Seismicb

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/346271435
Seismic Analyses of Pine Flat Dam: Simpliﬁed Use of Viscous Spring

Boundaries and Anisotropic Damage in the Principal Directions with the
Ability of Cracks Re-Closure
Chapter · January 2021

DOI: 10.1007/978-3-030-51085-5_12
CITATIONS READS
0 51
3 authors, including:
Emmanuel Robbe G. Devésa

Électricité de France (EDF) Électricité de France (EDF)
21 PUBLICATIONS 36 CITATIONS 20 PUBLICATIONS 185 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Arch dams behavior View project
DamQuake View project
All content following this page was uploaded by Emmanuel Robbe on 15 January 2021.
The user has requested enhancement of the downloaded file.

Seismic analyses of Pine Flat Dam
Simplified use of Viscous Spring boundaries and anisotropic damage
in the principal directions with the ability of cracks re-closure
E. Robbe1, E. Grimal1 and G.Devesa2
1
EDF Hydro, Le Bourget du lac, FRANCE
2
EDF R&D, Palaiseau, FRANCE
E-mail: [email protected]
ABSTRACT: This paper presents the analyses performed by EDF regarding the seismic analyses of
PineFlat dam proposed by the 2019 ICOLD Benchmark. The results will not be fully presented as they
will be summarized and compared with other participants by the formulator. The paper will focus on
the numerical analyses methods instead.
1 Introduction
The seismic analyses of Pine Flat dam proposed by the formulator provide an excellent
opportunity to qualify numerical method to evaluate the behavior of dams under earthquake.
Since several years, EDF improved its own Finite-element Software Code_Aster [1] to take
into account mass of the foundation, absorbing boundaries and fluid element in order to
correctly assess the safety of dams under earthquake. Comparisons with simplified analyses
(massless foundation and Westergaard added masses) showed that, advanced finite-element
analyses provided more accurate results compare to records on dams for example [2][3].
The subject proposed by the formulator includes non-linear analyses taking into account
concrete’s damage. It was a good opportunity to evaluate one of the damage model proposed
by Code_Aster and usually used in the case of swelling dams.
The following cases-studies have been performed by EDF:

 Case A: evaluation of the natural frequencies of Pine Flat dam taking into account
Fluid Element and the mass of the foundation and response of the dam-foundation-
reservoir system under an eccentric-mass vibration generator, as conducted in 1971.
The 4 subcases have been performed.
 Case B: foundation analysis using impulsive loads: allow to verify the efficiency of
the non-reflecting boundary conditions. Case B1 and B2 have been performed with the
700 m length foundation block.
 Case D: dynamic analyses of the dam-foundation-reservoir system under the Taft
earthquake records with 3 different water levels,
 Case E: dynamic analyses of the dam-foundation-reservoir system under the Taft
earthquake records and ETAF record combined with static load and taking into
account concrete non-linear material properties.
As the results will be presented and compared by the formulator, this paper will focus on
some lessons learned from the analyses.
2 Use of viscous – spring boundaries for dynamic analyses of

dam-foundation-reservoir system
In order to take into account the mass of the foundation, absorbing boundaries have to be used
in the foundation boundaries in order to avoid wave’s reflection. In Code_Aster, a viscous
spring boundary model is implemented as proposed and well described in [4] and [5], and
briefly summarized in Figure 1. It is employed to absorb the wave energy radiating away from
the dam and the foundation.
Figure 1 Viscous spring boundary model

In this method, earthquake input is introduced as compression and shear waves, vertically
propagating from the bottom to the top of the foundation. Side boundaries should not be
neglected using free-field column providing the propagation of the wave in an unbounded
foundation:
 If the foundation is homogenous, without damping, the acceleration of every point of
the free-field column can be theoretically known and the acceleration at the top of the
column will be twice the input at the bottom, with a delay due to propagation speed of
the wave from the bottom to the crest. This can be seen through the results of the case
B2 proposed by the formulator (Figure 2).
 If the foundation is non homogenous and/or with damping, acceleration of every point
of the free-field column has to be computed first. In addition, deconvolution should be
perform to evaluate the input at the bottom to get the wanted signal at the top.
Figure 2 case B2 results: propagation of a shear wave in the homogenous foundation without
damping
In addition the use of spring in the formulation of the viscous-spring boundaries is quite
useful to provide adequate boundaries conditions for static analysis when non-linear analysis
has to be performed (static and dynamic analysis can’t be run independently and results
solved)
3 Evaluation of natural frequencies (case A)

The case A propose to evaluate the first natural frequencies and the modal shapes for the dam-
reservoir system. This step is particularly useful from the engineering point of view because it
allow to compare the natural frequencies of the dam that can be measured by eccentric-mass
vibration generator (as done in Pine Flat in 1971) or nowadays by ambient vibration tests.
There are several ways to perform such analyses depending on the chosen numerical method:
 Finite-element analyses software are usually able to perform modal analysis. This
easy-to-use method requires to restrain displacements at the base (and eventually
side) of the foundation block as proposed by the benchmark formulator.
 Using viscous spring boundaries, modal analysis cannot be performed any more in the
conventional way. It is then proposed to perform harmonic analyses with either a
compression wave or a shear wave. It is then possible to visualize the imaginary part
of the structure displacements function of the frequency (Figure 3): this allows to
point-pick the natural frequencies of the dam.
Figure 3 evaluation of the natural frequency with viscous-spring boundaries and harmonic
analysis
top left: mode shape from the imaginary part under P wave at 2.3 Hz
bottom left: mode shape from the imaginary part under S wave at 2.3 Hz
right: imaginary part of the crest displacement (function of the frequency)
The Table 1 compare the natural frequencies evaluated by the 2 methods proposed
(conventional modal analysis and harmonic analysis with compression and shear waves) for
the case B1 of the benchmark. Some values are quite close but the harmonic analysis miss
some values found by the modal analysis.
Considering that in the engineering practice the comparison between natural frequencies
computed and measured is mainly based on the first natural frequency, the usual modal
analysis with restrain boundaries remain the easiest way.
Table 1 natural frequencies from modal analysis and harmonic analysis
Mode Modal Harmonic
analysis analysis
(Hz) (Hz)
1 2.35 2.3–2.4(*)
2 3.42
3 3.94 3.8
4 4.38
5 4.92 4.9
6 5.49
7 6.0
(*): value slightly different from the S-Wave or from the P-Wave
4 Simplified use of viscous-spring boundaries (Case B & D)

The evaluation of the dynamic response of Pine Flat dam under the Taft earthquake records
has been used to assess and simplify the introduction of the seismic input in FE analyses with
absorbing boundaries around the foundation.
For such analyses, there are several ways to introduce the seismic input as already described
in the §2:
 In a homogenous foundation without damping, the acceleration of the free field

columns can be solved analytically from the wave propagation equation,
 Considering that the formulators request the use of 2% damping in the foundation, the
introduction of the input should be in theory more complex :
o Analysis of a ‘free-field’ column only with 2% Rayleigh damping
o Deconvolution in order to get a new input at the bottom of the foundation to get
the ‘right’ input at the surface of the foundation
o Analysis of the 2D model considering for the free-field the results of the single
column under the new input.
The second method request to perform additional analyses, time-consuming from an

engineering point of view (and source of additional potential mistakes). A short sensibility
study is then proposed here to evaluate the effect of several simplification considering the low
value of damping in the foundation: 3 analyses are performed based on case D of the
benchmark with the following assumptions:
 No damping for the foundation material,
 2% Rayleigh damping in the foundation but no damping for the free field (directly
evaluated from the analytical solution)
 2% Rayleigh damping in the foundation including the free field (computed first as a
soil column with input at the base).
Figure 4 shows the displacement of the crest of the dam under the TAFT earthquake for the 3
analyses. Results are almost the same and show that for low value of damping in the
foundation, simplified seismic input procedure is suitable.
Figure 4 comparison of the displacement at the crest of the dam under TAFT earthquake
5 Non-linear dynamic analysis (Case E)

Case E of the benchmark proposes to evaluate the non-linear behavior of the dam under an
increasing earthquake input (ETAF) reaching more than 5 m/s2 at the end of the signal (Figure
5).
Figure 5 ETAF acceleration record

The damage model ‘ENDO_PORO_BETON’ is embedded in the ‘KIT_RGI’ group of laws
initially developed in CODE_ASTER for the swelling of concrete. This damage law can
nevertheless be used by itself for dynamic analyses of dams. More details can be found in
Erreur ! Source du renvoi introuvable.. This law included several mechanisms:
 Damage affecting tensile stresses:
o a pre-peak isotropic damage can occur if tensile peak strain is greater than the
elastic strain at the peak
o an orthotropic damage following Rankine criteria occurs during the post peak
phase corresponding to the localization of tensile cracks. Localized tensile
damages are computed in the principal directions of localized crack and lead to
opening of cracks. The model is able to re-close these localized cracks
depending to compression stress level.
 Isotropic damage under compression stresses using Drucker-Prager criteria.
With the concrete properties proposed by the formulator (tensile strength of 2 MPa, fracture
energy of 250 N/m, tensile strain at peak load of 0.00012), and the seismic methods describes
in the previous chapters (viscous-spring boundaries, fluid elements), the behavior of PineFlat
dam under the ETAF signal is computed (case E2 of the benchmark). Figure 6 show density
of cracks in the concrete due to tensile stresses for several instants of the signal (6, 9, 12 and
15s). This allow to visualize the progression of the damage during the earthquake.
Figure 7 compare the crest displacement computed with linear analysis (no damage) and non-
linear analysis (with damage mechanism) for the ETAF signal. Crest displacement computed
by the non-linear model remain in the same range than with the elastic model but starting
from about 12s, higher discrepancies start to appears. In addition, a decrease of the main
frequency is observed probably due to the important damage at the top of the dam.
Figure 6 maximum of the density of cracks du to tensile stresses in the main directions: blue =
no cracks; red fully cracked. t= 6s, 9s, 12s, 15s
Figure 7 crest displacement under ETAF signal: comparison between linear and non-linear
analyses
6 Conclusions
The use of viscous-spring boundaries for the analyses of concrete dams seems to be a good
compromise: it allows to correctly take into account the unbounded foundation and remain
quite simple to use from an engineer point of view. In addition, analyses showed than for low
value of damping in the foundation, simplified analytical free-field can be used : analyses
remain simple, there is no need of additional analyses to perform deconvolution.
The non-linear case of the benchmark demonstrate in addition that it can be easily combined
with damage law: springs embedded in the viscous-spring boundaries allow to easily start
with the static analysis, followed by dynamic analysis.
7 References
[1] Code_Aster – http://www.code-aster.org
[2] Robbe E. (2017). Seismic analyses of concrete dam, comparison between finite-element
analyses and seismic records, 16WCEE2017, Santiago, Chile.
[3] Robbe E. (2017). Seismic back analysis of Monticello arch dam – blind prediction
workshop and additional analyses, USSD annual conference, Anaheim
[4] Liu YS, Chen DH (2013), Earthquake response analysis of a gravity dam considering the
radiation damping of infinite foundation, APCOM & ISCM
[5] Zhang C, Pan J (2009), Influence of seismic input mechanisms and radiation damping on
arch dam response, Soil Dynamic and Earthquake engineering, 29 (9), 1282-1293
[6] Sellier, A. (2018). Anisotropic Damage and Visco-Elasto-Plasticity Applied to
Multiphasic Materials. Toulouse: LMDC - Laboratoire Matériaux et Durabilité des
Constructions de Toulouse ; Université de Toulouse III - Paul Sabatier ; INSA de
Toulouse. Retrieved from https://hal.insa-toulouse.fr/hal-01710289
View publication stats

Seismicb

Uploaded by

Copyright:

Available Formats

Seismicb

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Seismicb

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Seismic Analyses of Pine Flat Dam: Simpliﬁed Use of Viscous Spring

Chapter · January 2021

Emmanuel Robbe G. Devésa

SEE PROFILE SEE PROFILE

Arch dams behavior View project

DamQuake View project

The user has requested enhancement of the downloaded file.

The following cases-studies have been performed by EDF:

2 Use of viscous – spring boundaries for dynamic analyses of

Figure 1 Viscous spring boundary model

3 Evaluation of natural frequencies (case A)

4 Simplified use of viscous-spring boundaries (Case B & D)

 In a homogenous foundation without damping, the acceleration of the free field

The second method request to perform additional analyses, time-consuming from an

5 Non-linear dynamic analysis (Case E)

Figure 5 ETAF acceleration record

View publication stats

You might also like