Sie sind auf Seite 1von 15

Computers and Geotechnics 116 (2019) 103187

Contents lists available at ScienceDirect

Computers and Geotechnics


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

Research Paper

Rapid drawdown on earth dam stability after a strong earthquake T


a,⁎ b a
Stefania Sica , Luca Pagano , Federica Rotili
a
Università degli Studi del Sannio, Italy
b
Università degli Studi di Napoli “Federico II”, Italy

ARTICLE INFO ABSTRACT

Keywords: The paper focuses on the rapid drawdown of earth dams in the hypothesis that a prompt reservoir lowering is
Earth dam needed soon after a strong earthquake. The interest on such a topic comes from the fact that in Italy and
Earthquake worldwide there are many large dams placed very close to active faults. In this case, the dam may be asked to
Drawdown face the earthquake first and the rapid drawdown later to face an emergency state. The problem has been
Stability
numerically investigated with reference to a zoned earth dam (Campolattaro Dam) placed in a highly seismic
area of Southern Italy. To solve the overall boundary value problem, encompassing several stages of the dam
lifetime, a 2D finite difference model has been solved through the FLAC2D code, implementing a coupled
transient seepage formulation, which accounts for partial saturation of the soil during the drawdown stage.
Unlike past literature studies, in which the drawdown stage was modelled without accounting for the past
loading history of the dam, in the performed simulation the drawdown stage was placed at the end of a quite real
sequence of events, encompassing the embankment construction, the reservoir impounding, and earthquake
scenarios compatible with the dam site seismic hazard. The performed analyses pointed out that a seismic stage
previously experienced by the dam could further contribute to decreasing dam stability during the drawdown
especially during the first stages of reservoir lowering. In addition, the faster the drawdown, the smaller the dam
safety factor against stability (FOS) with more prominent effects of the initial (pre-drawdown) soil conditions.

1. Introduction variation in the reservoir due to the huge precipitation at the site.
As well known, a rapid lowering of the water level may temporarily
Lowering the water level of a reservoir may be critical for the sta- decrease the safety conditions of wetted slopes as the external stabi-
bility of an earth dam and of the natural slopes surrounding the re- lizing water pressure reduces in the meanwhile that the internal-to-
servoir, as testified by Sherard et al. [32], Morgenstern [19], Lawrence slope pore water pressures, unfavourable to stability, delay their de-
Von Thun [15], Pauls et al. [21], and Dai et al. [9] who listed several crease to the updated steady-state values.
cases of slope failures all over the world, caused by rapid drawdown. In Establishing the maximum drawdown rate that may safely be car-
most of the documented cases, at collapse the drawdown had not yet ried out is a crucial matter especially for those dams placed in earth-
reached the entire water level diminution but only half of it (approxi- quake-prone areas. A typical practice by most of the handling autho-
mately, dam midheight). An iconic case was that of San Luis Dam in rities is that of emptying the reservoir soon after the occurrence of a
California, which was one of the largest earthfill dams in the world strong earthquake to check whether the embankment suffered seismic-
(100 m high; 5500 m long) built in the last century. An upstream slide induced damages or, simply, under an unjustified believe to proceed at
developed in 1981 after several years of successful operation because a the safest. With the dam and the reservoir slopes sequentially subjected
very fast drawdown (around 0.3 m/day) was applied, with a change in to two severe events, the earthquake and the rapid drawdown, unstable
reservoir level of 55 m [1]. effects induced by both could act in synergy. In particular, the seismic-
More recently, Pinyol et al. [22] back-analysed the large landslide induced excess pore water pressures might unsafely sum up to those
occurred on the left bank of the Canelles reservoir in Spain. Their generated by the drawdown.
analysis indicated that the most likely reason for the instability was the From a theoretical point of view, predicting these combined effects
rapid drawdown that took place during the summer of 2006. requires the analysis of all transient stages experienced by the dam.
Yin et al. [31] analysed the landslides in Three Gorges reservoir Preliminary, the simulation of both construction and operation stages to
area, showing that the landslide was regulated by the water level initialize the field variables at the beginning of the seismic stage is


Corresponding author.
E-mail address: stefsica@unisannio.it (S. Sica).

https://doi.org/10.1016/j.compgeo.2019.103187
Received 13 December 2018; Received in revised form 29 May 2019; Accepted 29 July 2019
0266-352X/ © 2019 Elsevier Ltd. All rights reserved.
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

needed (e.g., [35,27,5]). Afterward, the sequential simulation of the reservoir operation (typically of the order of a few decimetres per days
seismic and post-seismic stages, with the latter incorporating the and not exceeding 1 m per day), in many cases the ratio R/k is domi-
drawdown operation, has to yield the pore water pressure and effective nated by the hydraulic conductivity of the soil. Fine-grained slopes tend
stress distributions required to characterize the evolution of dam safety. to generate excess pore water pressures typical of rapid drawdown
In former literature studies, the combined effect of earthquake and processes, while coarse-grained slopes tend to accommodate anytime
drawdown has never been considered. The main objectives of this paper the pore water pressure to the current steady-state distribution as it
are to shed light on such a topic by theoretically investigating whether happens in slow drawdown processes.
and how the effects induced by these two loading events may un- During the drawdown following a previous static stage of the dam,
favourably sum up, and characterizing the factors of major influence. the following physical variables at any time t should be considered:
These objectives are accomplished by analysing a real case history with
simulation of different scenarios for the seismic and post-seismic se- (1) the hydrostatic pressure Δp exerted by the reservoir water along the
quence. Unfortunately, the lack of monitoring data on earth dam per- slope surface, which will be reducing during the drawdown;
formance on such a chain of events prevented any kind of comparison (2) the hydraulic pore-water pressure along the slope surface Δuwb
between predicted and observed response. (=Δp), which will be reducing during the drawdown;
The aim of the work is twofold. On one hand, it could provide (3) the internal-to-slope stresses, Δσij, relaxing due to the static pres-
general criteria or procedures, which could be followed to quantify the sure removal, Δp;
lowering rate threshold below which a safe post-seismic drawdown is (4) the internal-to-slope pore-water pressure, which will be reducing in
assured for an earth dam. On the other, it could promote new items for any point of the slope due to both static, Δuw1 = f(Δσij), and hy-
future technical regulations or guidelines on the specific subject of dam draulic, Δuw2 = g(Δuwb), changes; as schematically represented in
maintenance and handling. Fig. 1, a rapid drawdown maximizes Δuw1 - which is equal to Δuw0
The paper initially outlines the foremost literature works on the occurring in undrained conditions - making Δuw2 nil; conversely, a
topic of rapid-drawdown effects on wetted slopes under only static slow drawdown produces an opposite effect making Δuw1 nil and
loading conditions due to the mentioned lack of works referring to maximizing Δuw2 at the steady-state level, Δuw2 = Δuw∞;
drawdowns carried out after strong earthquakes. Subsequently, it il-
lustrates the mathematical-numerical model adopted in the analysis of Re-equilibrium processes characterized by Δuw2, in any point of the
the selected case-study (Campolattaro Dam) to simulate the different embankment are related in part to degree of saturation changes, Δuw2Sr,
stages of the dam lifetime. Finally, the theoretical results and their and in part to porosity changes, Δuw2n:
critical implications are presented and discussed.
uw2 = uw2Sr + uw2n (1)

2. Literature overview on drawdown under static operation The removal of the stabilizing static pressure Δp, acting with in-
ternal pore water remaining far away of an amount Δuw-inst from
In a wetted slope, characterized by hydraulic conductivity k and equilibrating boundary changes Δuw∞, is what regulates the slope
subjected to steady-state conditions, as the water level reduces at a rate safety factor reduction. In short, Δuw-inst may be expressed as (Fig. 1):
R, the internal pore water pressures adjust with a time lag depending on uw = uw uw1 uw2Sr uw2n (2)
inst
the ratio, R/k. The above two parameters should be jointly considered
in order to establish if the slope reacts to the imposed drawdown in a Within this simplified conceptual framework, the effects of an
drained, partly drained or undrained manner [1]. earthquake stage preceding the drawdown may be comprised in Eq. (2)
The extreme cases of very high or very low R/k are respectively by adding the seismic-induced pore water pressure, Δuws, to the pore
associated with undrained and drained conditions in the slope and may water pressure changes needed to reach the hydraulic equilibrium
be referred to as “rapid” or “slow” drawdown processes. According to under static conditions:
Alonso and Pinyol [1], drawdown rates of 0.1 m/day are common,
uw inst = ( uw + uws) uw1 uw2Sr uw2n (3)
0.5 m/day are quite significant, 1 m/day and higher rates are rather
exceptional. This means that for usual lowering rates adopted in In past literature works on drawdown of submerged slopes, the

Initial reservoir level

L
Drawdown level
P H

Drawdown
drained partially drained undrained

uws
Uw(H)
uw1 uw1= uw0
Pressure (uw)
Pore water

uw∞ uw2Sr
uw2
uw-inst

uw2n Uw∞+ uw-inst

uw-inst Uw (H)- uw1


Uw∞(H-L)= Uw(H)- uw∞

R/k
Fig. 1. Δu vs R/k in a generic point P of a submerged slope

2
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

seismic induced effects (Δuws) have never been considered and the With reference to the case of the Glen Shira Dam, a homogeneous
drawdown was analysed only in a static regime. By tradition, two earth dam made of compacted morainic soils, Alonso and Pinyol [2]
simplified approaches have been adopted to predict the pore water focused on the results yielded by different continuum approaches,
pressure distribution after the drawdown: analysis of undrained con- whose differences lie in the complexity in predicting the pore water
ditions of a deformable medium (referred to as “undrained analysis” in pressure distribution in the embankment after an imposed drawdown.
the following) and/or analysis of transient flow stages through a rigid- Both elastic and elastoplastic constitutive soil models were adopted.
unsaturated domain (referred to as “pure flow method” or also as Analysis reliability was assessed by comparing predicted and measured
“uncoupled approach”). pore water pressures within the dam embankment. Since the reservoir
The pure flow analysis implicitly and invariably assumes that Δuw1 level change followed a pumping storage scheme, the Glen Shira Dam
nullifies, because of the rigid soil skeleton hypothesis. In the early was expected to experience fast drawdown rates. The following ap-
stages of a rapid drawdown process, Δuw-inst tends to be overestimated, proaches were investigated:
by an amount that should increase with increasing soil skeleton de-
formability and approaching undrained conditions. On the other side, • a pure flow analysis, in which the soils were considered rigid (Case
the same hypothesis implicitly assumes that the whole Δuw2n develops 1);
instantaneously, being hence independent from the drawdown rate • an instantaneous drawdown at the maximum rate, followed by pore
(this is a non-conservative assumption). According to Alonso & Pinyol water pressure dissipation, with the soils simulated as elastic ma-
[1], this may yield underestimation of pore water pressures after a terials (Case 2);
while from the beginning of the drawdown process. • a coupled unsaturated analysis, in which the soils were considered
This approach has been adopted, for instance, by Lane and Griffiths elastic (Case 3);
[14] who carried out an uncoupled 2D plane strain finite element • a coupled unsaturated analysis, in which the soils were considered
analysis of a partially submerged slope to identify critical values of the elastoplastic and modelled by the Barcelona basic model (BBM,
lowering ratio L/H in the case of rapid drawdown (Fig. 1). The factor of Alonso et al., [33]) (Case 4).
safety (FOS) of the slope was computed by the strength reduction
method, assuming the Mohr-Coulomb criterion for characterizing soil The Glem Shira Dam case study showed that the classical analysis
resistance. methods could not catch the available measurements. The method of
Simplified evaluation of Δuw0 (i.e., the pore water changes in the instantaneous drawdown (Case 2) or undrained approach resulted
undrained condition) may be carried out according to the procedure conservative, but very unrealistic. On the opposite extreme, the pure
suggested by Morgestern [19], who implemented the well-known flow analysis (Case 1) leaded to a systematic and unsafe under-
Skempton [28] formulation linking the pore water pressure changes to estimation of pore water pressures during the drawdown. Coupled
the total stress variation. analyses proved to capture well the measurements, independently from
The common case of drawdown that is intermediate between rapid the sophistication of the adopted constitutive soil law. In the analysed
and slow, thus generating partly drained conditions in the slope, im- case, plastic strains developed during the drawdown stage were found
plies that during the drawdown stages Δuw1 is below the undrained negligible compared to those induced by the construction stage of the
level, Δuw0, and that all terms of Eq. (1) develop over time (Fig. 1). The dam embankment and first filling, so that the simpler elastic model
analysis of this more general case requires the adoption of coupled provided a good approximation of the recorded pore water pressures
hydromechanical approaches of a three-phase (unsaturated) medium. during the drawdown stage.
Berilgen [4] and Alonso and Pinyol [2] provided reference examples
on the importance of exhaustive coupled analyses. The former in-
3. The Campolattaro Dam
vestigated the global stability of a homogeneous submerged slope
considering different slope geometries (3:1 slopes with two different
The Campolattaro Dam is a zoned earth dam placed in Benevento,
heights, H = 7 m and 14 m), hydraulic conductivities (k = 10−4 cm/s
which is one of the most seismic areas in Southern Italy. The 60 m high
and 10−6 cm/s), drawdown rates R (1 m/day and 0.1 m/day) and ratios
embankment was built between 1986 and 1992. The maximum water
L/H (values of 0.2, 0.4, 0.6, 0.8, 1.0). The Hardening soil model [25]
storage of the reservoir is of about 125 Mm3 for a plan dimension of
was adopted as stress-strain relationship. It was implemented however
256 km2. The main cross section of the dam is shown in Fig. 2. The
in a coupled saturated approach, so that the term Δuw2Sr was implicitly
geotechnical characterization of the foundation and embankment soils
assumed to develop at all instantaneously. The strength reduction
was based on the documents of the original design-construction stages
method was used to assess the slope stability safety factor. Undrained,
and on the results of two recent in situ investigations carried out by the
drained and partly drained drawdowns were simulated. It was found
dam owner (Province of Benevento). In Fig. 3 the grain size distribution
that for a very fast drawdown the steeper and less permeable slope
of the core (a), the filters (b) and the shells (c) are shown. The core is a
could achieve instability for L/H = 0.6 while the 7 m high slope was
medium-plasticity silt with clay (permeability of the order of 10−9 m/
always stable (FOS > 1) independently from drainage conditions.
s); the shells (k = 1.7 × 10−5 m/s) are mostly gravel while the drains

Fig. 2. Cross section of the Campolattaro Dam.

3
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Clay Silt Sand Gravel Table 1


100
Main parameters of the Campolattaro Dam soils.
90
Parameter γd [kN/m3] PI% n c' [kPa] φ' [°] k [m/s]
80
range of all tests carried Core 17.38 30 0.35 50 23 1.49 ∙ 10−9
out during the construction
70 Shells 21.09 0 0.25 0 43 1.68 ∙ 10−5
Drains 19.75 0 0.25 0 30 5.27 ∙ 10−6
60
Foundation 17.21 33 0.30 130 22 10−9
P [%]

50

40
Vs (m/s)
30 0 500 1000 1500
20 0
Vs
10 5
Vp
0
0.0001 0.001 0.01 0.1 1 10 100 10
d [mm]
Park & Kishida
a) 15 (2018)
Clay Silt Sand Gravel
100 20

Z (m)
90 25
80
range of all tests carried
out during the construction
30
70

60 35
P [%]

50 40
40
45
30 (a)
Vs (m/s)
20 0 1000 2000 3000
10 0
Vs
0 5
0.0001 0.001 0.01 0.1 1 10 100 Vp
d [mm] 10
b)
15

Clay Silt Sand Gravel


100 20
Z (m)

90 25
80
30
70

60 35
P [%]

50 range of all tests carried 40


out during the construction
40
45
30 (b)
20
Fig. 4. P and S seismic wave velocity vs depth measured by Down-Hole tests in
10 the Campolattaro Dam core (a) and foundation soils (b).
0
0.0001 0.001 0.01 0.1 1 10 100 crest level in both core and shells, Down-Hole (DH) tests in the core and
d [mm] foundation soils, laboratory tests on undisturbed samples of the core
c) and of the foundation clay (oedometric tests, drained and undrained
Fig. 3. Grain size distribution of the core (a), filters (b) and shells (c) of the
triaxial tests, resonant column and cyclic torsional shear tests RCTS).
Campolattaro Dam, such as verified during the construction stage (grey area) or Table 1 summarizes the main physical and mechanical parameters of
obtained from the core samples retrieved during the 2009–2016 investigation. the dam soils (dry unit weight γd, plasticity index PI, porosity n, co-
The grain size, d, and the percent finer by weight, P, are reported on the x and y hesion c’, friction angle φ', permeability k).
axis, respectively. Fig. 4 plots the profiles of both S and P-wave velocity measured in
the dam core and downstream foundation by DH tests. Along the ver-
(k = 5.3 × 10−7 m/s) are made of sand with gravel; the foundation soil tical axis of the core, below a first layer of stiffer soil corresponding to
consists of silty clayey sandstone dating back to the Miocenic Age. the coarse grained materials covering the core head (thickness of about
The more recent in-situ investigations (2009 and 2016) consist of 5 m), the Vs is around 280 m/s up to a depth of 35 m and then starts
boreholes combined to SPT carried out at different depths below the increasing between the depths of 35 m and 45 m below the crest level,

4
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

1
0.9
Experimental - core
0.8
0.7 G/G0 (Core - TC) Experimental - foundation
0.6 G/G0 (Foundation - TC)
Numerical - core
PI=15 % (Vucetic&Dobry, 1991)
G/G0

0.5
PI=30 % (Vucetic&Dobry, 1991)
0.4
PI=50 % (Vucetic&Dobry, 1991)
0.3
PI=100 % (Vucetic&Dobry, 1991)
0.2
0.1
0
0.00001 0.0001 0.001 0.01 0.1 1
γ [%] Fig. 7. Experimental pore water pressure from cyclic torsional shear test [11]
on the core and foundation samples consolidated at p′ = 270 kPa.
Fig. 5. G/G0-γ curves from cyclic torsional shear test [11] and literature data.

Consistently with the fact that the foundation clay (PI = 33%) is a
reaching the maximum value of 500 m/s at the end of the borehole. The
natural material, a better agreement with the reference trends by
Vs measurements obtained on the Campolattaro Dam core by means of
Vucetic and Dobry [30] was found.
DH tests carried out from the crest are consistent with the experimental
Initial damping (Fig. 6) is at 2.2% for both core and foundation
data collected by Park and Kishida [20] on similar dams built in Korea
soils. Damping then rises up to 16–17% at a shear strain level of
and Japan (Fig. 4a).
γ = 0.5%. Also for damping, the behaviour of the core material arises
Measurements of VP almost follow the same trend of Vs down to
inconsistent with the references curves, crossing these with increasing
35 m, where the abrupt increase in VP may be related to a higher degree
the shear strain, while the behaviour of the foundation clay shows
of saturation of the material [8] (when the DH test was carried out, the
better agreement with the trends by Vucetic and Dobry [30].
reservoir level was around 370 m a.s.l., which corresponds to the ele-
Finally, Fig. 7 shows the excess pore water pressures against the
vation of the first upper bank in Fig. 1).
shear strain, obtained from the cyclic torsional shear tests carried out
As the foundation soils regard (Fig. 4b), the shear wave velocity Vs
on both core and foundation samples. Excess pore water pressures start
varies between 315 m/s and 540 m/s in the upper 24 m below the
accumulating when the cyclic shear strain exceeds the threshold values
ground level (on the downstream side of the embankment). At a depth
of 0.1% in the core and 0.2% in the foundation.
of 24 m, the bedrock was intercepted (Vs = 1200 m/s). At a depth of
10 m, the VP measurements abruptly increase due to the interception of
4. Numerical procedure
the water table.
For the dam core and the foundation soils, Figs. 5 and 6 respectively
4.1. Simulation of the dam lifetime stages
show the G/G0-γ and D-γ experimental curves obtained from cyclic
RCTS laboratory tests. The experimental curves were compared to the
To simulate the different stages of the dam lifetime (embankment
well-known literature curves that Vucetic and Dobry [30] associated
construction, first impounding, operation, earthquake) and obtain the
with natural clays having different plasticity indexes (PI). According to
initial conditions needed to subsequently characterize the drawdown
the cited literature, the G/G0 curve of the dam core having PI = 30%
effects, the main cross-section of the Campolattaro Dam was numeri-
would have expected to be placed below the curve of a clay with
cally analysed. The model adopted in simulating the static pre-seismic,
PI = 50%. This is not the case of the Campolattaro dam core, which
seismic and static post-seismic (drawdown) stages was different in some
shows a more marked linear response than the clay with PI = 30%
features to better consider, on one side, the specific factors affecting
reported in Vucetic and Dobry [30]. The linear threshold of the core soil
each dam stage and, on the other, to overcome some intrinsic limits and
– detected as usual in correspondence of the ratio G/G0 = 0.95 - is
complexities of the adopted predictive tool. Since the performed dy-
around 0.018% and hence much higher than expected. According to
namic analyses require the full saturation hypothesis in the wet portion
Santucci et al. [23], compaction effects could actually induce important
of the spatial domain to be assured, all dam stages beforehand and
changes in material response with respect to the same soil tested in its
propaedeutic to the dynamic analysis (from the construction to the
natural state. Typically, the G/G0 curves are less influenced by the
operational stages) were solved in this hypothesis, while the three-
plasticity index than natural soils and characterized by a more marked
phase formulation was adopted only after the seismic stage to simulate
linear response. This statement is consistent with what recently pub-
the drawdown. The governing equations regulating the coupled un-
lished in Park and Koshida [20], who collected the results of shear
saturated state of the soil are reported in Appendix A.
modulus reduction and damping ratio curves on undisturbed samples of
Common features were the 2D-plane strain hypothesis and the
cores of different zoned earth dams.
adopted discretised geometry. In addition to the embankment, the
geometrical domain also includes a portion of the dam foundation of
20
D (Core -TC)
rectangular shape and size 1000 m × 170 m in the horizontal and ver-
17.5 tical direction, respectively. The discretized model consists of 3678
D (Foundation -TC)
15 elements and 3847 nodes. The maximum element height was selected
PI=15 % (Vucetic&Dobry, 1991)
to guarantee a proper propagation of the seismic signal in the model,
12.5 PI=30 % (Vucetic&Dobry, 1991)
i.e. the element size was selected not higher than one-eighth of the
D [%]

10 PI=50 % (Vucetic&Dobry, 1991) wavelength associated to the highest frequency of the input wave [17].
7.5 PI=100 % (Vucetic&Dobry, 1991) In particular, in the foundation soils the element height ranges between
5 3 and 12 m with the lower values in the upper alluvial layer; in the dam
2.5
embankment, the element height ranges between 1 and 3 m with the
smaller value towards the dam top.
0
0.00001 0.0001 0.001 0.01 0.1 1
γ [%] 4.2. Construction and impounding stages

Fig. 6. D-γ curves from cyclic torsional shear tests [11] and literature data. The simulation of the construction and impounding stages was

5
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Table 2 Table 3
Parameters characterizing the nonlinear elastic response of the Campolattaro Elastic parameters of the foundation soils.
dam soils.
z [m] Vs [m/s] G0 [MPa] v [–]
Soil A [kPa] B [kPa] C [kPa] ν
0–10 316 196 0.3
Core −187054 2693.1 0 0.3 10–24 540 580 0.3
Shells −396776 11,797 −17.57 0.3 24–170 1200 2528 0.3

carried out by assuming a coupled saturated (two-phase) approach for Steady state conditions were found to occur over about 4 months
the embankment soils and the foundation. (0.5 m/day) after the end of the construction.
An elastic-perfectly plastic stress-strain relationship was assigned to
the different embankment zones. In particular, a non-linear elastic be- 4.3. Seismic stages
haviour was assumed within a stress domain bounded by the Mohr-
Coulomb yield criterion. At yielding, a non-associated flow rule with The simulation of the seismic stage was carried out by assuming a
zero dilatancy angle was adopted. The elastic modulus G0 was assumed coupled saturated approach under the hypothesis of undrained condi-
to increase with the mean effective stress p’, according to the following tions with dynamic excess pore water pressures computed from the
equation: formulation of Byrne [7] as hereinafter specified.
2 During the dynamic stages of analysis, the cyclic soil behavior was
p' p' described using a hysteretic model consisting in an extension of the one-
G0 = A + B +C
pref pref (4) dimensional non-linear models that make use of the Masing [18] rules
to produce hysteresis loops. The model requires the small-strain shear
where pref is the reference pressure (1 kPa), A, B and C are constitutive modulus G0 and a modulus decay curve that describes the reduction of
parameters (Table 2). For the core (Fig. 8a) and foundation materials, the secant shear modulus with the shear strain amplitude. In the ana-
the elastic parameters were calibrated referring to the experimental lyses, the hysteretic model was used to update at each calculation step
results illustrated above in terms of Vs. For the shell material, the lack the shear modulus of the elastic- plastic soil model. The normalized
of experimental data was overcome by referring to literature data for secant modulus reduction (Ms = G/G0) was approximated using the
soils with similar grain size distribution and relative density (Fig. 8b). following equation in dependence of the shear strain γ:
In particular, reference was made to the nonlinear G0-p’ relationship
a
found for the Camastra dam (Pagano et al., 2008), characterized by the MS = yo +
same typology, geometry and construction procedures of the Campo- 1 + exp ( (log10
b
xo )
) (5)
lattaro dam shells.
The stress-strain relationship of the foundation soil was assumed where a, b, xo and yo are parameters regulating the curvature and po-
linearly elastic during the static stages of the dam lifetime. For the sition of the G/G0 – γ curve, to be calibrated on the experimental data
foundation soils, a stratified configuration with depth was assumed, as (Table 4). For the core, these parameters were calibrated on the results
deduced from the Vs profile previously shown (Fig. 4, see Table 3). shown in Fig. 4. For the filters and shells, the best-fitting with the lit-
The embankment construction was simulated by progressively ac- erature curves specific for coarser soils [22,24] was researched. All
tivating 11 horizontal layers. Each layer activation was conducted by: sigmoidal curves adopted in the computation are shown in Fig. 9.
(i) applying almost instantaneously the layer weight with nearly null In addition to the hysteretic damping, a small amount of Rayleigh
material stiffness to compute weight-induced stresses; (ii) im- damping (2%) was introduced in the overall model to simulate soil
plementing the stress-strain relationship with calibrated parameters damping at very small strains and to reduce numerical noise at high
and quantifying the G modulus based on the previously computed p’ frequencies.
values; (iii) consolidating the layer over a time consistent with the In a hierarchical way, an excess pore water pressure (p.w.p.) model
actual layer activation time, derived from the real embankment-time was added to the stress-strain laws assigned to the core and shell soils.
construction history. At the boundary of the embankment, a hydraulic The excess pore water pressure Δu induced in undrained conditions
condition of seepage surface was assumed. under seismic loading may be linked to the volumetric change in
The impounding stages were simulated by increasing the water drained conditions Δεvd by the following compatibility equation:
table up to the maximum service level and leaving it constant until the
vd
steady-state conditions were attained. A hydraulic condition of seepage u= 1 n
+ (6)
surface was assumed at the not-wetted boundary of the embankment. K Kw

p' [kPa] p' [kPa]


0 100 200 300 400 0 200 400 600
0.0E+00 0.0E+00
(a) (b)
1.0E+05
5.0E+05
2.0E+05

3.0E+05
G0 [kPa]
G0 [kPa]

1.0E+06
4.0E+05

5.0E+05 1.5E+06
6.0E+05 y = 2693.1x - 187054
R² = 0.9157 y = -17.571x2 + 11797x - 396776
7.0E+05 2.0E+06 R² = 0.9515

Fig. 8. G0-p′ relationships for the core (a) and the shells (b).

6
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Table 4 4.4. Drawdown stages


Parameters adopted to characterize dam soil behaviour under cyclic loading:
hysteretic and pore water pressure response. The simulation of the drawdown stages was carried out by assuming
Sigmoid curve Byrne a coupled unsaturated approach (3-phase formulation) for the em-
parameters p.w.p. bankment soils and a two-phase approach in the foundation. The soil
model skeleton stress-strain relationships adopted for the different dam zons
Parameter A B x0 y0 C1 C2 γv [%]
were the same adopted for the simulation of the static stages (con-
Core 0.85 −0.28 −0.96 0.15 0.13 2.96 0.02 struction and impounding). The 3-phase formulation is written in the
Shells 0.99 −0.6 −1.85 0.01 0.09 4.62 0.01 hypothesis that the liquid (w) and air (g) phase behave as two im-
Drains 0.95 −0.26 −1.55 0.05 – – – miscible fluids (see Appendix A) with constitutive laws:
Foundation 0.91 −0.36 −1.15 0.09 – – –
Sw u w Sw qiw
n + = Sw
Kw t t xi t (10)
1
0.9 Sg ug Sg qi g

0.8 drains n + = Sg
core Kg t t xi t (11)
0.7 shells
0.6 coupled to the momentum balance equation:
G/G0

0.5 foundation
ij ui
0.4
+ gi =
xj t (12)
0.3
0.2
and Darcy’s laws:
0.1
qiw = kijw krw (u w w gk xk )
0 xj (13)
0.0001 0.001 0.01 0.1 1
γ [%] µw g
qig = kijw kr (u g g gk xk )
Fig. 9. Numerical G/G0-γ curves adopted to characterize all dam materials. µg xj (14)
In the above equations, and qw
are the specific flow vectors of
i qgi
where n is soil porosity, K and Kw are the bulk modulus of the soil water and air, respectively; uw and ug are the pore water and gas
skeleton and water, respectively. The Byrne [7] formulation was pressure; kw
ij is the absolute permeability of the soil in its saturated state;
adopted to compute Δεvd according to: μw and μg are the dynamic viscosity of water and air, respectively; Kw, Kg
are fluid bulk moduli, ε is the volumetric strain, r w and r g are the
vd
= C1 exp C2 vd relative permeability for water and air. These latter two variables are
(7) related to the degree of saturation Sw by the van Genuchten [29] for-
mulation:
where
r
w
= Seb [1 (1 Se1/ a )a]2 (15)
– Δεvd is the volumetric strain increment per each cycle of shear strain;
r
g
= (1 Se )c [1 Se1/ a ]2a (16)
– εvd is the accumulated volumetric strain from previous cycles;
– γ is the amplitude of shear strain for the current cycle; where a, b and c are constant parameters and Se is the effective sa-
– C1 and C2 are internal parameters, which may be related to the turation defined as:
normalized standard penetration (SPT) blow counts (N1)60 by means Sw Srw
of the following relation: Se =
1 Srw (17)
C1 = 8.7[(N1 60 )] 1.25
(8) where is the residual wetting fluid saturation.
Srw
The soil-water retention curves adopted for the embankments soils
0.4 of the Campolattaro Dam are shown in Fig. 10 while the corresponding
C2 = parameters of the van Genuchten model, quantified on the basis of
C1 (9)
1
The comparison between the excess pore water pressures predicted
0.9
by the Byrne [7] p.w.p. model and the experimental data from the
torsional shear test on the undisturbed sample of the dam core is shown 0.8
in Fig. 7. 0.7
The seismic stage was simulated by retrieving the stress distribution 0.6
at the end of the overall static analyses as initial stress distribution for 0.5
Sw

the dynamic stage and then applying a time-history of shear stress [13] 0.4
at the base of the analysis domain (compliant base). The horizontal 0.3 core
input signals were selected according to the procedure described in 0.2 shells
Appendix B, for the two limit states of Life Safety (LLS) and Collapse drains
0.1
(CLS), required for the seismic safety assessment of the dam according
0
to the current Italian code (NTC 2018; NTD 2016).
0.01 0.1 1 10 100 1000 10000
Since the bottom boundary of the 2D analysis domain was placed
ug - uw [kPa]
175 m below the actual ground level, a deconvolution analysis was
carried out to transfer the selected natural accelerograms (recorded on Fig. 10. Retention curves assigned to the dam soils in the simulation of the
rock outcrop) at the base of the analysis domain. drawdown stage.

7
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Table 5 LLS CLS


Parameters adopted for the soil water characteristic curves of the embankment |amax|/|amax,b| a|max|/a|max,b|
soils (Appendix A). 0 1 2 3 0 1 2 3
1.0 1.0
Parameter a b c P0 [Pa] Sr

Core 0.145 0.5 0.5 6.45 · 104 0 0.9 0.9


Shells 0.904 0.5 0.5 1.24 · 104 0.306
Drains 0.507 0.5 0.5 5.83 · 103 0.436 0.8 0.8

0.7 0.7

literature indications [38,2], are provided in Table 5.


0.6 0.6
In addition to Eqs. (10)–(17), the overall mathematical formulation
describing soil response in its partial saturated state is completed by the

z/H
z/H
0.5 0.5
equations reported in Appendix A.
The reservoir lowering was simulated by parametrising the draw- 0.4 0.4
822xa
down rate, R, and the drawdown ratio, L/H. In particular, five draw-
0.3 822ya 0.3
down rates were selected between 0.5 m/day (slow drawdown) and 182xa
6 m/day (rapid drawdown). These drawdown rates were combined to 0.2
982xa
0.2
182ya
different values of the L/H ratio. In each analysis, the water lowering 4674xa 4674xa
was carried out into several steps, adjusting static and hydraulic 0.1 4674ya 0.1 6277ya
boundary conditions to the current water level. 6332xa 7142ya
0.0 0.0
Finally, slope stability analyses were performed for any analysis set
with computation of the dam safety factor (FOS). The shear strength Fig. 11. Profiles of |amax|/|amax,b|.
reduction technique proposed by Duncan [12] was adopted. In this
approach, the FOS represents the factor by which the shear strength of
core axis; the horizontal and vertical permanent displacements in the
the soil should be divided to bring the slope to the verge of failure. For
middle point of the dam crest (ux,r and uy,r). For further details on the
further details on the specific procedure implemented in FLAC2D, re-
adopted input signals, reference could be made to Appendix B.
ference can be made to Dawson et al. [10].
For each input motions, Fig. 11 plots the acceleration amplification
factor along the core axis, computed as the ratio between the peak
acceleration at the dam crest |amax| and base |amax,b|. As expected on
5. Results and discussion
the basis of former computations on other earth dams [26,34] and on-
site monitoring of accelerations developing throughout dams (e.g.,
5.1. Dam response to seismic actions
[35–27]), the seismic accelerations amplify moving upward throughout
the embankment, with amplification factors ranging between 1.25 and
To characterize the combined effects of the earthquake first and the
2.7 for the Life Limit State (LLS) and 1.5–2.25 for the Collapse Limit
rapid drawdown later, and individuate which factors mostly regulate
State (CLS). The lower amplification factors resulting from CLS analyses
dam stability under this complex but plausible scenario, several ana-
is due to the higher soil damping mobilized under the effects of stronger
lyses were carried out. In all runs, simulation of the dam construction,
earthquakes. From Table 6 it is worth noting is that the maximum
first filling, operation stages and earthquake provided the initial state in
permanent displacements predicted at the dam crest are provided by
terms of stresses, pore pressures and model internal variables for the
the signal 4674xa (South Island) not associated with significant am-
final drawdown analysis.
plification factors due to the diffuse soil plasticization in the upper part
Since propaedeutic to the interpretation of drawdown analyses,
of the embankment. Finally, Fig. 12 shows the pore water pressure
before presenting the final results on the drawdown issue some out-
contours inside the dam embankment before (Fig. 12a) and after
comes of the seismic simulation are hereinafter provided.
(Fig. 12b) the seismic analysis corresponding to the most severe input
Table 6 contains some of the key input and output parameters of the
motion (4674xa). At the end of the analysis, excess pore water pressure
dynamic analyses. Specifically are reported: the design horizontal ac-
Δu of the order of 100 kPa were computed at the base of the core
celeration, as provided by the current national zonation map (ag,NTC) on
(points A and D) while lower but still significant values at 40 kPa were
rock outcrop at the dam location; the input acceleration applied at the
found at higher elevations around the core mid-height (points C and F).
domain base after the decovolution procedure (amodelb); the maximum
acceleration at the embankment base (amax,b) and crest (amax,crest) at the

Table 6
Main input and output data of the dynamic analyses for the two ultimate limit states (LLS and CLS) considered.
Earthquake ag,NTC [g] amodelb [g] amax,b [g] amax,crest [g] ux,r [cm] uy,r [cm]

Life-Safety Limit State (LLS)


822xa 0.314 0.217 0.147 0.244 −0.81 −0.27
822ya 0.314 0.206 0.184 0.344 0.09 −0.39
982xa 0.314 0.133 0.133 0.239 −0.34 −0.19
4674xa 0.314 0.241 0.338 0.324 −24.73 −8.65
4674ya 0.314 0.225 0.216 0.375 −7.08 −1.01
6332xa 0.314 0.191 0.141 0.352 −0.12 −0.52

Collapse Limit State (CLS)


182xa 0.415 0.232 0.266 0.410 23.40 −0.83
182ya 0.415 0.249 0.199 0.298 26.56 −1.31
4674xa 0.415 0.321 0.512 0.760 −42.42 −15.1
6277ya 0.415 0.234 0.171 0.352 −9.41 −1.51
7142ya 0.415 0.307 0.242 0.320 −10.30 −1.28

8
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Fig. 12. Pore water pressure contour before (a) and after (b) the seismic stage simulating the 4674xa input signal for the LLS scenario.

5.2. Dam response to drawdown operations 5.2.1. Stress-path and pore water pressure evolutions
Dam response to the loading history of all simulated stages is pro-
Table 6 clearly indicates that the most severe input motion in terms vided in terms of total and effective-stress paths in Fig. 13 and of pore
of seismic-induced permanent deformations and pore water pressures is water pressures (Fig. 14) at two different control points selected at the
the signal (4674xa). Since pore water pressures levels at the beginning mid-height of the core axis (point #5) and of the upstream shell (point
of the drawdown process are crucial for the evolution of stability con- #3). In the shell, the progressive growth of the dead load during the
ditions (see §2), drawdown effects were investigated by assuming in the construction stages makes both deviatoric and mean stresses increasing
simulations the effects produced by this signal. After that, different (Fig. 13, point #3, black line) under null pore water pressure, thus
drawdown ratios (L/H) and rates (R) were numerically simulated. implying a perfect match between the total (TSP) and the effective
(ESP) stress paths. Converseley, in the core (Fig. 13, point #5, black
lines) the effective stress-path departs from the total one due to the

400 400 #5
#3

300 300
q [kPa]

200 200

100 100

0 0
0 200 400 600 0 200 400 600
p, p' [kPa] p, p' [kPa]

Fig. 13. Effective (ESP) and total (TSP) stress paths at point #3 (shell) and #5 (core) during the different stages of the dam lifetime.

9
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

400 400
350 #3 350 #5

300 300
250 250
q [kPa]

q [kPa]
200 200
150 Dam construction (PP) 150
Reservoir filling (PP)
100 100
Earthquake (LLS) (PP)
50 Drawdown (R = 0.5 m/day) (PP) 50
Drawdown (R = 4 m/day) (PP)
0 0
0 200 400 600 0 200 400 600
u [kPa] u [kPa]
Fig. 14. Pore water pressure (PP) at point #3 (shell) and #5 (core) during the different stages of the dam lifetime.

progressive development of excess pore water pressures. total and effective stress paths obviously tend to converge each other
During the impounding stages, the increase of the static load on the towards a nullification of pore water pressures. This effect is marked in
wet boundary of the dam determines a generalized increase in mean the shell under both slower (R = 0.5 m/s) and faster (R = 4 m/s)
total stress, contextually to deviatoric stress lowering (point #3 and #5, drawdown scenarios due to its high hydraulic conductivity, which de-
red lines), both changes associated with the minimum principal stress termines almost drained conditions during the whole drawdown pro-
increase (due to its direction close to that of the static load) and max- cess. This effect is less noticeable in the core, where it is even hardly
imum principal stress scarcely changing. In the shell, the progressive appreciable for the faster drawdown case. It is evident (Fig. 14) that the
increase in pore water pressure reduces p’, pushing back the effective residual pore water pressures acting in the core during the drawdown
stress state (ESP) along the path followed during the construction for are related to the pore water pressures acting at the beginning of the
almost one-third of its length. In the core, hydraulic re-equilibrium of process itself, which in turn are affected by the severity of the seismic
the excess-pore water pressures that grew during construction caused p’ shaking.
increments, so that differently from what happens in the shell, the ef- Fig. 15 shows the pore pressure profiles computed along a control
fective stress state departs considerably from the path followed during line crossing both the upstream shell and the core of the Campolattaro
construction. Dam, at the end of a drawdown following the simulation of the most
The seismic stages produce diffuse deviatoric plastic strains and a severe input motion (4674xa). The two plots refer to different draw-
generalized tendency for the dam body to spreading. This phenomenon down rates (0.5 and 4 m/day). In any point of the line placed above the
since inhibited at the dam foundation contact, results into an increase horizontal line corresponding to a fixed L/H value, pore water pressures
in mean total stresses across the dam. These increases accompany with at steady state should result less or at most equal to zero. For fixed L/H,
reductions (point #3, light blue lines) or increases (point #5, light blue the plot shows that this condition results satisfied for the shells. The
lines) of deviatoric stress in the shell and in the core, respectively. The core, instead, shows positive residual values, which increase with in-
seismic loads also cause an increase of excess pore water pressures creasing R.
(Fig. 14), so that the distance between the effective and total stress
paths increases. 5.2.2. Slope stability factor evolution
During the drawdown stages (Fig. 13, green and violet lines), the Dam stability was ascertained by computing the global safety factor

Fig. 15. Pore water pressure at the end of a drawdown along a control line crossing the upstream shell and the core. In (a) R = 0.5 m/day while in (b) R = 4 m/day.

10
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

(a) 2.4
0.5 m/day
2.2
1 m/day
2 2 m/day

1.8 4 m/day
FOS

1.6 0.5 m/day (LLS)

1.4 1 m/day (LLS)

2 m/day (LLS)
1.2
4 m/day (LLS)
1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
L/H
(b) 2.4
0.5 m/day
2.2
1 m/day
2 2 m/day

1.8 4 m/day
FOS

1.6 0.5 m/day (CLS)

1.4 1 m/day (CLS)

2 m/day (CLS)
1.2
4 m/day (CLS)
1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
L/H
Fig. 16. Dam stability (FOS) vs the drawdown ratio L/H for different drawdown rates R and two earthquake scenarios corresponding to Life Safety (a) and Collapse
(b) limit states.

(FOS) soon after the drawdown simulation. It is worth pointing out that changes in the meanwhile that small drawdown ratios (L/H ≪ 0.4) are
for the Campolattaro Dam the drawdown investigation had only a imposed to the slope, so that a very shallow critical slip surface of the
speculative aim since the global stability factor at the end of the seismic type shown in Fig. 17a takes place. The global instability phenomenon
analysis (i.e., at the beginning of the drawdown operation) was higher develops entirely within the upstream shell, where nearly drained
than 1. The final goal of the drawdown analysis for the specific case- conditions, only slightly affected by R, take place with almost null pore
history was to define the maximum drawdown rate that may be safely water pressures above the new water table. This stabilizing effect
applied to empty the reservoir in case of emergency after a strong slightly prevails on the non-stabilizing one represented by the static
earthquake. This is tantamount to assessing how much the safety factor pressure removal, Δp, on the upstream boundary. With increasing the
against global instability may further drop due to a post-seismic drawdown depth beyond L/H = 0.28, the amount of Δuw∞ delaying
drawdown. within the core (Δuw-inst) increases, thus determining, in addition to Δp
Fig. 16 plots the FOS against L/H for different drawdown rates and removal, core strength reduction. This relates to a critical slip surface
for the two considered seismic intensities (LLS and CLS) preceding the partly crossing the core (Fig. 17b). With furtherly increasing L/H, the
lowering of the reservoir. To better understand how and to what extent FOS progressively decreases because Δuw∞ is increasing. Conversely,
the seismic stage affects the FOS computed after an imposed drawdown, the drawdown rate R again poorly affects the results in terms of FOS as
the FOS was also recomputed under the hypothesis that no earthquake the attainment of Δuw∞ pertains only to the core which is only partly
had occurred (Fig. 16, dashed lines), to discriminate in this way the crossed by the slip surface.
seismic-induced changes. An earthquake preceding the drawdown yields unaltered behaviour
If a strong earthquake had occurred before starting the reservoir in terms of FOS for rates R up to 1 m/day. For L/H higher than 0.6, the
lowering (continuous lines in Fig. 16), for drawdown ratios L/H < 0.4 curves referred to R = 2 and 4 m/day, instead, depart substantially
the FOS coefficients are higher than the value (FOS = 2) computed from their static counterparts (dashed lines). This happens because the
before starting the drawdown (L/H = 0) operation. This happens for amounts of pore water pressures away from steady state conditions at
both LLS and CLS earthquake motions (Fig. 16 (a and b)) preceding the the end of the drawdown process (Δuw-inst) now increases dramatically
drawdown. With exception of a very fast drawdown characterized by combining with the seismic-induced effects, Δus. These changes are
R > 2 m/day, which is an almost unrealistic rate for an earth dam, an important since the FOS is now decreasing towards values that identify
increasing trend of FOS may be envisaged for 0 < L/H < 0.28, that is instability conditions. Worth of note is the minimum attained for L/H
stability is increased in the first stage of reservoir lowering. Conversely, around 0.7 rather than at the full drawdown level (L/H = 1), as it
the condition L/H = 0.4 marks the drawdown depth over which the would be expected. It is worth remarking that for these drawdown
drawdown effects result no longer beneficial in terms of FOS, i.e. the rates, instability is predicted only if an earthquake had occurred before
safety factor coefficients are lower than the value corresponding to L/ the drawdown stage.
H = 0 (no drawdown). The drawdown rate, R, provides slight FOS Fig. 18 rearranges the same results provided in Fig. 16 in the time

11
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Fig. 17. Critical slip surfaces corresponding to: (a) L/H = 0.2, (b) L/H = 0.4.

domain to show when the drawdown-induced instability arises in re- level has almost reached the dam midheigth (L/H > 0.5) and - as ex-
lation to the drawdown rate, R. The water lowering with R = 4 m/days pected - the slower the drawdown rate, the more stable the dam (higher
yields 10 days while the drawdown with R = 2 m /days yields almost FOS values). Conversely, a sudden change of the external water level,
the double. Finally, values of R lower than 1 m/day produce a FOS without allowing the time for slope drainage, generate unsafe condi-
reduction that does not induce any instability in the dam. tions also in a highly permeable soil, such as the upstream shell of a
zoned earth dam.
6. Concluding remarks For the selected case study, a threshold drawdown rate to carry out
a safe emptying of the reservoir (L/H = 1) after a strong earthquake
The paper illustrated the combined effects of two possible sequen- was identified in 1 m/day. Alternatively, a combination of different
tial stages an earth dam could experience in its lifetime, i.e. a strong drawdown ratios and rates might be selected. Initially, even a very
earthquake first and a rapid drawdown later. This latter may be re- rapid drawdown (R > 1 m/day) could be applied up to L/H = 0.3;
quired to face an emergency condition after a severe seismic event. The then to avoid slope instability, the reservoir drawdown should be
main factors, which regulate dam stability under this complex scenario, completed with a very slow rate (R ≤ 0.5 m/day).
were highlighted with simulating the real sequence of events, which It is worth mentioning that the above conclusions were figured out
encompasses dam construction, first reservoir filling, operation stages by a mathematical-numerical formulation based on a two-phase ap-
and earthquake, all together providing the initial state of the drawdown proach applied from dam construction up to the end of the seismic stage
analysis in terms of stresses, pore pressures and model internal vari- and a three-phase approach adopted only during the drawdown simu-
ables. lation. A further improvement of the work could be the implementation
For the investigated case study of Campolattaro Dam, the occur- of a coupled three-phase formulation from the very beginning of the
rence of an earthquake could reduce the safety of the upstream shell computation process.
during the intermediate stage of the drawdown (L/H around 50%) As a final remark, the present study provided only a theoretical
when an exceptionally fast rate (R > 1 m/day) is adopted. investigation on the rapid drawdown imposed to an earth dam soon
The drawdown rate represents a key factor when the drawdown after a strong earthquake. Due to the lack of monitoring data on the

2.4

2.2

1.8
FOS

1.6

1.4

1.2

1
1 10 100
days

0.5 m/day 1 m/day 2 m/day 4 m/day


0.5 m/day (LLS) 1 m/day (LLS) 2 m/day (LLS) 4 m/day (LLS)
0.5 m/day (CLS) 1 m/day (CLS) 2 m/day (CLS) 4 m/day (CLS)

Fig. 18. FOS vs time.

12
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

specific chain of events (drawdown preceded by a seismic event), the existing dams, with collection of suitable monitoring data, could cor-
mutual validation between experimental and predicted results was not roborate the obtained theoretical findings.
possible. To overcome this limitation, future experimental activities on

Appendix A

A.1. Mathematical formulation of the 3-phase problem

The three-phase formulation implemented in the FLAC code embodies the following laws.

A.1.1. Transport laws


Wetting (w) and non-wetting (g) fluid transport is described by Darcy’s law:

qiw = kijw krw (u w w gk xk )


xj (A.1)
µw g
qig = kijw kr (u g g gk xk )
µg xj (A.2)
where qi is the specific flow vector, uw and ug are the pore water and air pressure, respectively; ρw is water density, gk is k-th component of the gravity
vector g, kij is saturated mobility coefficient, κr is the relative permeability for the fluid, which is a function of saturation Sw, μ is the dynamic
viscosity. The mobility coefficient is defined as the ratio of intrinsic permeability to dynamic viscosity.

A.1.2. Relative permeability laws


Relative permeabilities are related to saturation Sw by empirical laws of the van Genuchten form [29]:

r
w
= Seb [1 (1 Se1/ a )a]2 (A.3)

r
g
= (1 Se )c [1 Se1/ a ]2a (A.4)
In those laws, a, b and c are constant parameters and Se is the effective saturation.
The effective saturation is defined as
Sw Srw
Se =
1 Srw (A.5)
where Srw is residual wetting fluid saturation.

A.1.3. Capillary pressure law


The capillary pressure law relates the difference in fluid pore pressures to saturation:
ug u w = Pc (Sw ) (A.6)
The empirical law is the following van Genuchten equation:
Pc (Sw ) = P0 [Se 1/ a
1]1 a
(A.7)
The parameter P0 was derived by dimensional analysis and has the form:

P0 =
/n (A.8)
where σ is surface tension, a property of the matrix, κ is intrinsic permeability, and n is porosity. The “alpha coefficient” is sometimes introduced in
the literature in place of P0. The relation between those two scaling parameters is
wg
P0 =
(A.9)
where ρw is wetting fluid density and g is gravity.

A.1.4. Saturation
The two fluids completely fill the pore space so that:
Sw + Sg = 1 (A.10)

A.1.5. Fluid balance laws


For slightly compressible fluids, the balance relations are:

w qiw
= + qvw
t xi (A.11)
g
g qi
= + qvg
t xi (A.12)

13
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

where ζ is the variation of fluid content (variation of fluid volume per unit volume of porous material), and qv is the volumetric fluid source intensity.

A.1.6. Fluid constitutive laws


The constitutive laws for the fluids are:

uw K Sw
Sw = w w
n Sw
t n t t t (A.13)

ug Kg g Sg
Sg = n Sg
t n t t t (A.14)
where Kw, Kg are fluid bulk moduli, and ε is the volumetric strain.
Finally, after substitution of Eq. (A.11) in Eq. (A.13), Eq. (A.12) in Eq. (A.14) and some rearrangement of terms, we obtain:

Sw u w Sw qiw
n + = Sw
Kw t t xi t (A.15)
g
Sg ug Sg qi
n + = Sg
Kg t t xi t (A.16)
For a coupled fluid-mechanical calculation, there following equations should be added.

A.1.7. Balance of Momentum


The balance equation is:
ij ui
+ gi =
xj t (A.17)
where ρ is bulk density, and u is velocity. For two-phase flow calculations, we have
= d + n (S w w + Sg g ) (A.18)
where ρw, ρg are fluid densities, and ρd is the matrix dry density.

A.1.8. Mechanical constitutive laws


The incremental constitutive response for the porous solid has the form:

ij = H ( ij, ij, ) (A.19)


where ij is the change in effective stress, H is the functional form of the constitutive law, and κ is a history parameter. For two-phase flow, the
change in effective stress is defined as:
'
ij = ij + ¯u ij (A.20)
where
¯u = Sw u w + Sg ug (A.21)
As may be seen from the above formulation, provided the total stress remains constant, deformation will only occur if a change of pore pressure
takes place. For constitutive models involving plasticity, Bishop effective stress is used to detect failure. This effective stress is defined as.
ij + ū ij (A.22)
with
u¯ = Sw u w + Sg ug (A.23)

A.1.9. Compatibility equation


The relation between strain rate and velocity gradient is given as usual and here reported for completeness:

1 ui uj
ij = +
2 xj xi (A.24)

Appendix B

B.1. Definition of the input signals for the seismic analysis of Campolattaro Dam

For the assessment of the seismic response of the Campolattaro Dam, the input signals were selected in accordance with the Italian Technical code
(NTC2018). After the definition of the return period (TR) for each limit state and the estimation of peak ground accelerations (PGA) through a
disaggregation of the seismic hazard for the reference site, natural accelerograms were chosen. The selection was carried out checking the similarity
between the elastic response spectrum of the selected accelerogram and the target response spectrum provided by the code. The check was per-
formed in terms of compatibility of the peak horizontal acceleration, quantified through an acceleration scale factor FS, and in terms of deviation

14
S. Sica, et al. Computers and Geotechnics 116 (2019) 103187

Table A1
Result of the accelerogram selection for the LLS and the CLS.
ID Earthquake Date M Epicentral distance [km] PGA [g] FS Scaled PGA [g] Drms

Life-Safety Limit State (LLS): VN = 50 years; TR = 710 years; PGA = 0.314 g; 5.0 ≤ M ≤ 7.0 ; 0 km ≤ R ≤ 10 km
822xa Umbria Marche (aftershock) 03/10/1997 5.3 5 0.156 2.01 0.314 0.058
982xa Friuli (aftershock) 16/09/1977 5.4 9 0.191 1.65 0.314 0.067
4674xa South Iceland 17/06/2000 6.5 5 0.318 0.98 0.314 0.069
6332xa South Iceland (aftershock) 21/06/2000 6.4 6 0.529 0.59 0.314 0.071
822ya Umbria Marche (aftershock) 03/10/1997 5.3 5 0.187 1.67 0.314 0.077
4674ya South Iceland 17/06/2000 6.5 5 0.338 0.93 0.314 0.129

Collapse Limit State (CLS): VN = 50 years; TR = 1460 years; PGA = 0.415 g; 5.5 ≤ M ≤ 7.5 ; 0 km ≤ R ≤ 20 km
7142ya Bingol 01/05/2003 6.3 14 0.297 1.40 0.415 0.037
6277xa South Iceland 17/06/2000 6.5 15 0.518 0.80 0.415 0.062
4674xa South Iceland 17/06/2000 6.5 5 0.318 1.31 0.415 0.067
182ya Tabas 16/09/1978 7.3 12 0.385 1.08 0.415 0.085
182xa Tabas 16/09/1978 7.3 12 0.338 1.23 0.415 0.127

between the selected and the target spectrum, measured by the average root-mean square deviation Drms [6]. In the selection of the input accel-
erograms, the reference period of the signals was centred on the fundamental period of the dam (Tdam = 0.233 s) evaluated by means of the transfer
function TF between the Fourier amplitude of the acceleration signals computed at the dam crest and base. The TF was obtained numerically by
exciting the overall model by a very low input motion (PGA = 0.03 g) so that only low-strain stiffness of the dam soils was activated.
The seismic analyses were carried out considering the Life Safety Limit State (LLS) and the Collapse Limit State (CLS), thus the two sets of
accelerograms with FS ≤ 2 and Drms ≤ 0.16 were selected (Table A1).
The natural input signals correspond to rock outcrop accelerograms. The input signals at the bottom of the discretized 2D model were computed
through a “deconvolution” analysis using the 1D wave propagation code, EERA [3]. The deconvoluted signals were applied at the base of the 2D
model as time-history of normal and shear stresses, linked to the velocity time-history of the input motion via the P and S wave velocity of the soil. In
this way a quiet boundary was applied at the bottom of the f.d.m. model to avoid wave reflection inside the analysis domain.

References 1969;95(EM4):859–77.
[18] Masing G. Eignespannungen und verfestigung beim messing. Second international
congress on applied mechanics, Zurich, Switzerland; 1926. p. 332–35.
[1] Alonso E, Pinyol N. Slope stability under rapid drawdown conditions. Barcelona: [19] Morgenstern N. Stability charts for earth slopes during rapid drawdown.
Universitat Politècnica de Catalunya; 2009. Géotechnique 1963;13(1):121–31.
[2] Alonso E, Pinyol N. Numerical analysis of rapid drawdown: applications in real [20] Park DS, Kishida T. Shear modulus reduction and damping ratio curves for earth
cases. Water Sci Eng 2016. https://doi.org/10.1016/j.wse.2016.11.003. core materials of dams. Can Geotech J 2018;2018. https://doi.org/10.1139/cgj-
[3] Bardet JP, Ichii K, Lin CH. EERA: a computer program for equivalent-linear earth- 2017-0529 (on line since March.
quake site response analysis of layred soil deposits. Department of Civil [21] Pauls GJ, Sauer EK, Christiansen EA, Wigder RA. A transient analysis of slope sta-
Engineering, University of Southern California; 2000. bility following drawdown after flooding of highly plastic clay. Can Geotech J
[4] Berilgen MM. Investigation of stability of slopes under drawdown conditions. 1999;36(6):1151–71. https://doi.org/10.1139/t99-073.
Comput Geotech 2007;34:81–91. [22] Pinyol NM, Alonso EE, Corominas J, Moya J. Canelles landslide: Modelling rapid
[5] Bilotta E, Sica S, Pagano L. Effect of ground-motion asynchronism on the equivalent drawdown and potential sliding. Landslides 2012;9(1):33e51. https://doi.org/10.
acceleration of earth dams. Soil Dyn Earthquake Eng 2010;30(7):561–79. 1007/s10346-011-0264-x.
[6] Bommer J, Acevedo A. The use of real earthquake accelerograms as input to dy- [23] Santucci de Magistris F, Silvestri F, Vinale F. The influence of compaction on the
namic analysis. J Earthquake Eng 2004;8(Special Issue 1):43–91. mechanical behaviour of a silty sand. Soils Found 1998;38(4):41–56.
[7] Byrne PM. A cyclic shear-volume coupling and pore pressure model for sand. [24] Seed HB, Wong RT, Idriss IM, Tokimatsu K. Moduli and damping factors for dy-
International conferences on recent advances in geotechnical earthquake en- namic analyses of cohesionless soils. J Geotech Geoenviron Eng ASCE
gineering and soil dynamics. Paper 1. 1991. 1986;112(11):1016–32.
[8] Cosentini RM, Foti S. Evaluation of porosity and degree of saturation from seismic [25] Schanz T, Vermeer PA, Bonnier PG. The hardening soil model: formulation and
and electrical data. Geotechnique 2014;64(4):278–86. verification. Beyond 2000 Comput Geotech 1999:281–96.
[9] Dai FC, Deng JH, Tham LG, Law KT, Lee CF. A large landslide in Zigui County, Three [26] Sica S, Pagano L, Modaressi A. Influence of past loading history on the seismic
Gorges area. Can Geotech J 2004;41(6):1233–40. https://doi.org/10.1139/t04-049. response of earth dams. Comput Geotech 2008;35(1):61–85.
[10] Dawson EM, Roth WH, Drescher A. Slope stability analysis by strength reduction. [27] Sica S, Pagano L, Vinale F. Interpretazione dei segnali sismici registrati sulla diga di
Géotechnique 1999;49(6):835–40. Camastra. Rivista Italiana di Geotecnica 2008;4:97–111. [in Italian].
[11] d’Onofrio A., 2016. Prove geotecniche di laboratorio - Rivalutazione della sicurezza [28] Skempton AW. The pore pressure coefficients A and B. Géotechnique
sismica dello sbarramento, opere accessorie e sponde del serbatoio della Diga di 1954;4(4):143–7.
Campolattaro sul fiume Tammaro – Provincia di Benevento [in Italian]. [29] van Genuchten MTh. A closed-form equation for predicting the hydraulic con-
[12] Duncan JM. State of the art: limit equilibrium and finite-element analysis of slopes. ductivity of unsaturated soils. J Soil Sci 1980;44:892–8.
J Geotech Eng Div Am Soc Civ Eng 1996;122(7):577–96. [30] Vucetic M, Dobry R. The effect of soil plasticity on cyclic response. ASCE Geotech J
[13] Itasca Consulting Group Inc. FLAC - fast lagrangian analysis of continua. Version 5. 1991;117(1):89–107.
0, User’s manual; 2005. [31] Yin Y, Huang B, Wang W. Reservoir-induced landslides and risk control in Three
[14] Lane PA, Griffiths DV. Assessment of stability of slopes under drawdown conditions. Gorges project on Yangtze River, China. J Rock Mech Geotech Eng
J Geotech Geoenv Eng 2000;126(5):443–50. 2016;8(5):577–95.
[15] Lawrence Von Thun J. San Luis dam upstream slide. Proceedings of the 11th in- [32] Sherard JL, Woodward RJ, Gizienski SF, Clevenger WA. Earth and Earth-rock Dams.
ternational conference on soil mechanics and foundation engineering. San New York: John Wiley and Sons; 1963.
Francisco: CRC Press; 1985. p. 2593–8. [33] Alonso EE, Gens A, Josa A. A constitutive model for partially saturated soils.
[16] Lirer S. Caratterizzazione meccanica di materiali a grana grossa in condizioni di Geotechnique 1990;40(3):405e430. https://doi.org/10.1680/geot.1990.40.3.405.
carico ciclico. Rapporto di ricerca, Dipartimento di Ingegneria Idraulica, Geotecnica [34] Sica S, Pagano L. Performance-based analysis of earth dams: procedures and ap-
ed Ambientale, Università di Napoli “Federico II”; 2008. plication to a sample case. Soils Found 2009;49(6):921–39.
[17] Lysmer J, Kuhlemeyer L. Finite dynamic model for infinite media. J Eng Mech Div

15

Das könnte Ihnen auch gefallen