....

...

3-D analysis of gravity dams

G. Lombardi, Lombardi Engineering Ltd, Switzerland
The traditional structural analysis of gravity dams, both concrete and fill, is based on an oversimplified twodimensional scheme, which is largely misleading as soon as the valley slopes start to become steep. It is
postulated that such dams should be analysed in a way that takes into account their real three-dimensional
behaviour if unpleasant mishaps are to be avoided, such as have occurred at a number of concrete faced
fill dams. A simplified way to study this problem is presented here, to get a first feeling of the various
aspects related to the real conditions of equilibrium of gravity structures, especially in narrow valleys.

Fig. 1. Traditional
assumption for the
analysis of a straight
gravity dam.
Independent 2-D
slices across the dam
body resting each
one on a
hypothetical
horizontal
foundation (W =
weight; H = water
pressure; U = uplift).

H-.

lw
uj

98

rom the physical point of view, any dam represents a three-dimensional body and should thus
be analysed taking into account this aspect of the
reality [Lombardi, 1993 1; 19942; 20043].
In fact, it can be noticed that, mainly for traditional
reasons, straight gravity structures, (conventional or
RCC gravity dams, but also fill dams with a core or a
concrete face), are generally analysed as two-dimensional structures. That means that only the stability of
thin 2-D cross-slices is investigated. The foundation
surface for each slice is implicitly assumed to be horizontal in the direction of the dam axis, that is crossvalley. In other words, possible problems in the third
dimension, that is, in the direction of the dam axis, or
of the slope of the valley flanks, are simply overlooked
or ignored (Fig. 1)
The historical reason for this way of thinking probably goes back to a period when the weak structural
element was the concrete dam itself. The main scope
of the analysis was thus to define the stresses in the
dam body. Today, however, the weakest element is
generally no longer the dam body itself, but its foundation. This change is the consequence of significant
improvements in construction techniques, but also
mainly the fact that the better dam sites have been
already developed, and those now available are much
less favourable as regards their geotechnical conditions.
The aim of the present paper is simply to show that,
when the traditional approach is followed, risks may
be assumed of which many designers are apparently
not really aware. The reasons for focusing attention on
the three-dimensional behaviour of the structure are
manifold: risks are steadily increasing from decade to
decade for various reasons. For example, the continuous technical and economical improvements of the
RCC technique create a temptation to adopt this type
of structure even where a fill dam could be the best
solution, or even the only correct one from a technical
and safety point of view.
Sometimes, even medium quality rock foundations
are considered so as not to allow for the construction
of an arch dam and thus to favour a gravity structure,
giving the wrong impression that any foundation problem can then be automatically solved. Sometimes such

a structure is selected, instead of an arch dam, which

could possibly be more appropriate for the geological
and geotechnical conditions.
A number of powerful computation codes for gravity structures, frequently used without sufficient caution, may lead designers to put 'blind trust' in their
results, with insufficient critical sensitivity or adequate cross-checking.
It is surprising to note that some quite fundamentally simple questions, such as the one to be discussed
next, often escape the attention of designers. This can
lead to dangerous situations [Bustamente and Tradisic,
20064 ; Lombardi, 20065].

1. Variable geotechnical conditions

According to Fig. 2, we may consider the simple case
of a concrete gravity dam of constant height, built on
ground of variable quality, that is, on rock zones of
different shear strengths and deformabilities. This is a
quite frequent occurrence. To compensate for possible
differential settlements, as well as to take into account
the contraction of the concrete by shrinkage and cooling, vertical joints are provided to form independent
blocks, each one resting by itself on a homogeneous
but different foundation.
The factor of safety (FoS) against sliding of each of
the three blocks, considered as independent, thus differs; the lowest value of FoS may be sufficient or not,
and the block itself may thus be sufficiently safe or
not.
Should the decision then be taken to close the joints,
to achieve an averaged overall factor of safety, thus to
compensate for the too low local ones (or for any other
reason), then the behaviour of the structure becomes
three-dimensional: for the usual load cases (when the
deformabilities are different); or just for the extreme
ones (when the strengths are different), or at least
when carrying out a limit equilibrium analysis.
To ignore these particular conditions may have costly and possibly even serious consequences.

FoS >

q><

FoS <

J,

FoS >

Fig. 2. Concrete gravity dam of constant height resting on a

ground of variable quality. (FoS = factor of safety; qJ = friction
angle), 1 to 3: block numbers: J/, h: vertical joints.
Hydropower & Dams

Issue One. 2007

~~'

'.I'
li"1

:i
i~
!~

:~,;

(a)

B~'::

(el

:~"

I W-cos 11
B
u

W'Slnl1

Fig. 3. CFRD dam. The imposed 2-D analysis for a 3-D

problem led to additional useless expenses.

A real, but anonymous, case is shown by Fig. 3. It

represents a quite large concrete faced rockfill dam
resting on an exceptionally well compacted morainic
material of the best quality. In a narrow irregular
canyon below that material, a layer of fine sand was
found, which could possibly liquefy in the event of a
strong earthquake.
For anybody with sound geothecnical and structural
judgement, the situation appears clearly to be a three
dimensional one, as it is evident that a thin vertical
slice cut in the massive dam cannot move out without
pulling some part of the nearby volume of material,
which of course would resist the movement and thus
increase the factor of safety of the thin two-dimensional slice considered.
In spite of these simple considerations, a two-dimensional equilibrium analysis was specified and imposed
on the owner. This led to the construction of an expensive but totally useless additional berm at the downstream dam heel.
This case is a very clear, but not unique, example of
a 'simplification' of the problem. Unfortunately, some
codes of analysis follow this way of thinking based on
the same traditional 2-D considerations, as imposed at
the project described.

2. The downhill stability

2.1 General aspects
The most important of the often overlooked factors in
the analysis of gravity structures appears to be the
reduced stability of the dam elements (or blocks) resting on the inclined valley flanks. It is therefore worthwhile to discuss this aspect and its influence on the
equilibrium and on the factor of safety of the entire
dam structure.

effect) and of the downhill component of the weight

(W sin a), giving thus an 'inclined' action on the foundation surface.
On the other hand, the shear resistance available on
the foundation surface depends on the force acting on
it in the normal direction, while obviously taking into
account the corresponding inclined uplift force.
Because of the inclination of the foundatIon, the
'uplift force' is increased in the proportion of l/cosa,
corresponding to the actual width of the foundation in
the downhill direction, with respect to the usual case.
On the other hand, the weight component acting normally on the foundation is reduced in inverse proportion, thus being (W cosa).
As can be seen in Fig. 5, the angle of friction theoretically required on the foundation for a given factor
of safety is greatly increased as a function of the crossvalley inclination. There is no doubt that in many real
cases, and for quite a number of existing dams, the
concrete blocks on the valley flanks would not be stable by themselves, or at least would not show a sufficient factor of safety. The actual factor of safety, of
course, stays well below the value obtained by the
usual computations based on the assumption of a
cross-valley horizontal foundation.
In many cases these blocks will need some support
from the nearby ones, in fact, from the adjacent ones, on
the way down to the central part of the valley, where
presumably a horizontal foundation for the central
blocks will exist, or where the compensation of the
forces from the opposite valley flank may be found.

2.0

Hydropower & Dams

Issue One, 2007

~(n)

n'" 1.5
~ (n)

25"

35"

30'

40

40

50

n"'1

~'" F(a.,9,n)

a",20: ,9"'0

-..... '1:::.-------~30; n=1.0

--

2.2 The single block on the slope

Let us first consider an independent concrete block
resting on a valley flank. For the sake of simplicity, we
will analyse only the case of sliding on the foundation
surface, while possibly deeper, differently oriented
discontinuities in the rock mass could obviously be
more critical and should also be considered. To avoid
additional useless complications at this stage, the
foundation surface is simply assumed to be horizontal
in the up/downstream direction.
According to Fig. 4, the foundation surface should
resist the combined force R resulting from the vectorial composition of the downstream oriented hydrostatic pressure H (plus possibly a hydrodynamic

1/cosa

.....

..... ~

"-es:>

,~

. . . , :.:~7,:',~=35; ,9=213

t:: ~ 't(

I'~I'

~60'

. . . ,ii('~-\,..-'~-'=.!"~

1.0

1
I
1

o
o

60"
'I'
"
'
)\ \
\
\\
I \ \
\
\
70"
1\\
\
\SO'

10 20

50~

30

40

50

60

60

90'

70 80 90 ~ =

a'" Slope of the foundation [oJ

70
80

90~-t

Friction
angLe
--'

Fig. 4. Equilibrium
of a single
"independent" damblock. (a) block on
an inclined
foundation; (b)
foundation surface;
and, (c) downhill
cross section with
main forces. a =
downhill inclination;
W = weight; U =
uplift; H =
horizontal up/downstream driving force;
R = resultant.
f3= multiplier ofthe
uplift (U) (f3 = 1 for
triangular uplift at
no drains);
n = requiredfactor
of safety; qJ ~
corresponding
friction angle.

11

Fig. 5. The
equilibrium of a
blockfounded on an
inclined surface.
Only weight, water
pressure and uplift
are taken here into
account.
Example: if the cross
valley slope is 35
and uplift equals 2/3
(of the triangular
diagram), a friction
angle qJ of60 would
be required (at no
cohesion) to obtain a
factor ofsafety of
15 (qJ = 49for a
factor of safety of 1).
99

Clearly the situation becomes more critical, the

steeper the valley flanks are.
It may be recalled that in former times it was quite
usual to form the foundation of gravity dams by stairlike steps, so to place each single concrete block on a
horizontal surface. The aim of this design was sometimes related to the old theoretical wrong vision dealt
with here, but mainly the aim was to improve the stability of the single blocks at the time of construction.
However, such a design was based on a misinterpretation of the actual geotechnical situation, because
inclined discontinuities paralleling the ground surface,
which generally do exist, could easily cut off the tiers,
and thus be a determinant for the stability. It should
also be recalled that the geotechnical conditions were
often better in the past than those found today. At present, the foundation line is thus, as a rule, shaped as an
inclined smooth one, while the blocks are usually cast
from the valley bottom up (especially for RCC dams).

the whole structure can be carried out. This must, in

any case, rely of a number of possibly arbitrary
assumptions about various elements of the global
structure. In addition, the shrinkage and the long-term
cooling of the concrete may change the stress and
stiffness conditions over the years, causing a modification to the stability conditions ofthe structure.
As the main aim of this paper is to define the overall
conditions of equilibrium and of the factor of safety to
be expected for the structure, rather than to compute
the exact value of some local stresses (which might be
analysed at a second stage), some simplified procedures can be considered. The later stress analysis can
thus be confined to the points of stress concentration,
which may appear, for example at the shear keys, as a
result of the large forces to be transferred through the
joints from one block to the next.

2.3 Restoring the continuity of the dam body

In the proposed simplified analysis, which represents

at least a first step, the way to follow will be the socalled 'load factor method'. The actions, like water
pressure, uplift, silt load, earthquake (for example,
considered as pseudo-stl!tic), will be multiplied by the
required factor of safety, while the resisting forces
(mainly the self ~eight, the friction and the cohesion
on the foundation) will be considered with their actual values.
Another way could obviously be to keep all the loads
at their actual value, but reduce the geotechnical
strength factors, dividing them by the required factor
of safety. The reason for doing this could be the intention to obtain a clearer definition of the points of
impact of the forces.
The method of partial factors of safety could also be
used. This is in fact a combination of the two methods
mentioned above. For the sake of simplicity, the 'load
factor' method is used here.
The absolute value of the resisting shear force along
the foundation surface can quite easily be defined on the
basis of the friction angle and the cohesion assumed on
surface. On the contrary, its direction is not known in
advance, and depends on a number of factors.

As a result of the conditions described above, it is

often unavoidable to restore the continuity of the
structure in closing the joints and implementing special devices (like shear-keys) to increase the factor of
safety by making possible a transfer of the required
forces from the valley flanks down to the central part
of the dam, which will supposedly resist them.
In this case, from the crest abutment down to the valley bottom, a number of factors may intervene which
could increase the stability of the blocks and of the
entire structure as, for example:
the presence of downstream rock or concrete masses, which can at least partly resist the sliding forces of
single blocks, and thus stabilize the entire wing of the
dam; or,
sections of the dam with a transversally flat foundation, which can interrupt the downhill flow of forces in
resisting them directly; or,
keys and specially shaped foundation surfaces with
increased shear strength.
Such situations must obviously be taken into account
in the design, but, for the sake of simplicity, are not be
considered in the following presentation.
The analysis of the transfer of the forces, required to
achieve a given factor of safety may be complex, as
shown by Fig. 6. The three components X, Y and Z
must obviously be taken into account at each step of
the computation.
The way the forces will be transferred to the next
block is influenced by the stiffness of the structures
implemented in the joints, as well as by the rigidity of
the foundation versus that of the structure itself.
Obviously, a complete 3-D finite element analysis of

3. Simplified definition of the factor

of safety

".,
a)

100

(b)

~"
H

o
n- 1

\R",

Block n

n+1

Fig. 6. Forces
normally to be
transferred by the
joints from block to
block.
(a) Downhill crosssection. Transmission
of compressive forces
from block to block;
(b) Plan view with
the forces transferred
from block to block
through the joints

b)

R".1~

H_

R\

Fig. 7. Foundation surface. Three typical cases for the

equilibrium ofa single block.
R,_! = action from the uphill block;
H = horizontal driving force;
D = downhill component of the weight;
F = resistance by friction and cohesion on the foundation; and,
R, = reaction by the downhill block.
Obviously the vertical components are considered accordingly
(a) first limit, (b) intermediate case, (c) second limit.
Hydropower & Dams

Issue One. 2007

As regards the direction of these shear forces on the

foundation, two limits as well as intermediate values
may be considered, as shown at Fig. 7.

lal

(dl

Thefirst case - Fig. 7(a)

It is assumed that, for each block, the downhill component will be directly supported by the next block by
a compressive force. The frictional resistance of the
foundations is thus entirely devoted to resisting the
downstream component, that is the driving force H. If
this resistance were insufficient, the difference is then
supposed to be transferred to the next downhill block
by a transversal shear force acting in the joint.
The second case - Fig. 7(b)
It is assumed that, for any block, the shear resistance of
the foundation is oriented in the opposite direction to
the resultant of the forces acting directly on it. Again,
if required, to restore the equilibrium, the lacking
forces will be transferred to the next block by the joint.
Other assumptions - Fig. 7(c)
These can be assumed, depending on the nature and
configuration of the joints, for example, if the joints
are 'open' in their normal direction, the shear resistance of the foundation will act, in the first place, to
support the downhill component of the forces. If necessary the shear and even the compressive resistance
of the joint must intervene to ensure the downstream
stability of the block with the required safety factor.

Obviously the computation is carried out, in all

cases, from the crest elevation down to the valley bottom, and each block is analysed taking into account
the forces coming down from the upper ones.
For the sake of the computation, additional 'virtual'
joints may be introduced where a change in the conditions is given, as for example, if the assumed values
for the friction angle or the cohesion, as well as the
inclination of the foundation from a section to the next
one will change.
It appears that, in general, the first case considered
above will be excessively favourable, while the second
one, which may look more logical, is somewhat detrimental to the equilibrium. It is felt that the most probable solution will be in between these limits. It should
therefore be worthwhile to fulfil the computations on
the basis of various assumptions, even in the case of a
finite element analysis, so as to obtain a better and
more complete overview of the situation.
Indeed, the choice between the various assumptions
about the direction of the shear on the foundation surface depends on a number of factors, but first and foremost on the nature of the joints, their opening, as well
their relative stiffness and strength both in the radial
and the tangential directions. In an actual dam, various
additional assumptions may apply for the various
blocks.
It goes without saying that, at the end of the day, the
central part of the dam, at the valley bottom, should be
able to resist, at least with the required factor of safety, the forces coming down from both valley flanks.

In case cohesion is taken into account, for example,

according to the Mohr-Colomb law, it should also be
considered that possibly even a small sliding movement which could be caused by a seismic event, could
destroy it. Therefore, pre-"as well as post-sliding conditions must be analysed.
The question of the geotechnical parameters is particularly important because in many cases the friction
angle to be considered may be only slightly greater
than the slope of the foundation surface.

5. An example
Following the concepts presented above, a real case
was computed and the result are given in Fig. 8. This
shows the main data of the left flank of an RCC dam
96 m high, with its actual joints, with the assumption
that the joints cannot transfer normal forces but only
shear forces.
It is assumed that the shear forces are transferred
from block to block at a short distance above the foundation line. The results of the computation are repreResults of the computation for the dam shown by Fig. 8,
for normal loading; case (c) of Fig. 7
Block

FoS

number

I.B.

1.65

2
3
4

5
6

4. Geotechnical parameters

The question of the geotechnical parameters to be

selected as a basis for the computations is obviously
of the greatest importance.
In particular, it should be considered whether a displacement by sliding takes place along the foundation
or along any somewhat deeper discontinuity in the
rock mass. In this case, the corresponding geotechnical values should possibly be used, for simplicity, at
the foundation surface itself.

Hydropower & Dams

Issue One, 2007

Fig. 8. Example of a
dam. (a) view from
upstream; (b) plan
view; (c) cross-section
lateral blocks 1 to 6;
(d) cross-section
central blocks 5 to 9
Load case: normal
water elevation 90 m,
sediments to 60 m.
In the plan view (b)
the shear forces in the
joints are shown. They
are required to obtain
a safety factor of 15,
in the case the joints
cannot resist normal
forces. These shear
forces reach 970 MN
(97000 tonne).

Joint
number

Shear forces
Y (in MN)

-44

-81

+ll

+264

+640

+970

+640

+310

10

FoS

C.B.
1.5

1.20

1.5

1.03

1.5
1.5

0.99

1.5

1.02
1.l0

1.5
1.5

1.85

1.5

1.85
1.85

1.5

FOS = Factor of Safety; LB. = Independent block; C.B. = 'connected'

block (by shear keys); Forces in MN (I MN = 100 metric tonne).
Note: For I.B. downhill FoS = downstream FoS; for C.B. downhill.
FoS= 1

101

sented in the Figure as well as by the Table below.

These results call for two main comments.
First, the overall factor of safety is only 1.5, while
the 2-D analysis showed a value of 1.85 for a horizontal foundation (Blocks 7 to 9).
Second, it should be mentioned that the forces to be
transferred through the joints are very significant, and
therefore adequate devices must be designed to resist
them. In addition, it should be noted that the factor of
safety depends greatly on the dimensions of the block
on the valley bottom. For example, if block 9 did not
exist, then the factor of safety would be about 1.4
instead of 1.5.

6. Summary and conclusions

According to a quite old tradition, gravity dams (including fill darns) are generally computed as two-dimensional structures, with a study only of cross sections
normal to the dam axis; usually even only the highest
one, at the centreline of the valley bottom is analysed.
It must be considered that in reality a gravity dam
exhibits three-dimensional behaviour as soon as the
height of the structure is variable along its axis or as
soon the geotechnical conditions are not uniform all
along the foundation line.
It is thus postulated that instead of continuing with
traditional habits, these dams should be analysed taking into account the real conditions, that is the actual
three-dimensional behaviour of the structure.
A simplified calculation method has been presented
here as a first step in doing this, but here, as a first
step, it is mainly intended to raise awareness.
The actual state of equilibrium, of the various possible assumptions shown, depends strongly on the
nature of the vertical joints and on their stiffness in the
normal and tangential directions. More attention
should be devoted to this problem than is usually
done, at least to avoid unpleasant situations.
The forces to be transferred from one block to the
next to ensure a sufficient factor of safety can be
extremely high and adequate provisions must be taken
if significant mishaps are to be avoided.
Similar conditions may be observed also for fill
dams and some recent incidents which have occurred
at concrete faced rockfill dams also appear to be related to 3-D conditions of equilibrium.
Finally, it can be said that in fact gravity dams, both
concrete and fill ones, were originally (at least implicitly) conceived for wide valleys with gently sloping
flanks, that is, for a slowly changing height.
The development of some construction techniques is
leading to the implementation of this type of dam also
on steep flanks, that is, in narrow valleys.
Often the consequences of this evolution are overlooked and this could have serious consequences, for
example in the event of heavy earthquakes.
The question dealt with here is quite important,
because the actual factor of safety is often significantly smaller than may be believed on the base of the
usual two-dimensional analysis.
This problem is not limited to the question of structural safety, but may also have an economic impact
and may thus lead to different conclusions in the
choice of darn type, especially if the fact is taken into
account that the available dam sites are becoming
ever more challenging from the geotechnical point of
view.
Finally, it should be recalled that sliding on the foundation surface is only one of many possible failure
modes. In particular, sliding on deeper discontinuities
102

in the foundation mass must be duly taken into

account, as well as other failure modes, such as toppling or overstressing of the shear keys in the joints. 0

References
1. Lombardi, G., "Concrete Dams and their Foundation Evaluation for static Loading", Vol. 4. Proceedings,
International Workshop on Dam Safety Evaluation,
Grindelwald, Switzerland; April 1993.
2. Lombardi, G. "Fondations de barrages en heton et leurs
traitements", International Symposium of ISRM, IV South
American Congress on Rock Mechanics, Santiago de Chile;
May 1994.
3. Lombardi, G., "La cimentaci6n de presas de fabrica",
Sinergia Congress 2004, Argentine Committee on Dams;
October 2004.
4. Bustamente and Tradisic, "Structural design and RCC
Zoning in Ralco dam" 22nd International Congress on Large
Dams, (ICOLD) Barcelona, Spain; June 2006.
5. Lombardi, G., "Algunos desarrollos en el an3lisis de presas
de fabrica", IV Congress on Dams and Hydropower,
Argentine Committee on Dams, Posadas (Misiones),
Argentina; August 2006

G.Lombardi

Dr Giovanni Lombardi is a consulting engineer

specializing in design, construction and research in the fields
of dams, underground works, and rock mechanics. He
obtained his first degree in Civil Engineering from the Swiss
Federal Institute of Technology in Zurich, Switzerland, and
was awarded his PhD from the same Institute in 1955 for his
thesis 'Les barrages en vofite mince" (Thin arch dams). He .
was then awarded the title Dr Eng h.c. at the Swiss Federal
Institute of Technology of Lausanne, Switzerland, and Eng
a.H. at the Politecnico of Milano, Italy.
He began his career as a Design Engineer in the offices of
Consulting Engineer H. Gicot, where he was involved in
projects in African and Europe. He then became Chief
Engineer in the offices of A. Kaech in Berne, Switzerland
(subsequently Kaech & Lombardi). He served as President
of the International Commission on Large Dams from 1985
to 1988. and is now Honorary President. Throughout his
career he has designed or given expert advice on dams in
many parts of the world, and he has authored more than 120
papers.
Since 1989 Dr Lombardi has been President of Lombardi
Engineering Ltd, based in Minusio-Locamo, Switzerland,
and also President of Lombardi Ita1ia in Milan. Italy.
He is a Member of the Swiss Academy of Engineering and
Sciences, and Honorary Member of the Swiss Society of
Engineers and Architects.
Lombardi Engineering Ltd, Via R. Simen 19, PO Box 1535,
CH-6648 Minusio-Locamo. Switzerland.

Hydropower & Dams

Issue One, 2007