Joint apertures and water pressures are
The sliding verification of Pedrógão dam is carried out updated at every time step, as described in Lemos (2008).
with a 2D HM discontinuum model, using the discrete It is assumed that rock blocks are impervious and that flow
element method (DEM), involving the cross section of the takes place only through the set of interconnecting
central area of the dam. The discontinuous model discontinuities. Effective normal stresses are obtained at
developed to study the ULS concerning sliding is shown the mechanical contacts. Flow is simulated by means of
in Figure 9. The model presented here is based on a the parallel plate model, and the flow rate per model unit
previously developed model, which had been calibrated width is thus expressed by the cubic law. The medium is
taking into account the quantity of water collected at the assumed to be deformable and the flow is dependent on the
dam’s drainage curtain (Farinha, 2010). In this model, two state of stress within the foundation.
sets of discontinuities were simulated: the first joint set is Both dam concrete and rock mass blocks are assumed
horizontal and continuous, with a spacing of 5.0 m, and the to follow a linear elastic behaviour, with the properties
second set is formed by vertical cross-joints, with a shown in Figure 9. The Mohr – Coulomb failure criterion is
spacing of 5.0 m normal to joint tracks and standard assigned to discontinuities. As previously mentioned, in
deviation from the mean of 2.0 m (for the sliding safety ULS verifications the most important ground properties
assessment it is not necessary to consider the actual joint are those related to strength, deformability generally being

Concrete:
Unit weight = 24 kN/m3
Young´s modulus = 30 GPa
33.8 m α Poisson’s ratio = 0.2
Foundation blocks:
Unit weight = 26.5 kN/m3
Young´s modulus = 10 GPa
Poisson’s ratio = 0.2
Foundation discontinuities:
80 m
kn = 10 GPa/m
ks = 0.1 kn
ϕ’k = 29.7; 32.3°; 33.9º
α varying from 8º to 35º

200 m drainage curtain

grout curtain

Figure 9. Discontinuum model of Pedrógão dam foundation and material properties.

 1316 M.L.B. Farinha et al.

second in importance. For the present case, a sensitivity ares ¼ 0:02 mm. The value of a0 was defined taking into
numerical analysis showed that the stiffness variation has a account both field data and the results of numerous tests
very small influence on water pressure build up. Therefore, performed at US dam sites in the depth range 0– 60 m,
values of 10 GPa/m for the normal stiffness (kn) and which indicated that most conducting apertures were in the
1 GPa/m for the shear stiffness (ks) were assumed at the range of 50 –150 mm at this shallow depth (Barton, Bandis,
foundation discontinuities and at the dam –foundation & Bakhtar, 1985). The maximum aperture was limited
interface. In these discontinuities, as mentioned in Section to 20 £ ares . It was assumed that the grout curtain was
4.2, three different values of the characteristic friction 10 times less pervious than the surrounding rock mass.
angle in effective stresses (f0k;mean ¼ 32.38, f0k;inf ¼ 27.78
and f0k;var ¼ 33.98) were considered. In rock joints,
cohesion in effective stresses was assumed to be zero, 4.4. Identification of the foundation failure mechanism
while at the dam lift joints this parameter was assigned In order to identify the foundation failure mechanism,
2.0 MPa. As the aim of this study is the ULS foundation numerical analysis was carried out using the DEM and
analysis, the strength reduction method was only applied several different geometries, each one of them with the
to the foundation rock mass properties. This procedure is rock mass joint downstream from the dam dipping towards
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

adequate in discrete element analysis and has the upstream at a different angle a (as shown in Figure 9).
advantage of enabling the evaluation of displacement Values of a varying from 88 to 358 were considered. For
indicators during the process of strength reduction. In rock each calculation, the strength of the foundation disconti-
joints, dilatancy may be relevant for small displacements, nuities was gradually reduced, in a sequence of analysis.
but it should not be considered in failure analysis. In this study, the reduction factor was applied to the
Analysis was carried out in two loading stages. First, tangent of the friction angle in effective stresses (tan w0 )
the mechanical effect of gravity loads with the reservoir only. The critical failure mechanism is that for which
empty was assessed. In the UDEC model, an in situ state of instability is achieved for the lowest value of the SRF.
effective stress with K 0 ¼ 0:5 was assumed in the rock It was concluded that a constant value of SRF was
mass. The water table was assumed to be at the same level obtained for an inclined rock mass joint dipping from 108
as the rock mass surface upstream from the dam. Second, to 158 towards upstream. In the study presented here, a
the hydrostatic loading corresponding to the full reservoir rock mass joint dipping 118 towards upstream was
(RWL) was applied to both the upstream face of the dam considered.
and reservoir bottom. Hydrostatic loading was also applied
to the rock mass surface downstream from the dam. In this
second loading stage, mechanical pressure was first 4.5. Numerical results and verification of the ULS of
applied, followed by HM analysis. In both stages, vertical sliding
and horizontal displacements at the base of the model and
Analysis was carried out with the reservoir at the RWL,
horizontal displacements perpendicular to the lateral
both with constant joint hydraulic aperture and taking into
model boundaries were prevented. Regarding the hydrau-
account the HM interaction. Figure 10 shows a detail of a
lic boundary conditions, joint contacts along the bottom
dam and foundation deformation due to the simultaneous
and sides of the model were assumed to have zero
effect of dam weight, hydrostatic loading and flow, in one
permeability. On the rock mass surface, the head was
of the studied load cases of the persistent design situation.
33.8 m upstream from the dam, and 0.5 m downstream.
In this figure, in which block deformation is magnified
The water head of 33.8 m upstream from the dam
3000 times, both horizontal and vertical crest displace-
simulates the water in the reservoir at its RWL. The
ments are represented.
drainage system was simulated by assigning domain water
Figure 11 shows the total head contours in the
pressures along the drain axis, assuming the atmospheric
foundation, in both the persistent and accidental (clogging
pressure at the drains’ head.
of the drainage system) design situations. The main aim of
In UDEC, the hydraulic aperture of the discontinuities
drainage is to reduce the hydraulic head, therefore,
is given by
clogging of the drainage system is a serious problem, as
a ¼ a0 þ Da; ð9Þ this leads to an increase in water pressures within the dam
foundation, and thus along the base of the dam.
where a0 is the aperture at nominal zero normal stress and The variation in water pressures along the dam –
Da is the joint normal displacement taken as positive in foundation joint and along the rock mass joint downstream
opening. A maximum aperture, amax , is assumed, and a from the dam dipping 118 towards upstream is shown in
minimum value, ares , below which mechanical closure Figure 12 (the failure surface is highlighted in black).
does not affect the contact permeability. In the study Figure 12 also shows a comparison of water pressures
presented here it was assumed that a0 ¼ 0:1 mm, and that along the base of the dam with both bi-linear and linear
 Structure and Infrastructure Engineering 1317

3.050 mm pressures calculated in the hydraulic analysis in which the

HM effect is not taken into account (constant joint aperture)
4.039 mm
are very close to those given by the bilinear distribution of
water pressures along the base of the dam, and are the same
regardless of a0 , because the joint hydraulic aperture
remains constant. Previous studies in which it was assumed
that a0 could vary between 0.05 and 0.2 mm showed that,
upstream from the drainage line, water pressures increase as
a0 increases, and are lower than those obtained with
constant joint aperture (Caldeira et al., 2013).
Downstream from the drainage line, water pressures
obtained when the HM interaction is taken into account are
higher than those obtained with constant joint aperture,
and are thus, in this case, slightly higher than those given
by the bilinear distribution of water pressures, which
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Figure 10. Block deformation (magnified 3000 times) due to highlights the relevance of carrying out numerical
dam weight, hydrostatic loading and flow [Combination 2 (C2); analysis. Figure 12(b) shows the water pressures along
characteristic value of the friction angle in effective stresses of the base of the dam with the drainage system clogged.
32.38]. Water pressures along the inclined rock mass joint
downstream from the dam vary almost linearly between
uplift distribution, usually assumed in stability analysis the pressure at the toe of the dam and that corresponding to
with and without drainage systems, respectively. Figure the water height downstream from the dam.
12(a) illustrates that variations in water pressures are highly Figure 13 depicts the variation in dam crest horizontal
dependent on the pressure at the drainage line. Water displacements during the process of reduction of the

Hydraulic head (m)

(a)

drain

grout
curtain

Hydraulic head (m)

(b)

grout
curtain

Figure 11. Total head contours for full reservoir in both design situations: (a) with the drainage system operating properly and (b) with
the drainage system clogged.
 1318 M.L.B. Farinha et al.
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Figure 12. Water pressure along the dam– foundation joint and along the rock mass joint downstream from the dam: (a) dam with both
grout and drainage curtains and (b) dam with the drainage system clogged.

tangent of the friction angle in effective stresses the latter accidental situation, failure is reached for a lower
(f0k;mean ¼ 32.38), in which, for ease of analysis, friction friction angle. Results obtained considering both perma-
angles in the x-axis are shown in reverse order. Figure nent actions (dam weight and hydrostatic pressure on the
analysis leads to the conclusion that, in this particular case, upstream face of the dam) as unfavourable (C1.2) are very
the effect of dam weight reduction (C1.1) on dam crest close to those obtained in Combination 2 of the persistent
horizontal displacements is very close to that due to the design situation (C2). It should be noted that load
increase in water pressures owing to complete clogging of Combination C1.2 does not need to be taken into account
the drainage system of the dam foundation (A), although in in sliding stability analysis, as the dam weight always has a
 Structure and Infrastructure Engineering 1319
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Figure 13. Variations in dam crest horizontal displacements during the SRF process, for the different load cases and design situations
(characteristic value of the friction angle in effective stresses of 32.38).

stabilising effect. Figure analysis shows evidence that a indicating an unsafe design, as shown in Caldeira et al.
dam’s SLS is not reached before the sliding ULS is (2013), because it must be considered to be a persistent
attained. This conclusion is based on the very small design situation.
horizontal displacement calculated for the failure indicator It can be observed that ODF absolute values obtained
at the crest of the dam, when the foundation ULS is for this particular dam are higher than that prescribed in
reached. EC7. The conclusion could be drawn that the safety levels
Table 2 presents the results of the LSD according to the associated with dam design are larger than those used in
DA1 of EC7, in both design situations (see Figure 1), for other structures. However, this dam presents an unusual
the different load cases (Figure 2) and for the different cross section as mentioned in Section 4.1. A large dam is a
characteristic values of the friction angle in effective special structure, the failure of which may have severe
stresses. Figure 14 shows the variation of the ODF with the consequences [class CC3 according to EN 1990 (CEN,
friction angle in effective stresses. In all the analysed 2002b)]. For this type of structure, a special reliability
cases, the ODF is greater than 1, indicating a safe design level is required (class RC3), for which a penalising factor
(the base sliding stability is ensured). In the example of must be applied to actions or resistances. A value of 1.1 is
Pedrógão dam, sliding stability is governed by the suggested in EN 1990 only for actions. Adopting this
persistent design situation and load case C1.1. In this procedure, the minimum ODF value associated with f0k;inf
case, an increase in the ODF of 11.8% is obtained when decreases from 1.19 to 1.08.
f0k;mean is used instead of f0k;inf . When spatial variability is However, instead of applying factor 1.1 to gF for
taken into account this increase is 24.4%. However, if the unfavourable actions, EN 1990 states that it is normally
scale of fluctuation was higher than 20 m, the increase in preferable to require high levels of quality control during
the ODF due to spatial variability would be lower than construction, maintenance and operation. This is the usual
11.8% as a result of the decrease of f0k;var (see Figure 8). procedure in large concrete dams and thus the ranges
It should be noted that in a dam without an operating presented in Table 2 may be considered as final values.
drainage system, the ODF values would be lower than 1, In EC7 DA1, the ULS is usually governed by Combi-

Table 2. Results of the LSD according to DA1of EC7.

Design situation Permanent design situation Accidental design situation

Combination 1 Combination 2 Combination 2
Characteristic value C1.1 C1.2 C2 A
of the friction angle
in effective stresses SRF ODF SRF ODF ODF ODF
32.38 1.80 1.33 2.50 1.85 2.00 1.50
29.78 1.60 1.19 2.30 1.70 1.80 1.40
33.98 2.00 1.48 2.70 2.00 2.20 1.60
 1320 M.L.B. Farinha et al.
Over design factor -ODF (–)
2.4 rock mass foundations. This study presents in detail the
C2 LSD procedures, proposed in Eurocodes, for concrete
2.2
gravity dam sliding stability assessment. The application
2.0 C1.2
of these procedures to Pedrógão dam is presented, taking
1.8 into account information regarding the characteristics of
1.6 A the foundation rock mass provided from tests carried out
C1.1 both in in situ and in the laboratory. The DEM and a two-
1.4
dimensional nonlinear HM discontinuum model were used
1.2 to analyse the sliding safety. The critical failure
1.0 mechanism was identified using the SRF method. The
29.0 30.0 31.0 32.0 33.0 34.0 35.0
same method was used to verify the ULS for both
Friction angle in effective stresses (°) persistent and an accidental design situation involving the
Figure 14. Variation of the ODF with the friction angle in complete clogging of the drainage system of the dam
effective stresses. foundation. Results led to the conclusion that the LSD
methodology may be followed for dam foundation design
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

with the partial factor values prescribed in EC7.

nation 2. Nonetheless, results analysis showed that in this LSD requires the use of characteristic values of both
case, despite the geotechnical character of the problem, loads and material strength parameters, which have to be
the safety is controlled by Combination 1, considering the carefully set. In the geotechnical context, the definition of
gravity load as favourable. This can be explained by the the strength parameters is not unique and depends on the
high hydrostatic pressures on the upstream face of the dam failure surface extent and on the spatial variability of the
and by the high uplift pressures when compared with the analysed property. Two different values are proposed in
geostructure’s self-weight. This result can be considered EC7, and the rational choice between them involves the
relevant for further applications of the procedure for other evaluation of the scale of fluctuation of the property
gravity dams. variation based on both in situ and laboratory data. Spatial
variability is introduced in the study presented here as a
means of obtaining an estimate of the characteristic value
of the strength parameter. In fact, this value of the strength
5. Conclusions parameters should be averaged over the extent of the
LSD is the current trend in both structural and potential failure surface. The determination of the scale of
geotechnical engineering, not only in Europe, but also in fluctuation is often complex; however, as shown in the case
Australia, Canada, Japan and the USA. It is true that the of Pedrógão dam, the consideration of spatial variability
modus operandi for justifying both the deterministic and could allow optimisation of the design as well as reduction
LSD approaches is basically the same, that is the of costs. This issue is an innovative approach in the rock
examination of loads, strength parameters and the safety mechanics field, which merits further research.
verifications. However, in accordance with the LSD Dam stability analysis is accurately carried out with
analysis, partial factors are used to weight actions, HM models in which the rock mass discontinuities are
strengths and model factors, which take into account the explicitly represented. The results of these models,
uncertainties involved in the analysis. however, depend mainly on the fracture pattern and on
The main advantage of this approach is that it provides HM characterisation data, and therefore, detailed field data
a unique safety concept which may be applied to any kind are required. The fracture pattern in the foundation of
of structure. It is significantly different from the traditional Pedrógão dam was assumed as a deterministic parameter
allowable stress and overall safety factor DAs, as it allows and an unfavourable situation was considered, after a
different limit states to be defined regarding safety and detailed study regarding the identification of the
serviceability. In addition, LSD imposes a rigorous and foundation failure mechanism. However, the application
systematic formalism on calculations, which will hope- of this methodology to other dams may require carrying
fully lead to standardising analytical procedure and out of numerical analysis, taking into account different
unequivocal vocabulary. This formal structure will ensure foundation fracture patterns.
that all analyses are conducted in the same pattern, leading The numerical analysis raises problems regarding the
to a standard practice that will allow the expected direct application of load partial factors due to the
introduction of the main requirements for dam safety in the difficulty in distinguishing the favourable and unfavour-
Eurocodes. able effects caused by different actions. The procedure to
Regarding the foundations of concrete gravity dams, overcome this difficulty is presented in this paper. This
ULSs include overall and sliding instabilities; however, it study is part of an ongoing project regarding the
is well known that the latter limit state controls safety in application of a semi-probabilistic approach to the safety
 Structure and Infrastructure Engineering 1321

evaluation of large concrete dams, in accordance with the CFBR (2010). Recommandations pour la justification de la
safety philosophy on which Eurocodes are based. stabilité des barrages et des digues en remblai [Guidelines for
the justification of the stability of embankment dams]. Goup
de Travaiul “Justification des Barrages et des digues en
Remblai”. Comité Franc ais des Barrages et Réservoirs.
Acknowledgements CFBR (2012). Recommandations pour la justification de la
Thanks are due to the dam owner, EDIA, Empresa de stabilité des barrages-poids [Guidelines for the justification
Desenvolvimento e Infra-Estruturas do Alqueva, SA for of the stability of gravity dams, published in November
permission to publish data related to Pedrógão dam. 2013]. Ad-hoc working group. Comité Franc ais des Barrages
et Réservoirs.
Deroo, L., & Boris, J. (2011). Notes sur l’accidentologie des
Notes barrages-poids [Notes on gravity dams accidentology].
In Proceedings of the colloque CFBR-AFEID: Etudes de
1. Email: [email protected] dangers, Lyon, France, 28 – 20 November 2011 (pp. 1 – 18).
2. Email: [email protected] Comité Franc ais des Barrages et Réservoirs.
EDP (2004). Aproveitamento hidroeléctrico de fins múltiplos de
Alqueva. Escalão de Pedrógão. Cartografia geológica-
References geotécnica das superfı́cies escavadas [Multi-purpose
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

Alonso, E., Carol, I., Delahaye, C., Gens, A., & Prat, P. (1996). Alqueva Hydroelectric Scheme. Pedrogão gravity dam.
Evaluation of safety factors in discontinuous rock. Geological and geotechnical mapping of excavated sur-
International Journal of Rock Mechanics and Mining faces]. Technical report, Porto, Portugal (in Portuguese).
Science & Geomechanics Abstracts, 33, 513– 537. European Club of ICOLD (2004a). Sliding safety of existing
Asadollahi, P., & Tonon, F. (2010). Definition of factor of safety gravity dams – Final report. Report of the European
for rock blocks. International Journal of Rock Mechanics Working Group.
and Mining Sciences, 47, 1384– 1390. European Club of ICOLD (2004b). Uplift pressures under
Australian National Committee on Large Dams (ANCOLD). concrete dams – Final report. Report of the European
(2013). Guidelines on design criteria for concrete gravity Working Group.
dams (September 2013). Australia: Author. Farinha, M.L.B. (2010). Hydromechanical behaviour of concrete
Barla, G., Bonini, M., & Cammarata, G. (2004). Stress and dam foundations – In situ tests and numerical modelling
seepage analyses for a gravity dam on a jointed granitic rock (Ph.D. thesis. Technical University of Lisbon, Lisbon,
mass. In Proceedings of the 1st International UDEC/3DEC Portugal).
Symposium, Bochum, 29 September – 1 October 2004 FERC (2002). Engineering guidelines for the evaluation of
(pp. 263– 268). Rotterdam: A.A. Balkema. hydropower projects. Chapter 3 – Gravity dams.
Barton, N., Bandis, S., & Bakhtar, K. (1985). Strength, Frank, R., Bauduin, C., Driscoll, R., Kavvadas, M., Krebs
deformation and conductivity coupling of rock joints. Ovesen, N., Orr, T., & Schuppener, B. (2004). Designer’s
International Journal of Rock Mechanics and Mining guide to EN 1997-1 Eurocode 7: Geotechnical design –
Science & Geomechanics Abstracts, 22, 121– 140. General rules. London: Thomas Telford Ltd.
Bernstone, C., Westberg, M., & Jeppsson, J. (2009). Structural Gimenes, E.Á., & Fernández, G. (2006). Hydromechanical
assessment of a concrete dam based on uplift pressure analysis of flow behaviour in concrete gravity dam
monitoring. Journal of Geotechnical and Geoenvironmental foundations. Canadian Geotechnical Journal, 43, 244– 259.
Engineering, 135, 133– 142. Gulvanessian, H., Calgaro, J.-A., & Holichy, M. (2002).
Caldeira, L., Farinha, M.L.B., Maranha das Neves, E., & Lemos, Designer’s guide to EN 1990. Eurocode: Basis of structural
J.V. (2013). Limit state design of the foundations of concrete design. London: Telford Ltd.
gravity dams – A case study. In P. Arnold, G.A. Fenton, M. International Commission on Large Dams (1988). Dam design
A. Hicks, T. Schweckendiek, & B. Simpson (Eds.), Modern criteria. The philosophy of their selection. Bulletin 61. Paris:
geotechnical design codes of practice – Implementation,
Author.
application and development (pp. 143– 156). Amsterdam:
International Commission on Large Dams (1993). Rock
IOS Press BV.
foundations for dams. Bulletin 88. Paris: Author.
CEN (2002a). Eurocode 1: Actions on structures – Part 1-1:
General actions – Densities, self-weigth, imposed loads for International Commission on Large Dams (1995). Dam failures.
buildings. EN 1991-1-1. Statistical analysis. Paris: Author.
CEN (2002b). Eurocode 0: Basis of structural design. EN 1990. Itasca (2004). UDEC – Universal distinct element code, Version
CEN (2004). Eurocode 7: Geotechnical design – Part 1: General 4.0, User’s Manual. Minneapolis, MN: Itasca Consulting
rules. EN 1997-1. Group.
CEN (2005). Eurocode 8: Design of structures for earthquakes Kovari, K., & Fritz, P. (1989). Re-evaluation of the sliding
resistance. EN 1998. stability of concrete structures on rock with emphasis on
CEN (2006). Eurocode 1: Actions on structures – Part 1-7: European experience (Technical report REMR-GT-12).
General actions – Accidental actions. EN 1991-1-7. London: U.S. Army Research Development and Standard-
CEN (2007). Eurocode 7: Geotechnical design – Part 2: Ground ization Group.
investigation and testing. EN 1997-2. Kovari, K., & Fritz, P. (1993). Reevaluation of the stability of
CFBR (2006). Recommandations pour la justification de la large concrete structures on rock. In J.A. Hudson (Ed.),
stabilité des barrages-poids [Guidelines for the justification Reevaluation of the stability of large concrete structures on
of the stability of gravity dams]. Groupe de Travail “Calcul rock (Vol. 5, pp. 653– 700). Oxford: Pergamon Press.
des barrages-poids”. Comité Franc ais des Barrages et Lemos, J.V. (1999). Discrete element analysis of dam
Réservoirs. foundations. In V.M. Sharma, K.R. Saxena, & R.D. Woods
 1322 M.L.B. Farinha et al.

(Eds.), Distinct element modelling in geomechanics Peyras, L., Royet, P., Deroo, L., Albert, R., Becue, J.-P., Aigouy,
(pp. 89 – 115). Rotterdam: Balkema. S., Bourdarot, E., Loudiere, D., & Kovarik, J.-B. (2008).
Lemos, J.V. (2008). Block modelling of rock masses. Concepts French recommendations for limit-state analytical review of
and application to dam foundations. European Journal of gravity dam stability. European Journal of Environmental
Environmental and Civil Engineering, 12, 915– 949. and Civil Engineering, 12, 1137 –1164.
LNEC (2004). Geomechanical characterization tests of the Rocha, M. (1978). Analysis and design of the foundations of
foundation rock mass of the Pedrógão dam (Report 44/04- concrete dams. General report, theme III. In Proceedings of
NFOS). Lisbon: LNEC (in Portuguese). the International Symposium on Rock Mechanics Related to
Londe, P. (1973). The role of rock mechanics in the
Dam Foundations, Rio de Janeiro, Brasil, 27 –29 September
reconnaissance of rock foundations. The Quarterly Journal
of Engineering Geology, 6, 56 – 74. 1978 (Vol. 2, pp. III.11-III.70). Rio de Janeiro: Gráfica
Lupoi, A., & Callari, C. (2012). A probabilistic method for the Editora.
seismic assessment of existing concrete gravity dams. Royet, P., & Peyras, L. (2013). French guidelines for structural
Structure and Infrastructure Engineering, 8, 985– 998. safety of gravity dams in a semi-probabilistic format.
Miranda, M.P., & Maia, M.C. (2004). Main features of the In Proceddings of the 9th ICOLD European Club
Alqueva and Pedrógão Projects. The International Journal Symposium, Venice, Italy, 10 – 12 April 2013 (pp. 1-8).
on Hydropower and Dams, 11, 95 – 99. European Club of ICOLD.
Mostyn, G., Helgstedt, M.D., & Douglas, K.J. (1997). Towards Schneider, H.R. (1999). Determination of characteristics
Downloaded by [University of Nebraska, Lincoln] at 20:03 28 October 2015

field bounds on rock mass failure criteria. International soil properties. In Proceedings of the 12th European
Journal of Rock Mechanics and Mining Sciences, 34, 208. Conference on Soil Mechanics and Foundation Engineering
e201 – 208.e218. (Vol. 1, pp. 273– 281). Rotherdam: A.A. Balkema.
Noorishad, J., Ayatollahi, M.S., & Witherspoon, P.A. (1982). A USACE (1983). Design of gravity dams on rock foundations:
finite-element method for coupled stress and fluid flow Sliding stability assessment by limit equilibrium and
analysis in fractured masses. International Journal of Rock selection of shear strength parameters (Technical report
Mechanics and Mining Science & Geomechanics Abstracts,
GL-83-13). Washington, DC: Author.
19, 185– 193.
NPB (1993). Normas de projecto de barragens [Portuguese USACE (1994). Engineering and design. Rock foundations
guidelines for dam design]. Portaria 846/1993 de 10 de (Engineer Manual 1110-1-2908). Washington, DC: Author.
Setembro. Diário da República, Lisbon, Portugal (in USACE (2000). Evaluation and comparison of stability analysis
Portuguese). and uplift criteria for concrete gravity dams by three Federal
Orr, L.L. (2012). How Eurocode 7 has affected geotechnical Agencies (ERDC/ITL TR-00-1). Washington, DC: Author.
design: A review. Geotechnical Engineering. Proceedings of USACE (2005). Stability analysis of concrete structures
the Institution of Civil Engineers, 165, 337– 350. (Engineer Manual 1110-2-2100). Washington, DC: Author.