Advances in Engineering Software 34 (2003) 403–420

www.elsevier.com/locate/advengsoft

Computer aided stability analysis of gravity dams—CADAM

Martin Leclerc, Pierre Léger*, René Tinawi
Department of Civil Engineering, École Polytechnique de Montréal, University of Montreal Campus, P.O. Box 6079,
Station CV, Montreal, Que., Canada H3C 3A7
Received 6 March 2002; accepted 3 March 2003

Abstract
This paper presents the main features and organisation of CADAM, a computer program, freely available, that has been developed for the
static and seismic stability evaluations of concrete gravity dams. CADAM is based on the gravity method using rigid body equilibrium and
beam theory to perform stress analyses, compute crack lengths, and safety factors. Seismic analyses could be done using either the pseudo-
static or a simplified response spectrum method. CADAM is primarily designed to provide support for learning the principles of structural
stability evaluation of gravity dams. It could also be used for research and development on stability of gravity dams. In adopting several
different world-wide published dam safety guidelines, a large number of modelling options have been implemented. These include (i) crack
initiation and propagation, (ii) effects of drainage and cracking under static, seismic, and post-seismic uplift pressure conditions, and (iii)
safety evaluation procedures using deterministic, allowable stresses and limit states probabilistic analyses (Monte-Carlo simulations).
Structural stability evaluation of a 30 m dam is presented to illustrate the use of CADAM.
q 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Concrete gravity dams; Stability analysis; Computer aided design; Gravity method; Floods; Seismic response; Monte-carlo simulations

1. Introduction the reservoir must be contained to avoid a catastrophe

downstream.
There are over 4800 large concrete gravity dams in A progressive methodology is normally adopted starting
existence throughout the world outside China. In North with the gravity method based on rigid body equilibrium
America, in particular, the average age of these dams is and beam theory before considering linear or nonlinear
about fifty years. The static and seismic safety of existing finite element models, if necessary [1,2]. FERC [3,4], CDA
concrete gravity dams is therefore a continuous concern to [5], USACE [6], Ancold [7], and USBR [8] present
dam owners owing to the ageing processes altering their guidelines for dam safety assessment based on the gravity
strength and stiffness, as well as revised predictions of the method. Nevertheless, even the gravity method, which is
maximum loads associated to severe floods and earth- relatively simple to understand and apply, can be lengthy
quakes. It is thus required to perform periodic reassessment when evaluating crack length, especially for inclined failure
of their static and seismic structural stability under extreme planes, or when using a pseudo-dynamic technique. On the
loads for which these dams were not designed. In addition, other hand, finite elements in the linear or nonlinear range
owners with a large number of dams need to assess the have their share of difficulties related to stress singularities
or crack propagation particularly with discrete cracks.
safety of these structures and prioritise their investment
Therefore there was a need to develop an interactive user-
when undertaking expensive rehabilitation works to
friendly software, such as CADAM [9], to assess very
accommodate a probable maximum flood or a maximum
quickly for a given dam the safety margins under extreme
credible earthquake. It is obvious that if under these
loads (Fig. 1). Alternatively, CADAM risk analysis
extreme conditions structural damage can be tolerated,
capabilities are useful to classify which structures are the
* Corresponding author. Tel.: þ1-514-340-4711x3712; fax: þ 1-514- most vulnerable within a portfolio of dams.
340-5881. In addition to safety issues, there are several differences
E-mail address: leger@struc.polymtl.ca (P. Léger). among adopted guidelines regarding:
0965-9978/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0965-9978(03)00040-1
404 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

Fig. 1. CADAM loading conditions for static and seismic analyses: (a) basic static analysis conditions; (b) pseudo-static seismic analysis; (c) pseudo-dynamic
seismic analysis.

(a) cracking initiation and propagation criteria, noted above, there is also a growing interest in performing
(b) static and seismic uplift pressures along joints and risk based safety evaluation where the probability of failure
cracks, and of a dam is evaluated considering explicitly uncertainties in
(c) safety evaluation format (allowable stress, limit state strength and loading modelling parameters through suitable
method). probability density functions [12].
In most engineering offices, in-house spreadsheets are
Moreover, the response of spillways and gravity dams developed and adapted on a case by case basis to perform
during the 1996 Saguenay flood in Quebec, Canada dam stability analysis following particular safety guidelines.
emphasised once more the need to consider overtopping This is due to the very lengthy and tedious computations,
and floating debris while performing flood safety assess- particularly when pseudo-dynamic seismic analyses are
ment [10]. Seismic safety evaluations are very frequently considered. Moreover, there are no widely available
conducted using the pseudo-dynamic (response spectrum) computational tools:
method as presented by Chopra [11] in complement to the
pseudo-static seismic coefficient (rigid body) method. (a) for learning the principles of stability analysis in the
Although these calculations are well documented, they are academic or professional environment, and,
complicated due to the iterative nature of crack length (b) for performing research and development on the
calculation and its consequence on the uplift pressures. As structural safety of gravity dams.
 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 405

We have thus identified needs to develop, and put in the † The dam body is divided into lift joints of homogeneous
public domain, a comprehensive computer program, properties along their length; the mass concrete and lift
CADAM, to perform stability evaluation of gravity dams joints are initially assumed uniformly elastic (uncracked
based on the gravity method. state).
This paper presents the organisation and computational † All applied loads are transferred to the foundation by the
features of CADAM to provide a fully integrated computing cantilever action of the dam without interactions with
environment with output reports and graphic support to adjacent monoliths.
visualise input parameters and output performance indi- † There is no interaction between the joints, that is, each
cators as required in practice (stresses, crack length, joint is analysed independently from the others.
resultant position, safety factors). Several modelling options † Normal stresses are linearly distributed along horizontal
have been included allowing users: planes.
† The uplift pressures intensity could be modified along a
† to perform static, pseudo-static, pseudo-dynamic, and crack, depending upon the drainage condition and rate of
probabilistic safety assessment, crack opening (static vs. seismic conditions).
† to corroborate hand calculations with computer calcu-
lations to develop the understanding of the computational 2.1. Stress analysis
procedures,
† to conduct parametric analysis on the effects of CADAM uses an iterative procedure summarised in
geometry, strength of materials and load magnitudes on Fig. 2b to compute the crack length, Lc : Once the crack
the structural response, initiation criterion indicates the formation of a crack, the
† to compare uplift pressures, crack propagation, and shear iterative computation begins. The crack length is computed
strength (peak, residual) assumptions from different dam using a bi-section method and the uplift pressures are
safety guidelines [3 –8], and, updated according to the selected drainage options until the
† to study different strengthening scenarios including post- crack propagation criterion indicates stress equilibrium and
tensioning. crack arrest. To consider stress concentration at the crack
tip, the criterion for crack initiation may be different than
After presenting an overview of CADAM main that for crack propagation. Closed form formulations to
features and analysis options, specific modelling tech- compute crack length while updating uplift pressures as
niques adopted for basic static analyses, flood, and seismic water penetrates the crack are only available for simple
as well as probabilistic analyses are discussed. Appli- cases where the crack is considered horizontal, a no-tension
cations related to the structural response of a 30 m high criterion is used and without drainage. The iterative
gravity dam are described for illustrative purposes. The computation of crack length when concrete tensile strength
paper ends with perspectives for future CADAM is nonzero or for an inclined plane requires the compu-
developments. tational power of a computer.
In most guidelines, uplift pressures are considered as
external load acting on the surface of the joint. The stress at
2. Basic principles of the gravity method the crack tip, sn; is computed while including uplift
pressures in the force resultant (in the crack propagation
The evaluation of the structural stability of a dam against iterative procedure) [3,6,8,13]. This calculation produces a
sliding, overturning and uplifting along concrete lift joints is linear effective normal stress distribution, sn ; even in the
performed considering two distinct analyses: case where a nonlinear uplift pressure distribution is present
along the base due to drainage or cracking:
† A stress analysis to determine eventual crack length and P P
compressive stresses (Fig. 2). V My
sn ¼ ^ ð1Þ
† A stability analysis to determine: (a) the safety margins A I
against sliding along the joint considered, and (b) the
position of the resultant of all forces acting on the joint. where
P
The gravity method considers the dam as a cantilevered V ¼ sum of all vertical load including uplift pressures,
structure and is based (a) on rigid body equilibrium to P ¼ area of uncracked ligament,
A
determine the internal forces acting on the potential failure M ¼ moment about the centre of gravity of the
plane (joints and concrete-rock interface), and (b) on beam uncracked ligament of all loads including uplift press-
theory to compute stresses. The use of the gravity method ures,
requires several simplifying assumptions regarding the I ¼ moment of inertia of the uncracked ligament,
structural behaviour of the dam and the application of y ¼ distance from centre gravity of the uncracked
loads [4]: ligament to the location where the stresses are computed.
406 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

Fig. 2. (a) Existing dam vs. idealized structural models. (b) Iterative procedure for crack length computations.

Alternatively, uplift pressures could be considered as an specify a minimal value of compressive stress, spn ; to
internal load along the joint [4]. The stresses at the crack tip determine the compressed area upon which cohesion
are computed from total stresses without uplift pressures. could be mobilised,
Uplift pressures are then subtracted from total stress to AC ¼ area in compression (a function of the crack
obtain effective stresses, sn to be used for crack initiation length)
P and
(propagation) criterion [4]. These effective stresses may not H ¼ sum of horizontal forces.
exhibit a linear distribution along a joint.
For a stability analysis, the basic shear-friction sliding Eqs. (1) and (2) have been enhanced in CADAM to
safety factor (SSF) formula along a horizontal plane is given consider inclined lift joints and all relevant seismic load
as: components.
P
ð V
2 UÞtan f þ cAC
SSF ¼ P ð2Þ
H 3. CADAM—overview of main features and analysis
where options
P
V
¼ Sum of vertical forces excluding uplift pressures, 3.1. Programming and computing environment
U ¼ uplift pressure force resultant,
f ¼ friction angle (peak value or residual value), Developers tend to divide along language boundaries.
c ¼ Cohesion (apparent for rough unbonded joint or real Once they know a programming language, they identify
for bonded joint); for apparent cohesion, the user may themselves by it: ‘a Cþ þ programmer,’ ‘a Delphi
 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 407

developer’, etc. The key is applicability and each program- Delphi shares the compiler back end with Cþ þ Builder
ming language is as a specialized tool. A hammer specialist compiler, so the efficiency of the generated codes is
does not make a good carpenter. The authors find Cþ þ , comparable. In reliable benchmarks [14], Microsoft Visual
Delphi, and Java to all be useful languages, and even a little Cþ þ rated tops in speed and size efficiency in many cases.
Visual Basic applies when appropriate. Although these small advantages are unnoticeable for

Fig. 3. CADAM overall organization.

408 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

general applications. Visual Basic operates in an interpreted distributed applications, including client/server, multi-
mode and is quite reactive. However, Visual Basic speed tiered, and Web-based systems.
rates well behind Delphi and Cþ þ tools. Java-based tools Delphi has been adopted at first for its object oriented
such as Borland JBuilder and Microsoft Visual Jþ þ programming environment. The other reason for this choice
approach compile times of Delphi. However, Java speed is that often the complete design methodology is only
efficiency leaves something to be desired, because Java is an worked out in full detail at the programming stage.
interpreted language. Therefore understanding of the design engineering problem
CADAM has been developed using Borland Delphie 6 is more important than knowledge and experience of
compiler [15]. Borland Delphi is an object-oriented, visual programming [16]. Finally, the ease of programming,
programming environment to develop 32-bit applications support, and suitability for engineering design software,
for deployment on Windows and Linux platforms. Using are all additional reasons for adopting this programming
Delphi, the programmer can create highly efficient appli- language for this particularly complex engineering analysis
cations with a minimum of manual coding. Delphi provides and design problem.
all the tools needed to develop, test, debug, and deploy
applications, including a large library of reusable com- 3.2. Overall program organisation and analysis options
ponents, a suite of design tools, applications and form
template, and programming wizards. These tools simplify Fig. 3 presents CADAM overall organisation. The dam
software coding and shorten development time. Delphi can geometry (Fig. 4a,b), material properties (Fig. 4c) the
be used to write Windows and Linux graphical user various load conditions, cracking options, and load
interface (GUI) applications, console applications, service combinations are first specified as input data for subsequent
applications, dynamic-link libraries (DLLs), and other structural analyses, outlined in Fig. 1. The following
programs. Delphi includes features to write easily analysis options are currently available: (1) static analyses,

Fig. 4. Definition of dam model: (a) dam-foundation-reservoir system and CADAM interface; (b) dam geometry; (c) material properties.
 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 409

(2) seismic analyses, (3) post-seismic analyses, (4) incre- pointing u/s and d/s to estimate the cumulative damage by
mental load analysis, and (5) probabilistic safety analysis. reducing the cohesion that could be mobilised along the
joint considered. The cohesion is considered null along the
seismically induced crack length to compute the SSFs in
4. Static loading conditions seismic and post-seismic conditions.
Since the pseudo-static method does not recognise the
The load conditions supported by CADAM are shown in oscillatory nature of earthquake loads CADAM performs
Fig. 1. Some particular features are described in the the safety evaluation in two phases:
following. Various dam safety guidelines equations pre-
sented to compute the uplift pressures according to the (a) a stress analysis using peak ground acceleration (or
position of the drain from the upstream (u/s) face, the drain spectral acceleration) values to compute the crack
effectiveness and the elevation of the drainage gallery have length, and
been implemented (Fig. 5). It is interesting to note that (b) a stability analysis using sustained acceleration
Federal agencies in the US (FERC, USACE and USBR) are values to compute SSFs.
currently evaluating the need for unified Federal criteria for
the calculation of uplift pressures as well as crack initiation A single acceleration peak might be sufficient to induce a
and propagation criteria in the stability of concrete gravity crack but it may not be of sufficient duration to induce
dams [17]. It is believed that a computational tool like significant sliding displacement. For stability evaluation an
CADAM could be of great assistance to conduct extensive ‘effective’ acceleration equals to 0.67 –0.5 the peak value
parametric analyses for various dam geometry and drainage has often been used in practice [18]. The stress analysis is
conditions to study the effects of modelling assumptions on
therefore used to determine the length over which cohesion
computed performance indicators. During a severe flood, it
will be applied in the stability analysis.
is possible that a section of the dam be overtopped. In this
case, water pressure may be considered on the crest surface
as well as floating debris. 5.1. Pseudo-static analysis

In a pseudo-static seismic analysis, the inertia forces

5. Seismic and post-seismic safety analysis induced by the earthquake are computed from the product of
the mass and the acceleration. The dynamic amplification of
Some original features that have been included for inertia forces along the height of the dam due to its
seismic and post-seismic safety analyses are presented flexibility is neglected. In a pseudo-static analysis, it is
below. Existing cracks computed from the initial static required to specify the peak ground horizontal and vertical
conditions may close depending on the intensity and accelerations as well as the sustained accelerations.
orientation of the earthquake forces. Separate analyses Westergaard added mass [6] is used to represent the
could be performed successively with the base acceleration hydrodynamic effects of the reservoir on the dam. Options

Fig. 5. Uplift pressures and drainage system options.

410 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

are provided to account for water compressibility effects and modes and then apply the modal combination rule to
inclination of the u/s and d/s faces. stresses. However, this approach, adopted in linear analysis,
is not suitable to estimate crack length, especially if uplift
5.2. Pseudo-dynamic analysis pressures are to be varied within the seismic crack (e.g. with
no uplift pressure in an opened crack). Moreover, it is
A pseudo-dynamic seismic analysis is based on the assumed that the period of vibration of the dam is unaffected
response spectrum method. A pseudo-dynamic analysis is by cracking which is obviously an approximation that might
conceptually similar to a pseudo-static analysis except that be overcome only if transient nonlinear dynamic analyses
it recognises the dynamic amplification of the inertia forces are considered [1]. However, consistency in modelling
along the height of the dam. However, the oscillatory nature assumptions, as implemented in CADAM, ensures that the
of the amplified inertia forces is not considered. That is the results of a pseudo-dynamic analysis converge towards the
stress and stability analyses are performed with the inertia results of a pseudo-static analysis as the period of vibration
forces continuously applied in the same direction. The basic tends to zero in the pseudo-dynamic analysis. This novel
input data required to perform a pseudo-dynamic analysis, consistent approach is made possible and simple to use
using the simplified response spectrum method proposed by within CADAM.
Chopra [11], are:

(a) peak ground and spectral accelerations (peak and

6. Cracking options and evolution of uplift pressures
effective values),
in cracks
(b) dam and foundation stiffness and damping properties,
(c) reservoir bottom damping properties and velocity of an
impulsive pressure wave in water, CADAM provides various options for specification of:
(d) modal summation rule.
(a) tensile strengths for crack initiation and propagation,
In a pseudo-dynamic analysis, the moment and axial (b) dynamic amplification factor for the tensile strength,
force acting on the lift joint considered are computed from (c) incidence of cracking on static uplift pressure distri-
the selected modal combination rule. The resulting moment butions and drain effectiveness (Fig. 6),
and axial force are then used to compute the related stresses (d) effect of cracking on the transient evolution of uplift
and crack length. This approach is generally conservative. pressures during earthquakes (Fig. 7),
In linear (uncracked) analysis, it is more appropriate to (e) evolution of uplift pressures in the post-seismic
compute stresses separately for the first mode and the higher conditions.

Fig. 6. Cracking options: effect of cracking on the drainage system.

 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 411

† USACE [3] and FERC [3,4] assume that uplift pressures

are unchanged by earthquake load (i.e. at the pre-
earthquake intensity during the earthquake).
† USBR [8] mentions: ‘When a crack develops during an
earthquake event, uplift pressures within the crack are
assumed to be zero’. This is based on the assumption that
a rapid crack opening reduces the uplift pressures and
that the cyclic crack motions are too fast to allow
reservoir water to penetrate and build-up the pressure.
† CDSA [13] mentions: ‘In areas of low seismicity,
uplift pressures prior to the seismic event are normally
assumed to be maintained during the earthquake even
if cracking occurs. In areas of high seismicity, the
assumption is frequently made that uplift pressures on
the crack surface are zero during the earthquake when
the seismic force are tending to open the crack’.
Fig. 7. Transient evolutions of uplift pressures in seismically induced crack.
To model these various assumptions CADAM allows
When cracking is allowed, a distinction is made between three options for transient evolution of uplift pressures in
the criteria for crack initiation and crack propagation. After seismic cracks:
crack initiation, say at the u/s end of a joint where stress
concentration is minimal, it is likely that stress concentration (i) no change from pre-earthquake pressure conditions,
will occur near the tip of the propagating crack [4]. For (ii) full pressure, and
example the crack initiation criterion could be set to a tensile (iii) zero pressure.
strength of 1000 kPa but once the crack is initiated, it should
be propagated to a length sufficient to develop compression at To reach equilibrium in the post-earthquake condition
the crack tip (no-tension condition for crack propagation). two options are provided. First, a conservative assump-
The allowable tensile strengths for crack initiation and tion for post-seismic uplift pressures is to use the full
propagation are specified for different usual, flood, seismic reservoir pressure in earthquake-induced cracks in the
and post-seismic load combinations. Allowable tensile post-seismic safety assessment. However, if seismic
strengths for crack initiation and propagation are specified cracks are closed after the earthquake the uplift pressures
as the tensile strength divided by appropriate coefficients. to be used for the post-seismic condition could be
Upon cracking when drainage is considered, four options relieved by the drainage system. In this case, the pre-
are offered (Fig. 6): earthquake uplift pressure intensity is used immediately
after the earthquake.
(1) no drain effectiveness under any cracking condition,
(2) no drain effectiveness when the crack reaches the drain
line, 7. Load combinations and safety evaluation format
(3) full drain effectiveness, but with full uplift pressures
applied between the reservoir and the drain line, Five load combinations are supported by CADAM
(4) full drain effectiveness with a linear decrement in (Fig. 8):
uplift pressures starting from full reservoir pressure
at the reservoir level to the drainage pressure at the (1) normal operating,
drain line. (2) flood
(3) seismic 1 (e.g. Design Base Earthquake),
In a design office, the consideration of these options is (4) seismic 2 (e.g. Maximum Credible Earthquake), and
often studied to obtain upper and lower bounds values due to (5) post-seismic.
uncertainties related to drainage conditions.
Due to the lack of historical and experimental evidences, For each load combination, multiplication factors could
there is still a poor knowledge of the transient evolution of be specified for each basic load conditions. This option is
uplift pressures in a crack due to its cyclic movements very useful when an applied load is increased until a safety
during earthquakes (Fig. 7). factor equal to 1 is reached and thus determines the ultimate
strength of the dam. For each load combination, the required
† ICOLD [19] states: ‘The assumption that pore pressure safety factors to ensure an adequate safety margin for
equal to the reservoir head is instantly attained in cracks structural stability are specified. These values are not used in
is probably adequate and safe’. the computational algorithm of the program. They are
412 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

Fig. 8. Load combinations, SSF computation and required safety factors.

reported in the output results to facilitate the interpretation strength parameters to reflect related uncertainties. By
of the computed safety factors in comparison with the adjusting the input material parameters, and applying the
corresponding allowable values. specified load multiplication factors, CADAM could be
Different strategies have been adopted to study the safety used to perform limit analysis of gravity dams as
margin of concrete dams as a function of the uncertainties in described by ANCOLD [7].
the applied loading and material strength parameters. By
proper definition of basic loading condition parameters and
multiplication factors to form load combinations, a variety 8. Probabilistic and risk analyses
of loading scenarios could be defined to assess the safety of
a dam-foundation-reservoir system. In some cases, the The objectives of CADAM probabilistic analysis module
applied loads are increased to induce failure (e.g. u/s, d/s is to compute the probability of failure of a gravity dam as a
water levels are increased, ice loads, water density, etc.). function of the uncertainties in loading and strength
The safety margin is then assessed by comparing the parameters that are considered as random variables
magnitude of the load inducing failure with that of the (Fig. 9). A probabilistic analysis requires more information
applied load for the combination under study. CADAM can than a deterministic analysis. For example, probability
be used effectively to perform this type of study using a density functions (PDF) (uniform, normal, log-normal or
series of analyses while increasing the applied loads either user defined) are to be selected for the friction coefficient
through the basic loading input parameters, by applying and cohesion; the mean values, and the standard deviation
appropriate load condition multiplication factors while must then be specified.
forming the load combinations, or by using the incremental CADAM probabilistic analysis module could be used:
load analysis option.
In a different approach, the specified material strengths † For educational purpose to develop a basic understanding
are reduced while inputting basic data (friction coefficient of the concepts and procedure required to perform a risk
ðtan fÞ; cohesion, tensile strength, etc.). Series of analyses analysis, where risk is evaluated as the product of the
are then performed until a safety factor of 1 is reached for a probability of failure ðpf Þ and the related consequences.
particular failure mechanism. Comparing the material † To actually perform probabilistic (risk) analysis for a
strength inducing failure to the expected material strength particular dam. It is then possible to construct a fragility
could then assess the safety margin. curve, that defines the probability of failure as a function
The Australian National Committee on Large Dams [7] of an applied load level and compute reliability indices
presented a dam safety evaluation format based on a limit (as a function of ð1 2 pf Þ).
state approach. Various magnification and reduction † To perform R&D in risk based dam safety assessment
factors are applied to basic load conditions and material such as calibration of nominal strength (resistance R),
 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 413

Fig. 9. Probabilistic analysis input data.

reduction factor, f; and load ðLÞ factor, g; to develop where N ¼ total number of simulations and nf ¼
limit state based safety evaluation of the form: fR $ gL: ¼ number of failures.
† To study different safety approaches (e.g. strength
requirements to ensure uniform risk during the service The output results can also be analysed statistically to
life of a dam). define the mean, the variance, the PDF and the cumulative
density function (CDF).
Due to concrete cracking, and related modifications in
uplift pressures, the stress and stability analysis of a dam is
in general a nonlinear process. Monte-Carlo simulation is 9. User feedbacks and comments
used as the computational procedure to perform the
probabilistic ‘nonlinear’ analysis in CADAM. Monte- During the development process of CADAM, developers
Carlo simulation technique ‘involve sampling at random consulted dam engineers to ensure that appropriate technical
to simulate artificially a large number of experiments and to terms were used and that every feature of the software is in
observe the results’ [20]: respect with the state-of-the-practice. On the other hand,
practicing engineers were invited to comment on the overall
(1) a large number (up to 250,000) of loading and strength organisation and characteristics of the software as well as on
parameters are ‘sampled’ at random within bounds of the user’s manual.
user specified PDF to perform a large number of Practicing engineers suggested us to improve the
possible strength-loading scenarios. For the reason that software in a hierarchical way, similar to a wizard window
Monte-Carlo simulations require thousands of non- that is asking the user to progressively define its problem.
linear analysis, a program like CADAM is almost the This suggestion was implemented in the software by
only and the most effective tool to execute such ordering efficiently the menu items and shortcut buttons
computations. For example, performing 250,000 non- on the tool bars, thus leading to a more efficient definition of
linear analyses with three independent variables take the input data. Moreover, user’s feedback leaded to a better
about 65 s on a 1.7 GHz Pentium 4;
organization in the software and a more detailed illustration
(2) stress and stability analyses are performed;
in CADAM dialog windows of the uplift pressure
(3) statistics are performed on the results (e.g. sliding
distributions of the most important North American dam
safety factors, SSF) to determine the probability of
safety guidelines. Even so, all cracking related options were
failure, pf :
restructured and regrouped to ensure a precise definition of
the nonlinearities to be considered in the analysis. CADAM
nf nðSSF , 1Þ was enabled to provide detailed output results of all
pf ¼ ¼ ð3Þ
N N intermediate computational steps. The output reports were
414 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

also designed to allow users to trace clearly all input PGA). It is then possible to evaluate for which loading
features that were activated in a particular analysis. In intensity, safety factors will fall below allowable values
addition, direct connection to Microsoft Excel spreadsheet such that proper action could be planned. CADAM allows
was developed to allow users to further perform comp- proceeding with an incremental load analysis, which will be
lementary post-processing of CADAM results. illustrated here to evaluate the evolution of cracking and
Another feature of the program is the instantaneous safety factors in function of the flood upstream reservoir
verification in consistency of the input data. At any time in elevation for the 30 m dam shown in Fig. 4a. The effect of
the definition of the problem, the user’s input data reservoir overtopping on the applied crest vertical water
consistency is verified by CADAM. Upon inconsistency, pressure has been taken into account. Floating debris are not
CADAM generates a warning message explaining how the considered in this incremental flood analysis.
user’s inputs are in conflict. For example, if the user selects Fig. 12 presents the results from an analysis where the
a certain guideline to activate drainage and select another flood reservoir elevation was incrementally raised from
guideline to model the effects of cracking on drainage elevation 27.432 m (90 ft) to 33.528 m (110 ft) by
efficiency, CADAM warns the user without forcing him to increment of 0.001 m (0.003 ft). Fig. 12 shows the
change his choice. Finally, CADAM website (http://www. evolution of both peak and residual SSFs as well as
struc.polymtl.ca/cadam) offers an electronic form allowing cracking at the base as a function of the flood reservoir
users to comment on the software. elevation. Cracking initiates at a reservoir elevation of
30.166 m (98.97 ft). This crack reaches the drain line for a
reservoir elevation of 30.546 m (100.22 ft). As soon as the
10. Application examples crack reaches the drain, the drain effectiveness is
completely lost [3]. This results in an increase in the
10.1. System analysed uplift force for finally propagating the crack to 43.7% of
the base length. At the same time, SSFs are dramatically
The 30.48 m (100 ft) high gravity dam to illustrate some reduced. Shear failure (sliding) of the dam at the base
of CADAM potentials is shown in Fig. 4a. This dam was occurs for a reservoir elevation of 31.4 m (103 ft), when
used in USACE [12] to evaluate and compare stability considering the peak shear strength.
analysis and uplifting criteria for gravity dams by three US The structural analysis results along a particular joint
Federal agencies. The analyses are performed considering could be displayed providing a visual feedback of
the material properties shown in Fig. 4b and c. The usual u/s interrelations between the position of the resultant,
and d/s reservoir elevations are set to 27.432 m (90 ft) and cracking, uplift pressures, and normal stresses distri-
1.524 m (5 ft), respectively. Lift joints are spaced at every butions. In an incremental flood analysis, Fig. 13 shows
3.048 m (10 ft) in elevation from the base. The drainage the results corresponding to a full reservoir elevation of
system is initially considered according to USACE [3] 30.48 m. The resultant force in this case is outside the
guideline, the drain position, efficiency and the elevation of third median. Individual force components (normal,
the drainage gallery are given in Fig. 5. moments, and shear) are indicated to allow independent
validation of results.
10.2. Usual load combination
10.4. Probabilistic safety analysis
Figs. 10 and 11 show the global stability drawings,
generated by CADAM, for the usual load conditions. Fig. 10 In dam safety guidelines, it is customary to define safety
is related to the stress analysis indicating normal and factors in terms of allowable stresses (forces). The
principal stresses on the u/s and d/s faces as well as uplift calculations are performed using a deterministic model of
pressures. Fig. 11 is related to the stability analysis the dam assuming specific numerical values for the loads
presenting resultant forces and safety factors. and the strength parameters. A probabilistic safety analysis
considers explicitly the uncertainties in the loading and
10.3. Incremental flood safety analysis strength parameters that are treated as random variables.
The uncertainties in input parameters are then transformed
In dam safety evaluation there is most often high in probability of failure of a dam.
uncertainties with the loading intensity associated with As an illustrative example, Fig. 14 shows the results of
extreme flood and earthquake events with very long return probabilistic safety analyses where the peak shear strength
periods such as the reservoir elevation corresponding to the parameters (cohesion and friction angle) are considered as
10,000 years event (or Probable Maximum Flood (PMF)). It random variables following a normal PDF with means and
is very useful to know the evolution of typical SSFs (for variances indicated in Fig. 9. To construct the fragility
peak and residual strengths) as well as other performance curve, shown in Fig. 14, twenty probabilistic analyses were
indicators (e.g. crack length) as a function of a progressive performed by increasing the flood reservoir elevation. A
increase in the applied loading (i.e. reservoir elevation or probability of failure is obtained for each reservoir
M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 415

Fig. 10. Stability drawing: stress analysis.

416 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

Fig. 11. Stability drawing: stability analysis.

 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 417

Fig. 12. Incremental load analysis (Flood analysis).

elevation, thus defining a point of the curve. By definition, 10.5. Seismic safety analysis
the dam failure occurs when the peak SSF is less than one.
Annual exceedance frequency of particular reservoir Fig. 15 shows the results from a pseudo-static analysis
elevations are also plotted in Fig. 14 for illustrative (Fig. 15a), and a pseudo-dynamic analysis (Fig. 15b).
purposes, allowing to compute the frequency of dam failure Reservoir u/s and d/s elevations are set to the same normal
per year due to hydrological events [21]. operating levels as the usual combination. Uplift pressures

Fig. 13. Normal stress and uplift distribution at the base (reservoir at 30.48 m).
418 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

forces near the crest is around 3.8 times that of the pseudo-
static analysis. This could lead to cracking of joints near the
crest as it has been observed on several occasions in the field
[22].
CADAM provides three options to consider the transient
evolution of uplift pressures in cracks (Fig. 7) during
earthquakes. Fig. 17 shows the effect of seismic water
pressure assumptions in cracks. Based on the chosen uplift
assumption, the following results are obtained:

(a) insignificant cracking (Fig. 17a) occurs when there is

no uplift pressures in the opened crack,
(b) slightly longer crack (Fig. 17b) occurs when uplift
pressures remain unchanged, and
(c) longer crack with a small SSF (Fig. 17c) is obtained
when full uplift pressures are applied to the crack
Fig. 14. Probabilistic safety analysis.
section.

This example shows the versatility of CADAM to answer

typical engineering question in seismic safety analysis:
are assumed to remain unchanged during seismic analyses. What if?
The peak ground horizontal acceleration is set to 0.15 g. The
peak ground vertical acceleration is set 0.1 g. The peak
horizontal spectral acceleration is set to 0.2 g at the 11. Perspectives for future developments
fundamental period of the dam ðT1 ¼ 0:1 sÞ: Sustained
accelerations are taken as half of the peak acceleration There are almost endless possibilities for further
values. developments of a computer program like CADAM for
As shown in Fig. 15, both seismic analyses indicate structural safety assessment of gravity dams. Currently,
almost the same crack length along the base, while the work is progress to add the following features:
pseudo-dynamic analysis indicates more cracking in the
upper joints of the dam. This is due to the dynamic † From a pseudo-static or a pseudo-dynamic seismic
amplification of inertia forces along the height of the dam in analysis, the lift joint most susceptible to cracking can
the pseudo-dynamic analysis. Fig. 16 presents the amplifi- be easily obtained using CADAM. Calculation of
cation of the seismic loads (shear forces and moments) of seismic sliding displacements and rocking response of
the pseudo-dynamic analysis over the pseudo-static anal- cracked dam components using transient dynamic
ysis. For this dam, the dynamic amplification of seismic analysis of rigid body is envisaged.

Fig. 15. Seismic analyses (cracking and uplift pressures): (a) pseudo-static analysis; (b) pseudo-dynamic analysis.
 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420 419

Fig. 16. Amplification of pseudo-dynamic over pseudo-static analysis.

Fig. 17. Incidence of uplift pressures in cracks during pseudo-static analysis.

† Computation of displacements using beam theory for † Definition of more complex 2D geometry, spillway and
the dams and Boussinesq coefficients for the semi- water intake sections, eventually 3D sections.
infinite elastic foundation. † Arbitrary user defined uplift pressure distributions.
† Thermal analysis will be performed along lift joints † Link with finite element programs: automatic transfer
using finite differences to evaluate the thermal field of models data to finite element programs for detailed
required for thermal displacement and stress compu- static, thermal, seepage and seismic analyses.
tations. The displacement response of a 2D model
could be calibrated against that of a preliminary 3D
finite element model to determine the fraction of the 12. Conclusions
hydrostatic load that is resisted in a pure cantilever
mode. Unit thermal loads could also be used for CADAM provides a very versatile computing environ-
calibration purposes. The computation of displacement ment to learn or investigate modelling assumptions and
using beam theory will allow simple and effective computational processes related to the static and seismic
coupled thermo-mechanical analyses to link the structural stability of gravity dams based on the gravity
deterministic model of a dam with its statistical method. It has been shown in this paper that several
model derived from field measurements of pendulum assumptions related to load conditions, cracking criteria,
displacements. This can be viewed as an intermediate uplift pressures intensities and analysis procedure could be
step before undertaking detailed coupled thermo- used for static, seismic, and post-seismic safety assess-
mechanical finite element analyses, which requires ments. In general, the computations are complex to perform
large resources. due to the coupling between uplift pressures and crack
420 M. Leclerc et al. / Advances in Engineering Software 34 (2003) 403–420

length. In an actual situation, parametric analyses are most [6] USACE (US Army Corps of Engineers). Engineering and design:
often performed to cover uncertainties in strength and Gravity dam design. Report EM 1110-2-2000, Washington, DC; 1995
[7] ANCOLD. Guidelines on design criteria for concrete gravity dams.
loading parameters to take appropriate decision concerning
Australian National Committee for Large Dams; 1991
a particular structure. The authors have successfully used [8] USBR (United States Bureau of Reclamation). Design of small dams.
CADAM as a computational laboratory in seminars, to Denver, Colorado; 1987
engineers from practice, involved in dam safety evaluation. [9] Leclerc M, Léger P, Tinawi R. CADAM User’s Manual. Version 1.4.3
CADAM is also used for industrial applications and R&D in http://www.struc.polymtl.ca/cadam/ Department of Civil Engineer-
dam engineering and has been extensively validated, using ing, École Polytechnique de Montréal, Qué., Canada; 2001
[10] Léger P, Larivière R, Palavicini F, Tinawi R. Performance of gated
extensive lengthy manual calculations or linear and non-
spillways during the 1996 Saguenay flood (Québec, Canada) and
linear Finite Element Methods, during the past years. The evolution of related design criteria. Proceeding of ICOLD 20th
organisation of the program and the particular features that Congress Beijing, China, Q.79–R.26; 2000. p. 417–38
have been presented herein are useful for those interested in [11] Chopra AK. Earthquake response analysis of concrete dams. In:
the development and application of computer aided stability Jansen RB, editor. Advanced dam engineering for design, construc-
analysis of gravity dams. tion, and rehabilitation. New York: Van Nostrand Reinhold; 1998. p.
416 –65.
[12] Kreuzer H. The use of risk analysis to support dam Safety decisions
and Management. Proceedings ICOLD 20th Congress Beijing, China,
Acknowledgements Gr. Q. 76; 2000. p. 769–834
[13] Canadian Dam Safety Association (CDSA). Dam safety guidelines
The development of the computer program CADAM was and commentaries, Edmonton, Alberta, Canada; 1997, 1995
funded by NSERC (Natural Sciences and Engineering [14] Teixeira S, Pachero X. Delphi 6 Developer’s Guide. Sams Publishing;
2000. p. 1556.
Research Council of Canada), Hydro-Québec and Alcan.
[15] Borland Inprise Corporation, Delphie 6 professional. Part
The support of these organisations is gratefully #HDB1360WW1518S-12193, Scotts Valley, California, USA: Bor-
acknowledged. land Inprise Corporation; 2000.
[16] McCombie PF, Penman J. The production of interactive engineering
design software using Borland Delphi. Adv Engng Software 2001;32:
References 789 –96.
[17] USACE (US Army Corps of Engineers). Evaluation and comparison
of stability analysis and uplift criteria for concrete gravity dams by
[1] Ghrib F, Léger P, Tinawi R, Lupien R, Veilleux M. Seismic safety
three federal agencies. Engineering Research and Development
evaluation of gravity dams. Int J Hydropower Dams 1997;4(2):
Center-Information Technology Laboratory. Report ERDC/ITL TR-
126–38.
00-1, Washington, DC; 2000 (document from the web: http://www.
[2] Léger P, Tinawi R, Rheault S, Leclerc M. A progressive methodology
wes.army.mil/ITL/itlpubl.html)
for structural safety evaluation of gravity dams subjected to floods.
Proceedings of Canadian Dam Safety Conference, Niagara Falls, [18] Tinawi R, Léger P, Leclerc M, Cipolla G. Seismic safety of gravity
Ontario; 1996. p. 2–16 dams: from shake table experiments to numerical analyses. ASCE J
[3] FERC (Federal Energy Regulatory Commission). Engineering Struct Engng 2000;126(4):518–29.
guidelines for evaluation of hydropower projects—Draft Chapter III [19] International Commission on Large Dams (ICOLD). Earthquake
Gravity Dams. Federal Energy Regulatory Commission, Office of analysis for dams. Bulletin 52, Paris; 1986
Energy Projects, Division of Dam Safety and Inspections, Washing- [20] Melchers RE. Structural reliability analysis and prediction, 2nd ed.
ton DC, USA; 2000 New York: Wiley; 1999.
[4] FERC (Federal Energy Regulatory Commission). Engineering [21] McCann MW, Franzini JB, Kavazanjian E, Shah HC. Preliminary
guidelines for evaluation of hydropower projects—Chapter III safety evaluation of existing dams, vol. 1. Report No. 69, The John
Gravity Dams. Federal Energy Regulatory Commission, Office of A. Blume Earthquake Engineering Center, Department of Civil
Hydropower Licensing, Report No. FERC 0119-2, Washington DC, Engineering, Stanford University; 1985.
USA; 1991 [22] Hall JF. The dynamic and earthquake behaviour of concrete dams:
[5] Canadian Dam Association (CDA). Dam safety guidelines. Edmon- review of experimental behaviour and observational evidence. Soil
ton, Alberta; 1999 Dynamics Earthquake Engng 1988;7(2):58 –117.