proceedings of the 2016 sinaps@ plenary session€¦ · plenary session of sinaps@, held at...
TRANSCRIPT
0
20 16 Proceedings of the 2016 SINAPS@
Plenary session
Plenary Session 2016
November, 21 & 22 - EDF LAB PARIS SACLAY
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 1
SINAPS@ Project – "Earthquake & Nuclear Plant:
Ensure and Sustain Safety"
Explore uncertainties, Quantify margins to ensure and maintain the safety of nuclear facilities against earthquake
C. BERGE-THIERRY, SINAPS@ Coordinator, WP6 leader
M. NICOLAS, WP1 leader,
F.L. CABALLERO, WP2 leader,
F. RAGUENEAU, WP3 leader,
F. VOLDOIRE, WP4 leader
A. LE MAOULT, WP5 leader
COMMISSARIAT À L’ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
ELECTRICITE DE FRANCE
ECOLE NORMALE DE CACHAN
CENTRALE SUPELEC
ECOLE CENTRALE NANTES
INSTITUT POLYTECHNIQUE DE GRENOBLE
AREVA
INSTITUT DE RADIOPROTECTION ET DE SURETE NUCLEAIRE
EGIS-INDUSTRIES
UNIVERISTE JOSEPH FOURIER - ISTERRE
IFSTTAR
CEREMA
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 2
TABLE OF CONTENTS
WP 1: SEISMIC HAZARD ................................................................................................................ 5
PROGRESS OF THE WP1 “SEISMIC HAZARD” IN 2016 .................................................................................................... 6
STOCHASTIC SOURCE, PATH AND SITE ATTENUATION PARAMETERS AND ASSOCIATED VARIABILITIES FOR SHALLOW CRUSTAL
EUROPEAN EARTHQUAKES ......................................................................................................................................... 8
S.S. BORA 1, F.COTTON 1, F. SCHERBAUM 2, B. EDWARDS 3 and P. TRAVERSA
FOCAL MECHANISMS IN WESTERN FRANCE: A KEY TO PROBE THE STRESS FIELD AND BETTER UNDERSTAND THE ORIGIN OF THE
REGIONAL SEISMICITY? ............................................................................................................................................ 11
L. BOLLINGER1, D. DO PACO1, A. GUILHEM1, P. ROUDIL1, M. GRUNBERG2
USE OF THE SIHEX CATALOG TO CONSTRUCT STRONG MOTION DATA BASE ......................................................................... 13
D. LÉOBAL, O. SÈBE, F. SCHINDELÉ, S. MARIN, B. HERNANDEZ, CEA DAM DIF
COMBINATION OF GMPE'S WITH A BAYESIAN APPROACH ............................................................................................. 15
M. BERTIN 1, S. MARIN 2
EPISTEMIC VARIABILITY IN DETERMINISTIC HAZARD ASSESSMENT .................................................................................... 17
M. LANCIERI, A. DELVOYE, L. PROVOST, C. GÉLIS, S.DEL GAUDIO, IRSN, PRP-DGE/SCAN/BERSSIN
RESPONSE-SPECTRUM COMPATIBLE RECORD SELECTION FOR NONLINEAR STRUCTURAL ANALYSIS ......................................... 20
L. ISBILIROGLU 1, 2, M. LANCIERI 1, P. GUEGUEN 2
WP 2: NON LINEAR INTERACTION BETWEEN NEAR & FAR SEISMIC WAVE FIELDS, THE SOIL & THE
STRUCTURE ................................................................................................................................ 23
PROGRESS OF THE WP2 “NON LINEAR INTERACTION BETWEEN NEAR & FAR SEISMIC WAVE FIELDS, THE SOIL & THE STRUCTURE”
IN 2016 .............................................................................................................................................................. 24
F. LOPEZ-CABALLERO, CentraleSupélec, Université Paris-Saclay (former ECP)
MODELING OF NONLINEAR SOIL STRUCTURE INTERACTION: "DOMAIN REDUCTION METHOD" APPROACH ............................... 27
A. FRAU1, E. FOERSTER1, F. LOPEZ-CABALLERO2
EVALUATION AND PROPAGATION OF EPISTEMIC UNCERTAINTIES IN 1-D NON-LINEAR SITE RESPONSE ANALYSIS: CONTRIBUTIONS OF
THE PRENOLIN PROJECT. .......................................................................................................................................... 30
J. RÉGNIER 1, P.-Y. BARD 2, E. BERTRAND 1, L.-F. BONILLA 3 & F. LOPEZ-CABALLERO 4.
THE ARGOSTOLI TEST-SITE: DATA AND FIRST RESULTS FROM THE POST-SEISMIC SURVEY AND THE ARGONET ARRAY ................ 33
F. HOLLENDER 1, V. PERRON 1, S. SBAA 1, 2, A. SVAY 3 , A. IMTIAZ 4, 5, A. MARISCAL 4, P.-Y BARD 4 and the T2.3 working
team
3D SEISMIC WAVE PROPAGATION IN HETEROGENEOUS NON-LINEAR MEDIA BY SPECTRAL ELEMENT METHOD ................ 37
F. GATTI, L. DE CARVALHO PALUDO, A. SVAY, F. LOPEZ-CABALLERO, R. COTTEREAU, CentraleSupélec
WP 3: SEISMIC BEHAVIOR OF STRUCTURES AND COMPONENTS .................................................. 41
PROGRESS OF THE WP3 “SEISMIC BEHAVIOR OF STRUCTURES AND COMPONENTS” IN 2016 ................................................ 42
F. RAGUENEAU, ENS-Cachan
ENHANCEMENT OF MULTIFIBER BEAM ELEMENTS IN THE CASE OF REINFORCED CONCRETE STRUCTURES FOR TAKING INTO
ACCOUNT THE LATERAL CONFINEMENT OF CONCRETE DUE TO STIRRUP .......................................................................... 44
N. KHODER 1
, S. GRANGE 2
, Y. SIEFFERT 1
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 3
A NEW FINITE ELEMENT BEAM TO SIMULATE THE BEHAVIOR OF REINFORCED CONCRETE STRUCTURES TILL FAILURE ..................... 47
I.BITAR1, P. KOTRONIS1, N. BENKEMOUN2, S. GRANGE3
MODEL-ORDER REDUCTION FOR THE PARAMETRIC ANALYSIS OF DAMAGE IN REINFORCED CONCRETE STRUCTURES ................ 51
M. VITSE, D. NÉRON, P.-A. BOUCARD – LMT Cachan, ENS Paris-Saclay
DETERMINATION OF ULTIMATE CAPACITY OF ELECTRICAL EQUIPMENT FOR SUBSEQUENT INDUSTRIAL USE ................................. 55
N. MOUSSALLAM, C. GRAF, B. BOUDY, AREVA NP
UNDESIRABLE EFFECTS OF SEISMIC BASE ISOLATION ..................................................................................................... 57
I. POLITOPOULOS, CEA
WP 4: SEISMIC RISK ASSESSMENT ............................................................................................... 60
PROGRESS OF THE WP4 “SEISMIC RISK ASSESSMENT” IN 2016 ...................................................................................... 60
F. VOLDOIRE, EDF, R&D Division
PRELIMINARY CALIBRATION OF THE NUMERICAL LARGE-SCALE SCENARIO OF THE NIIGATA-KEN CHŪETSU-OKI EARTHQUAKE ........ 63
F. GATTI 1,2, F. LOPEZ-CABALLERO 1, R. PAOLUCCI 2, D. CLOUTEAU 1
IMPLEMENTATION OF KARISMA DEMONSTRATIVE CASE STUDY: INITIAL PHASE ..................................................................... 68
F. WANG, C. FEAU, Laboratoire EMSI, CEA-Saclay
GENERATION OF SYNTHETIC ACCELEROGRAMS IN AGREEMENT WITH CONDITIONAL SCENARIO SPECTRA FOR THE COMPUTATION OF
FLOOR RESPONSE SPECTRA. - APPLICATION TO KARISMA BENCHMARK. .............................................................................. 71
I. ZENTNER, EDF R&D
TASK 4.3.2 : DEMONSTRATIVE NUMERICAL CASE STUDY, FINAL STEP. ............................................................................... 74
François Voldoire, EDF R&D
WP 5: EXPERIMENTAL SUPPORT TO SINAPS@’S ISSUES AND BUILDING TO BUILDING INTERACTION
.................................................................................................................................................. 78
PROGRESS OF THE WP5 “EXPERIMENTAL SUPPORT TO SINAPS@ ISSUES AND BUILDING TO BUILDING INTERACTION” IN 2016 . 79
A. LE MAOULT, CEA/DEN/SEMT
IDEFIX: IDENTIFICATION OF DISSIPATION IN RC STRUCTURAL ELEMENTS ........................................................................... 81
T. HEITZ(1,2), B. RICHARD(2), C.GIRY(1), F. RAGUENEAU(1), A.LE MAOULT(2)
INTERACTION BETWEEN BUILDINGS: SENSITIVITY ANALYSIS AND EXPERIMENTAL CAMPAIGN ................................................... 84
V. CROZET, I. POLITOPOULOS, CEA, DEN, DANS, DM2S, SEMT, EMSI : Laboratoire d’études de Mécanique sismique
WP 6: TRAINING AND KNOWLEDGE DISSEMINATION .................................................................. 87
PROGRESS OF THE WP6 “TRAINING & KNOWLEDGE DISSEMINATION” IN 2016 ................................................................. 88
C. BERGE-THIERRY, CEA/DEN/DM2S
APPENDICE ................................................................................................................................. 90
TOWARD AN INTEGRATED SEISMIC RISK ASSESSMENT FOR NUCLEAR SAFETY IMPROVING CURRENT FRENCH METHODOLOGIES
THROUGH THE SINAPS@ RESEARCH PROJECT .............................................................................................................. 91
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 4
Introduction
These proceedings contain the abstracts of the scientific presentations delivered during the third
plenary session of SINAPS@, held at Electricité de France, on the new Saclay site, November 21 & 22,
2016. The plenary session involved the SINAPS@ monitoring committee which finally provides
recommendations regarding the scientific contents and/or interactions between activities. For each
work package, the abstracts are preceded by a brief synthesis remaining the main issues and a progress
report of activities. The figure below reminds the structure of the SINAPS@ project through the
structure of 5 scientific work packages (WP1 to 5) and the last one (WP6) dedicated to the knowledge
dissemination through training sessions. In appendix a global description of SINAPS@ is provided as
presented in an article proposed following the SMIRT-2015 Manchester conference: this article
presents the project objectives, discusses the state of practices and exhibits the key issues for each
step of the risk analysis. This paper has been accepted for publication in Nuclear Engineering and
Design (in press, 2016).
SINAPS@: towards an integrated approach of seismic risk analysis from fault to structures and components propagating
the uncertainties.
Demonstrative test case:
Kashiwazaki Kariwa NPP site
+ WP 6 Knowledge Dissemination
« WP1 » Seismic Hazard
« WP2 » N.L. Site Effects & SSI
« WP3 & WP5 »
Structural & SSC’s responses
« WP4 » Seismic Risk Assessment
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 5
WP 1: Seismic Hazard
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 6
PROGRESS OF THE WP1 “SEISMIC HAZARD” IN 2016
Context and main objectives of WP1 :
WP1 is focused on the seismic hazard assessment (S.H.A.) on the French metropolitan territory,
characterized by a spatially heterogeneous and low rate of seismicity. A key scientific issue addressed
in the WP1 is to analyse if the methods used for assessing the seismic hazard are suitable in terms of
knowledge or lack of knowledge about the seismicity and associated physical processes. After the
Fukushima accident, the development of the probabilistic approaches are encourages, even in
countries with a deterministic tradition such as the France. Complementary safety studies (ECS)
performed in France have been reviewed by European experts and they recommended a completion
of the classical deterministic method by a probabilistic seismic hazard assessment.
Beyond methodological debates the true issues of WP1 are:
to systematically identify, quantify all uncertainties and clarify their treatment in the S.H.A;
to provide to WP2, 3 and 4 researchers and engineers reliable and adapted seismic inputs in
coherence with the methodologies used to predict site effects, SSI and risk.
The WP1 objectives are:
to characterize the typical French data by the most appropriate and validated methods to
generate metadata and their uncertainties,
to suggest a ranking of the key parameters in the seismic hazard assessment methodology and
estimate the associated uncertainties to guide future research,
to assess the sensitivity of deterministic approaches such as the RFS 2001-01 and probabilistic
approaches according to the known input data,
make recommendations for the advancement of the French regulatory standards,
to provide a relevant description of the seismic hazard for the engineering needs.
Main advancements 2016:
During the 2016 plenary session the following presentations are expected:
1. S. BORA GFZ, Germany & P. TRAVERSA EDF CEIDRE/TEGG: An Original way to perform host-to-
target adjustment of ground motion in Seismic hazard Studies.
This work focuses on a European strong database study regarding source, site and propagation
parameters that have a huge importance when deriving strong motion empirical prediction equations.
This work takes place in the frame of “site specific” approaches. It is performed in the WP1 context
but obviously presents connections with WP2 as taking into account the site effect characterization
and prediction.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 7
2. L. BOLLINGER, CEA: Focal mechanisms in western France: A key to probe the stress field and
better understand the origin of the regional seismicity?
This work investigates and determines the variability of the fault plane solutions in Western
metropolitan France through an extensive database of events. The robustness of all solutions
previously published is tested and the existing database is completed with focal mechanisms of events
that occurred within the last ten years, a period still uncovered by the publications. This contribution
concerns the characterization of basic data used in the S.H.A. and their uncertainties.
3. D. LEOBAL, CEA: Use of the SIHEX catalog to construct strong motion data base.
This work investigates the possibility to complement the French strong motions recorded by the
permanent national accelerometric network, using the of the LDG velocimetric network composed by
43 stations, in order to enlarge the French near-field records database, at least since 1997 when the
digital network has been replaced with numerical equipment. In addition to increase the number of
available records, the interest of LDG velocimetric stations consist in their “rock site condition” of
implementation.
4. M. BERTIN, CEA & ENS Cachan: Combination of GMPE's with a Bayesian Approach.
This work aims evaluating the feasibility of using Bayesian Model Averaging (BMA) techniques when
applied with Ground Motion Predictive Equation's to describe the mainland France's seismicity. The
model averaging calibration is computed with records from the RAP database. First results show a good
predictive performance of the resulting model average, and a strong discrepancy between GMPE's
BMA weights, which distribution differs according to the frequency domain considered.
5. M. LANCIERI, IRSN: Investigation of variability in deterministic seismic hazard assessment: the
impact of GMPE and physic based ground motion simulation methods.
This work investigates the epistemic uncertainties in deterministic hazard assessment due to the use
of different GMPEs. In particular, the application of the “extended fault” GMPEs requires an excellent
knowledge of the fault structure and site location with respect to the fault. Since this is not always the
case, the spectral amplitude variation is explored in function of different parameters and combination
of parameters. The aim is to quantify the variability and understand which parameters (or combination
of parameters) mainly control the amplitude of predicted ground motion.
6. L. ISBILIROGLU, ISTerre-IRSN: Response-Spectrum compatible record selection for nonlinear
structural analysis.
The presented work in part of the PhD thesis performed in WP1-SINAPS@ and linked to the
users from WP2, WP3 and WP4 of the project. This study has a broader goal of quantifying the
effects of different input Ground Motions (GMs)–such as real, stochastic, seismic-source based, and
matched GMs–on nonlinear structural responses of simple and complex models. In the current work,
variability in input GM sets and the corresponded structural responses are discussed.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 8
STOCHASTIC SOURCE, PATH AND SITE ATTENUATION PARAMETERS AND ASSOCIATED VARIABILITIES
FOR SHALLOW CRUSTAL EUROPEAN EARTHQUAKES
S.S. BORA 1, F.COTTON 1, F. SCHERBAUM 2, B. EDWARDS 3 and P. TRAVERSA
1 GFZ German Research Centre for Geosciences, Potsdam 2 Institut für Erd- und Umweltwissenschaften, Universität Potsdam 3 University of Liverpool 4 EDF TEGG
We have analyzed pan-European strong motion database (RESORCE-2012) of acceleration traces
compiled across Europe and Mediterranean regions. As the majority of the earthquakes are of small
to moderate magnitude, a point source ω2 model is assumed to be appropriate. The selected dataset
exhibits a bilinear distance-dependent Q model with average κ0 value 0.0308 s. However, strong
regional variations in inelastic attenuation were also observed. For instance, frequency-independent
Q0 of 1462 and 601 were estimated for Turkish and Italian data respectively. In addition, apparent site
attenuation parameter (κ0) values also indicate strong trade-off with regional variations in Q0 (with
value 0.0457 and 0.0261 s for Turkey and Italy respectively). The linear site amplification factors were
constrained from residual analysis at each station and siteclass type. The moment magnitudes
determined from the Fourier amplitude spectra of acceleration traces were found comparable with
the database values. Stress-parameters (Δσ) did not exhibit magnitude dependence. The median Δσ
value was obtained as 5.75 and 5.65 MPa from inverted and database magnitudes respectively. We
provide κ0 values for 45 European stations and source parameters (i.e., fc, stress parameter and seismic
moments) for 43 well-recorded earthquakes. The analysis presented can be considered as an update
of that undertaken for the previous Euro-Mediterranean strong motion database presented by
Edwards & Fäh (Edwards & Fäh, 2013a). A reasonably good comparison of response spectra from the
stochastic model (derived herein) with that from (regional) ground motion prediction equations
(GMPEs) suggests that the presented seismological parameters can be used to represent the
corresponding seismological attributes of the regional GMPEs in a host-to-target adjustment
framework.
Figure 1: Station κ0 plotted against Vs30 values for Vs30 > 360 m/s: (a) when earthquakes located at less than 40 km
(from a station) are used, (b) when all the earthquakes recorded at a station are used. Markers (empty circles, disks
and empty squares) indicate the median while the extent of vertical bars indicates the values corresponding to16 and
84 percentiles at each station, i.e., within-station variability. The horizontal solid line indicates the median value of all
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 9
station κ0 in the sample, while two dashed lines indicate 16 and 84 percentile values in the sample, i.e. between-station
variability. In both cases stations which have recorded at least 10 records (including both the components) are used.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 10
Figure 2: Main metadata features of the selected dataset. (a) Distribution of earthquake epicenters. (b) MW-hypocentral
distance distribution; (c) MW-depth distribution; (d) light-shade: high-pass and dark-shade: low-pass frequency; (e)
number of records per station versus number of stations.
Figure 3: Stress parameters (Δσ) (obtained from site-class specific amplification corrected spectra) plotted against
database MW in panel (a) against depth in panel (b). Disks indicate the Δσ values when both fc and MW were obtained
from inversion while the empty circles indicate those when only fc was obtained from inversion keeping the M0 fixed
from database MW. Again, events recorded at least at three stations (six records including both the components) are
shown.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 11
FOCAL MECHANISMS IN WESTERN FRANCE: A KEY TO PROBE THE STRESS FIELD AND BETTER
UNDERSTAND THE ORIGIN OF THE REGIONAL SEISMICITY?
L. BOLLINGER1, D. DO PACO1, A. GUILHEM1, P. ROUDIL1, M. GRUNBERG2
1 CEA, Bruyères Le Chatel, DAM/DIF/LDG, 91297 Arpajon. 2 Université de Strasbourg, EOST, UMR 7516, 5 rue R. Descartes, 67084 Strasbourg cedex
Scope within SINAPS@
This project investigates the relations between faults and seismicity in metropolitan France through
the compilation of an extensive database of focal mechanisms and its confrontation with stress build
up models.
The scientific context
In slowly deforming regions of continental plate interiors, regional sources of stress and strain can
result in transient deformation rates comparable to or greater than the background tectonic rates (e.g.
Craig et al., 2016). As a result, several tectonic and external forcing models need to be tested in order
to properly understand the mechanisms that lead to the generation of local earthquakes. Because the
stress build up on a fault is highly dependent of the regional stress and the fault plane orientation, all
these models require as a prerequisite a good knowledge of the local geology and geophysics, as well
as of the earthquake focal plane orientation.
However, French metropolitan territory is poorly covered by global studies of focal mechanisms. In
western France however, a few regional studies documented the largest earthquakes focal
mechanisms (e.g. Nicolas et al. 1990; Mazabraud et al., 2005). These focal solutions mostly cover the
period from 1964 to 2002 and lack more recent events, latest events that eventually benefit from
better seismic networks coverage.
Objectives of this work
In this work, we determine the variability of the fault plane solutions in Western metropolitan France
through an extensive database of events. We compile and test the robustness of all solutions
previously published and complement the existing database with focal mechanisms of events that
occurred within the last ten years, a period still uncovered by the publications.
The current work progress
One hundred and forty three focal mechanisms of local earthquakes were compiled and tested for
robustness of the parameters published. In addition, thirty six new focal mechanisms were determined
from the study of the first motion directions. Two of them are compared with a set of other solutions
derived from full waveform inversion. Several maps of the local and regional stress tensor orientations
are derived from this dataset.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 12
Figure: Map of the 143 focal mechanisms compiled and tested for robustness (Black beach balls). The 36 new focal
mechanisms determined from the study of first motion directions are in red. Inset covers “Ile d’Oléron” and its local
focal mechanism solutions.
References
Craig, T. J., Calais, E., Fleitout, L., Bollinger, L., & Scotti, O. (2016). Evidence for the release of long‐term
tectonic strain stored in continental interiors through intraplate earthquakes. Geophysical Research Letters,
43(13), 6826-6836.
Mazabraud, Y., Béthoux, N., Guilbert, J., & Bellier, O. (2005). Evidence for short-scale stress field variations
within intraplate central-western France. Geophysical Journal International, 160(1), 161-178.
Nicolas, M., Santoire, J. P., & Delpech, P. Y. (1990). Intraplate seismicity: new seismotectonic data in Western
Europe. Tectonophysics, 179(1-2), 27-53.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 13
USE OF THE SIHEX CATALOG TO CONSTRUCT STRONG MOTION DATA BASE
D. LÉOBAL, O. SÈBE, F. SCHINDELÉ, S. MARIN, B. HERNANDEZ, CEA DAM DIF
Scope within SINAPS@
Seismic hazard studies require near-field ground motion records of big earthquakes, but in regions of
moderate seismicity like France, such measurements are rare. We propose here a contribution to the
improvement of the seismic databases, based on data from the RAP (French permanent accelerometric
network) and from LDG.
The scientific context and objectives
The RAP has recorded near-field accelerations since the mid 90’s. Recently, the SiHEX project proposed
a catalogue of the French seismicity (1962-2009), including a systematic estimation of moment
magnitude calibrated on recent high-quality instrumental data, especially the LDG seismic network.
The 43 stations of the LDG velocimetric network could help to enlarge the French near-field records
database, at least since 1997 when the digital network has been replaced with numerical equipment.
The current work progress
A preliminary step consisted to construct a strong motion data-base on French metropolitan territory,
based on RAP records selected to perform an automatic analysis of signals such as PGA estimation or
PSA calculation, and based on SiHEX catalog. Over the period 1997-2013, 83 events with magnitude
higher than 3.4 have been selected. As the current accelerometric records may present several
discrepancies in the signal acquisition (high noise level, inadequate time signal windows, cutting
relevant data, or too short to fully study the informative content of seismogram), an automatic
algorithm has been developed to select signal relevant to the type of analysis performed. This
procedure provides a subset according to time, frequency and signal-to-noise ratio criteria, allowing a
stable estimation of seismic parameter such as PGA or PSA.
Second, in order to assess the interest of LDG velocity records to extend the accelerometric data base,
a detailed comparison of records at 5 RAP and LDG colocated stations has been carried out. The
objective is to determine to what extent records of both networks are similar, and under which
conditions acceleration derived from velocimetric records can be used. For strong motion studies
focusing their analysis on the frequency range [0-20]Hz, the interest of velocimetric record has been
clearly highlighted for moderate seismic activity such as in France. Indeed non saturated velocimetric
signals can provide information similar to the accelerometric ones, and the velocimetric signal presents
better signal-to-noise ratio with a longer time range. In addition, the LDG records database concern
rock site records which represent less than 25% of RAP recording sites.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 14
Figure: Map of the colocated RAP and LDG stations (center). For each location, the spectrum for the common events is
displayed in black for the velocimeter (VEL) and the accelerometer (ACC), and the average noise spectrum is shown in
blue. The red dots indicate the maximum values of the spectrum, while the green curves show the models of seismic
ambient noise (New Low and New High Noise Model, Peterson, 1993).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 15
COMBINATION OF GMPE'S WITH A BAYESIAN APPROACH
M. BERTIN 1, S. MARIN 2
1 ENS Cachan – CMLA/LRC MESO, France 2 CEA-DAM-DIF, France
Scope within SINAPS@
The present postdoctoral work is consistent with the first SINAPS@’s work package that deals with the
evaluation of seismic hazard. In particular, this study is associated with the work package 1.2.2 since
we propose a bayesian strategy to enhance the predictive quality of GMPE models in a random context,
when used to study the mainland France's seismicity.
The scientific context
In France's context, where seismicity is moderate, records don't cover the whole range of variables
configurations useful for the evaluation of seismic hazard. Thus, a set of models established in a similar
context (Italy, the Mediterranean Basin, U.S.A., Japan, etc.) is considered through a model selection
process and with the help of a decision tree. Ultimately, this approach is mainly based on the scientist's
expertise.
But there is a way to address the issue of model's predictive quality evaluation and of model selection
without any additional hypothesis: the Bayesian Model Averaging approach (BMA). This method is an
extension of classical Bayesian Calibration techniques and allows to take into account a set of several
models instead of a unique one. By making these models encounter a dataset of observations through
a statistical framework, BMA approach provides an unbiased evaluation of each model likelihood and
moreover, it produces a weighted average formula using every model considered to get the best
predictive result.
Objectives of this work
The main objective of this postdoctoral work is the evaluation of the feasibility and of the contribution
of BMA techniques when applied with GMPE's to describe the mainland France's seismicity.
The current work progress
Several BMA algorithms are implemented, compared and used with 8 GMPE's. The model averaging
calibration is computed with around a hundred records of seismic events from the RAP database
(French Accelerometric Network). First results show a good predictive performance of the resulting
model average, and a strong discrepancy between GMPE's BMA weights (that can be seen as
confidence criteria) is also noticed, specifically, the distribution of these weights differs according to
the frequency domain considered.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 16
PDF of each GMPE's and of the BMA model calibrated with 50% of the observation database (frequency is 0.5 Hz). The
grey area is the 95% confidence interval.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 17
EPISTEMIC VARIABILITY IN DETERMINISTIC HAZARD ASSESSMENT
M. LANCIERI, A. DELVOYE, L. PROVOST, C. GÉLIS, S.DEL GAUDIO, IRSN, PRP-DGE/SCAN/BERSSIN
Scope within SINAPS@
One of the primary scopes of SINAPS@ project is to understand the impact of incertitude on hazard
assessment. In this work, we will focus on the variability of ground motion estimation when little
information is available on seismic sources in a given area of interest.
The scientific context
The current practice in deterministic and probabilistic hazard assessment is to estimate the spectral
amplitude associated with a given seismic scenario (described by magnitude and source-site distance)
using ground motion prediction equations (GMPE). GMPEs express the correlation between the
spectral amplitudes, issued by natural or synthetic records of the soil acceleration, and the metadata
describing records properties in terms of source, fault- site geometry and soil characteristic.
Among the pure empirical GMPEs we can distinguish three categories following different degrees of
complexity in the seismic source description. The “point source” GMPEs are characterized by simple
functional forms based on event magnitude, hypocentral distance, and, in some cases, the style of
faulting (taken into account using the rake angle), and a parameter describing soil conditions. The
“simple extended fault” GMPEs have functional forms similar to those in the point source category,
but they use the Joyner and Boore distance (RJB), defined as the distance from the site and the
projection of the bottom of the fault plane. The “extended fault” GMPEs include a detailed definition
of the fault geometry expressed by the fault width, three different distance metrics (RJB, distance from
the rupture plane and distance from the top) and the site-source azimuth angle.
Scope of the work
In the present work, we investigate the epistemic uncertainties in deterministic hazard assessment
due to the use of different GMPEs (selected among the three categories described above). Indeed, the
application of the “extended fault” GMPEs requires an excellent knowledge of the fault structure and
site location with respect to the fault. Since this is not always the case, we also explore the spectral
amplitude variation in function of different parameters and combination of parameters. The aim is to
quantify the variability and understand which parameters (or combination of parameters) mainly
control the amplitude of predicted ground motion.
Work in progress
For a fixed scenario, expressed in terms of magnitude, hypocentral distance and Vs30, we investigate
nine GMPEs selected by publication date and according to a criterion of increasing complexity (see
table 1).
Under the hypothesis of poor knowledge of the seismic source, we explore the spectral variability in
function of: focal mechanism (dip angle, rake angle), fault depth and source-site azimuth (equivalent
to the fault strike). The distance parameters in table 1 are geometrically related with the focal angle
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 18
and the fault dip – these relationships are described in Kaklamanos et al. (2011) – and used in this work
to deduce the unknown parameters.
Table 1: The investigated GMPEs and the input parameters.
The epistemic uncertainty has been investigated using a logic tree approach, and by attributing the
same weight to each GMPEs, as shown in figure 1.
Figure 1: Logic tree used in the study
A first glance on the obtained variability is given in figure 2, where the grey lines are the spectra
corresponding to each logic tree branch and the red curves describe the median and several quantiles
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 19
of the spectral distribution. The blue line is the Berge Thierry et al. 2003 GMPE currently used in France
to evaluate the seismic hazard at nuclear sites following a deterministic approach.
As expected, the complex GMPEs are characterized by an important variability. In the specific case of
Mw 5.8 and hypocentral distance of 20 Km, the higher ground motion amplitudes are related with
smaller RJB. The spectra for RJB = 0 are not plotted on the figure. This particular configuration will be
studied in the next actions, and the obtained results will be compared with near fault simulation.
Figure 2: spectral distribution
On the basis of this preliminary results we conclude that, today a deterministic study cannot be based
on a single GMPEs. The complex GMPEs, better describe the fault-site geometry but their application
requires a deep knowledge of the sources characteristic. This level of knowledge is not currently
available in France, their application requires to investigate the epistemic uncertainties.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 20
RESPONSE-SPECTRUM COMPATIBLE RECORD SELECTION FOR NONLINEAR STRUCTURAL
ANALYSIS
L. ISBILIROGLU 1, 2, M. LANCIERI 1, P. GUEGUEN 2
1 Institut de Radioprotection et de Sûreté Nucléaire, IRSN (BERSSIN) 2 ISTerre, University Joseph Fourier (Grenoble)
Scope within SINAPS@
The part, herein, aims at bridging the necessary components of seismic analysis from defining seismic
hazard scenario to nonlinear structural analysis. The main focus is to understand the effect of different
ground motion selection and modification (GMSM) methods on nonlinear structural analysis while
considering ‘natural’ variability, which necessitates linking two disciplines: seismology and structural
engineering.
The scientific context
Nonlinear dynamic analysis is used to determine seismic behavior of structures under given input
ground motions (e.g., seismic loading) that are selected appropriately with seismic hazard. Nonlinear
analyses are very sensitive to the selected input GMs; however, there has not been a consensus among
engineers and seismologists on how to select and scale GMs for nonlinear analysis.
One of the factors to judge the GMSM methods is to analyze whether they conserve the ‘natural’
variability in their input GMs or not (Pousse et al. 2006). An earthquake (EQ) scenario, defined by
magnitude and distance (M&R), can have a large variability in spectral accelerations (SA). It can be due
to the uncertainty involved in earthquake-based physics, directivity, near-source effects, and different
GMSM methods.
The nonlinear structural analyses can be high-cost and the number of structural analyses is limited in
the current practice. For that reason, the input variability is reduced drastically although ‘natural’
variability exists in GMs. Spectrum matching (Al Atik and Abrahamson 2010) is a commonly used
method to reduce the variability and to satisfy the strict requirements stated in building codes. For
example, ASN Guide 2-01 (2006), French nuclear safety guide, allow ±5% of variability around the
target spectra.
We believed that it is highly necessary to tie the variability in input GMs with the variability in structural
responses and to understand the impact of input GM variability on output structural responses as a
function of followings: type of structure (whether it is a residential building or a power plant or a
bridge), goals of engineering application (performance assessment, retrofitting, or design) and
approaches to model the structure (modeling in a basic way as 1D or in a complex way as 3D).
Objectives of this work
This study is a part of the ongoing PhD project having a broader goal of quantifying the effects of
different input GMs–such as real, stochastic, seismic-source based, and matched GMs–on nonlinear
structural responses of simple and complex models. In the current work, variability in input GM sets
and the corresponded structural responses is discussed.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 21
The current work progress
The work, herein, is aligned with the question of whether we ignore an important phenomenon by
reducing the input GM variability or not. In order to gain better insight, the following questions will be
discussed:
What is the level of variability in input GM sets?
How are the structural responses affected in the existence of the GM variability?
Will it introduce any biases on the structural responses?
In this work, four earthquake (EQ) scenarios with two different M&R cases and two different soil
conditions are considered. The spectral shapes are obtained from seven different ground motion
prediction equations (GMPE) with 5% damping. Then, two different families of GMs are obtained: real
unscaled EQ records and spectrum matched waveforms. GM selection is performed based on spectral
compatibility with upper amplitude ratio, lower amplitude ratio and frequency range. Figure 1.a shows
an example of spectrum compatible GM selection. Lines in red represent the GM selection criteria
(upper and lower amplitude ratios and frequency range). Black line is the target spectrum obtained
from a GMPE. Green line is the average of five GM spectra (shown in gray lines) falling within upper
and lower amplitude ratios for the whole frequency range. The variability within one GM set is called
intraset variability. There can be several other eligible GM sets as shown in Figure 1.b. The variability
between all eligible GM sets is called interset variability, which is our main focus.
(a) Single eligible GM Set
(b) All eligible GM Sets (> 1 500)
Figure 1: (a) Intraset Variability and (b) Interset Variability
The first spectrum compatible selection is carried out for the real unscaled records. None of the GM
sets including real unscaled records can be retrieved based on the original ASN Guide 2-01 (2006),
which is ±5% around the target spectrum. GM sets including real unscaled records are obtained with
various upper amplitudes of 30%, 40% and 50%. Lower amplitude is kept constant as -5%. The
frequency range of [0.5 Hz, 20.0 Hz] is used in all analyses. In order to obtain eligible GM sets in
compliance with ASN Guide 2-01, the spectrum matching technique (Al Atik and Abrahamson 2009) is
performed to match waveforms. Spectrum compatible GM selection is then made with the matched
waveforms for upper amplitudes of 5% and 10%. An example of spectrum matched waveforms is
shown in Figure 2.
0.03 g – 0.004g
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 22
Figure 2: Five GM spectra (with colorful lines) before (on the left) and after (on the right) applying spectrum matching with respect to a given target spectra (black line)
The variability is defined by Metric 1 given in Equation 1. It is the distance from 16th percentile to 84th
percentile when distribution of GM sets is normalized to median. The GM sets with real unscaled data
have input variability changing from 14% to 25% in terms of PSA at 1 Hz; whereas, the GM sets with
matched waveforms have input variability less than 5%.
𝑀𝑒𝑡𝑟𝑖𝑐 1 =84th Percentile
Median−
16th Percentile
Median
Equation 1
All of the eligible GM sets are then injected into single-degree-of-freedom (SDOF) oscillator with a
fundamental frequency (fo) of 1 Hz and elastic-plastic material with strain hardening behavior (+5%).
Constant strength approach is used with a strength reduction of 8. GM sets with unscaled real records
have the variability in maximum displacement between 15% and 50%. GM sets with matched
waveforms have variability larger than 15%.
We conclude that
A good number of GM sets is needed to sufficiently represent the interset variability,
Variability does not depend on GMPEs, and
For GM sets including matched waveforms, the input variability is reduced drastically but
variability in maximum displacement is relatively large.
The future work includes the analysis of sufficient amount of GM sets for a stable distribution of
structural responses and the comparison of median responses. The study will also be extended to
different material behaviors and complex structural models.
References:
Al Atik, L. and N. Abrahamson, 2010, “An improved method for nonstationary spectral matching,”
Earthquake Spectra, 26 (3), 601-617.
Autorite de Surete Nucleaire (ASN) Guide 2-01, Prise en compte du risque sismique à la conception des
ouvrages de génie civil d'installations nucléaires de base à l'exception des stockages à long terme des déchets
radioactifs, 2006.
Berge-Thierry C., Cotton F., Scotti O., Griot-Pommera D. A., Fukushima Y., « New empirical response spectral
attenuation laws for moderate European earthquakes », Journal of Earthquake Engineering, vol. 7, n° 2,
2003, p. 193-222.
Pousse G., Bonilla L. F., Cotton F., Margerin L., « Stochastic Simulation of Strong Ground Motion Time
Histories Including Natural Variability: Application to the K-Net Japanese Database », Bulletin of the
Seismological Society of America, vol. 96, n° 6, 2006, p. 2103–2117.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 23
WP 2: Non Linear Interaction
between Near & Far Seismic Wave
Fields, the Soil & the Structure
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 24
PROGRESS OF THE WP2 “NON LINEAR INTERACTION BETWEEN NEAR & FAR SEISMIC WAVE
FIELDS, THE SOIL & THE STRUCTURE” IN 2016
F. LOPEZ-CABALLERO, CentraleSupélec, Université Paris-Saclay (former ECP)
The scientific context
In the framework of the wave propagation from the source to the equipment at the structure, this
work-package is placed at the interface between the soil and the structure, the seismology and the
structure dynamics, the hazard and the structure vulnerability. Even if the soil-structure interaction
effects are well known from the 70’s, they have often been tackled under simplified assumptions:
Winkler springs, uniform incident wave field, shallow rigid footings or linear equivalent soil behavior,
among others. Some of these assumptions have been improved in recent years highlighting the
intrinsic safety margins. Moreover, those works showed their high sensitivity to the uncertainty on
both the seismic loading and the soil surrounding the structure. Moreover, so as to take into account
extreme events in the post-elastic behavior of structures, it is necessary to have a more detailed
description of the seismic loading, in both time and space, exceeding the given maximum acceleration
or code spectrum. Finally, the instrumental and theoretical seismology has highlighted the complexity
and variability of the incoming seismic waves: near field effect, site effects, non-linear filtering strong
movements, spatial variability. These advances build now a big picture, which combines divers
methods with difficulties to be associated and sometimes inconsistent with the regulations and
common methods used in the world.
The Work-package 2 is divided in 3 axes with tasks of research, which define its principal objectives.
The partners involved are: CEA, CEREMA, ECP, EDF, IFSTTAR, IRSN, ISTerre.
Objectives of this work package
WP2 has the following three objectives:
(i) Improvement of traditional methods defining the input motion at structure base, a. Based on obtained results of work-package 1 b. Including spatial variability of signals c. Quantification of uncertainties of divers soil materials
(ii) Development of new methods, a. From the fault to the equipment: including non-linearity and variability of soil
properties. b. Coupling structural and wave propagation codes
(iii) New seismic data acquisition to validations: a. In the seismicity framework of France (low to moderate). b. Validation site (Argostoli site, Kephalonia island, Greece)
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 25
The current work progress
The presentations by WP2 partners will focus on progress of:
Task 2.1.3: Soil-structure coupling:
(i) Modeling of Nonlinear soil structure interaction: "Domain Reduction Method" approach.
Task 2.2.3: Validation of non-linear soil models for the strong ground motions:
(ii) Evaluation and propagation of epistemic uncertainties in the simulation of seismic site response: Contributions from PRENOLIN benchmark.
Task 2.3.1: Validation site: Data acquisition:
(iii) Argostoli test site: review of the post-seismic survey and ArgoNet network.
Task 2.2.1: Development of a non-linear and large-scale probabilistic model from the source to the
structure and Task 2.3.2: Spatial variability at free field, validation and explanation:
(iv) Development of a non-linear and large-scale probabilistic model from the source to the structure.
The goal of these presentations is to show how the different tasks are related one another and also
how they feed the objectives of the WP2. It is important to keep in mind that the deliverables of each
task are the gear of the final product, which is to perform a large-scale probabilistic model from the
source to the structure, taking into account the non-linear site effect, the Soil-Structure Interaction
and the propagation of uncertainties (e.g. material properties, type of sources) for the demonstrative
case-study of WP4.
In this context, presentation (ii) shows the performance of different numerical models to represent
the non-linear soil behavior. The results were compared to the observations on sites of the Japanese
accelerometric network. Thus, two numerical models presented in this benchmark were implemented
in the numerical codes used in the works of presentations (i) and (iv). These numerical models allow
simulating the dynamic soil behavior in a simple way with an acceptable level of accuracy.
Concerning the presentation (i), the objective is to compare different strategies for fully nonlinear
analysis for SSI problems using fictive boundaries to represent the infinite domain. One of the
strategies to simulate the incoming waves is based on the computation of the equivalent force field at
the boundaries of the reduced model from a large numerical model, which includes for example the
earthquake source (i.e. work of presentation (iv)).
Presentation (iv) will display a brief overview of results obtained from a 3D regional scale non-linear
and probabilistic model using the SEM3D code. A numerical application using the implemented
libraries (i.e. generation of the gaussian random-field and a non-linear constitutive relationship to
represent the soil behavior) is done. In this work, a special attention is given to the overall set-up of an
efficient numerical workchain to solve large wave propagation problems over tens of hundreds of
processors on massively parallel supercomputers. As example, Figure 1 displays the evolution of the
overall performance of the SEM3D code from the begining of the project until today. It concerns the
diminition of the CPU time for a same numerical model of wave propagation (size 600m x 600m x
600m) and also the improvements in the capability of the code to generate a random-field for a large
3D model, reducing the computational time.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 26
Figure 1. Examples of the evolution of the overall performance of the SEM3D code from the begining of the project until
today
Finally, the aim of presentation (iii) is to provide an overview of the measures recorded in the
experimental campaigns in the Argostoli site. This database will be used to compare the recorded
seismological data and the simulated one with the code SEM3D among others.
The future works of this WP concern principally both the improvements of a kinematic model of the
source in order to increase the range of generated frequencies (e.g. 10Hz) and the development of a
code coupling SEM3D with a structural FE code so as to perform Soil-Structure Interaction
computations using other approach than the one shown in presentation (i).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 27
MODELING OF NONLINEAR SOIL STRUCTURE INTERACTION: "DOMAIN REDUCTION METHOD"
APPROACH
A. FRAU1, E. FOERSTER1, F. LOPEZ-CABALLERO2
1 DEN/DANS/DM2S/SEMT/EMSI 2CentraleSupélec, Université Paris-Saclay (former ECP)
Scope within SINAPS@
This work is carried out in the framework of the Axes 1 of Work-package 2. Stating from the approach
widely used by the engineers for the SSI (Soil-Structure Interaction) analysis, the main purpose of this
axis is improve several aspect of them in order to define a more realistic models. The task 2.1.3 focus
the attention on two problems:
1. The effects of the foundation embedment on the structure response taking in into account in the SSI models;
2. Transition from the classical linear equivalent approach to nonlinear ones. Several observations and analysis are carried out using a simple constitutive law for the soil developed in the Cast3M FEM Code. An approach based on the Domain Reduction Method (DMR) is formulated in order to reduce the computational time. The study is carried out using the KARISMA benchmark model.
The first part was presented at the end of 2014. Thus, this communication concerns only the second
part.
The scientific context
One of the main techniques, used for taking into account the nonlinear behavior of soil in the design
stage, is based on the linear equivalent method. This one is widely used by the engineers in the past.
Otherwise, the limits of this approach are well described in the literature ([1], [2]). In general, an upper
limit is represented by the maximum shear deformation of soil. Less than 10−4 (corresponding to low
seismic input motion) the equivalent linear method give an acceptable response of soil-structure
system. Thus, for the strong input motion, this value is reached very fast and the nonlinear behavior
must be take into account using a fully nonlinear model.
Usually a Non Linear Soil Structure (NL-SSI) problem are solved using Direct Methods, which could be
very expensive in terms of computation time due to treatment of infinite domain (i.e. fictive
boundaries for a large scale domain). In order to reduce the computation time, a possible strategy is
to reduce the computational domain (i.e. soil domain) and getting close the soil boundaries to the
structure. In this case two aspects are very important:
1) In the full FEM approach the incident waves must be imposed in according to the hypothesis of soil behavior at the fictive boundaries;
2) Moving close the fictive boundaries to the structure means that the influence of the outgoing waves induced from them are important. Then, absorbing boundaries are needed so as to satisfy a radiation condition for the incompatible outgoing waves. Thus, the efficiency of the absorbing layer is key point in order to reduce the size of the problem;
The points mentioned above are the main highlights concerning the DRM methods (Domain Reduction
Method) as proposed by ([3]).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 28
Objectives of this work
In this study, using the input data for the benchmark KARISMA, the objective is to compare different
strategies for fully nonlinear analysis for SSI problem using fictive boundaries to represent the infinite
domain. Usually to perform SSI simulations, periodic conditions are used in order to approximate the
lateral boundaries. Moreover, the seismic loading is imposed as vertical incident waves at the bottom
of the model. This approach is accurate under two conditions: fictive boundaries are distant from the
structure and the problem respects the period conditions (eg. symmetry of geometry and loading).
Due to the presence of the structure, in the idea to reduce the size of the model, this is not the best
accurate technique and other kind of boundaries model are needed to eliminate the outgoing waves
from the structure toward infinity. Hence, a parametric study is carried out using some strategies to
absorb all outgoing waves for symmetrical and asymmetrical systems. In addition, a numerical
procedure is defined to simulate the incoming waves. It is based on the computation of the equivalent
force field at the boundaries from a model without the structure. In this case, for the soil close to the
structure, the computations are performed considering a nonlinear behavior, which is described by a
simple constitutive law developed in Cast3M FEM Code.
The current work progress
For the task 2.1.3 the analysis are carried out using an equivalent 2D model of Soil-Structure problem
for the KARISMA benchmark problem. The goal for this task is to show an alternative methodology for
NL-SSI analysis. The full 3D NL-SSI KARISMA model will be computed in the Work-package 4 as a
demonstrator. This task could define some guidelines for this phase.
Figure 1: Comparative study regarding the spurious wave generated in SSI problem using two different absorbing
boundaries: a) Lysmer boundary et b) Multi-Absorbing Layer.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 29
Figure 2: Response of an asymmetric SSI problem for different configurations. Three conditions are used for the lateral
fictive boundaries: periodic conditions (blue line); Lysmer (red line) and multi-layer absorbing boundaries (cyan line). A
reference solution (black line) is also computed for comparison.
Figure 3: Difference in the frequency content of the response between two configuration of soil (Case1 : larger of soil L
equal to 200 m ; Case2 : L=360 m) using two type of conditions for the lateral boundaries in the SSI model.
Fatahi, B., & Tabatabaiefar, S. H. R. (2013). Fully nonlinear versus equivalent linear computation method for
seismic analysis of midrise buildings on soft soils. International Journal of Geomechanics, 14(4).
Byrne, P. M., Naesgaard, E., & Seid-Karbasi, M. (2006, October). Analysis and design of earth structures to
resist seismic soil liquefaction. In Proceedings of the 59th Canadian Geotechnical Engineering Conference,
Hardy Lecture, Vancouver, BC (pp. 1-4).
Bielak, J., Loukakis, K., Hisada, Y., & Yoshimura, C. (2003). Domain reduction method for three-dimensional
earthquake modeling in localized regions, Part I: Theory. Bulletin of the Seismological Society of
America, 93(2), 817-824.
10-1
100
101
0
10
20
30
40
50
60
Freq [Hz]
Ax [%]
Period.
Lysmer
10-1
100
101
0
10
20
30
40
Freq [Hz]
Az [%]
Period.
Lysmer
10-1
100
101
0
20
40
60
80
Freq [Hz]
Ax [%]
Period.
Lysmer
10-1
100
101
0
10
20
30
40
Freq [Hz]
Az [%]
Period.
Lysmer
+12.00 m Interface +12.00 m Interface
Top of structure -13.70 m Basemat
Multi-Layer Soil
Asymmetric rigid massless
foundation Point A
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 30
EVALUATION AND PROPAGATION OF EPISTEMIC UNCERTAINTIES IN 1-D NON-LINEAR SITE
RESPONSE ANALYSIS: CONTRIBUTIONS OF THE PRENOLIN PROJECT.
J. RÉGNIER 1, P.-Y. BARD 2, E. BERTRAND 1, L.-F. BONILLA 3 & F. LOPEZ-CABALLERO 4.
1 CEREMA 2ISTerre 3 IFSTTAR 4CentraleSupélec
Scope within SINAPS@
Site effects can locally amplify the seismic waves compared to rock reference sites and are dependent
on the site configuration and geology, subsurface geometry and the incident motion. Their evaluation,
therefore, associates epistemic and aleatoric uncertainties. These uncertainties are one link within the
propagation of the uncertainties from the seismic source to the building damage assessment.
The scientific context
The evaluation of site effects can be performed through the analysis of earthquake recordings but is
limited to a specific location, and, most often for moderate seismic hazard region, to weak motion
recordings only. Evaluation using site classification (Vs30 or fundamental resonance frequency) is low
cost and can be executed on a large scale but will be associated with strong uncertainties. Numerical
simulations are the only method to be site-specific and to be able to predict the non-linear site
response during a strong motion. This requires (1) a detailed definition of the soil parameters (2) a
constitutive model of soil that mimics the real soil behavior during earthquakes and (3) an input
motion. The natural soil variability, the measurements errors, the model approximations and the
earthquake variability are the four sources of uncertainty when evaluating site amplification.
Objectives of this work
One of the main objectives of the PRENOLIN benchmark was to evaluate the uncertainties in 1-D non-
linear site response analysis. A large amount of codes were tested on canonical cases (Régnier et al.,
2016) and the results were compared to observations on sites of the Japanese accelerometric network.
The benchmark presently involves 21 teams and 21 different non-linear computations
The current work progress
We selected two sites, Sendai and KSRH10. Sendai is a shallow vertical array with a down-hole sensor
located at GL-8m. KSRH10 is a deeper site with a down-hole sensor located at GL-255m. To constrain
the linear and non-linear soil parameters, in-situ measurements and multiple laboratory
measurements were conducted on disturbed and undisturbed soil samples. Three alternative sets of
non-linear parameters were proposed for the soil models. The first (called SC1) comes from the use of
parameters defined in the literature and based on soil type and depth, here we used the Darendeli’s
formulation (Darendeli, 2001). The second (SC2) and third (SC3) soil columns come from the
interpretation of the laboratory data. The residuals (equation 1) and the code-to-code variability
(equation 2) were calculated on two sites and compared to the part of the uncertainty in GMPE’s
(Ground Motion Prediction Equation) that is associated to the site amplification. The next figure shows
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 31
the Misfit and code-to-code variability averaged over the whole response spectra at the surface. There
is therefore only one value per input motion and soil column.
Eq 1
Eq 2
For Sendai, we compare the results for four input motions (with PGA at the surface from 481 to 8
cm/s2) for which all calculations were performed and for the two soil models SC1 and SC2. At KSRH10,
5 input motions (with PGA from 558 to 54 cm/s2) and 3 soil models (SC1, SC2 and SC3) were used. The
misfit is generally higher than the code-to-code variability. We also observe that contrary to what was
expected, the code-to-code variability and the misfit are quite equivalent for the weakest input
motions at Sendai site, while in the previous exercise, performed on canonical cases, the predictions
were closer one-to-another when no non-linear soil behaviour was involved. (2) On Sendai, the SC1
soil model provides closer results to the observation compared to the SC2 whereas for KSRH10’s SC2
and SC3 models provide lower misfit values. The misfit between observations and simulations lies
between 0.1 to 0.15 (log10 scale) for Sendai and from and 0.1 to 0.27 for KSRH10.
Many authors recently attempted to breakdown the variability associated to GMPEs into between
event and within event (e.g. Al Atik et al., 2010; Rodriguez-Marek et al., 2011; Strasser et al., 2009).
The site amplification contributes to the within-event variability. The amplification factor is defined as
an analytical form that depends on site explanatory parameters (i.e. Vs30 or the fundamental
resonance frequency) and described by a median and two residuals decomposed in (a) site-to-site
residual (δS2Ss) and (b) a site amplification residual δAmpes. The values found by Rodriguez-Marek et
al. (2011) are higher than in PRENOLIN (0.213 at PGA). This suggests that even if the results of the
predictions in PRENOLIN have a large residual, its values is reduced improving the surface ground
motion evaluation compared to site classification methods.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 32
References
Al Atik, L., Abrahamson, N., Bommer, J.J., Scherbaum, F., Cotton, F., Kuehn, N., 2010. The Variability of
Ground-Motion Prediction Models and Its Components. Bull. Seismol. Soc. Am. 81, 794–801.
Darendeli, M.B., 2001. Development of a new family of normalized modulus reduction and material damping
curves.
Régnier, J., Bonilla, L.-F., Bard, P.-Y., Bertrand, E., Hollender, F., Kawase, H., Sicilia, D., Arduino, P., Amorosi,
A., Asimaki, D. et al 2016. International Benchmark on Numerical Simulations for 1D, Nonlinear Site Response
(PRENOLIN): Verification Phase Based on Canonical Cases. Bull. Seismol. Soc. Am. 106, 2112–2135.
Rodriguez-Marek, A., Mantalva, G. 1, Cotton, F., Bonilla, F., 2011. Analysis of Single-Station Standard
Deviation Using the KiK-net Data. Bull. Seismol. Soc. Am. 101, 1242–1258.
Strasser, F.O., Abrahamson, N.A., Bommer, J.J., 2.009. Sigma: Issues, insights, and challenges. Seismol. Res.
Lett. 80, 40–56.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 33
THE ARGOSTOLI TEST-SITE: DATA AND FIRST RESULTS FROM THE POST-SEISMIC SURVEY AND
THE ARGONET ARRAY
F. HOLLENDER 1, V. PERRON 1, S. SBAA 1, 2, A. SVAY 3 , A. IMTIAZ 4, 5, A. MARISCAL 4, P.-Y BARD 4 and the
T2.3 working team
1 CEA 2 EOST 3 CentraleSupélec 4 ISTerre 5 BRGM
Scope within SINAPS@
The present work concerns the instrumentation of a new test-site in a highly active seismic zone, in
order to acquire high quality data with high quality metadata for future validation of NL simulation
codes in 2D-3D environments. This work corresponds to SINAPS@ subtask 2.3.2.
As the site prior to the project (the Koutavos Park in the Argostoli area, Cephalonia Island, Greece) was
shaken by an earthquake sequence in early 2014, it was decided to carry a post-seismic survey in order
to acquire new data, including data about spatial variability and rotational motion on a nearby rock
outcrop, in order to feed SINAPS@ subtask 2.3.3 with complementary data sets.
The scientific context
Numerical simulation tools, such as those developed within the framework of the SINAPS@ project,
always need to be checked (validated) by comparing their predictions to real data. This is particularly
true for the modeling of non-linear soil behavior (2.3.2) and small scale spatial variability (2.3.3), as
existing models have many, often poorly constrained parameters. There exist however only very few
data sets with good enough metadata to keep the epistemic uncertainty of numerical simulation
approach small enough to be able to actually assess the performance of the simulation tool.
Considering the past benchmarking exercises (3D linear with SIGMA / CASHIMA projects, 1D Non-
Linear with SIGMA / SINAPS@ "Prenolin" projects), and the scope of the whole Task 2, the next step is
the validation of numerical simulation in 2D / 3D media, which is the goal of this new test–site located
in a seismically very active area.
Another capability of the new numerical simulation tool developed in this task is the possibility to
include randomly distributed heterogeneities with flexible correlation lengths and intensities. In order
to better tune the scattering properties, it is also needed to compare predicted and observed
characteristics of short scale spatial variability. The occurrence of a significant seismic sequence in
Cephalonia at the very beginning of the SINAPS@ project offered the opportunity to gather dense array
data, together with direct rotation data on rock and soft soil conditions.
Objectives of this work
The objective here is thus first to present an overview of the data sets which have been recorded by
both the temporary, post-seismic survey, and the semi-permanent vertical array installed in Summer
2015, and in addition to present the various analyses performed on the various data sets and the
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 34
results obtained so far. The temporary post-seismic array consisted of a/ 5 accelerometers installed
for about 17 months in the Koutavos Park area on quaternary sediments hunting for non-linear effects,
b/ a dense array of 21 sensitive velocimeters installed for about 5 weeks on a southeastern rock edge
of the valley in 4 concentric circles with radius 10, 30, 90 and 180 m, and c/ a (much less sensitive)
rotation sensor that has been collocated with the rock array center for a few weeks, and with the
Koutavos Park accelerometer for many months.
The current work progress
The large aftershock activity allowed to gather a very large number of high quality recordings from the
temporary network with signal-to-noise ratio exceeding 10 for at least one frequency, which have been
organized in a common data base with P and S wave pickings, earthquake ID, etc. to ease all possible
further processing and analysis work :
Accelerometric array, February 2014 – June 2015 : around 6000 events (Figure 1)
Velocimetric dense rock array, February – March 2014: around 2000 events
Rotation sensor : around 1400 events (as a whole, at three distinct locations)
On the other hand, for the semi-permanent Argonet vertical array, which is recording continuously
since July 2015 (see info on the http://www.institut-seism.fr/category/projet/sinaps/), it has been
decided not to systematically archive all the events, but only those exceeding some pre-established
threshold (0.1 cm/s2 at depth), which corresponds to about 300 events during the first year, including
3 with PGA exceeding 1 m/s2 at the surface sensor (Figure 2, Figure 3) .
Various processing and analysis has already been performed on these data bases, and several examples
and outcomes will be presented, amongst which:
Standard site / reference spectral ratios as a first hunt for NL behavior (Figure 4, which also exhibits significant amplification on the vertical component)
The frequency and distance dependence of coherency values on rock, and their comparison with soft soil coherency
The analysis of the composition of the rock wavefield through advanced array processing The relationship between peak rotation data and peak acceleration data, the comparison between
rock and soft sites, and other datasets at other places of the world (
Figure 5)
The ability of dense arrays to provide reliable estimates of the spatial derivatives (= rotational motion)
Further, ongoing, analyses will also be shortly described.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 35
Figure 1 : Location of events in the post-seismic survey database (regional scale on left), and corresponding distribution
in the magnitude – hypocentral distance plane (right. The red color intensity corresponds to the PGA.
Figure 2 : Distribution of the main surface recordings (exceeding 0.1 cm/s2 at down-hole sensor) in the magnitude –
distance plane for the first year of Argonet array. In the left frame, the color scale corresponds to the down-hole PGA,
on the right frame it corresponds to the surface PGA value.
Figure 3 : Example recordings of the vertical accelerometric array form the largest earthquake: M=6.5 in Lefkadas,
located around 60 km to the North-North East. The red time series correspond to the outcropping rock sensor on valley
eastern edge
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 36
Figure 4 : Example site to reference spectral ratios for the valley center (with a reference on the outcropping rock site
on valley eastern edge). Horizontal component (quadratic mean of the two horizontal components) on the left, and
vertical component on the right. The color code corresponds to the density of results (from dark blue: only one data
from one earthquake, to bright red: many data form many earthquakes).
Figure 5: Comparison of the Argostoli rotation data (blue dits for rock and red dots for soft soil site), with the (few) other
available data sets in the world, in terms of relationships between PGR (rad/s) and PGA (g): the Argostoli site already
has the world highest rotation value (around 0.007 rad/s) !
Analysis of rotational data recorded during the Sinaps@ Kefalonia post-seismic survey 37/73
Figure 25: Comparison of all the Argostoli data with the bibliographic data in terms of relationship between PGR and PGA.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 37
3D SEISMIC WAVE PROPAGATION IN HETEROGENEOUS NON-LINEAR MEDIA BY SPECTRAL
ELEMENT METHOD
F. GATTI, L. DE CARVALHO PALUDO, A. SVAY, F. LOPEZ-CABALLERO, R. COTTEREAU, CentraleSupélec
Abstract
A proper estimation of the earthquake ground motion is required at the site of interest to assess the
seismic structural response and draw hazard maps. Nowadays, due to the increased computational
power, complex structural time-domain analysis is possible. This requires several realistic time-
histories as input motions. Unfortunately, the number of recorded seismic signals of a good quality,
albeit increasing, is still limited. The 3D numerical simulation of strong ground motions at a regional
scale has therefore become the leading and most reliable tool to reproduce artificial wave-forms [6].
Those earthquake scenarios are very appealing to better understand the physics of strong ground
motions (especially in near-field conditions). Indeed, they usually encompass the causative fault
mechanism, the geological configuration and the regional topography.
Advances in informatics have made large 3D simulations within reach but demands the use of memory-
distributed clusters. The efficient use of these resources requires a numerical implementation tailored
to use a massively-parallel architecture (as described in [5]). That is the case of our simulation code,
SEM3D, by CentraleSupélec, the Institut de Physique du Globe de Paris (IPGP) and the Commissariat à
l’Énergie Atomique et aux énergies alternatives (CEA). Recently we extended the capabilities of SEM3D
to deal with two more ingredients: heterogeneous materials and non-linear behaviour.
Soil Heterogeneity (HET)
The heterogeneous nature of the earth-crust is usually described by means of a statistical approach
([1, 3, 8]). In this sense, the heterogeneous mechanical property X is point-wise randomly generated.
We suppose its statistics are fully defined by its Average, Coefficient of Variation, Correlation Model,
Correlation Length and First-order marginal. In this work, an heterogenous shear-wave velocity field
(VS (𝑥)) was considered in the uppermost layer, resulting in a heterogeneous behaviour tensor
(assuming constant Poisson’s ratio). This description is able to generate 3D elasticity tensor fields even
of a random anisotropic material [10]. In doing so, we faced the difficulty of sampling 3D
heterogeneous fields when the size of the domain is much bigger than the characteristic size of the
fluctuations. To overcome this difficulty the generation of the random-field was efficiently
implemented into a FORTRAN95 library exploiting MPI protocols. In this implementation we generate
Gaussian random fields using a spectral technique [9] and we apply a point-wise non-linear map (e.g.,
the Rosenblatt transform [7]) into the desired first-order marginal density, for example log-normal, for
the mechanical property at hand. The library applies domain subdivision techniques to keep a high
parallel scalability on massively-parallel architectures. The time spent to generate the heterogeneous
properties is much smaller than the simulation time.
Soil Non-linearity (NL)
A non-linear constitutive relationship was implemented into a dynamic time-marching scheme
featuring the spectral element semi-discretization of the domain. The hysteretic material behaviour of
the uppermost soil layer was modelled by exploiting a J2-plasticity framework characterized by a von
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 38
Mises failure criterion and a non-linear Frederick-Armstrong kinematic hardening rule [4]. Stress-strain
relationship is determined by means of three major parameters: the von Mises yield limit σyld and two
parameters defining the non-linear kinematic hardening, called respectively C and κ. The isochoric
elastic-plastic constitutive relationship schematically represented in Figure 1.
Figure 1: Schematic representation of the non-linear Frederick-Armstrong hardening rule. On the left: norm of the
deviatoric stress //σD// (green) and back-stress //X// (blue) as a function of the plastic-strain norm //εpl//. On the
right: representation of the von Mises yield surfaces in the deviatoric plane. The orange circle represents the active yield
surface, interested by a rigid translation in the deviatoric plane; the green circle limits the orange one, representing the
critical state. At this point, the norm of the deviatoric stress tensor //σD// equals 𝜎𝑦𝑙𝑑 + 𝐶
𝜅. The blue circle represents
the back-stress saturation surface.
Figure 2 shows an example of comparison between experimental and numerical𝐺
𝐺𝑚𝑎𝑥− 𝛾 − 𝐷 curves,
obtained with the Frederick-Armstrong hardening model.
The non-linear hardening rule defines its center shift in the deviatoric plane: critical state (i.e.
hardening saturation) is achieved when the stress point lies on the green surface (yield limit surface),
i.e. when the active yield surface intersect the limit one.
A first use of the SEM3D along with heterogeneous material properties is presented in [2]. Several
simulations were performed to obtain the spatial coherency curves. A classical case study was
considered: a seismic wave generated by a point-wise non-spherical radiative source (i.e. a localized
double-couple representing a pure-shear dislocation, in far field approximation) propagating towards
the surface throughout a 3-layered domain is studied (see Figure 3). The bottom layer represents the
bedrock, the top layer represent the shallow soil and the middle layer is a transition between the two.
Some results in terms of coherency and presented in Figure (4): good agreement between simulated
and analytical results is shown.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 39
Figure 2: Experimental 𝐺
𝐺𝑚𝑎𝑥
− 𝛾 − 𝐷 curves (at the Kashiwazaki-Kariwa Nuclear site (Japan), Tokyo Electric Power
Company - 2008) compared to the numerical results obtained by the described non-linear model.
Figure 3: Reference 3-layered domain considered in analysis in the homogeneous (left) and heterogeneous (right) case.
From bottom to top, layers height: 2000m, 600m and 300m. Cross section 600m × 600m. Correlation length 20m in all
directions
Figure 4: Coherency curves obtained using SEM3D in heterogeneous media.
To extended the mentioned analysis, we focused on the effect of the soil heterogeneity and non-
linearity of uppermost layer onto the surface wave-field and therefore on the coherency curves. To
this end, four simulations were performed and compared herein, by either considering the top-layer
as (1) homogeneous linear-elastic (HOM-EL), (2) heterogeneous linear-elastic (HET-EL), (3)
homogeneous non-linear (HOM-NL) and (4) heterogeneous non-linear (HET- NL) material.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 40
Acknowledgement
This work, within the SINAPS@ project, benefited from French state funding managed by the National
Research Agency under program RNSR Future Investments bearing reference No. ANR-11-RSNR-0022-
04. The research reported in this paper has been supported in part by the SEISM Paris Saclay Research
Institute.
References
Keiiti Aki and Bernard Chouet. Origin of coda waves: Source, attenuation, and scattering effects, 1975.
C. BergeThierry, A. Svay, A. Laurendeau, T. Chartier, V. Perron, C. Guyonnet-Benaize, Kishta, R. Cottereau, F.
Lopez-Caballero, F. Hollender, B. Richard, F. Ragueneaud, Voldoire, F. Banci, I. Zentner, N. Moussallam, M.
Lancieri, P.Y. Bard, S. Grange, S. Er- licher, P. Kotronis, A. Le-Maoult, M. Nicolas, J. Regnier, F. Bonilla, and N.
Theodoulidis. Toward an integrated seismic risk assessment for nuclear safety improving current french
methodologies through the sinaps@ research project. Nuclear Engineering and Design, 2016.
Y. Capdeville, L. Guillot, and J.-J. Marigo. 2d non-periodic homogenization to upscale elastic media for P-SV
waves. Geophys. J. Int., 182(2):903–922, 2010.
C. O. Frederick and P. J. Armstrong. A mathematical representation of the multiaxial Bauschinger effect.
Materials at High Temperatures, 24(1):1–26, 2007.
Dominik Göddeke, Dimitri Komatitsch, and Matthias Möller. Finite and Spectral Element Methods on
Unstructured Grids for Flow and Wave Propagation Methods, chapter 9, pages 183–206. Springer, July 2014.
R. Paolucci, I. Mazzieri, C. Smerzini, and M. Stupazzini. Physics -Based Earthquake Ground Shaking Scenarios
in Large Urban Areas. In Atilla Ansal, editor, Perspectives on European Earthquake Engineering and
Seismology, volume 1 of Geotechnical, Geological and Earthquake Engineering - Vol.34, pages 331–359.
Springer, 2014.
Murray Rosenblatt. Remarks on a multivariate transformation. The annals of mathematical statistics,
23(3):470–472, 1952.
L. Ryzhik, G. Papanicolaou, and J. B. Keller. Transport equations for elastic and other waves in random media.
Wave Motion, 24:327–370, 1996.
Masanobu Shinozuka and George Deodatis. Simulation of stochastic processes by spectral representation.
Applied Mechanics Reviews, 44(4):191–204, 1991.
Q.-A. Ta, D. Clouteau, and R. Cottereau. Modeling of random anisotropic elastic media and impact on wave
propagation. Europ. J. Comp. Mech., 19(1-3):241–253, 2010.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 41
WP 3: Seismic Behavior of
structures and components
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 42
PROGRESS OF THE WP3 “SEISMIC BEHAVIOR OF STRUCTURES AND COMPONENTS” IN 2016
F. RAGUENEAU, ENS-Cachan
Context and main objectives of WP3
The behaviour analysis of structures and equipment necessarily needs the modelling of the different
mechanisms involved in the seismic wave’s transmission from the soil interaction to the equipment.
During these last two decades, important and numerous improvements occurred within the field of
modelling the material and structural elements degradation due to earthquake and cyclic loading.
When 3D behaviours are expected, the numerical cost is often very high. The necessity to account for
non-deterministic features of loadings (hazard) and material scattering guides one to the development
of original approaches allowing parametrical analysis needing a large number of simulations, keeping
a refine description of local mechanical behaviours. In such a way, three main tasks have been
scheduled regarding the modelling of structures and equipment.
The first task is dedicated to the analysis of experimentations allowing for models calibration or
validation. The second task is focused on the development of numerical models themselves. At last,
the third task handles the industrial structures behaviour regarding the seismic risk.
Concerning the advancements of 2015-2016, three PhD students defended their thesis during this
period (enhanced plate modelling, enhanced beam modelling and reduced order models for reinforced
concrete modelling).
Main works focus on task 2 for the development of simplified models and task 3 to estimate the
drawbacks of seismic base isolation.
Main advancements (2015/2016)
Regarding the three major tasks, the following enhancements have been performed:
Task 1: Experimentations comparisons and model updating
To improve the interaction with the work package 5, an experimental work is concerned in this task to
a common issue: the identification of dissipation in RC structures. This work (T. Heitz PhD thesis) will
help the WP3 teams to better identify or validate their models. The description of this work is deeply
presented in WP5. The experimental campaign has begun (23 RC beams subject to forced vibration on
different modes and their couplings).
Task 2: Numerical modelling for structures g
Within the development of numerical models adapted to perform structures computations, two main
issues need to be tackled: the development properly-said of mechanical models incorporating different
nonlinear mechanisms likely to occur and the ability for existing tools to account for variability inherent
in structures and materials allowing the interaction of the different sources of uncertainty.
To ensure good representation of models regarding the failure criteria (frequency decrease, inter-
storey drift, crack openings), different kind of models are developed.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 43
Based on 1D multifibre assumptions framework, the PhD thesis of I. Bitar (common work between UJF
Grenoble and ECN Nantes) aims at describing cracks and failure within the beam theory by adding
higher order kinematic descriptions.
For more complex situations, the 2D description of structural elements is a mandatory step.
Concerning the beam representation, the necessity to account for shear and torsion is an important
task that has been tackled in the project (PhD thesis defence last October of S. Capdevielle). If
reinforced concrete material is used, the lattice representation of steel reinforcement bars has to be
introduced in the mechanical model. The main issue of the PhD of N. Khoder is to introduce the
presence of transversal reinforced within the framework of beam theory. New developments will be
presented during the session.
Model reduction techniques such as PGD framework is used in the context of reinforced concrete
structures to allow parametrical studies. The variability of materials and loadings will be included to
evaluate and order them regarding their influence on the global structure vulnerability. The PhD of M.
Vitse (ENS-Cachan) solves this problem by building virtual charts (similar to abaci) of solutions
associated to the parametric problem, using a two steps scheme: the first step is the (possibly
expensive) computation of the abaci, for each set of parameters (offline step). The second step consists
in a particularization of the solution for a given set of parameter values (online step), as part of an
optimization study or dimensioning of a structure. The main point of this work lies in the construction
of those charts during the offline step. The PhD defence will take place next December, the last results
will be presented during the session.
Task 3: Risk mitigation
Seismic base isolation is one of the most efficacious seismic mitigation methods. Though there
is a large number of base isolated structures worldwide, only very few applications of the
methods to nuclear plants exist. Up to now, these applications have been realized in sites with
medium seismicity. In general, seismic isolation, results in seismic response significantly lower
than that of conventional structures, in some cases, undesirable effects may occur which reduce
the expected benefit. These effects are studied and pointed out in SINAPS@ and attention is drawn
to aspects which should not be overlooked during the design and analysis of base isolated nuclear
plants.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 44
ENHANCEMENT OF MULTIFIBER BEAM ELEMENTS IN THE CASE OF REINFORCED CONCRETE
STRUCTURES FOR TAKING INTO ACCOUNT THE LATERAL CONFINEMENT OF CONCRETE DUE TO
STIRRUP
N. KHODER 1, S. GRANGE 2, Y. SIEFFERT 1
1 Université de Grenoble Alpes, Laboratoire 3SR (Sols, Solides, Structures-Risques), Grenoble, France 2 Université Lyon, INSA de LYON, SMS-ID, F-69621, Lyon, France
Scope within SINAPS@
The present research is carried out within the framework of the third work package of the SINAPS@
project, and is related to Task 3.2.1.2.𝛽 dealing with numerical modelling for structures. It focuses on
the development of a multifiber beam element formulation able to account for transversal
reinforcements and correctly reproduce the behavior of reinforced concrete beam elements with the
effect of confined concrete.
The scientific context
To assess the seismic vulnerability of existing reinforced concrete structures, a large number of degrees
of freedom is involved. Consequently, efficient numerical tools are required. An alternative to full solid
models, which are too costly, is the use of multifiber beam elements. The latter one combines the
advantages of high computational speed with an increased accuracy for nonlinear materials. It consists
on adding a two dimensional section at the Gauss point of the element. In addition, a variety of
approaches have been developed to try to introduce shear effects, such as those proposed by [Le
Corvec et al., 2012], but whose model can’t be applied to reinforced concrete elements, as well as the
2D plane formulation of [Mohr et al., 2010]. More recently, [Capdevielle, 2016] and [Capdevielle et al.,
2016] developed a nonlinear multifiber beam model adapted to reinforced concrete, and which
provides robust results by the introduction of warping.
In the above mentioned works, the transverse steel is sometimes taken into consideration with
approximated manner or often not at all. However, as shown by some experimental tests conducted
by [Cusson and Paultre, 1995], the amounts of transverse reinforcement pilot significantly the behavior
of beam elements, especially under cyclic loading. Thus the next step would be to correctly model the
stirrups and transverse steel by enhancing the multifiber elements.
Objectives of this work
The main goal of this work is to investigate solutions for an enhanced multifiber beam element
accounting for vertical stretching of the cross section, occurring due to the presence of transverse
reinforcement. In this respect, the following steps have to be carried out:
1) Reach the transverse equilibrium of the section with our enhanced numerical model and
validate its efficiency in linear elastic phase.
2) Introduce the longitudinal and transverse steel reinforcements, then conduct the study in
the framework of elasticity and extend it to non-linear behavior using an adequate dilatant
constitutive law for reinforced concrete under monotonic and cyclic loadings.
3) Simulate the 3D behavior of structure concrete elements under cyclic loadings with the
enhanced model.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 45
The current work progress
A 2D multifiber Timoshenko beam, displacement-based, element has been developed. It takes into
account deformations due to shear and has higher order interpolation functions, developed by
[Caillerie et al., 2015] to avoid any shear locking phenomena. The main hypothesis considered herein
is that the full displacement of any fiber in the cross-section can be approximated by the sum of the
plane section displacement field (up), obtained from Timoshenko’s beam theory, and a new
displacement field that enables the section to distort and warp (uw). The distortion field has one
transverse component which stands for the distortion of the section.
Figure 1 shows the beam element introduced in our model, with internal degrees of freedom, while
on the right, Figure 2 presents the working principle of our enhanced multifiber beam element.
Figure 1: [Caillerie et al., 2015] element Figure 2: Principle of the enhanced beam element
The efficiency of the proposed modeling strategies is tested with results obtained from tension and
flexion tests conducted on pure concrete. Figure 3 shows the variation of transversal stresses 𝜎𝑦𝑦
computed at each fiber, with respect to the length of the beam. It’s a cantilever beam subjected to
vertical loading at its free end. Points in red are the results obtained with 2D FE model and those in
blue are the one obtained with our enhanced multifiber element. These results prove that the
equilibrium state is reached and the enhanced developed model works well in the linear elastic phase.
Figure 3: Variation of the transversal stresses 𝜎𝑦𝑦 with respect to the length of the beam: Comparison between 2D finite element model (red) and our enhanced multifiber beam element (blue).
Figure 4: Transversal stresses 𝜎𝑦𝑦 computed at each fiber of section 1
If we take a cross section situated near the fixed end, then the variation of the transversal stresses 𝜎𝑦𝑦
computed at each fiber of this section can be presented on figure 4.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 46
The ongoing works aim at modelling transverse steel as well as implementing an existing constitutive
law for concrete under monotonic and cyclic loadings and hence study the non-linear response of
structural elements subjected to transverse shear.
References
D. Caillerie, P. Kotronis and R. Cybulski. A Timoshenko finite element straight beam with internal degrees of
freedom. International Journal For Numerical And Analytical Methods In Geomechanics (2015)
D. Cusson and P. Paultre. Stress-Strain model for confined high-strength concrete. American Society Of Civil
Engineers (1995)
S. Capdevielle. Introduction du gauchissement dans les éléments finis multifibres pour la modélisation non
linéaire des structures en béton armé. PhD Thesis, Université de Grenoble Alpes, 2016
S. Capdevielle, S. Grange, F. Dufour, C. Desprez. A multifiber beam model coupling torsional warping and
damage for reinforced concrete structures. European Journal of Environmental and Civil Engineering, 20(8),
2016
S. Mohr, J. M. Bairán and A. R. Marí. A frame element model for the analysis of reinforced concrete structures
under shear and bending. Engineering structures, 32(12): 3936-3954, 2010
V. Le Corvec. Nonlinear 3d frame element with multi-axial coupling under consideration of local effects. PhD
Thesis, University of California, Berkeley, 2012.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 47
A NEW FINITE ELEMENT BEAM TO SIMULATE THE BEHAVIOR OF REINFORCED CONCRETE
STRUCTURES TILL FAILURE
I.BITAR1, P. KOTRONIS1, N. BENKEMOUN2, S. GRANGE3
1 Ecole Centrale de Nantes, Institut de Recherche en Génie Civil et Mécanique (GeM) 2 IUT Saint Nazaire, Université de Nantes, Institut de Recherche en Génie Civil et Mécanique (GeM) 3 Univ Lyon, INSA de Lyon, SMS-ID, F-69621, Lyon, France
Scope within SINAPS@
The work presented herein contributes to the work package 3 of SINAPS@ project (task 3.2.1.2 (α)). It
focuses on the development of a new finite element beam able to describe the behavior of reinforced
concrete beam-like structures till failure.
The scientific context
Different kinematic assumptions are used in beam analysis in order to simplify the global equilibrium
equations and to reduce the required number of degrees of freedom. The Timoshenko beam theory
considers that plane sections remain plane but not necessary normal to the deformed axis. The
advantage of this theory is that it takes into account the influence of shear strains (contrary to Euler-
Bernoulli assumption).
In a multi-fiber finite element beam, the beam section is divided into several fibers with specific
stress/strain relations (Owen & Hinton, 1980). Finite elements of this type are efficient for various
applications in civil engineering: nonlinear analysis of beam type or bearing wall structures with non-
homogenous sections (e.g. reinforced concrete) (Kotronis, Ragueneau, & Mazars, 2005), (Grange,
Kotronis, & Mazars, 2008), arbitrarily geometrical plane or hollow shape sections (Grange, Botrugno,
Kotronis, & Tamagnini, 2011), (Desprez, Kotronis, & Mazars, 2015) submitted to bending, shear or
torsion (Mazars, Kotronis, Ragueneau, & Casaux, 2006), Soil Structure Interaction problems (Grange,
Botrugno, Kotronis, & Tamagnini, 2011), vulnerability assessment cases (Desprez, Kotronis, & Mazars,
2015) and Fiber-Reinforced Polymer retrofitting (Desprez, Mazars, Kotronis, & Paultre, 2013).
Objectives of this work
This work aims to develop a new multi-fiber displacement based beam finite element based on the
Timoshenko theory to simulate failure of reinforced concrete structures subjected to static or dynamic
loadings. The new displacement-based finite element Timoshenko beam proposed by (Caillerie,
Kotronis, & Cybulski, 2015) is adopted. This formulation uses shape functions of order three for the
transverse displacements, two for the rotations and an additional internal node. This results to a finite
element free of shear locking. In (Bitar, Grange, Kotronis, & Benkemoun, 2016), a comparative study
of this novel formulation with other formulations found in the literature (Guedes, Pegon, & Pinto,
1994), (Friedman & Kosmatka, 1993) is shown. For a reinforced concrete beam, the multi-fiber section
is divided into fibers of concrete and steel. The concrete fibers behave according to a damage
mechanics law and the steel fibers have an elasto-plastic behaviour.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 48
The Embedded Discontinuity Approach (EDA) based on the strong discontinuity is adopted to simulate
failure and get information concerning cracks. The EDA allows the incorporation of localized failure
into standard displacements-based finite elements using discontinuities variables (Simo, Oliver, &
Armero, 1993). The fibers are enhanced in order to describe concrete cracks and the development of
plastic hinges (Jukic, Brank, & Ibrahimbegovic, 2014). The strong discontinuity is introduced by adding
a jump in the displacement field. Accordingly, additional shape functions are needed to interpolate the
displacement jump within the enhanced finite element. Hence, the enhanced axial displacement 𝑢𝑥
reads:
where 𝑑𝑥 is the continuous nodal displacement, 𝐻𝑥𝑑(𝑥) the Heaviside function, 𝜙(𝑥) an additional
function to preserve the displacement continuity at the nodes between elements and �̿�(𝑥) the
displacement discontinuity. Therefore, the axial strain 휀𝑥 can be deduced as:
where 𝛿𝑥𝑑 is the Dirac function.
A cohesive law linking the axial stress and the displacement jump by a linearly decreasing relation
characterizes the materials behaviour at the discontinuity, which allows capturing the released
fracture energy. The variational formulation is presented in the context of the incompatible modes
method. It leads to the following system of equilibrium equations:
Where �̿� the number of elements containing at least one enhanced fiber and �̿�𝑓(𝑒) the number of
enhanced fibers of the element 𝑒. The first equation refers to the global level of the structure and the
other occurs for each enhanced fiber. The additional modes are statically condensed at the fiber level
in order to keep unchanged the architecture of the finite element code. The reader is referred to (Bitar,
Benkemoun, Kotronis, & Stéphane, 2016) for more information.
The current work progress
A cantilever concrete beam of 1m length submitted to an imposed transversal displacement at its free
end is given hereafter as a first example. The rectangular section (20 × 50 cm2) is divided into 10 fibers.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 49
A Timoshenko finite element beam with linear shape functions is adopted for the numerical
applications (Guedes, Pegon, & Pinto, 1994). The mesh dependency of the global response is studied
as long as the behavior till failure. Figures 1 and 2 show that the fiber enhancement can limit the mesh
dependency at global scale.
Another example, considering this time a similar concrete beam with two steel bars (2Φ32𝑚𝑚) at the
top and the bottom of the section is presented. Again, Figures 3 and 4 depict the global response till
failure before and after the fiber enhancement.
Currently, we are working on the introduction of the fiber enhancement in the new finite element
beam formulation (Caillerie, Kotronis, & Cybulski, 2015). The corresponding discontinuity kinematics
is also studied. This latter is controlled by the choice of the enhancement function associated to the
discontinuity variable.
Bibliography
Bitar, I., Benkemoun, N., Kotronis, P., & Grange, S. (2016). A novel multi-fiber Timoshenko beam finite
element formulation with embedded discontinuities to describe reinforced concrete failure under static
loadings. Proceedings of the 9th International Conference on Fracture Mechanics of Concrete and Concrete
Structures. Berkeley.
Bitar, I., Grange, S., Kotronis, P., & Benkemoun, N. (2016). {A comparison of displacement-based Timoshenko
multi-fiber beams finite element formulations and elasto-plastic applications}. European Journal of
Environmental and Civil Engineering, 8189(July), 1-27.
Caillerie, D., Kotronis, P., & Cybulski, R. (2015). {A Timoshenko finite element straight beam with internam
degrees of freedom}. International Journal for Numerical and Analytical Methods in Geomechanics, 39(16),
1753-1773.
Desprez, C., Kotronis, P., & Mazars, J. (2015). {Seismic vulnerability assessment of a RC structure before and
after FRP retrofitting}. Bulletin of Earthquake Engineering, 13, 539-564.
Desprez, C., Mazars, J., Kotronis, P., & Paultre, P. (2013, mar). {Damage model for FRP-confined concrete
columns under cyclic loading}. Engineering Structures, 48, 519-531.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1-20
0
20
40
60
80
100
120
Imposed transversal displacement
Fle
xure
mom
ent
Figure 1: Concrete beam
n=2
n=5
n=10
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
10
20
30
40
50
60
70
80
90
Imposed transversal displacement
Fle
xure
mom
ent
Figure 2: Enhanced conrete beam
FLI NE=2 avec DISC
FLI NE=3 avec DISC
FLI NE=5 avec DISC
FLI NE=10 avec acier
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
50
100
150
200
250
300
350
Imposed tranversal displacement
Fle
xure
mom
ent
Figure 3: RC multi-fiber beam
FLI NE=2
FLI NE=5
FLI NE=10
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
50
100
150
200
250
300
350
Imposed transversal displacement
Fle
xure
mom
ent
Figure 4: RC multi-fiber enhanced beam
FLI NE=2
FLI NE=5
FLI NE=10
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 50
Friedman, Z., & Kosmatka, J. B. (1993). {An imrpved two-node Timoshenko beam Finite Element}. Computer
& Structures, 47(3), 473-481.
Grange, S., Botrugno, L., Kotronis, P., & Tamagnini, C. (2011). {The effects of Soil-Structure Interaction on a
reinforced concrete viaduct}. Earthquake Engineering {\&} Structural Dynamics, 41(11), 1549-1568.
Grange, S., Kotronis, P., & Mazars, J. (2008). {Numerical modelling of the seismic behaviour of a 7-story
building: NEES benchmark}. Materials and Structures, 42(10), 1433-1442.
Guedes, J., Pegon, P., & Pinto, A. V. (1994). {A Fibre/Timoshenko beam element in Castem 2000}. Special
publication Nr. I.94.31, Applied Mechanics Unit, Institute for Safety Technology, Joint Research Centre,
Commission of the European Communities, I-21020 ISPRA (VA).
Jukic, M., Brank, B., & Ibrahimbegovic, A. (2014). {Failure analysis of reinforced concrete frames by beam
finite element that combines damage, plasticity and embedded discontinuity}. Engineering Structures, 75,
507-527.
Kosmatka, J. B. (1995). {An improved two-node finite element for stability and natural frequencies of axial-
loaded Timoshenko beams}. Computers and Structures, 57(1), 141-149.
Kotronis, P., Ragueneau, F., & Mazars, J. (2005, jul). {A simplified modelling strategy for R/C walls satisfying
PS92 and EC8 design}. Engineering Structures, 27(8), 1197-1208.
Mazars, J., Kotronis, P., Ragueneau, F., & Casaux, G. (2006, nov). {Using multifiber beams to account for shear
and torsion}. Computer Methods in Applied Mechanics and Engineering, 195(52), 7264-7281.
Owen, D. R., & Hinton, E. (1980). {Finite elements in plasticity}. Pineridge press.
Pegon, P. (1994). {A Timoshenko simple beam element in Castem 2000}. Special publication Nr. I.94.04,
Applied Mechanics Unit, Institute for Safety Technology, Joint Research Centre, Commission of the European
Communities, I-21020 ISPRA (VA), Italy.
Simo, J. C., Oliver, J., & Armero, F. (1993). {An analysis of strong discontinuities induced by strain-softening
in rate-independent inelastic solids}. Computational Mechanics, 12(5), 277-296.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 51
MODEL-ORDER REDUCTION FOR THE PARAMETRIC ANALYSIS OF DAMAGE IN REINFORCED
CONCRETE STRUCTURES
M. VITSE, D. NÉRON, P.-A. BOUCARD – LMT Cachan, ENS Paris-Saclay
Scope within SINAPS@
The study of the variability in materials is an open subject in computational mechanics when
dimensioning structures. This variability can be the result of the manufacturing of the material
(choice of the constitutive elements, manufacturing process), but also can come from the
description its physical model. Hence, it is easy to understand that the stronger the uncertainties, the
more difficult the computations can get, whichever the scale considered (from a micro-scale point
of view to the scale of an engineering structure). This is especially true for reinforced concrete,
which has been (and still is) extensively used for civil engineering purposes, even if its long-
term mechanical behavior and the degradation mechanisms that result from the aging of the
structures are not perfectly known or even understood. Chemical reactions, mechanical
degradations can occur along the life of a given structure, which makes it even more difficult,
even with the advanced numerical tools available nowadays, to predict the behavior of a building
to a certain loading.
Our interest lies in the fact that when studying the response of a reinforced concrete structure to
seismic loading, one can consider several time scales, all having an important impact on the final
behavior of the structure: (i) early-age concrete’s mechanical properties evolve quickly due to the
hydration of the concrete paste, with highly exothermic reactions, which can impact the stress and
strain state of the structure. The long-term effects of those mechanisms on the global mechanical
response of a structure are not well known in particular for massive constructions, for which the
high gradients of temperature can initiate cracks within the structure; (ii) inner and desiccation
creeps, which are independent from the hydration, have long term effects which impact the strain
state of the structure; (iii) chemical and mechanical degradations may occur over the life of the
structure, weakening it and changing its mechanical properties.
The seismic hazard itself is also not well understood and modeled. The problem faced here is the
following: one has to compute the mechanical response, at time t, of a structure for which the
loading conditions are defined over a short time scale, but has to take into account the whole
history of the structure (sometimes several decades in the case of massive civil engineering
structures). Some models do exist to do so, however their numerical implementation can often be
complex. Yet, this works does not intend to provide advances in the modeling of those long-term
mechanisms previously exposed, but rather focuses on the numerical strategy to solve such
problems. To do so, we propose to consider a reinforced concrete structure at a given instant for
which some of the mechanical properties face a variability on some of their coefficients such that
whichever the degradation mechanisms that may have occurred over the life of the structure, we
will only consider their potential effects on (loading or material) parameters solely defined by an
interval of variation. The aim here is not to compute the response of the structure over several
months, years or decades but to assess the mechanical response of such family of structures
under cyclic loading conditions, with the behavior of the material highlighting both strong
nonlinearities and variabilities.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 52
The scientific context
Despite the rise of algorithms and technologies for high-performance computing, the resolution of
such parametric problems is still an issue as the cost of the computation (both CPU and storage)
increases exponentially with the number of parameters in the formulation of the problem (dozens,
potentially hundreds of parameters). Among the classes of methods that arose over the last
decades to circumvent that issue, we focus on mode-reduction techniques, based on a separated-
variable representation of the unknown fields (let say the displacement field), which show the
double advantage of reducing the computation cost of parametric problems and providing a
good framework for the storage of the solution (and its post-treatment). Those methods, unlike meta-
modeling techniques, do not intend to simplify the model itself but rather to provide a framework
for approximating the solution of the mechanical problem associated with a rather complex physical
model. The choice of such techniques is especially motivated by the fact that the number of
uncertain parameters, as well as their intervals of variation, can be very large, which might
make probabilistic approaches too computationally expensive to carry out.
Model-order reduction methods usually rely on a two phase approach: the first step consists in
building the reduced-order model, seen in our case as a database of solutions (which will also be
referred to as “virtual charts” of solutions, an analogy to the engineering charts used in the
past), which can be expensive when dealing with nonlinear parametric problems but which
encompasses all the different possible occurrences, whereas the second step consists in
particularizing the solution for a given purpose, a strategy which makes conceivable the possibility
of obtaining rapidly (in real time) the mechanical response of the system. Among those techniques,
the proper generalized decomposition (PGD) method [Ladevèze, 1999; Chinesta et al., 2010] has
been quite extensively used over the last decade in numerous areas such as fluid dynamics, study
of composite structures, real- time surgery,… , with a recent emphasis on verification and validation
aspects. It enables to directly build a separated-variable approximation of the solution fields using
a greedy algorithm, without any prior information on this solution. However, such approach used
alone is not well-suited for solving nonlinear problems as the ones encountered in civil engineering
calculations. Linearization strategies, such as Newton-Raphson techniques, the asymptotic numerical
method or the LATIN method [Ladevèze, 1999] can be used to solve this issue. The LATIN method
relies on a simple but powerful concept: a separation of the linear/nonlinear problems and an
alternative resolution of the two sub-problems. This approach, coupled with PGD, is very attractive
as it iteratively provides an approximation of the solution field under a separated representation,
which can be enriched until reaching a given quality criterion. It has been extensively used in the
last couple of decades in numerous research fields, such as the study of multi-scale or multi-
physics problems for example. Considering the parametric aspect of the equations, the LATIN–
PGD association is fairly natural in this context and provides a great framework for parametric
studies, based for example on successive enrichments of solutions (multi-parametric strategy
[Boucard and Ladevèze, 1999]).
Objectives of this work
This LATIN–PGD approach is at the core of the developments presented in this work, as it
enables to tackle the two problems at stake in our study: on the one hand, damage mechanisms
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 53
are often modeled by strongly nonlinear equations, whereas on the other hand some equations
of the mechanical model depend on parameters which may have an important variability.
The strategy chosen here to take the parametric dependency into account in the LATIN–PGD
framework is however different than [Boucard and Ladevèze, 1999] and two contributions are
presented in this work. The first one is a new extension of the classical LATIN–PGD algorithm to
parametric studies for which, unlike the multi-parametric strategy, the parameters are considered
as extra-variables for the PGD decomposition. This enables to work on larger parametric spaces,
as the basis is iteratively enriched to take into account all the different occurrences (sometimes
even nonphysical / irrelevant ones). The second contribution is the numerical implementation of
this strategy, applied to a damage model with unilateral effect.
Regarding the objectives of the SINAPS@ project, this work does not intend to provide criteria related
regarding the mechanical strength of civil engineering structures under cyclic loading conditions. Our
objective is to provide a tool to accurately and quickly compute databases of solutions, which take
into account all the different sets of parameters, which can be afterwards used by engineers for
design purposes for example.
The current work progress
The feasibility of this extension was shown in [Vitse et al., 2014] on a simple 1-D heat evolution
problem, with a variability on the thermal conductivity. This present work provides a more general
framework for the resolution of nonlinear parametric problems and presents its application to the
analysis of reinforced concrete structures, with an isotropic damage model and unilateral effect
[Richard and Ragueneau, 2012]. The algorithmic developments are implemented into a 3-D Matlab
demonstrator. Numerical examples are carried out on reinforced concrete beams for different types
of loading (tension, 4-points bending – see Fig.1, cyclic bending) and some of those results have
been submitted in [Vitse et al, 2016]. The work done during this PhD will be defended in December
2016.
Fig. 1: 4-points bending test – damage map – particularization of the damage field associated with a variability on
the amplitude of the prescribed displacement (µ2) as well as the Young modulus of the concrete medium (µ3) for a
cyclic prescribed displacement (at a time step (µ1=90 s) )
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 54
References
[Boucard and Ladevèze, 1999] Boucard, P.-A. and Ladevèze, P. A multiple solution method for
non-linear structural mechanics. Mechanical Engineering, 50(5):317–328, 1999.
[Chinesta et al., 2010] Chinesta, F., Ammar, A., and Cueto, E. Recent advances and new challenges in
the use of the proper generalized decomposition for solving multidimensional models. Archives of
Computational Methods in Engineering, 1–24, 2010.
[Ladevèze, 1999] P. Ladevèze. Nonlinear computational structural mechanics: new approaches and
non- incremental methods of calculation, Springer, 1999.
[Richard and Ragueneau, 2012] B. Richard, F. Ragueneau. Continuum damage mechanics based
model for quasi-brittle materials subjected to cyclic loadings: formulation, numerical implementation
and applications, Engineering Fracture Mechanics, Elsevier, 383-406, 2012.
[Vitse et al., 2014] M. Vitse, D. Néron, P.-A. Boucard. Virtual charts for solutions of nonlinear
parametrized equations, Computational Mechanics, Springer, 1529-1539, 2014.
[Vitse et al, 2016] M. Vitse, D. Néron, P.-A. Boucard. Damage prediction and its variability through PGD
models: application to reinforced concrete structures, International Journal for Numerical Methods
in Engineering, Springer, submitted, 2016.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 55
DETERMINATION OF ULTIMATE CAPACITY OF ELECTRICAL EQUIPMENT FOR SUBSEQUENT
INDUSTRIAL USE
N. MOUSSALLAM, C. GRAF, B. BOUDY, AREVA NP
Scope within SINAPS@
This work is part of task 3.3.1 of the SINAPS@ project, aimed at identifying equipment sensitive to
earthquake loads and determining its ultimate behavior.
An identification of generic systems and components driving the overall fragility of a Nuclear Power
Plant (NPP) was performed in 2014. Subsequent developments were made to improve analyses
methods for equipment such as cranes, fuel racks and other sliding, rocking or nonlinearly supported
equipment in 2015.
The present work focuses on qualification methods for electrical equipment, and on determination of
seismic demand to capacity ratio for this equipment.
The scientific context
Electrical equipment of NPP is generally qualified by tests on a shaking table. Most of the time,
complete electrical cabinets are tested at once and tests are not performed beyond the project
requirement. The equipment behavior beyond tested levels then remains unknown. As a result, when
performing probabilistic seismic risk assessment, electrical equipment generally ranks high in the list
of most significant contributors to the overall seismic risk of the plant.
Objectives of this work
This work aims at improving test procedures, modeling practices and analyses tools for being able to
perform fast and reliable estimates of seismic demand to capacity ratio for a large number of NPP
electrical devices.
The current work progress
Preliminary tests have been performed on a single electrical device on a shaking table in Erlangen.
From this test, a failure map of the device, expressed in terms of frequency and amplitude of excitation,
was composed. This failure map represents the device capacity.
In parallel, documentation has been gathered on electrical cabinets that were either tested or modeled
with finite elements. From this documentation, the transfer function associated with each cabinet can
be retrieved. These transfer functions, in combination with the floor seismic motion, produce the
seismic demand.
The general architecture of two scripts aimed at (a) generating the demand and (b) comparing this
demand to the capacity has been sketched. Work is ongoing on defining the content and format of the
databases that are to be accessed by these scripts.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 56
Figure 1 Single device test on AREVA NP shaking table (Erlangen)
Figure 2 Example of electrical cabinet analytical model
Figure 3 Automation of determination of demand to capacity ratios
2. Support structure transfer function 1. Floor excitation
3. Inner device failure map (seismic capacity)
Get Support Response
Support structure excitation (seismic demand)
Compare Demand to Capacity
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 57
UNDESIRABLE EFFECTS OF SEISMIC BASE ISOLATION
I. POLITOPOULOS, CEA
Scope within SINAPS@
Seismic base isolation is one of the most efficacious seismic mitigation methods. Though there is a
large number of base isolated structures worldwide, only very few applications of the methods to
nuclear plants exist. Up to now, these applications have been realized in sites with medium seismicity.
However, seismic isolation is even more interesting in the case of severe seismic events, because of its
capacity to filter the seismic excitation. In such cases, it is the most efficient passive technology to
protect both buildings and equipment.
Nevertheless, though, in general, seismic isolation, results in seismic response significantly lower than
that of conventional structures, in some cases, undesirable effects may occur which reduce the
expected benefit. These effects are studied and pointed out in SINAPS@ and attention is drawn to
aspects which should not be overlooked during the design and analysis of base isolated nuclear plants.
The scientific context
The research devoted to the improvement and consolidation of the seismic isolation techniques is an
important part of the research in the field of earthquake engineering. However, most of the work done
on seismic isolation focuses mainly on the impact of the seismic isolation on the main structure itself
and much less on the behaviour of equipment or components. On the other hand, the proper function
of nuclear facilities, during and after an earthquake, depends, to a great extent, on the capacity of their
components and equipment to withstand the earthquake-induced forces. Therefore, the objective of
the base isolation method is, mainly, to reduce the equipment demand and secondly to reduce the
strength demand in the building itself.
In fact, floor response spectra of rather stiff structures such as nuclear plants buildings, subjected to
horizontal excitation, will exhibit a local amplification in the vicinity of the isolation frequency but
spectral values for higher frequencies will be on a horizontal plateau. The value of this plateau is
approximately the maximum acceleration corresponding to a rigid body mounted on the isolation
bearings. This particular shape of floor response spectra is very attractive since it significantly reduces
the equipment forces for frequencies higher than about two times the isolation frequency and
simplifies the analysis and design of equipment.
Though the aforementioned ideal floor response spectrum is a good approximation of actual floor
spectra for many real seismically isolated structures, in some cases, depending on the frequency
content of the earthquake excitation and the dynamic characteristics of the structure, an amplification
of the non-isolated modes response (modes other than the lower isolated modes) arises that changes
this ideal picture and considerably reduces the expected benefit from seismic isolation. Such
amplification phenomena have been observed, amongst others, in recent studies of two experimental
nuclear reactors (RJH, ITER) in south France.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 58
Objectives of this work
The objectives of this activity are:
a) Study the sources of the amplification of the response of the non-isolated modes;
b) Investigate possible remedies to the above undesirable amplification.
The current work progress
Regarding objective a) the following three main sources of the amplification of the non-isolated modes
response have been investigated:
i) High base energy dissipation (linear or nonlinear viscous dampers, elasto‐plastic of friction
dissipative devices, etc.).
ii) Base rocking‐induced excitation due to horizontally propagating waves or to the scattered
motion in the case of embedded foundation.
iii) Coupling between vertical excitation and horizontal response in the case of asymmetric
superstructures.
The above amplification mechanisms were investigated theoretically and numerically based on simple
yet representative models. These studies gave qualitative and quantitative results which allowed us to
draw clear and meaningful conclusions which are helpful to guide the design phase and, also, to gain
a further insight into numerical simulations’ results.
Furthermore, the above conclusions have been illustrated and confirmed considering more detailed
finite element models also. For instance, Figure 1 illustrates the influence of base rocking induced
excitation on the floor response spectra of a partially embedded base isolated reactor building.
a) b)
Figure 1. Floor response spectra for 5% damping at the lower basement floor a) 3D FE model for soil
and superstructure b) axisymmetric FE soil model + 3DOF superstructure model.
Regarding objective b) the following alternatives to the commonly used isolation devices were
examined:
i) Full 3D isolation devices (i.e. flexible in the vertical direction also) without anti-rocking devices.
Such devices are quite effective against rocking excitation but they may have adverse effects
0 2 4 6 80
2
4
6
8
10
12
14
16
Frequency (Hz)
pseudoaccele
ration (
m/s
2)
multilayer soil embedded
cs=719 m/s
embedded
cs=719 m/s
surface
0 1 2 3 4 5 6 7 80
2
4
6
8
10
12
14
16
Frequency (Hz)
pseudoaccele
ration (
m/s
2)
multilayer soilembedded
cs=719 m/s
embedded
cs=719 m/s
surface
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 59
even in the case of “classical” translational excitation due to unavoidable eccentricities inducing
coupling between vertical and horizontal directions. Hence they are not recommended.
ii) A relaxation isolator combining a device such as a classical low damping rubber bearing (LDRB)
in parallel with a Maxwell element. This kind of isolator is proposed for seismic isolation for the
first time in this work. As shown in Figure 2, it is capable to reduce the base displacement of
isolated structures (significant reduction of the peak at the isolation frequency) with no
considerable amplification of the non-isolated modes and thus it is superior to additional passive
damping (case LDRB 25.0b in Figure 2). To make the use of this kind of isolators more
appealing, issues related to their practical realization should be studied further but this out of
the scope of this work.
Figure 2. Normalized floor response spectrum (2% damping) at the base of an isolated structure for the
Ardal earthquake.
Besides the above passive alternatives, in the remaining part of SINAPS@, we will address issues
related to possible benefits on floor response spectra from the use of semi-active devices combined
with passive isolators (mixed isolation). Actually, nowadays, such techniques are not limited to small
structures and some civil engineering structures and buildings, mainly in Japan, are equipped with
semi-active devices.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 60
WP 4: Seismic Risk Assessment
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 61
PROGRESS OF THE WP4 “SEISMIC RISK ASSESSMENT” IN 2016
F. VOLDOIRE, EDF, R&D Division
The scientific context
This work package is centered on the KARISMA SINAPS@ demonstrative case-study, (already
considered for the benchmark organized by the IAEA in 2010, see reference [1] IAEA 2013) based on
the event experience feedback July 16, 2007 earthquake occurred at the Kashiwazaki-Kariwa NPP
(TEPCO), for which a rich set of data throughout the analysis chain is available.
The purpose of the WP4 demonstrative study is to implement and evaluate the methodology
leveraging various contributions of SINAPS@ project at each step of the seismic analysis on a concrete
case of nuclear plant, identifying phenomena contributing to a "best-estimate" response, enabling to
comment and prioritize sources of potential margins, from the fault to probabilistic floor spectra. The
ingredients are the following:
seismic loading hypothesis coming from seismologists of WP1,
numerical methods to compute nonlinear soil and structure interaction (SSI), specific analysis
on nonlinear site-effects, and finally coupling SSI and site effects with structural seismic
behaviour numerical simulation, with a strong interaction with WP2,
however, due to the specific design of Kashiwazaki-Kariwa NPP Unit 7 and the specific seismic
event in view, unlike SSI and site effect ingredients, the need of nonlinear reinforced concrete
structural predicting constitutive models, provided by WP3, is not crucial: indeed very few
cracking was observed, and elastic modelling seems to be sufficient in a first approach,
expertise on methods used to predict the seismic motion transferred from structure up to
equipment,
and methods used to propagate and hierarchize the uncertainties within the vulnerability and
probabilistic risk assessment analysis.
The partners involved are: ECP, CEA, EDF, AREVA, ISTerre. In 2017, IRSN will join the previous partners.
Objectives of this work package
WP4 has the following three objectives:
(i) set up the overall methodological approach using the various products of the project and
by comparing their implementation in a coordinated global validation work; based on the
demonstrative case study,
(ii) to consolidate methods of probabilistic risk assessment, including algorithmic
performance aspect for practical studies (statistical meta-models and sampling
techniques),
(iii) aiming at validating and disseminating advanced methodologies for practitioners and
structural engineers.
The current work progress
The Task WP4.1: “Ground motion parameters devoted to seismic risk assessment, parameters adapted
to the vulnerability analysis, involving SSI” has not started yet (it is planned for 2017). However, a first
version of the deliverable “Methodological guide on numerical approaches for Site response estimates
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 62
in low seismicity zones” produced in the framework of the task 4.1.2, has been submitted for discussion
among partners. The purpose is dedicated to the complex geometrical description of the site, and the
taking into account of soil nonlinearities, according to the characteristics of the input motions and the
given seismic hazard, the source type and the uncertainties to be tackled.
The Task WP4.2: “Simulation and propagation of uncertainties” included in 2015 a specific sub-task
4.2.1 entitled “Performance of probabilistic simulations for vulnerability assessment”, which is
completed (investigation about meta-models, Gaussian processes and neuronal networks), and
another sub-task 4.2.2 “Fragility curves computations: extreme statistics and Bayesian methods” which
is to be launched at the early beginning of 2017.
The following presentations by WP4 partners will focus on progress of Task WP4.3. Indeed, 2016 is the
last year for the first stage of the demonstrative study (sub-task 4.3.1), while the final stage (sub-task
4.3.2) has to be launched for the last two years of the SINAPS@ project.
Task WP4.3: “Demonstrative numerical case study”:
(i) “Preliminary calibration of the numerical large-scale scenario of the Niigata-Ken Chuetsu-Oki
earthquake”, by F. Gatti;
(ii) Implementation of KARISMA demonstrative case study: initial phase, by C. Feau and F. Wang;
(iii) Generation of synthetic accelerograms in agreement with conditional scenario spectra for
the computation of floor response spectra - Application to Karisma benchmark, by I. Zentner;
(iv) “Task 4.3.2: Demonstrative numerical case study, final step”, by F. Voldoire, in order to
present the first proposals by partners concerning the specifications of the final stage on the
Karisma case study: site response and fragility curves calculation, in the light of the results
completed during the first phase.
The following deliverables are done or expected for 2016:
Bard, P.-Y., Methodological guide on numerical approaches for Site response estimates in low seismicity zones. ISTerre, June 2016.
Berge-Thierry, C., et al. (2017). “Toward an integrated seismic risk assessment for nuclear safety improving current French methodologies through the SINAPS@ research project”. Submitted Nucl. Eng. Design, 2016.
Gatti, F., et al. (2016). Simulations numériques préliminaires du scenario sismique à large échelle du tremblement de terre MW6.6 de Niigata (2007).
Gatti, F., et al. (2016). Extension de la Méthode d'Éléments Spectraux pour la propagation d’onde sismique 3D : Intégration d’une loi constitutive non-linéaire pour géo-matériaux.
I. Zentner, Z. Wang, CR-T64-2016-88, Calculs de courbes de fragilité pour l’étude KARISMA (volet 4 du projet ANR SINAPS@) effectués par Zhiyi WANG dans le cadre de sa première année de thèse. 10/06/2016.
V. Alves-Fernandes, CR-T64-2016-180, Compte rendu des stages du projet OMARISI2016 autour du benchmark Karisma : Structure-soil-structure interaction, nonlinear soil-structure interaction, taking into account of the basemat uplifting, comparison of several modelling strategies.
References
IAEA- TECDOC-1722, Review of Seismic Evaluation Methodologies for Nuclear Power Plants Based on a
Benchmark Exercise. International Atomic Energy Agency, (2013). Altinyollar A. et al., VIENNA, IAEA.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 63
PRELIMINARY CALIBRATION OF THE NUMERICAL LARGE-SCALE SCENARIO OF THE NIIGATA-KEN
CHŪETSU-OKI EARTHQUAKE
F. GATTI 1,2, F. LOPEZ-CABALLERO 1, R. PAOLUCCI 2, D. CLOUTEAU 1
1 CentraleSupélec 2 Politecnico di Milano
Introduction
This study presents a first calibration of a large-scale seismological model of the July, 16th 2007 MW 6.6
Niigata- Ken Chūetsu-Oki earthquake (NCOEQ). The analysis is intended to clarify some aspects of the
recorded site-effects in the epicentral area (within the Japanese Niigata prefecture) and at the
Japanese nuclear site of Kashiwazaki-Kariwa (KKNPP), held by the Tokyo Electric Power Company
(TEPCO). A map of the KKNPP site is shown in Figure 1a.
(a) (b)
Figure 1: (a) Map of the KKNPP with the accelerometric stations installed that recorded the NCOEQ. (b) Map of the
Niigata area surrounding the NCOEQ epicenter. The slip model proposed by Shiba et al. [10] is shown along with the
KKNPP site. Several recording stations belonging to the Japanese KNET and KikNet accelerometric network are shown.
The main objectives of this study are (1) the choice of a suitable geological model and (2) the
quantitative yet preliminary assessment of the parameters required for the kinematic modelling of the
source mechanism (e.g. the fault plane location and dimensions originated, the slip pattern, rupture
velocity, rise time).
Due to the relative small source-to-site distance and shallow hypocenter depth, the mentioned seismic
scenario is extremely complex to be characterized, although very appealing due to the consistent
seismic record database available. From this point of view, different assumptions of the SINAPS@
project work-packages may be tested. For instance, several source-inversions were performed on the
NCOEQ [2, 4, 7, 10, and 3]. Most of the proposed slip models are reliable up to 0.5-1 Hz and poorly
constrained by near-source data. In this study, the slip contour proposed by Shiba et al. [10] is
considered (the fault plane is depicted in Figure 1b), duly calibrated upon KKNPP records. Moreover, a
physics-based 3D numerical scenario is herein built-up, to overcome some of the major simplifications
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 64
commonly made in the estimation of strong-ground motion time-histories (e.g. the disregarded
topography effect, the 1D geological profile).
In the following, a preliminary calibration of the mentioned scenario is presented. In this sense, the
work is presented in three steps: the semi-analytical validation of (1) a simplified 1D geological model
and of (2) an inverted source model; (3) the numerical simulation of the July, 16th 2007 MW 4.4
aftershock wave-propagation in the Niigata region, by means of a Spectral Element Method (SEM)
based numerical code.
Validation of seismological model
A numerical large-scale earthquake scenario requires the standalone identification of the source and path
effects on the radiated wave-field. This identification is performed by applying complex wave-form
inversion analyses on mainshock and aftershock data. In this section, the geological and source models
proposed in the literature are validated against the available set of recordings. Specifically, the analyses
were performed by means of the Wave-Number Integration (WNI) method, proposed and numerically
implemented by Hisada [5, 6]. The WNI is a semi-analytical method that simulates the complete 3D
wave propagation field radiated from an extended kinematic seismic source in an extended half-space.
This approach is based on the computation of static and dynamic Green’s functions of displacements
and stresses for a viscoelastic horizontally layered half-space (no topography considered).
Simplified geological model of the epicentral area: Several simplified 1D geological models are
available of the Niigata region. VP and VS profiles are depicted in Figure 2a and 2b respectively. Their
validation was performed by running a WNI analysis of the July 16th 2007 MW4.4 aftershock
(aftershocks are considered as point-wise seismic sources).
Figure 2: Several VP (a) VS (b) 1D geological models of the Niigata basin from literature.
The analyses carried out clarified that the two most reliable 1D geological profiles were the one
proposed by Aochi et al. [1] and the one proposed by Shinohara et al. [11] (tagged as Aochi2013 and
Shin.2008 respectively in Figure 2a, 2b). This is proved by the good fit provided by the time-histories
shown in Figure 3a, 3b (records filter at between 0.05 and 0.5 Hz).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 65
Figure 3: Comparisons between recorded (blue) and synthetic (red) velocity wave-forms at NIG018 (KNET station)
for the 16th July 2007 MW4.4. Both the time-histories were filtered within a 0.05-0.5 Hz frequency band. (a) Synthetics
were obtained by using the geology profile Aochi2013 [1]. (b) Synthetics were obtained by using the geology profile
Shin.2008 [11].
Seismic source characterization: Average co-seismic slip values barely estimate the ground motion
complexity, even in far-field conditions, because the area of strong motion generation usually
coincides with slip heterogeneities, i.e. asperities area. As depicted in Figure 1b, it composes of three
major asperities. To test latter slip model, the same approach presented in the previous section was
exploited although the fault plane is subdivided into rectangular sub-faults at constant slip and rake
angle. Each contribution is then convolved to obtain the wave-forms at each receiver’s location. The
analysis is limited to 0.5 Hz, since the Green’s functions are computed by means of a truncated wave-
number series.
Results at KKNPP are shown in Figures 4a, 4b. The synthetics are compared to the recordings at G.L.-
250m within the nuclear site (KSH-SG4, in Figure 1a). The velocity time-histories were obtained by
using the WNI, with the same geological profiles considered above.
Figure 4: Comparisons between recorded (blue) and synthetic (red) velocity wave-forms at KSHSG4 (TEPCO station) for
the NCOEQ. Both the time-histories were filtered within a 0.05-0.5 Hz frequency band. (a) Synthetics were obtained
by using the geology profile Aochi2013 [1]. (b) Synthetics were obtained by using the geology profile Shin.2008 [11].
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 66
(a) (b)
Figure 5: (a) Mesh of the Niigata region including surface topography. (b) Snapshot of the displacement field
numerically reproduced by Spectral Element Method.
Large-scale SEM wave-propagation of NCOEQ aftershock
The final yet preliminary phase of this study concerns the forward numerical wave-propagation of the
July 16th 2007 MW 4.4 aftershock. The simulation was performed by means of a SEM-based numerical
code implemented on massively parallel computer architecture. Figure 5a shows a detail of the
meshed domain. Figure 5b depicts the propagated displacement contour.
The numerical model is featured by the topography of the Niigata area, along with the 1D geological
profile calibrated in the previous section. The seismic source is represented by a point-wise double-
couple calibrated on the Hi-Net moment tensor solution [8, 9] of the July 16 2007 MW4.4 aftershock.
Time-histories at several locations (K-NET, KiK-net and KKNPP stations) were compared to test the main
difference between semi- analytical and SEM analyses, addressing the effect of topography and the
source time-function. Moreover, different degree of mesh refinement were tested to increase the
frequency range of reliability of the numerical model.
Acknowledgement
This work, within the SINAPS@ project, benefited from French state funding managed by the National
Research Agency under program RNSR Future Investments bearing reference No ANR-11-RSNR-0022-
04. The research reported in this paper has been supported in part by the SEISM Paris Saclay Research
Institute. Time histories and velocity profiles used in this study were collected from the KiK-net
website: http://www.kik.bosai.go.jp/kik/ (last accessed November 2011). The authors are very grateful
to the National Research Institute for Earth Science and Disaster Prevention (NIED) for providing such
high-quality earthquake recordings.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 67
References
[H. Aochi, A. Ducellier, F. Dupros, M. Delatre, T. Ulrich, F. de Martin, and M. Yoshimi. Finite difference
simulations of seismic wave propagation for the 2007 Mw 6.6 Niigata-ken Chuetsu-Oki earthquake: Validity
of models and reliable input ground motion in the near field. Pure and Applied Geophysics, 170((1-2)):43–
64, 2013. Springer Verlag (Germany).
Shin Aoi, Haruko Sekiguchi, Nobuyuki Morikawa, and Takashi Kunugi. Source process of the 2007 Niigata-
ken Chuetsu-oki earthquake derived from near-fault strong motion data. Earth, Planets Space, 60(11):1131–
1135, 2014.
a. Cirella, a. Piatanesi, E. Tinti, and M. Cocco. Rupture process of the 2007 Niigataken Chuetsu-oki earthquake
by non-linear joint inversion of strong motion and GPS data. Geophysical Research Letters, 35(16):1–5, 2008.
K. Hikima and K. Koketsu. Source process of the 2007 Chuetsu-oki earthquake inferred from far field
waveforms and strong motions. Fall Meeting 2007, pages 1–5. Seismological Society of Japan, 2007.
Y. Hisada. An efficient method for computing Green’s Functions for layered half-space with sources and
receivers at close depths. Bulletin of the Seismological Society of America, 84(5):1456–1472, October 1994.
Y. Hisada. An efficient method for computing Green’s functions for a layered half-space with sources and
receivers at close depths (part 2). Bulletin of the Seismological Society of America, 85(4):1080–1093, August
1995.
A. Nozu. Rupture process of the 2007 Chuetsu-oki, Niigata, Japan, earthquake - Waveform inversion using
empirical Green’s functions. Earth, Planets Space, 60:1169–1176, 2008.
K. Obara, K. Kasahara, S. Hori, and Y. Okada. A densely distributed high-sensitivity seismograph network in
Japan: Hi-net by National Research Institute for Earth Science and Disaster Prevention. Review of Scientific
Instruments, 76, 2005.
Y. Okada, K. Kasahara, K. Hori, S. Obara, H. Sekiguchi, S. Fujiwara, and A. Yamamoto. Recent progress of
seismic observation networks in Japan -Hi-net, F-net, K-NET and KiK-net. Earth, Planets and Space, 56:xv–
xxviii, 2004.
Y Shiba. Source Process and Broadband Strong Motions during the 2007 Niigata-ken Chuetsu-oki Earth-
quake. 1st Kashiwazaki International Symposium on Seismic Safety of Nuclear Facilities, 2010.
Masanao Shinohara, Toshihiko Kanazawa, Tomoaki Yamada, Kazuo Nakahigashi, Shin’ichi Sakai, Ryota Hino,
Yoshio Murai, Akira Yamazaki, Koichiro Obana, Yoshihiro Ito, et al. Precise aftershock distribution of the 2007
Chuetsu-oki Earthquake obtained by using an ocean bottom seismometer network. Earth, planets and space,
60(11):1121–1126, 2008.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 68
IMPLEMENTATION OF KARISMA DEMONSTRATIVE CASE STUDY: INITIAL PHASE
F. WANG, C. FEAU, Laboratoire EMSI, CEA-Saclay
Scope within SINAPS@
WP 4: SEISMIC RISK ASSESSMENT, TASK 4.3.1
The scientific context
As one of the participants to the international KARISMA Benchmark, EMSI laboratory (CEA) has
conducted soil-structure interaction analysis on the Unit 7 Reactor Building (RB7) of the Kashiwazaki-
Kariwa nuclear power plant (Japan). A large 3D finite element model including the Reactor Building
and the nearby soil was used to simulate the structural response under the NCOE earthquake of July
2007. In the current framework of SINAPS@ project, Task 4.3.1 focuses on the uncertainty propagation
through the soil-structure system.
Objectives of this work
The objective of this work is to construct the fragility curve of some selected plant equipment due to
the variability of the input seismic signals. For this purpose, a simplified model representing the largely
embedded reactor building has been implemented for computation efficiency. The soil non-linearity
caused by each seismic signal has taken into account using the equivalent linear method. The fragility
curve has been approximated by the cumulative distribution function of a lognormal random variable.
Its parameters (median and standard deviation) have been evaluated using the principle of maximum
likelihood. Moreover, a confidence interval has been determined using a bootstrap method [1].
The current work progress
A stick model (figures below) for the Reactor Building RB7, used by TEPCO, the owner of the power
plant has been adopted and implemented in the Code CAST3M for this work. This is a very simple
model but it can describe two interesting features of the building: the first one is the embedment (26
m over a total height of 63.4m) and the second is the flexibility of the soil-structure interface.
XZ section Stick model
Figure 1: Stick model for the Reactor Building RB7, used by TEPCO
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 69
To perform numerical calculations, ground motion signals were generated by EDF [2] using the
Campbell and Bozorgnia 2008 empirical Ground Motion Prediction Equation (GMPE) [3], considering
the NCOE 2007 scenario (Mw=6.6 and epicentral distance of 16 km). Two sets of synthetic signals were
considered. For the first set, 50 ground motion signals were generated at the ground surface on soft
soil condition (Vs30 = 250 m/s). For the second set, 50 ground motion signals were generated at the
ground surface for a bedrock site condition (outcropping bedrock, Vs30 = 720 m/s) in order to avoid the
“soil non linearity” phenomenon. Furthermore, to increase the seismic inputs number and to cover
strong motions, the “classical engineering scaling” process was applied to the two sets (with factors of
0.5, 1, 2, 2.5 and 3). In the numerical procedure, for each input signal, all the soil springs are re-
calculated using Novak method based on the equivalent linear soil model determined by soil column
deconvolution. The response of the structure to each ground motion is finally obtained by modal
superposition method.
In this work, no specific equipment is chosen, only its resonance frequency is postulated at 4 Hz. The
failure criterion is the exceedance of its 5% damped pseudo acceleration response at 4 Hz of a level of
acceleration; this is supposed unknown, and will be explored during the study.
The following figure summarizes the results of numerical calculations. It presents in ordinates the PSA
values corresponding to seismic inputs from set 1 (triangles) and to set 2 (circles) as functions of PGA
values at the RB basement: the color scale is related to the soil distorsion rate.
Figure 2: Pseudo acceleration response of the 4 Hz resonant equipment as a function of PGA (g) at the control point 3 – RB7 basement at - 25 m (triangles when initial seismic input is the “outcropping bedrock” point 2, circles when it is point 1 “Surface, soil”). The color scale illustrates the maximal soil distorsion reached during the deconvolution process (restricted to 0.8%).
The results produced by set 1 and set 2 seems to have trends that are almost the same in the very low
PGA range but, this study shows that, the medians of the fragility curves estimated with the results
related to the set 1 are always greater than the medians related to the set 2, whereas the
deconvolution problem is not well defined in the case of set 1. The following figures show two
examples of median fragility curves and confidence intervals estimated for set 1 and set 2 (blue and
red curves respectivley), for failure criteria of 0.2g and 0.7g (left and right respectively).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 70
Figure 3: Two examples of median fragility curves and confidence intervals @ 95% for the 4 Hz resonant equipment, computed (Left) for a failure criterion equal to 0.2 g and (Right) for a failure criterion equal to 0.7 g.
Through this example, this work illustrates the need of checking the coherency of all assumptions for
describing the seismic motion from the geological fault, including modelling of the seismic wave field
through complex geology, its transmission and interactions with the foundations of the structure and
equipment. Particularly, this work showed that defining the control point at the free field as required
in the French nuclear approach ([4] and [5]) condition is not appropriate and may conduct to “biased
results” when performing nonlinear soil-structure fragility analyses. The control point, interface
between the seismic hazard and the SSI, should be defined at the “outcropping bedrock” level.
References
I. Zentner (2010). Numerical computation of fragility curves for NPP equipment. Nuclear Eng. Design, 240
(6), 1614-1621.
Irmela Zentner, F. Allain, N. Humbert and M. Caudron (2014). Generation of spectrum compatible ground
motion and its use in regulatory and performance-based seismic analysis. Proceedings of the 9th
International Conference on Structural Dynamics, EURODYN. Porto, Portugal.
K.W. Campbell and Y. Bozorgnia (2008). NGA Ground motion model for the geometric mean horizontal
component of PGA, PGV, PGD and 5% damped linear elastic response spectra for periods ranging from 0.01
to 10s. Earthquake. Spectra, 24 (1).
“RFS2001-01, Règle fondamentale de sûreté n°2001-01 relatives aux installations nucléaires de base.
Détermination du risque sismique pour la sûreté des installations nucléaires de base”, (2001), Nuclear
Authority Safety.
“Guide/ASN/2/01, Prise en compte du risque sismique à la conception des ouvrages de génie civil des
installations nucléaires de base, à l'exception des stockages à long terme des déchets, radioactifs”, (2006),
Nuclear Authority Safety.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 71
GENERATION OF SYNTHETIC ACCELEROGRAMS IN AGREEMENT WITH CONDITIONAL SCENARIO
SPECTRA FOR THE COMPUTATION OF FLOOR RESPONSE SPECTRA. - APPLICATION TO KARISMA
BENCHMARK.
I. ZENTNER, EDF R&D
Scope within SINAPS@
The Seismic Probabilistic Risk Assessment (SPRA) methodology is the most commonly used approach
for the evaluation of seismic risk and is now applied worldwide. In current practice, a set of ground
motion is often simulated to comply in its mean with the target Uniform Hazard Spectrum (UHS). This
can be considered as conservative since the UHS constitutes an envelope: it combines various
scenarios to yield a target spectrum. Indeed, the UHS provides spectral accelerations as a function of
frequency such that all values have same exceedance probability (“same hazard level”). The spectral
content of time histories in agreement with such spectra does not correspond to any real earthquake
scenario. The conditional spectra methodology developed in the US defines ground motion by multiple
scenario spectra obtained by disaggregation rather than the UHS. It allows to define ground motion at
design level with realistic spectral shape. Indeed, the use of multiple scenario spectra rather than one
UHS avoids exciting a broad frequency range in each single structural analysis. Nevertheless, the choice
of earthquake scenario target spectra complying with design requirements and the selection or
generation of time histories in agreement with the latter is not straightforward.
The scientific context
Generally speaking, two different settings can be considered (e.g. ATC 2011) for the definition of
conditional spectra. This is illustrated in Figure 1 with the risk-based assessment (right) and intensity-
based assessment (left). For both settings, Conditional Spectra (CS) that represent the spectral shape
associated to single earthquake scenarios can be derived. In the intensity-based approach (assessment
at design level intensity), the CS can be defined for one UHS considering a list of conditioning
frequencies while in the risk-based approach, the CS are defined for a fixed frequency and increasing
risk in terms of return periods. The intensity-based assessment is the framework for structural analysis
that are performed for a given risk target spectrum such as UHS. However, in recent applications to
safety analysis of NPP, the choice was rather to consider only one conditioning frequency. In this
framework, conditional mean spectra (CMS) or CS are considered for a set of UHS at different return
periods (for example Renault et al 2015). Sets of time histories that best fit the set of target spectra
are then selected in a database, and possibly scaled in order to improve the matching. The risk-based
approach allows to reconstruct the full hazard curves and to obtain sets of fully “hazard-consistent”
ground motion.
Objectives of this work
The present work proposes to assess the feasibility of the CS approach for the computation of fragility
curves and floor spectra, combined with a stochastic ground motion simulation model. The
methodology used here is inspired by an innovative ground motion selection procedure that has been
recently proposed by Lin, Haselton & Baker. The original method is based on the simulation of sets of
conditional spectra (CS) that represent the spectral shape in median and sigma and the selection (and
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 72
possibly scaling) of recorded time history that best fit in the CS, one-by-one . Ground motion simulation
allows to obtain time histories without having to resort to scaling and modification. Moreover, there
is no limitation in the number of available appropriate time histories, matching various criteria such as
spectral shape, strong motion duration and other ground motion proxies since the latter can be
generated at low cost.
a) CMS at multiple conditioning frequencies (red asterisk) for one target spectrum (UHS – dark blue curve, left side)
b) CMS at one conditioning frequency (red asterisk) for multiple return period target spectra (UHS – dark blue curves, right side)
Figure 1: UHS and scenario target spectra
The current work progress
The intensity-based approach is pursued here. The UHS is broken down into a suite of conditional
scenario spectra at different frequencies. The envelope of the CMS reproduces the UHS (figure 1a).
Sets of CS are then simulated for each conditioning frequency and floor spectral accelerations are
computed as a function that conditioning frequency. This assures that predominant structural
frequencies are excited at (UHS) design level (target return period as defined by the regulator) while
keeping a realistic spectral shape. Floor response spectra are computed and compared to the ones
obtained with the classical procedure using UHS-compatible ground motion.
The present work was focused on the feasibility of an intensity-based CS approach combined with a
stochastic ground motion simulation model. The time histories are simulated for a number of
conditional target spectrum using a stochastic model. Figure 2 shows that a close fit to target
conditional spectra can be obtained by ground motion simulation. A great number of conditional
ground motion time histories can be simulated at low cost. Moreover, the simulation facilitates the
definition of 3D load that complies with the target. The methodology was assessed by the computation
of in-structure floor response spectra for some Karisma NPP benchmark structure.
Further work is required and in progress in order to
Implementation of the risk-based approach and comparison of methodologies and results
Implement more complex PSHA case studies with multiple sources and models
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 73
Evaluate the impact of CS-approaches on SPRA in terms of plant failure probabilities (by
convolution with hazard)
Figure 2: Target CS at frequencies 2Hz and 10Hz and spectra of simulated accelerograms (green)
References
ATC (2011). Guidelines for seismic performance assessment of buildings, ATC-58 100% draft. Technical
report, Applied Technology Council, Redwood City, USA.
Carlton, Abrahamson (2014), Issues and approaches for implementing conditional mean spectra in practice.
Bull. Seism. Soc Am 104(1).
Jayaram, Lin, Baker (2011), A computationally efficient ground-motion selection algorithm for matching a
target response spectrum mean and variance. Earthquake Spectra, 27(3), 797-815.
Lin, Haselton, Baker (2013), Conditional spectrum-based ground motion selection. Part I and II. Earth.
Eng.Struct. Dyn. 42(12).
Renault Ph., Proske D., Kurmann D., Asfura A. (2015), EVALUATION OF THE SEISMIC RISK OF A NPP BUILDING
USING THE CONDITIONAL SPECTRA APPROACH. Proceedings of SMIRT-23, Manchester, UK
Zentner, I. (2014), A procedure for simulating synthetic accelerograms compatible with correlated and
conditional probabilistic response spectra. Soil Dyn Earth Eng. 63(1), 226-233.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 74
TASK 4.3.2 : DEMONSTRATIVE NUMERICAL CASE STUDY, FINAL STEP.
François Voldoire, EDF R&D
Scope within SINAPS@
The SINAPS@ work package 4 includes a demonstrative study consisting in revisiting the KARISMA
international benchmark by integrating the modeling strategies and innovative methodologies
developed by the other work packages, at each step of the seismic analysis, contributing to a "best-
estimate" response, [3]. This study case is fully coherent in the post-Fukushima context where extreme
hazards have to be studied, with new methods of justification that will be implemented in the French
context.
The Kashiwasaki-Kariwa Nuclear Power Plant (KK-NPP) was submitted to Niigata Chuetsu-Oki strong
earthquake (NCOE) of magnitude Mw=6.6 on 2007, with recorded ground motions being beyond
assumed design criteria. These recordings were subsequently used as basis for the KARISMA
(KAshiwazaki-Kariwa Research Initiative for Seismic Margin Assessment) international benchmark
exercise on 2009, organized by the IAEA (International Atomic Energy Agency) and OECD/NEA, [6]. One
important result from the KARISMA benchmark was the difficulty to properly predict the unit 7 reactor
building (KK-RB7) response during the main shock by considering available standard engineering
methods. The imagined levers to enhance the engineering practice are to strengthen the integration
of the whole chain of analysis, in particular focusing on seismic scenario and ground motions, site
response and soil-structure interaction assumptions, in particular dealing with nonlinear effects, and
the uncertainty propagation to the calculation of fragility curves on specific equipment. However, it
has been justified that reinforced concrete RB7 structures behave elastically, very few cracking being
observed.
The scientific context
During the first stage of the demonstrative case study (task 4.3.1 of SINAPS@ project, 2014-2016) using
recently available best-estimate practices, the following topics were implemented by Areva,
ECP/LMSSMat, CEA/EMSI and EDF/AMA partners and discussed:
Comparative study between Code_Aster + Miss3D and Ansys-Sassi solutions on the KK-RB7
soil-structure interaction analysis under the viscoelastic soil behaviour assumption;
Calculations with Cast3m (3D FEA) on a linear stick-model for the KK-RB7 with flexible and
buried foundation, including soil-structure interaction by means of a full-FEM soil domain
modelling and equivalent linear viscoelastic properties calibrated on a soil column model and
Rayleigh damping for the building. Simplified Novak method of equivalent foundation
interface springs was used. In order to propagate uncertainties and calculate fragility curves,
considering characteristic equipment failure criteria, [4], a synthetic ground motions set based
on NGA – Campbell & Bozorgnia GMPE was implemented on the 2007 NCOE scenario [8], see
[10];
Site response calculations with the spectral elements solver SEM3D, with several soil profiles
and calibrated equivalent linear viscoelastic properties, at the vicinity of the KK Unit 5, for
which some geological data and ground motions sets are available, see also [5];
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 75
A fault model was used [8], which has been calibrated on the main shock free field ground
motions, the site response produced in depth signals similar to measured ones, in the low
frequency range (<5 Hz);
Soil-structure interaction calculations of the KK-RB7 with buried foundation with Code_Aster.
Assessment of the impact of different SSI modeling strategies on the unit 7 KK-NPP reactor
building response during the main shock, including soil non-linearity, by considering a cyclic
critical state elastic-plastic model, possibility of raft up-lifting (with comparison of two
methods: nonlinear junction finite elements and equivalent discrete springs elements), and
structure–soil–structure interaction (SSSI) influence by directly considering a simplified model
of the adjacent turbine building (in that case with a equivalent linear viscoelastic model). Two
different approaches are assessed: either Full-FEM with absorbing boundaries modeling or a
hybrid Laplace-time domain approach with FEM-BEM solution [7];
Direct uncertainties propagation and fragility curves calculation [11], for anchors of an electric
cabinet (by likelihood maximisation) with Code_Aster using the equivalent linear viscoelastic
model for soil domain, determined in coherence with seismic loading (synthetic motions
generation at outcropping bedrock) and BEM-FEM coupling for SSI simulation. These
calculations account for uncertainties on synthetic ground seismic motions set (thus on soil
equivalent degraded parameters), based on NGA – Campbell & Bozorgnia GMPE, and building
behaviour parameters; it has been observed that the latest have very few impact on the
results.
Figure 1: seismic ground motion generated at the outcropping bedrock.
All these deliverables have allowed to define important recommendations about an advanced
methodology for seismic margins NPP assessment, in particular the need to use probabilistic spectra
for seismic hazard, the need to define and control the seismic motion at the “engineering” outcropping
bedrock rather than at free field, for fragility curves computations, which reduces significantly the
uncertainty in the results of calculations (and therefore a noticeable reduction of the standard
deviation, and a thinner confidence interval), the need to prescribe the convoluted seismic motion on
the lateral boundaries of the finite element soil domain model… In particular, such investigation was
partially confirmed by referring to borehole G5 recordings, available on surface and in depth.
To conclude, comparisons of results with recorded signals within the KK-NPP reactor building allow to
emphasize some conclusions about the role of modelling assumptions. In particular, the effective site
effect and soil-structure interaction need three-dimensional modelling and nonlinearities modelling.
Nevertheless, it has been observed that difficulties to interpret field measurements (for instance,
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 76
sensors directivity, signal processing) produced troubles in the comparison with calculations results
and discrepancies on the quantities of interest.
Objectives of this work
In the final stage of the SINAPS@ demonstrative case study, major achievements of the project
SINAPS@ will be considered on the same KARISMA benchmark. The main issues involved in this task
4.3.2 of SINAPS@ project are therefore:
Produce a prioritization of uncertainties impacts (data and methods) on all key steps: seismic
hazard, site effects, soil-structure interaction, seismic behaviour of structures and equipment,
risk assessment;
Identify and quantify potential seismic margins within the analysis chain modelling steps;
Update deterministic and probabilistic approaches, and disseminate them among practice
engineering.
Several questions are remaining, in particular for soil-structure interaction modelling. The current
practice is based either (1) on simplified approach assuming a homogeneous soil domain and using
calibrated soil springs and dampers, from foundation impedance functions, taken from literature or
BEM numerical solution, often neglecting the buried part of it, or (2) the sub-structuring BEM-FEM
method assuming linear viscoelasticity in the soli layers within a frequency domain solution procedure.
The soil viscoelastic properties are updated using usual G-gamma – D-gamma degraded equivalent
properties method, from actual maximal strain solicitation in a one-dimensional representative soil
column, submitted to the 3D seismic ground motion at the bedrock level. So we have to investigate
some improvements: could the usual practice represent accurately the whole nonlinear behaviour
incursion on the whole ground motion duration? Is a full FEM modelling technique more appropriate
and efficient by comparison with sub-structuring BEM-FEM linear method? How to prescribe the
ground motion on the lateral buried faces of the foundation? How to process if the bedrock is not
sufficiently stiff: the diffracted signal having to be represented? How to prescribe the absorbing
boundaries (paraxial hypothesis for example) on the soil domain Finite Element model limits? How to
process in case of tilted soil layers?
The current work progress
Further work is planned for the next two years 2017-2018 by partners CEA/EMSI, ECP/LMSSMat and
IRSN, in the framework of task 4.3.2 of SINAPS@, in order to:
Calibrate and identify parameters of geological and seismic source models (LMSSMat);
Numerical simulations of a seismic scenario at local scale (LMSSMat);
Whole KK site response simulation (LMSSMat);
Fragility curves calculations introducing the soil nonlinearity modelled with the Iwan’s
constitutive relation by means of the NERA software, with the only variability taken on the
previous synthetic ground motion set; comparisons with previous equivalent degraded
viscoelastic soil properties model calculations (CEA/EMSI), and reference to field
measurements on KK Unit 5.
Perform fragility curves on electric cabinet by reference to SSI calculations including advanced
hypotheses from the former task 4.3.1 (CEA/EMSI), and using a functional approach (IRSN).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 77
Finally, a methodological synthesis can be planned in order to prepare the dissemination of these new
modelling tools and methods among engineering practitioners.
References
Alves Fernandes, V., et al. (2017). “Different SSI modelling strategies applied to the Kashiwasaki-Kariwa
Nuclear Power Plant”. Submitted to COMPDYN 2017 conference, Rhodes Island, Greece.
Berge-Thierry, C., et al. (2017). “The SINAPS@ French research project: first lessons of an integrated seismic
risk assessment for nuclear plants safety”. 16WCEE 2017, Santiago, Chili.
Berge-Thierry, C., et al. (2017). “Toward an integrated seismic risk assessment for nuclear safety improving
current French methodologies through the SINAPS@ research project”. Submitted Nucl.Eng.Design, 2016.
Feau, C., Wang F. (2016). “Implementation of KARISMA demonstrative case study: initial phase”. SINAPS@
plenary session, Nov. 2016, Palaiseau.
Gatti, F. et al. (2015). One-Dimensional Seismic Soil Response at the Nuclear Power Plant of Kashiwazaki-
Kariwa during the 2007 Niigata-Chuetsu-Oki Earthquake. 15th Civil-Comp Conference.
IAEA-Tecdoc-1722 2013, “Review of Seismic Evaluation Methodologies for Nuclear Power Plants Based on a
Benchmark Exercise”, 2013. International Atomic Energy Agency, IAEA-TECDOC-1722, (http://www-
pub.iaea.org/MTCD/publications/PDF/TE-1722_web.pdf).
Nieto, A., et al. (2012). On a hybrid Laplace-time domain approach to dynamic interaction problems, Eur. J.
Comp. Mech – REMN, 21.
Shiba, Y., K. Hikima, T. Uetake, H. Mizutani, K. Tsuda, T. Hayakawa, and S. Tanaka (2011). “Source model of
the 2007 Chuetsu-Oki earthquake based on precise aftershock distribution and 3-D velocity structure”, Japan
Geoscience Union meeting, SSS023-P13.
Zentner, I. (2014). “A procedure for simulating synthetic accelerograms compatible with correlated and
conditional probabilistic response spectra”. Soil Dyn Earth Eng. 63(1), 226-233.
Zentner, I. Génération de signaux sismiques. http://www.code-aster.org/V2/doc/default/fr/man_r/r4/r4.05.05.pdf.
Zentner, I., Wang Z. Calculs de courbes de fragilité pour l’étude KARISMA (volet 4 du projet ANR SINAPS@)
effectués par Zhiyi WANG dans le cadre de sa première année de thèse. EDF/AMA CR-T64-2016-088,
10/06/2016.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 78
WP 5: Experimental support to
SINAPS@’s issues and Building
to Building Interaction
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 79
PROGRESS OF THE WP5 “EXPERIMENTAL SUPPORT TO SINAPS@ ISSUES AND BUILDING TO
BUILDING INTERACTION” IN 2016
A. LE MAOULT, CEA/DEN/SEMT
Scope within SINAPS@
Work Package 5 (WP5) gives an experimental support to structural design of nuclear facilities. Using
the large Azalée shaking table of the EMSI laboratory in CEA/Saclay, all partners can combine numerical
and experimental simulation to better understand the seismic behaviour of structures. Two main
research fields were selected at the beginning of SINAPS@ project; both are linked to seismic safety
and margin evaluation in nuclear facilities. For each field, a state of the art and preliminary numerical
simulation permit to design a mock up able to exhibit the little known physical phenomena. The mock
up are then built, tested and analyzed.
The most important interaction with other WP of SINAPS@ project is then with the WP3 which is
dedicated to the study of the behavior of structure and equipment. Nevertheless, WP1 is also helping
for the selection of seismic excitation of experimental campaigns.
The scientific context
Concerning the project dedicated to the study of the interaction between buildings (IBB), one
important question for nuclear facilities is the effect of impacts between slabs. Indeed, high
frequencies generated by impacts can damage equipment and their anchorages. This complex
question needs experimental data but only few experimental studies already exist.
Regarding the project for the evaluation of damping in structures, previous experimental studies have
shown a gap between models and reality, especially for higher modes and various levels of damages.
The project will focus on the reduction of this gap and then a better evaluation of margin in nuclear
facilities.
Objectives of this work
Two large experimental campaigns are carried out to improve numerical simulation regarding
two important safety issues:
First, WP5 investigates energy dissipation in reinforced concrete (RC) structures: damping in RC
structure is directly connected to the amplitude of the response of nuclear facilities and thus affects
safety margins. Damping is commonly represented with a Rayleigh-type damping which may result in
an overestimation as soon as the degradation of the structure is initiated. An alternative numerical
approach is currently developed in the WP3 and the shaking table tests on 23 large RC beams has been
performed in 2016. This will permit to identify parameters and validate this dissipation model.
Second, WP5 will investigate pounding: the existing gap between buildings of nuclear facilities is
often small because the design earthquake input at the time of their construction was much less than
the earthquake intensities imposed currently by the safety authorities. The objectives of this activity:
a) to gain a further insight into the response of pounding buildings, b) study the efficacy of alternative
solutions preventing pounding (e.g. optimally designed inter-structural connections) and c) investigate
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 80
the capability of numerical simulation tools to accurately predict the dynamic response of pounding
buildings.
Both experimental campaigns are achieved on the large Azalée shaking table in CEA, Saclay. The
pounding campaign will be performed in 2017 for a period of 4 months. The IDEFIX campaign has been
carried out in 2016, for a period of 6 months.
The current work progress
Building to building interaction (IBB):
2014: first numerical and experimental state of the art and initial design of the experimental setup.
2015: a dimensional and sensitivity study of simple systems has been undertaken to evaluate the most
important structural and excitation parameters which should be taken into consideration for the
seismic tests. A PhD student, Vincent Crozet, started his thesis on IBB subject by December 2015.
2016: Design of the mock up and call for tenders for the manufacturing of the mock up.
Dissipation in RC structures (IDEFIX experimental campaign):
March 2014: first discussions regarding a possible experimental campaign (during the “interaction
between work package” meeting).
May 2014: decision to perform the experimental campaign.
October 2014 – March 2015: beginning of the thesis of Thomas Heitz (WP3), numerical simulations
and design of the experimental campaign.
March to December 2015: Tests specification report, manufacturer selection and mock up
manufacturing.
May to November 2016: test of 23 beams.
A dynamic test of an IDEFIX beam on the Azalée shaking table A quasi-static test of an IDEFIX beam
tested on the strong floor
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 81
IDEFIX: IDENTIFICATION OF DISSIPATION IN RC STRUCTURAL ELEMENTS
T. HEITZ(1,2), B. RICHARD(2), C.GIRY(1), F. RAGUENEAU(1), A.LE MAOULT(2)
1 LMT Cachan/CNRS/ Paris-Saclay University. 2 CEA, DEN, DANS, DM2S, SEMT, Laboratoire d’Etudes de Mécanique Sismique.
Scope within SINAPS@
The description of the seismic behaviour of reinforced concrete (RC) structures still remains a complex
issue due to the various phenomena involved, in particular when dealing with the nonlinear range. The
way of quantifying and modelling efficiently the energy dissipations are still a major concern. This work
falls between the Work Packages (WP) #3 and #5 following a call of interest from WP #3.
Scientific context
While the physics would suggest describing the dissipations by means of realistic constitutive laws and
well controlled boundary conditions, designers would rather reduce the computational cost of such an
analysis and therefore, often use simplified modelling strategies to describe not only the geometry of
the structure (shape and boundaries) but also its mechanical behaviour.
Within this framework, a Rayleigh-type damping model (also qualified as classical or proportional
damping) is a convenient and numerically efficient way to represent all or part of the dissipative
phenomena taking place in concrete at different scales. However, it requires a binding calibration
work, which should be performed after each material or structural modification stage to avoid a bad
description of the damping. Moreover, the physical meaning of the mass proportional term when
computing the damping matrix is not very clear and its control is generally ensured within a sharp
frequency range. Alternative approaches, such as the modal recombination strategy, allow building a
damping matrix exhibiting a better control of the damping ratios on a given frequency range but this
does not solve the concern about the damping modification along a nonlinear time-history analysis.
However, a trade-off between computational efficiency and physical meaning was proposed by
[Crambuer et al, 2013] with an updating strategy of the damping matrix depending on the damage
state based upon a simple 1-DOF oscillator framework. The same concept has been developed by
[Brun, 2002] to update the eigenfrequencies of structure elements depending on measurable
variables. An extension to a multi-DOF damping matrix – along with an updating law depending on
clearly identified material/structural parameters – would allow large scale structure time-history
analyses in accordance with the current structural state and characteristics of the structure with a
stronger physical content than the usual proportional damping.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 82
Figure 6. IDEFIX experimental setting on AZALEE shaking table (CEA, Saclay)
Objectives of this work
For a better understanding of the structural and material parameters influencing the energy
dissipation, a literature review has been carried out. It turned out that the explicit dependencies of the
dissipations depending on material and structural parameters and input signals are not given. A strong
experimental database is required to address this issue. To this end, an experimental campaign called
IDEFIX (French acronym for “Identification of damping/dissipation in RC structural elements”) is being
carried out from May to December 2016 on the AZALEE shaking table as pictured on Figure 6.
The current work progress
The IDEFIX campaign is divided in two main parts:
quasi-static tests performed on the strong floor of the TAMARIS facility and actioned by two
jacks acting symmetrically (four-points bending test) or anti-symmetrically on the beams;
dynamic tests on the AZALEE shaking table with the possibility to excite the beam in 2 axes: in
the transverse-axis and around the vertical rotation yaw-axis (see Figure 6).
The last tests allowed us to compare dissipations occurring during quasi-static tests versus dynamic
tests and to identify several factors of influence. Thus, different damping identification methods have
been selected and implemented in a home-made post-treatment routine using an optical full-field
displacement measurement technique. The main conclusions reached so far are that the energy
dissipated by an eigenmode depends on both the modal displacement amplitude and the modal
velocity.
The presentation of the IDEFIX tests and the first results will be organised as follows:
1) Preliminarily, the experimental campaign is described highlighting each key points that will be
addressed;
2) Then, the advancement of the experimental campaign and the first results and tendencies
concerning energy dissipation in the specimens will be presented;
3) Eventually, the remaining work and the prospects for a modal dissipation model will be
assessed.
1. Air cushions
2. Elastic hinges 3. Additional masses
Yaw
X
1
3
2
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 83
References
Brun, M. (2002). Contribution à l'étude des effets endommageants des séismes proches et lointains sur des
voiles en béton armé : approche simplifiée couplant la dégradation des caractéristiques dynamiques avec
un indicateur de dommage. Lyon: INSA.
Crambuer, R., Richard, B., Ile, N., & Ragueneau, F. (2013). Experimental characterization and modeling of
energy dissipation in reinforced concrete beams subjected to cyclic loading. Engineering Structures(56), 919-
934.
Heitz, T., Richard, B., Giry, C., & Ragueneau, F. (2016). Damping identification and quantification:
experimental evidences and first numerical results. 16th World Conference on Earthquake Engineering.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 84
INTERACTION BETWEEN BUILDINGS: SENSITIVITY ANALYSIS AND EXPERIMENTAL CAMPAIGN
V. CROZET, I. POLITOPOULOS, CEA, DEN, DANS, DM2S, SEMT, EMSI : Laboratoire d’études de
Mécanique sismique
Scope within SINAPS@
Pounding between buildings during earthquakes is one of the purposes experimentally investigated in
WP5. Due to the increase of the earthquake inputs reference accelerations by the safety authorities
numerous buildings for which pounding situations were neglected are now concerned. In addition, the
effect of pounding on structures is complex to characterize as only few experimental works [1], [2]
were performed on realistic scale structures subjected to pounding during earthquakes.
The scientific context
Despite profuse studies on pounding between buildings during earthquakes, conflictual conclusions
are found in both experimental studies, analytical works and reports on feedback from past
earthquakes. Moreover, previous analytical works did not provide a complete overview of the effect
of pounding since, in most cases, its influence on floor response spectra (i.e. on equipment) is not often
considered. Thus difficulties are still present to find efficient and reliable methods to evaluate and
mitigate the consequences of pounding between buildings.
Objectives of this work
The objectives of this study within WP5 are : a) to gain insight on the consequences of pounding
between buildings though both analytical and experimental works ; b) to review numerical methods
to evaluate the effect of pounding between buildings ; c) to investigate the capability of reducing the
effects of pounding or if possible to prevent buildings from pounding.
The current work progress
As mentioned, characterizing the consequences of pounding on buildings is not an easy task to achieve.
Despite numerous existing studies, difficulties remain to determine which are the parameters that
amplify the undesirable consequences of pounding. To this end, a comprehensive sensitivity analysis
has been carried out by means of Monte Carlo simulations on two single degree of freedom impacting
oscillators. Both linear elastic and nonlinear oscillators have been considered. The examined nonlinear
oscillators included elastoplastic, origin oriented and bilinear elastic mechanical behaviors. The study,
based on Sobol’ global sensitivity analysis [3], provides a consistent measure of the relative importance
of variables such as the dimensionless main excitation frequency Π 𝜔𝑒, the mass Π𝑚 and frequency Π𝜔
ratios of the oscillators and the coefficient of restitution Π𝜀. With respect to previous studies
performed in WP5 (M. Yang 2015), the dimensionless framework was reformulate to guarantee the
reliability of the applied sensitivity analysis. The consequences of pounding, on the structures
themselves, were analyzed in terms of maximum force and ductility demand amplification compared
to the case without pounding. The sensitivity analysis shows that the most influential parameter is the
mass ratio (Figure 1). Regarding the influence of pounding on the floor response spectra the quantity
of interest is the observed maximum impact impulse. For the linear elastic case the most influential
parameter is the coefficient of restitution with also high influence of the structures’ frequency and
mass ratios. However in the nonlinear cases the role of the coefficient of restitution becomes much
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 85
less important. In addition, the high influence of the structures’ frequency ratio must be tempered as
this influence on the maximum impact impulse is confined in a region where the oscillators have the
same frequency. Furthermore, as an alternative to Sobol’s method a sensitivity analysis making use of
chaos polynomial expansion was carried out and gave similar trends.
Sensitivity analysis underlined configurations of interest for the experimental campaign. Thus two mixt
steel/concrete structures (Figure 2) consisting of two storey moment resisting steel frames with
concrete slabs at each floor were designed. These structures were designed with the objective of
maximizing the detrimental consequences of pounding. The experimental campaign will divided in four
different stages. First, pull-out tests of impacting structures will be performed to characterize pounding
and provide information to identify simplified models characteristics. Second shake table tests under
earthquake excitation will be carried out to investigate pounding effects. Third, the two structures will
be linked together and tested again. The type and design of the link devices to be used will be
addressed in a later stage. Eventually, the effect of a nonlinear behavior of the impacting structures
will be studied.
Figure 7 : Sensitivity index Si for linear and nonlinear constitutive laws applied to the impacting oscillators.
Figure 2: Models on Azalee shake table
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 86
References
M. Papadrakakis, H.P. Mouzakis. Earthquake simulator testing of pounding between adjacent buildings.
Earthquake engineering and structural dynamics 1995; 24 : 811-834
A. Filiatrault, P. Wagner, S. Cherry. Analytical prediction of experimental building poundings. Earthquake
engineering and structural dynamics 1995; 24 : 1131-1154
I.M. Sobol. Sensitivity estimates for non-linear mathematical models. Mathematical modelling and
computational experiment 1993; 1 : 407-414
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 87
WP 6: Training and
knowledge dissemination
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 88
PROGRESS OF THE WP6 “TRAINING & KNOWLEDGE DISSEMINATION” IN 2016
C. BERGE-THIERRY, CEA/DEN/DM2S
Main objectives of WP6:
The objective of this WP is to offer trainings that cover all the scientific issues addressed in the
SINAPS@ project. Two sessions have been planned in the original proposal. The first training has been
delivered this year in French and for a broad audience, while the second training session in 2017 (in
English) should address an audience of professionals in the fields of research, building, earthquake
engineering, and people working for authorities and/or technical supports. For this second session, the
link with the regulations in force in France, but more generally to the international will be particularly
sought and exploited through the coordination of the ISSC-Component with IAEA (ISSC for
International Seismic Safety Center).
WP 6 was originally s organized as follows:
Phase 1 (End of 2015): Definition of the contents of the first training session. The
announcement of the training should be done mid of 2015.
Phase 2 (End of 2015 - beginning of 2016) Elaboration of the courses and tutorials.
Mid 2016: First training session
Phase 3 (End 2016 – beginning 2017) Lessons from the first training session. Definition of the
objectives and contents of the second session. Organization and realization of the second
session (advanced level) in collaboration with experts from ISSC-IAEA and other international
scientists.
Phase 4 (Mid or End 2017) : Second training session
Main advancements (2014-2016):
Phases 1 & 2: done. The first training session has been delivered as planned, May 29, 2016 –
June 3rd 2016, at the “Ecole de Physique des Houches”: it has been a great success!
SINAPS@ - First Training Session – Les Houches 2016
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 89
The pedagogical team made up of all 13 project partners had concocted a very rich program,
addressing all the scientific themes and concepts inherent in seismic risk assessment: from the
geological fault to the prediction of the seismic motion, the different approaches of estimation of the
seismic hazard were presented; then the phenomena of site effects (modifications of the incident wave
field by the superficial geological layers) and of interactions between the soil and the structures were
addressed. Finally, the calculation of the response of the structures and the equipment to the seismic
loading has been illustrated, allowing ultimately the estimation of fragility. Experimental approaches
came to an end with the presentation of the IDEFIX campaign carried out in the SINAPS @ project on
the TAMARIS platform.
This session proved really rich from the pedagogical point of view thanks to the quality of the media
and presentations, and original for several reasons: by illustrations drawn for most of the research
activities underway within SINAPS@, by the diversity of teachers being engineers, researchers and
doctoral students - academic teams but also applied organizations, and finally by the creation and
implementation of unprecedented interactive sessions. The invited lecture by P. Lestuzzi of the Ecole
Polytechnique Fédérale de Lausanne supplemented by case studies on the current building and on a
dam the problem of the quantification of margins and the seismic vulnerability of existing structures.
This training benefited to 54 participants, 70% from the SINAPS@ partnership and 30% from external
and/or foreign organizations (Algeria, Tunisia and Morocco): the audience targeted by this session was
young researchers in training (Masters, PhD students, Post-doctoral researchers), and young
engineers. Hosted by the prestigious Ecole de Physique des Houches, this first session proved to be
both scientifically and humanely rich, creating a real link between the participants in the SINAPS@
project (trainers and participants).
Information, full program and photo gallery on: http://www.institut-seism.fr/formation/sinaps2016/
Phase 3 (End 2016 – beginning 2017)
o Lessons from the first training session: done end of June 2016.
o Definition of the objectives and contents of the second session / Organization and
realization of the second session (advanced level) in collaboration with experts from
ISSC-IAEA and other international scientists: in progress.
Perspectives 2017-2018:
Organization & Realization of the second session with an international opening. The session should be
probably scheduled in 2018, as its preparation is really time consuming.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 90
Appendice
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 91
Article in Press, 2016 “Nuclear Engineering and Design”, Special Issue Post SMIRT23, Manchester 2015
TOWARD AN INTEGRATED SEISMIC RISK ASSESSMENT FOR NUCLEAR SAFETY IMPROVING CURRENT FRENCH
METHODOLOGIES THROUGH THE SINAPS@ RESEARCH PROJECT
Catherine Berge-Thierrya, Angkeara Svayb,e, Aurore Laurendeaua, Thomas Chartierc, Vincent Perrona,c, Cédric Guyonnet-Benaizea, Ejona Kishtaa,d, Régis Cottereaue, Fernando Lopez- Caballeroe, Fabrice Hollendera, Benjamin Richarda, Frédéric Ragueneaud, François Voldoireb, Fabien Bancib, Irmela Zentnerb, Nadim Moussallamf, Maria Lancieric, Pierre-Yves Bardg, Stéphane Grangeh, Silvano Erlicheri, Panagiotis Kotronisj, Alain Le Maoulta, Marc Nicolasa, Julie Régnierl , Fabian Bonillam and Nikolaos Theodoulidisn
a French Alternative Energies and Atomic Energy Commission, France b Electricité de France R&D-AMA, and Institute of Mechanical Sciences and Industrial Applications, EDF-CNRS-CEA-ENSTA UMR 9219, France c Institut de Radioprotection et de Sûreté Nucléaire, France d Ecole Normale Supérieure de Cachan, France e CentraleSupelec, Univ. Paris-Saclay, MSS-Mat CNRS UMR 8579 f AREVA, France gISTERRE, Grenoble, France h INP-Grenoble & Université Joseph Fourier, France i EGIS Industries, France jEcole Centrale Nantes, France l CEREMA, France
m IFSTTAR, France n ITSAK, Greece
ABSTRACT
The Tohoku earthquake and associated tsunami in March 2011 caused a severe nuclear accident at the Fukushima Daiichi Nuclear Power Plant, where level 7 (International Atomic Energy Agency (IAEA) - INES scale) meltdown at three reactors occurred. The underestimation of the seismic and tsunami hazards has been recognized and the seismic margins assessment of the nuclear plants remains a priority for the whole nuclear community. In this framework a five-year research project called SINAPS@ (Earthquake and Nuclear Installations: Ensuring and Sustaining Safety) is currently on-going in France. A reliable estimate of seismic margins is possible only if all uncertainties, epistemic and aleatory, are effectively identified, quantified and integrated in the seismic risk analysis. SINAPS@ brings together a multidisciplinary community of researchers and engineers from the academic and the nuclear world. SINAPS@ aims at exploring the uncertainties associated to databases, physical processes and methods used at each stage of seismic hazard, site effects, soil and structure interaction, structural and nuclear components vulnerability assessments, in a safety approach: the main objective is ultimately to identify the sources of potential seismic margins resulting from assumptions or when selecting the seismic design level or the design strategy. The whole project is built around an “integrating” work package enabling to test state-of-the-art practices and to challenge new methodologies for seismic risk assessment: the real case of Kashiwazaki-Kariwa Japanese nuclear plant, shocked by the severe earthquake in 2007 provided a rich dataset which will be used to compare with the predictions. The present paper proposes for each step of the seismic risk analysis a review of the state of practice in France in the nuclear field and then precise the objectives and research strategy of SINAPS@ to overcome identified limitations or weaknesses. Scientific issues are illustrated through preliminary results of the project.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 92
1. INTRODUCTION
1.1. International and French contexts
The SINAPS@ (Earthquake and Nuclear Plant – Ensuring and Sustaining Safety) research project is a contribution to the effort and challenges in earthquake engineering that national and international scientific community must reach. After the recent major natural events (2004 Indian Ocean magnitude (M) 9. 1-9.3 earthquake (hereafter EQ) and tsunami; 2007 Niigata Chuetsu Oki Japan, M6.8 EQ (hereafter NCOE) and 2011 Tohoku M9 EQ and tsunami) that have caused historical human disasters, huge financial losses, industrial plants and particularly nuclear ones are strongly challenged. Lessons learned from these events should allow increasing the level of safety of current and future nuclear facilities, especially through the improvement of risk assessment and associated mitigation methods.
Approaches for assessing the seismic hazard, seismic assumptions considered for the design of civil
engineering structures and equipment are associated to practices, codes, standards and rules adopted at the
national scale in countries with nuclear facilities. In France, the seismic risk is taken into account in the design
and reassessment of nuclear plants (NP) (the Fundamental Safety Rule RFS 2001-01 and the ASN Guide 2/01
published by the Nuclear Safety Authority in 2001 and 2006 respectively), whereas international references
exist such as the recommendations published by the International Atomic Energy Agency ( IAEA). However, the
March 2011 Tohoku Japanese earthquake followed by the mega-tsunami caused the major nuclear accident at
Fukushima Daiichi Nuclear Power Plant (NPP) and puts question on seismic risk practices for nuclear plant
safety. The complementary safety studies (CSS) conducted in France in the aftermath of the disaster highlight
the need for operators to implement "a hard core" of material and organizational measures to control the
fundamental safety functions in extreme situations," and in general, for their nuclear plant, "to increase as
soon as possible, beyond safety margins they already have, their robustness to extreme situations" (see ASN
CSS report (2012)). Such safety margins if any should come from the design stage or mitigation provisions if
necessary.
SINAPS@ aims to explore the uncertainties associated to data, knowledge of the physical processes and
methods that are used at each stage of the seismic hazard and seismic vulnerability of structures and
components assessment, as part of a nuclear safety approach: the main goal is ultimately to identify the
sources of potential seismic margins resulting from assumptions or when choosing the seismic design level
(i.e. taking into account uncertainties by conservative choices) or the design strategy (conservative assumptions,
choice of materials ...).
1.2. SINAPS@ : a French and multidisciplinary partnership
SINAPS@ aims to bring scientific demonstrations and to make recommendations to improve the seismic risk
management and then safety of current and future French nuclear facilities: the strength of the project is to bring
together the expertise of:
geologists, geophysicists, seismologists and statisticians to study the various components of the seismic hazard,
material modelling, soil mechanics and geotechnical specialists, civil engineers, to characterize the response of soil and structures to seismic loading,
researchers from both academic, industrial and the nuclear sector, scientists recognized as experts and heavily involved in regulatory body transcription of best practice
in seismic risk management (some strongly involved in the CSS), see Figure 1.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 93
Figure 1. SINAPS@ partnership.
2. SINAPS@ RESEARCH PROJECT PRESENTATION
SINAPS@ is a five-year project which began on September 2013, and structured around five “work packages (WP)”
that strongly interact (i) WP1, "Seismic Hazard Assessment" (SHA), (ii) WP2, "Non- linear interaction between both
near and far seismic fields, the soil and structures ", (iii) WP3, " Behavior of structures and equipment to seismic
loading, seismic isolation and reinforcement processes ", and (iv) WP4, devoted to the "Seismic Risk Assessment"
and aiming at integrating the first 3 WP’s findings. These four WP’s are mainly based on empirical and numerical
approaches. The WP5 is dedicated to the experimental laboratory approach, addressing issues for which the
databases poverty and/or lack of feedback and/or whose resolution by conventional simulation approaches
remains too uncertain. This WP 5 is based on the seismic experimental platform Tamaris CEA-Saclay (http://www-
tamaris.cea.fr/). Within the SINAPS @ project, tests involving interaction between buildings are planned. The test
results combined with simulations should allow providing important comments and/or recommendations regarding
the topic of building to building interaction.
Figure 2. SINAPS@ Scientific Work Packages Interactions
Finally, the dissemination of knowledge coming from SINAPS@ WP1 to 5 research is done through a training WP
promoting state of the art methods to assess the seismic risk for the safety of nuclear plants: one session in 2016
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 94
and a second one in 2017 under the International Seismic Safety Centre – IAEA umbrella are proposed for students,
researchers and people from earthquake engineering community.
Identifying the nuclear plant seismic margins requests a permanent dialogue between the three main domains of
seismic risk assessment i.e. (hazard, vulnerability and issues). We mention that, in the frame of the SINAPS@
project, we do not calculate the risk through an explicit convolution of the three 3 terms (whereas it is done in an
actual probabilistic safety analysis): in fact the” issues” term is only appreciated:
(i) first in the level of SHA considered (that means including a certain level of conservatism or
chosen specific return period for Probabilistic SHA (PSHA), depending on the issue and context associated
to the conducted study : for example for a conventional building without specific issue, seismic hazard will be
assessed for low return period (i.e. accepting quite high probability of exceedance of seismic intensity),
whereas for a NP usual seismic hazard assessment a 10 000 years return period at least is often considered,
and could be higher – up to 20 to 30000 years as in the complementary safety studies context, curves.
(ii) and through the choice of seismic indicators and damage criteria, used to assess the fragility
curves.
SINAPS@ governance project has established a monitoring committee, composed of national and international
experts in research and seismic engineering fields, which would assess, throughout the project duration, effective
communication between people in all work packages and consistency of research activities. The SINAPS@ project
addresses the issue of nuclear facilities (structures and components) safety only with respect to the seismic action.
This does not prejudge the dominant role of the seismic action regarding other loads (wind, snow, flood, static
constraints etc.).
3. SINAPS@’s KEY SCIENTIFIC ISSUES
In this section we present the key questions that currently exist in the frame of seismic risk analysis for nuclear
plants safety with respect to the regulation in force in France, and those raised by the Tohoku 2011 event and
Fukushima accident. This “state of practice” tentative review is not exhaustive but exhibits the topics that seem
either to need real improvement and/or to have a strong weight and impact in the risk analysis. We illustrate the
most important scientific issues addressed by SINAPS@ through some preliminary results. This review is conducting
following each step of the seismic risk analysis, from seismic hazard assessment, through site effects and soil-
structure nonlinear interactions, structural and equipment seismic behaviour assessment, and finally the fragility
evaluation. Finally the seismic margins characterization and quantification is discussed and we illustrate how the
topic is treated in the SINAPS@ project.
3.1. Key questions on the current French practice to assess the seismic hazard for nuclear plants, and associated SINAPS@ issues.
3.1.1. Current French practice characteristics and limits for SHA in nuclear field
As SHA is the input data for engineers to define the seismic loadings to design the plant or reassess its behaviour,
it is crucial to point out the strengths, weaknesses and limits of the current practice, especially in the “post-
Fukushima context”, which requires considering the occurrence of rare extreme events. SHA in France, in the
nuclear plant safety frame, is guided by a pure deterministic approach: this is the case since the first regulation
published in 1981, and maintained in its updated version in 2001 (RFS 2001-01). The RFS 2001-01 methodology
detailed in Berge-Thierry 2013 is a scenario-based approach: the predicted seismic ground motion is related to one
(or several) reference earthquake(s) characterized mainly through two parameters that are the magnitude and the
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 95
hypocentral distance. An intrinsic limitation of this deterministic approach is the absence of probability concept or
confidence level attached to the SHA result. Furthermore, the French metropolitan territory is characterized by a
spatially heterogeneous and low rate of seismicity. The SHA performed in such environment faces the scarcity of
seismic data, and when available they are highly uncertain. The way in which uncertainties are accounted in the
RFS 2001-01, mainly through “safety coefficients” does currently not allow appreciating the level of conservatism
of the assessment. Berge-Thierry et al., 2015 discussed in depth the crucial steps of the RFS 2001-01 and the sources
of uncertainties: without detailing, one can summarize as follows:
1. seismic sources definition (as identified faults or seismic zones),
2. characterization of reference events, either from historical catalog or instrumental one, from
macroseismic data or recorded motions,
3. selection of the reference events that are retained to define the seismic scenarios at the studied site,
4. geological site characterization, through a unique parameter (superficial 30 meters shear wave
velocity),
5. strong motion assessment (mean value) at the free field on site, using a unique specific Ground
motion prediction equation derived in 2003.
Although the fifth step clearly stated a mean prediction, this does not induce a final “best-estimate” or mean SHA,
because depending on how are performed the fourth previous steps, and especially depending on the uncertainties
treatment, the evaluations can strongly differ from one expert to another one. Recently, Scotti et al. (2014)
proposed to integrate and propagate epistemic and aleatory uncertainties in deterministic approach such as RFS
2001-01, but this has not yet been experienced in the frame of nuclear plants safety studies.
3.1.2. SINAPS@ issues and objectives on SHA
During the SINAPS@ project several scientific issues are addressed, among them the adequacy between the
methods used for assessing the seismic hazard and the level of knowledge about the seismicity and associated
physical processes. As previously mentioned, the deterministic approaches have been historically supported and
developed in the regulatory standards especially in countries with low seismicity, and for the nuclear power plant
safety (RFS 2001-01 in France). International references and practice (IAEA, 2010) standards and guides, the
implementation of Eurocode 8 are an incentive for the development of the probabilistic approaches, even in
countries with a deterministic tradition such as France. This tendency has been confirmed after the Fukushima
accident, leading us to question the assumptions and the methods adopted for the hazard assessment at this site
and the requirements of the Japanese regulatory standards. Complementary safety studies (CSS) performed in
France have been reviewed by European experts and they recommended to complement the classical deterministic
method by a probabilistic seismic hazard assessment. Furthermore, the probabilistic approaches are powerful and
popular especially because of their intrinsic ability to integrate the random and epistemic uncertainties. However,
the levels of hazard are stronger and are associated with large confidence intervals. The use of PSHA results then
address specific questions and new expert discussions are thus raised (recent feedbacks from the Swiss PEGASOS
project and the Thyspunt NPP PSHA study – see Bommer et al., 2014, these two project being in line with the SSHAC
procedure (1997, 2009)). The number of PSHA performed in France is quite small, and the experience and feedback
on this method remains limited. Beyond methodological debates the true issues addressed in SINAPS@’s are (i) to
systematically identify, quantify all uncertainties and clarify their treatment in the SHA (deterministic or
probabilistic), and (ii) to provide to “end-users” (researchers and engineers) reliable and adapted seismic inputs in
coherence with the methodologies used to predict site effects, soil-structure interaction (SSI) , vulnerability and
risk. Finally regarding the SHA topic, SINAPS@ objectives are to:
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 96
1. characterize the typical French data by the most appropriate and validated methods to generate
metadata and their uncertainties,
2. suggest a ranking of the key parameters in the seismic hazard assessment methodology and
estimate the associated uncertainties to guide future research,
3. assess the sensitivity of deterministic approaches such as the RFS 2001-01 and probabilistic
approaches according to the known input data,
4. provide a relevant description of the seismic hazard for the engineering needs,
5. make recommendations for the advancement of the French regulatory standards.
3.1.3. Two SINAPS@ research examples on SHA
We propose in this SHA section to illustrate SINAPS@ research through two recent works, the first related to the
“control point” and the associated "reference motion" for the site where the SHA is performed through empirical
ground motion prediction equations, and the second concerning the impact of the fault modelling (associated
parameters and uncertainties) in the predicted SHA level, using a probabilistic logic tree approach.
3.1.3.a. Characterizing the ground motion at the (hard) bedrock reference level instead of the current “standard rock (outcropping)” practice.
Among current studies conducted in SINAPS@, one should mention Laurendeau et al. (2016) ongoing works which
results should have a strong impact on seismic hazard assessment, whatever the retained approach (deterministic
or probabilistic). The final goal of Laurendeau et al. (2016) is to compare various methods to assess ground motions
for hard rock condition, and if possible to propose a specific Ground Motion Prediction Equation (GMPE). Currently,
huge uncertainties arise in the SHA from the use of classical GMPE’s through 2 sources: (i) the first one is related to
the ergodic assumption used to derive the GMPEs and (ii) the second comes from the very poor
geological/rheological of the site through the too simple Vs30m proxy. Recent works, (as the PSHA performed for
the Thyspunt NPP, Bommer et al. (2014)) propose to apply to the original GMPE’s several correcting coefficients in
order to be consistent with a fully “site specific” approach: these corrections are known as the “single station sigma”
(correcting the ergodic assumption), and the “Host to Target” correction introducing a new proxy κ0 which is a
seismic parameter representative of the high frequency attenuation site properties: this κ0 can be measured on
recorded data. Laurendeau et al. (2013) remind that in the engineering seismology community the classical model
used for high-frequency attenuation for a record at distance (r) from the source is characterized by the parameter
kappa (κ(r)). Anderson and Hough (1984) described κ(r) as being the slope of the high-frequency decay of the
acceleration Fourier amplitude spectrum in a log-linear space. The attenuation model is described as:
a(f)=A0 exp(−πκ(r)f) for f > fE
where A0 is a source and propagation-path dependent amplitude, fE is the frequency above which the decay is
approximately linear, and r is the epicentral distance. The distance dependence can then be eliminated by
extrapolating the κ(r) trend to r = 0, introducing a site-specific kappa, typically denoted as κ0, that is free of the
regional Q attenuation effect added by distance.
Laurendeau et al. (2016) consider that such an approach remains difficult to apply, especially due to the lack in
France for instance of real strong motions recorded at well characterized seismic stations, and because most of the
present GMPEs have been derived from surface strong motions which are affected by superficial soil alteration
even for “rock condition”. The strategy proposed by Laurendeau et al. (2016) to avoid these biases is: (1) in a first
way, to use strong motion recorded at depth on very hard rock condition (see star 3 on Figure 3) and to correct
them of the depth effects, i.e; correcting from "within" motion to "outcropping" motion; (2) in a second way, to
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 97
use strong motion recorded at surface on soft-soil to rock condition (stars 1 and 2, Figure 3) and to correct them
(by deconvolution) of the near surface materials response (Figure 3). Figure 4 shows the results obtained with these
two approaches, compared to results obtained using surface and downhole data without any correction. We see
that both correction approaches lead to very similar results (green and magenta spectra on Figure 4), which are
significantly lower than the “initial” dark blue spectrum which corresponds to the current practice. These
approaches may be used in a near future as an alternative methodology of the today standard “host to target”
correction involving “kappa corrections”. In the frame of SINAPS@ the computation of realistic site effect (i.e.
geometry 1D, 2D, 3D, linear or non-linear behaviour) will be specifically assessed considering the input motion
provided by the SHA for “very hard rock” which is most often found at large depth and controls the site response.
This strategy proposed by Laurendeau et al. (2016) and supported in SINAPS@, consisting in defining the SHA at
the hard bedrock level (hereafter “control point”) is clearly a strong difference with the current French practice in
the nuclear field, as the SHA is always defined for outcropping, "standard rock" condition (stars 1 or 2, Figure 3),
including the site effect roughly assessed through a coefficient included in the GMPE: for several years now, this
practice faces in France real deadlocks especially for nuclear sites exposed to “specific site effects” (that are poorly
or not predictable using the classical GMPE in force in the RFS2001-01). One should expect at the end of the
SINAPS@ project to formulate a recommendation to improve the regulation with respect to this “control point”
which connects the SHA with the other components of the risk analysis.
Figure 3. Schematic view of the site configuration and the definition of the « reference or control point » where the seismic motion is
defined in the S.H.A. study. In the current French nuclear pratice and regulation S.H.A. is always defined at the free-field (stars 1 and
2, respectively for soil and "standard, outcropping" rock condition). Laurendeau et al. (2016) propose a control point corresponding to
the deep, very hard bedrock level and elaborate specific GMPE for hard rock site conditions. The Vs30 values mentionned are those
related with the “rock and soil categories” in the RFS 2001-01.
An interesting study has been performed in SINAPS@ to check the influence of this control point definition for
ground motion on the SSI and fragility curves computations: this study exhibits the bias and even the error
introduced when considering a ground motion defined at the free field which is deconvolved at depth at the
foundations of the structure especially in case of soft soil non linear behavior under strong ground motions (Berge-
Thierry et al., 2016).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 98
Figure 4: Spectral acceleration response spectra for one given seismic scenario (Mw=6.5, Rrup=20 km, Vs30=1100 m/s) obtained using
GMPEs derived by Laurendeau et al. (2016) using Kik-net data. Dark blue line shows the results obtained using only surface data,
without correction. Light blue line shows the results obtained using only dowhole data, without correction (these results couldn’t be applied to free- surface purpose). Green line shows the results using downhole data, but corrected to be transposed to free surface
conditions. Purple line shows the results using surface data, corrected from site effect.
3.1.3.b. Impact on SHA of seismic faults parameters and uncertainties in the metropolitan French context using a probabilistic approach
Among the current practices at least in France that are challenged in SINAPS@, one of them is the way in which the
seismic source and its potential are defined to assess the hazard. Currently, in the frame of the RFS2001-01
approach the seismic source and associated ground motions are evaluated considering that the seismic energy is
concentrated on a unique point (“source point approximation”). The reality is different, a seismic source, whatever
the considered magnitude, consists in a geometrically extended fault, usually complex and segmented. Ignoring
this complexity leads to simplify the reality, and from the SHA point of view may conduct either to under or over
predict the expected strong motion, depending on the location of the site with respect to the source: some near
source effects, such as directivity (seismic waves focusing in the direction of the rupture propagation) are then
ignored while many earthquakes have experienced these effects associated to a huge variability of recorded strong
motions some of them being clearly damaging (“killer pulse”) due to a specific frequency content and large
amplitude. A first study was performed in SINAPS@ by Chartier et al. (2014) using a probabilistic approach
investigated the influence of faulting and zoning models and analyzed the impact on the hazard results of associated
hypotheses, which are (i) the geometry of the faults (ii) their seismic potential (i.e. slip rate), as shown Figure 5.
From this study located in eastern part of France, a strong variability in the PSHA results appears mainly induced by
the uncertainty on slip rate associated to the faults. This latter parameter (which is not directly used in the current
deterministic SHA French practice) is very poorly constrained in France, and should be improved through
paleoseismological studies and long term geodetic measurements. In this application the choice of the GMPE(s)
appears also crucial finally controlling the hazard level (Figure 8).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 99
Figure 5. Exploration of epistemic uncertainty through a logic tree in Chartier et al. 2014 study.
Figure 6. Sensitivity in the PSHA UHS results: influence of the slip rate value (red curves high value, green curves low value), and
considering 3 GMPEs, from T. Chartier et al., 2014.
We notice that Chartier et al., 2014 study used the NGA-1 (2008) GMPE’s because available at the date of the study,
whereas they have been upgraded by the next generation NGA-2 and published in 2014. We underline that in
France there is no requirement or guidance to select the appropriate GMPE’s among all published, especially with
respect to their publication date or feedback: this step of GMPE’s choice is crucial and impact strongly SHA results
and then induce debates between experts.
3.2. Key questions on the current French practice to account for site effects for nuclear plants, and associated issues in SINAPS@.
3.2.1. Current French practice characteristics and limits for site effects in nuclear field.
As mentioned in the previous section, the French nuclear practice and regulation require assessing the seismic
hazard on site at the free-field. Local site effects are accounted in the RFS 2001-01 by a site coefficient in the strong
motion attenuation relation prediction. This approach is valid only for simple geometry site e.g. site presenting only
vertical rheological shear wave velocity variation. Based on the Vs30m parameter two site categories are defined:
the "Rock" one being for site with Vs30m upper than 800 m/s, and "Soil" sites with Vs30m between 300 to 800 m/s.
The RFS 2001-01 stipulates that for low velocity sites (e.g. lower than 300 m/s) and for complex site geometries
(e.g. 2D and/or 3D) a specific site study is required. In such cases the nuclear operator has to perform a dedicated
site effect study to predict the strong ground motion representative of the complex behavior of soft soils and/or
complex wave field propagation and modification due to resonance or trapping phenomena. In France, and despite
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 100
a moderate seismicity, several nuclear or industrial sites are subjected to such "particular" site effects: it is the case
in ancient glacial valleys or recent sedimentary zones. At this time, several international benchmarkings have been
organized to check the capacity of scientific community to predict these effects, that are very complex due to their
dependence with the seismic level (non-consolidated soils exhibit nonlinear effects), their variability in frequency:
moreover, it is well demonstrated that the site effects will vary depending on the azimuthally position of the source
fault (in case of source close to the site). ESG 2006 focused on the Grenoble valley and a strong scattering of the
results between various numerical codes appeared (Chaljub et al., 2006). Finally, at this time, there is no consensus
neither in the scientific community nor in the nuclear one (e.g. including operators, technical expert and authority)
to practically account for and include such complex site effects, in a coherent way with respect to structural and
equipment design issues. To summarize and regarding the site effects phenomenon the current French nuclear
practice faces the following difficulties or limitations:
1. definition of two geological site categories through a unique parameter Vs30m (whose measure or
estimate, and uncertainty or variability are not specified by the rule),
2. local 1D linear site effect accounted through a site coefficient included in the GMPE (Berge-Thierry et
al. 2003),
3. site coefficients in the GMPE based on the SMDB (2000) whose site-meta data are considered has
poor, because very few stations have a Vs30m really measured on site. Especially, it can be though
that some stations initially qualified as rock sites where finally soil ones: this could introduce and over
or underestimate (depending on the frequencies) the strong motion,
4. limitations inherent to the use of an ergodic-type GMPE model (e.g. standard deviation of the GMPE
derived from various sites assigned to a specific site – when performed site-specific studies as done
in local PSHA, cf to the “host to target” and single-station sigma corrections discussed in § 3.1.3.a.),
5. for complex configuration with specific site effects suspected (2D, 3D and /or nonlinearity) no codified
approach proposed in the regulation, conducting to expert’s debates.
3.2.2. SINAPS@ issues and objectives on site effects
A key issue of SINAPS@ is to analyse the seismic risk in a more continuous way than it is currently done in the
regulation and practice. Indeed each step of the risk study is performed independently from the others, conducting
sometimes to restrictive or simplified assumptions, not necessary coherent from a step to the next, and the
uncertainties treatment is also not obvious to understand: as an illustration, we already pointed out the “control
point” problem, which is the reference where the S.H.A is performed (and then where the seismic motion through
most of the time the response spectrum is evaluated): we showed that the current practice considers this control
point at the free-field, whereas engineers need the motion at least at the foundations level or sometimes deeper
if specific large scales interaction between the soil, the wave field and the structure have to be accounted. This
practice conducts to perform deconvolutions using methods beyond their limits (e.g. equivalent linear approaches),
and that definitively should be proscribed in case of severe earthquake, for which the soil nonlinear behaviour will
be triggered. In SINAPS@, research on site effects strongly interacts with wave field generation and SSI activities
(thematic merged in SINAPS@ under “Non-linear interaction between both near and far seismic fields, the soil and
structures”) and addresses the following topics:
1. using strong motion databases that include data recorded by “downhole arrays” where the
geological/geotechnical properties are well defined (e.g. geological and wave velocity profiles, as is
the KiKNet Japanese database), so as to propose both statistical correlations and “correction
coefficients” to account for the nonlinear soil behavior,
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 101
2. depending on the selected model (especially for SSI analysis), the seismic loading can be applied: i) at
the outcropping rock, ii) at the downhole bedrock or iii) at the base of the structure. To perform non-
linear analysis of extended structures, it should be described in both time and space, and its
variability should be taken into account. Concerning the modelling of variability of the incident wave
field it is proposed:
a. the development of methods and tools for generating statistical distributions of seismic
fields in outcropping rock or in the site scale,
b. the characterization of the variability of strong motions and the influence of local site
effects by comparing the simulated signals in free field and those simulated at rock level and
propagated through a model of non-linear site, including the variability of soil mechanics
properties,
c. the development of response spectra, signals and measures of hazard at the foundation level,
accounting for the soil-structure interaction, with linear and nonlinear approaches.
3. the validation of nonlinear soil models for the strong ground motions.
The aim of this work is to validate the numerical codes that account for non-linear soil behaviour in the 1D wave
propagation with the responses of sites where strong motion were recorded (with priority for Japanese sites). The
goal is to quantify the uncertainties associated with the use of such codes for sites with moderate seismicity with
no strong motion records. This work will complete the first phase of verification/validation of nonlinear models
already performed under the SIGMA research program.
3.3. Key questions on the current French practice to account for soil and structure interaction, and associated issues in SINAPS@.
3.3.1. Current French practice and limits for SSI in nuclear field
In the French regulatory body, the SSI thematic is guided by the ASN Guide/2/01 updated in 2006. The present
standard nuclear industrial practice for soil-structure interaction calculation consists in:
the determination of the excitation signal at the foundation level from a deconvolution of a free field surface signal (signal compatible with the SHA output – usually a response spectrum), considering a vertical wave field incidence,
the modelling of the soil either as a homogeneous half space or as a media constituted of homogeneous horizontal layers. The soil characteristics are considered linear viscoelastic and are determined based on mechanical strain level computed in each layer during the deconvolution process. Several computations are usually performed varying the shear modulus, to cover for uncertainties in soil characteristics,
the determination of the structural dynamic response to the excitation using a soil model and a structural model coupled at their interface. Several sub-structuring approaches are possible at this stage. This resolution is generally performed in the frequency domain, so that interface impedances (stiffness and damping functions of frequency) can be used to model the dynamic interaction,
for few years, the representation of ground motion incoherency is also possible with the same SSI tools, e.g. open-source Code_Aster and MISS3D coupling.
This practice relies on some strong hypotheses:
the signal is well known at the surface,
the ground motion incoherency is only due to soil scattering (no wave passage effect),
the soil behaves linearly (equivalent viscoelasticity), without coupling between dilative and deviatoric
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 102
strains, though they can be considered up to 10-3,
the structure behaves linearly, no dynamic interactions between adjacent buildings are accounted for,
the soil-foundation interface is generally considered as perfect; only limited geometrical nonlinearities (uplift…) can be taken into account with a simplified approach, within transient analyses.
These hypotheses are acceptable for design or periodic reassessment surveys but are usually unable to cover the
soil-structure behavior under “extreme” seismic condition, which is necessary to assess the vulnerability of a plant
(aim of the SINAPS@ project). For example, the following physical processes are not accounted:
the soil or structure yielding (frequency shift due to stiffness degradation, energy dissipation increase and increased deformation, redistribution of efforts within the structure),
the building foundation sliding and uplifting (dissipated energy, frequency shift, breaking resonance but inducing impacts),
the nonlinear filtering (amplification and attenuation, wave conversions…) of the seismic excitation signal by the site effect, from the bedrock through the layered soil medium.
Therefore, there is a growing need, in the nuclear industry in particular, for being able to: (i) define the seismic
signal at bedrock instead of at the surface, (ii) describe the variability of ground motions characteristics by several
intensity measures. Doing so, site effect filtering would be automatically embedded into the analyses and (iii) build
robust industrial tools and validated methodologies allowing for the resolution of soil-structure problems in time-
domain – so that all kind of soil and/or structure nonlinearity could be represented – and finally (iv) propose an
easier analysis tool describing the propagation of uncertainties in the whole chain. These needs are the motivations
of the SINAPS@ “nonlinear interaction between the near and far wave fields, soil and structure” research and issues
described below.
3.3.2. SSI issues in SINAPS@
In the framework of the wave propagation from the source to the equipment at the structure, the soil and the
structure interaction topic is at the interface between the seismology and the structural dynamics, the hazard and
the structural vulnerability. Even if the soil-structure effects are well known from 70’s they have often been
considered in the design stage under simplified assumptions which are considered as conservative ones: Winkler
springs, uniform incident field, shallow, rigid foundations or linear equivalent soil behaviour …
Many of these assumptions have been improved in recent years, which allow highlighting safety margins (Mylonakis
and Gazetas 2008). These works also showed the sensitivity to high uncertainty attached both to seismic loading
and on the soil properties surrounding the structure. Moreover, so as to take into account extreme events in the
post-elastic behaviour of structures, it is necessary to have a more detailed description of the seismic loading, in
both time and space, exceeding the given maximum acceleration or regulation spectrum. Finally, the instrumental
and theoretical seismology has highlighted the complexity and variability of a field of seismic waves: near field
effect, site effects, non-linear filtering strong movements, spatial variability (Luco and de Barros 2004). These
advances build now a big picture, which combines various methods with difficulties to be associated and sometimes
inconsistent with the regulations and common methods used in the world (Pitilakis and Clouteau 2010, Cottereau
2007 and 2008).
In SINAPS@ and as mentioned in the previous section the SSI study strongly interacts with site effects and wave
field propagation modelling research. The main objectives in SINAPS@ regarding the nonlinear interactions
between near and far wave fields, the soil and the structures are to:
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 103
1. propose a global methodology based on both existing methods and tools : i) using large earthquake data bases, ii) using non-linear analyses for both the soil and structure, iii) defining the validity of simplified approaches, iv) accounting for the variability of motions and quantifying their uncertainties,
2. develop an advanced and unified computational tool to study the propagation from the source to the structure, and capable to reduce the uncertainties; a spectral element formulation (open source software SEM3D) is proposed, and will be coupled with a structural analysis software,
3. update the previous methodology incorporating this new simulation means,
4. compare these methods to the recent data so as to highlight the main sources of uncertainties; random and epistemic ones and finally propose experimental campaigns to reduce the epistemic component.
We mention that to reach these scientific objectives SINAPS@ has instrumented (since January 2014) a test site in
Greece (Argostoli): the weak and moderate motions already recorded allow to constrain the spatial variability of
motion, to propose new coherency functions, to characterized and quantify complex site effects, and will be useful
in order to verify and validate the numerical modelling of the nonlinear wave propagation in such heterogeneous
media.
3.4. Focus on the Greek, Argostoli SINAPS@ test-site: a complex sedimentary geological region, regularly producing weak and moderate seismicity. A way to strengthen nonlinear site effects, SSI effects and modelling.
In this section, we introduce the Argostoli test site and illustrate some preliminary results of SINAPS@ research in
the field of site effects, spatial variability empirical analysis, and complex full nonlinear 3D wave propagation
modelling in heterogeneous media.
3.4.1. Argostoli, Greece: an instrumented test-site of SINAPS@ project
Argostoli is the main town of the Kefalonia Island, in Ionian Sea. It is located in one of the most active seismic areas
of Europe, near a “triple point” in terms of plate tectonics. A basin, the "Koutavos" area, composed by Pliocene and
Quaternary deposits overlying Cretaceous limestones, is located a few kilometers South-East of Argostoli. This area
already benefitted of previous work conducted in the framework of the NERA European project that demonstrated
the presence of large site effects (Imtiaz, 2015; Cultrera et al., 2014; Bard et al., 2015). This basin was chosen by
the Sinaps@ project as a test-site in order to install a vertical accelerometer array, in the perspective of validation
of 3D non-linear simulation tools, since the a priori geotechnical properties of soils as well as the high seismicity of
the area are favourable to the recording of accelerograms that include nonlinear effects.
A first geophysical and geological campaign, organized in late 2013, aimed to propose an updated geological map
as well as geophysical information on the 3D structure of the basin. Methods involving ambient vibration were
used: H/V and ambient vibration arrays. This information led to a better understanding of the geology, a new
geological map and several 2D cross-sections (Figure 7), in the perspective of building a full 3D model.
Just at the beginning of the project, on the January 26th, 2014, a Mw=6.2 earthquake shook the island of Kefalonia.
It was immediately decided to launch a “SINAPS@ post-seismic survey” with two main objectives: (1) install
temporary accelerometers (“Array-1”) in anticipation to the setup of the planned permanent vertical array in order
to record possible strong after-shocks, (2) install a dense surface array in order to get a database to investigate
short-scale spatial variability (“Array-2”). This kind of database, even if it does not address the non-linearity issue,
is also essential for soil-structure interaction research performed within the Sinaps@ project, to characterize the
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 104
spatial variability and coherency of seismic motion. These databases constructed from these 2 arrays are already
available and are extensively used within the whole Sinaps@ program, as shown below.
3.4.2. Constraining site effects using the weak motions recorded at Argostoli
The temporary accelerometric network was in operation from February 3rd, 2014 (few hours after the second
strong earthquake with Mw=6.0) to July 2015 (date of installation of the permanent vertical array), and recorded
several thousands of events, with a maximum acceleration around 0,4 g. Figure 8 shows an example of signals
recorded for one earthquake and gives time domain illustration of the amplification due to site effect. A more
complete analysis of the database allowed computing standard spectral ratios (Figure 9) between the rock
reference site (station 1), three sites within the soft-soil area of the basin (station 2, 3 and 4) in order to optimize
the location of the permanent vertical array (chosen at the location of station 2 due to the largest site effect) and
two other stations along a 2D profile, located on stiff soil (station 5 and 6). Such site to reference spectral ratios
clearly exhibit the frequency band subjected to amplification. Here linear site effect is illustrated in 2D only along a
profile, but in the frame of the SINAPS@ project geometrical 3D site effects are currently studied.
Figure 7. (Top) New geological map of the Argostoli area, Kefalonia (Cushing et al., 2016), (Bottom) main cross-section section
(red line on the map above) of the Argostoli area and post-seismic accelerometric station location.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 105
Figure 8. Example of signals recorded through the basin by the temporary accelerometric network (magnitude 4.1 event located at
120km, exhibiting site effects). See the numbering of stations on Figure 7.
Figure 9. Mean Standard Spectral Ratio (SSR) on horizontal components computed on hundreds of earthquakes with a signal to noise
ratio greater than 100, using station 1 (rock site station) as reference. Stations 2, 3 and 4 are within the soft soil area of the site
(Koutavos park), station 5 and 6 are un the stiff soil area of the site.
3.4.3. Spatial coherency estimated from Argostoli dense array
Using the data recorded by the Argostoli “Array-2”, a statistical analysis of spatial coherencies of recorded seismic
ground motions was done. The dense “Array-2” was composed by 21 broadband seismometers, arranged on a five-
branch star with a maximum radius of 180 m. It was in operation over a five-weeks period. A database composed
by more than 1800 well-recorded earthquakes has been built. As example, Figure 10 displays the obtained average
incoherency of seismic ground motions for separations of 10 m, 30 m, 55 m, and 100 m for horizontal and vertical
components. According to this figure, it is interesting to note the small difference between the coherencies
estimated from horizontal components of signals and those from vertical ones. It could be explained by the fact
that the Argostoli dense array is situated on a rock site and it could be considered to have isotropic properties. It is
also remarked that the differences between the coherencies of two components are more significant for both when
the station distance increase and for frequencies greater than 6 Hz. Nevertheless, the uncertainty of coherency
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 106
increases when the coherencies decrease, i.e., when the frequency increases. For further details about these
results, refer to Svay 2015.
Figure 10. Average spatial incoherency (plane wave coherency) estimated from Argostoli database for 10 m, 30 m, 55 m and 100 m of
separations (Svay, 2015).
3.4.4. Modelling of seismic wave propagations in heterogeneous random media
In this part an overview of some results concerning the simulation wave propagation in heterogeneous
random media using SEM3D code (under development) and an application to study the spatial variability of
seismic ground motions for Argostoli site is done. For further details about these results, refer to Svay, 2015. So
as to identify the possible heterogeneity characteristic of the soils of Argostoli site, a comparison between of
the spatial coherency obtained from measured seismic ground motions with the ones from the numerical
model is performed. The profile of shear-wave velocity used to simulate Argostoli site is presented in Figure 11a.
Figure 11b. displays a schema of the numerical model of the studied case. In the model, the first layer is
considered to have heterogeneous random soil properties and the second layer (i.e. bedrock) is considered to
have homogeneous properties. The correlation model of random medium is considered to be a Gaussian one.
The algorithm used to generate the stochastic fields is based in the proposed by Shinozuka and Deodatis (1991)
and it was integrated in the code SEM3D (spectral element method) as a library and developed in the
framework of the SINAPS@ project (Paludo et al. 2016).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 107
Figure 11. a) Vs profile for Argostoli site and b) Schema of plane-wave propagation in random layer of soil (Svay, 2015). Figure 12 shows a comparison of spatial coherency of Argostoli data with the simulated one with spectral element method for two
separation distances (i.e. 10 and 30 m). It is noted that using a correlation length of medium (lc) equal to 30 m and a coefficient of
variance for the elastic shear wave velocity (CV) equal to 15% a good agreement between both coherency curves is found.
Figure 12. Comparison of obtained spatial variability of Argostoli site from measured and SEM3D simulations for a) 10 m and
b) 30 m (Svay, 2015).
3.5. Key questions on the current French practice to assess the seismic behaviour of structures and components, and associated issues in SINAPS@.
3.5.1. Current French practice and limits for structural and equipment seismic behaviour assessment in nuclear field.
In France, the ASN Guide 2/01 provides requirements and methods to design nuclear civil engineering buildings and
equipment. The key constrain is that all structure and component important for the plant safety have to remain in
the linear elastic range (q=1, q being the “behavior or reduction factor”), whatever the seismic loadings applied.
The consideration of inelastic behavior is permitted in this ASN Guide 2/01 only for structures whose behavior
requirement is the non-interaction.
Figure 13 illustrates the effect of the systematic consideration of inelastic behavior of structures allowed in the
Eurocode 8 (reference used in France for Risk Induced Industries such as chemical ones) in comparison with the
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 108
‘nuclear approach’, in which seismic action is not reduced (the nuclear and non-nuclear plants being located at the
same site). This figure demonstrates that despite an elastic design spectrum higher for non-nuclear plant (black
spectrum, Figure 13) than for the nuclear plant (red spectrum, Figure 13), the effective seismic action used to design
the structure is generally strongly lower for non-nuclear plant (blue green and magenta spectra – reduction of black
spectrum due to inelastic behavior coefficient) with respect to the nuclear plant (red spectrum not reduced).
Appreciation of seismic margins for nuclear plants cannot be restricted to the level of seismic hazard assessed at
the site using the RFS 2001-01, but also through others factors (SSI, structural behavior in the elastic linear domain)
(i) coming from the application of the ASN Guide 2/01, and (ii) from “realistic simulations” able to account for the
physical phenomena that really occur in the soil, structure and components, including nonlinearities and
interactions. The research performed in SINAPS@ regarding structural and components seismic behaviour
addresses the (ii) issue, and aims also to provide validated and efficient nonlinear structural models for the needs
of fragility curves computation in a probabilistic and uncertainties integration context.
Figure 13. Comparison, at the same site, of effective seismic actions used to design structures ('Chemical plant') using Eurocode 8 which
allows reduction of elastic response spectrum through inelastic behavior coefficients (q >1) and a nuclear plant using the ASN Guide
2/01 which requires to consider elastic spectrum.
3.5.2. SINAPS@ main issues for structural and equipment seismic behaviour assessment
The behavior analysis of structures and equipment necessarily needs the modelling of the different mechanisms
involved in the seismic wave’s transmission from the soil interaction to the equipment. In the frame of SINAPS@
and the need to consider potential extreme earthquakes (e.g. seismic loadings higher than those currently used for
design or reassessment) and to quantify the seismic margins, realistic structural modelling is unavoidable, that is
including complex 3D structural effects, nonlinearities and variabilities. During these last two decades, important
and numerous improvements occurred within the field of modelling the material and structural elements
degradation due to earthquake and cyclic loading. When 3D behaviors are expected, the numerical cost is often
very high. The necessity to account for non- deterministic features of loadings (hazard) and material scattering
guides one to the development of original approaches allowing parametrical analysis needing a large number of
simulations, keeping a refine description of local mechanical behaviors. In such a way, SINAPS@ identifies three
main tasks regarding the modelling of structures and equipment:
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 109
the first task is dedicated to the analysis of experimentations allowing for models calibration or validation.
A specific work is engaged in this project, to improve the understanding of the energy dissipation in
reinforced concrete structures; an experimental campaign is ongoing combining pseudo-static tests and
dynamic ones, using a shaking table. Interpretations and results should help one to better identify or
validate models developed in the frame of SINAPS@.
the second task is focused on the development of numerical models themselves, considering that two main
issues need to be tackled: the development properly-said of mechanical models incorporating different
nonlinear mechanisms likely to occur and the ability for existing tools to account for variability inherent in
structures and materials allowing the interaction of the different sources of uncertainty.
To ensure good representation of models regarding the failure criteria (frequency decrease, inter- storey drift, crack openings), different kind of models are developed:
o from 1D multifibre assumptions framework, describing cracks and failure within the beam theory
by adding higher order kinematic descriptions,
o to more complex situations, where the 2D description of structural elements is a mandatory step,
particularly for NP lateral shear walls. In this regard, the work of Kishta et al. 2015 aims at developing
a numerical model able to predict the different non-linear mechanisms at the structural scale. The
quantity of interest is cracking of quasi-brittle materials. This work focuses on the ability of the
numerical model in describing cracking features such as openings in an explicit and accurate way.
The present developments aim at capturing crack openings by using the Strong Discontinuity
Approach (SDA). This later is coupled with a continuous anisotropic damage model accounting
for different crack orientations and crack closure effects. A regularized version of the Dirac
distribution and the hardening parameter allows for the establishment of an enriched model
compatible with the continuous one. Numerical simulations at the integration point level and a three-
point bending test carried out on a single edge notched beam show the performances of the model.
A three-point bending test campaign on mortar beams, undergone in the LMT Cachan is used
(Oliver-Leblond et al. 2013). Square section specimens of dimension D = 70 cm and length 4D
have been tested. A single notch of depth D=2 and thickness 3 mm was sawed at the center of the
specimen before the test. The geometry and the boundary conditions are given in Figure 14. The
global results in terms of load-deexion given by the model are compared with the experimental
results are reported in Figure 15 (Left). A good agreement with the experiment is obtained. In order
to illustrate the performances of the model at capturing local information like crack openings,
the evolution of the height of the specimen versus crack opening is reported in Figure 15 (Right).
Results are given for different loading levels, at peak, 75% post-peak and 50% post-peak. The model
is able to capture quite well the local behaviour as well as the overall behaviour. Ongoing works
concern the responses of reinforced concrete elements such as ties, beams and shear walls whose
mechanical features are of main importance regarding the seismic vulnerability assessment of
buildings. In this scope, homogenised behaviour of reinforced concrete structural element
constitutive relations have been developed, in order to make feasible whole building seismic
nonlinear transient analyses, and implemented into the general purpose Finite Element Analysis
opensource software Code_Aster (Combescure et al., 2015, Huguet et al., 2016).
o Finally, model reduction techniques such as PGD framework are used in the context of reinforced
concrete structures to allow parametrical studies. The variability of materials and loadings will be
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 110
included to evaluate and order them regarding their influence on the global structure vulnerability
(Vitse et al. 2014).
at last, the third task handles the industrial structures behavior regarding the seismic risk. The general
objective is to investigate the participation of structural elements and equipment’s to the general
vulnerability of nuclear facilities and to address the different possibilities to mitigate them: the isolation
systems are particularly studied.
Figure 14: Geometry and Boundary conditions of the three point bending test
Figure 15. (Left) Global (Force and displacement) and (Right) local (crack opening and propagation) response of a concrete single
edge notch beam specimen subject to 3 point bending test (Kishta et al. 2015).
3.6. Key questions on the current French practice to assess the seismic fragility, to address the seismic margin assessment, and associated issues in SINAPS@.
3.6.1. Current French practice for fragility and seismic margin assessments in nuclear field.
From author’s knowledge, in the nuclear field, the computation of the structural and components fragility is mostly
performed during the design phase and obviously during probabilistic seismic analysis of existing plants, which can
be seen as a seismic margin approach. This part of the seismic risk analysis is not covered neither by the RFS 2001-
01 (only seismic hazard assessment) neither by the Guide ASN /2/01 which gives requirements and methods
acceptable for structural and equipment design. That means that engineers and researchers for these topics
(fragility, risk and seismic margin assessment) usually refer to international practices or texts, such as the
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 111
Performance Based Earthquake Engineering (ATC 58) for conventional civil engineering, and to EPRI guides for
nuclear facilities.
3.6.2. SINAPS@ strategy and issues on the identification of potential sources of seismic margins.
Risk assessment involves the aggregation of individual elements of the analysis chain in the convolution of the
seismic hazard and the various partial conditional probabilistic estimates of the damage, including the
vulnerabilities of structures and equipment (Figure 16), as defined in using the widespread concept of fragility
curves. It can be established during the design process as during a periodic safety review of the plant or component:
this safety plant review phase is clearly the purpose of SINAPS@ project. It can also be practiced in a systematic
framework in a probabilistic safety analysis, having defined by a preliminary study a fault tree system and associated
criteria, based on a multi-physics analysis. The approaches advocated in the committees of European experts aim
at building integrated methods of deterministic / probabilistic modeling (IDPSA) (see Zio 2014): these methods,
proposed for the safety of light water reactors, include (i) the interactions between safety related equipment and
unclassified material, and also the actions of the operators and (ii) the treatment of random and epistemic
uncertainties.
Figure 16. A continuous approach for probabilistic risk analysis from the seismic source, through site effects, soil-structure interaction
up to sensitive components, integrating uncertainties.
SINAPS@ aims to identify sources of seismic margins that could exist for an existing plant, considering they can
come from either some conservative assumptions or from the use of simplified methods or models that also include
safety coefficients. A seismic analysis integrating several conservatisms (in the methods and in the data
interpretation) results in a total seismic margin that would overestimate the “realistic” seismic response. To
evaluate potential seismic margin, a comparison between the design traditional approach and a more realistic
assessment of the plant seismic response requires taking into account nonlinear behavior of soil, materials and
structures and their interactions, but also to quantify and propagate coherently the uncertainties (from the fault to
the fragility curve computation). This issue is addressed in SINAPS@ project focusing on the demonstrative study
KARISMA, (already considered for the benchmark organized by the IAEA in 2010, see reference IAEA 2013) based
on the event experience feedback from the July 16, 2007 earthquake. This event occurred in the vicinity of the
Kashiwazaki- Kariwa NPP (hereafter KK) (see map on Figure 17), for which a rich set of data throughout the analysis
chain is available.
The purpose of this demonstrative study is to implement and evaluate the methodology leveraging various
contributions of SINAPS@ project at each step of the seismic analysis on a concrete case of nuclear plant, identifying
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 112
phenomena contributing to a "best-estimate" response (see Voldoire (2006)). The KK case concerns an event of a
higher level than those usually considered in France, but is quite comparable to some “maximal hypothesis” coming
from paleoseismicity and fully coherent in the post-Fukushima context where extreme hazards have to be studied.
It is then expected from this study demonstrative evidences (challenged against real seismic data) to establish new
methods of justification that will be implemented in the future in the French context, taking advantages of the most
relevant and recent research. This is fully in line with the French work done in 2012 by the group of experts from
Electricité de France, the Nuclear Safety Authority and its technical support the IRSN which formulated
recommendations on the mode of application of the justification methods in ASN Guide 2/01.
Works and sensitivity studies performed in the frame of the KK demonstrative case, should allow SINAPS@ partners
to comment and prioritize sources of potential margins, to check the efficiency of the new methodological
framework to be proposed by the project for the practitioners and engineering staffs in the safety margins
assessment, and to facilitate the dissemination of results.
3.6.3. The Kashiwazaki-Kariwa SINAPS@ test-case for seismic margin assessment
As introduced above, in order to advance in the seimic margin assessment of existing plant SINAPS@ has the following two objectives:
(i) set up the overall methodological approach using the various products of the project and by comparing their
implementation in a coordinated global validation work; a demonstrative case study on KK is proposed.
This benchmark has the advantage of corresponding to a real case of nuclear plant (unit 7 of Kashiwazaki-
Kariwa site) having undergone a strong earthquake beyond design criteria, for which there is a rich set of data
(measures, methods of analysis) on both the loading, the effect of site, soil-structure interaction and behaviour
of civil work structures [IAEA-TECDOC-1722 (2013)]. The exceedance of 2007 NCOE ground motions recorded
on the KK NPP with respect to the design levels (Figure 17, table on right) motivated in the international
community to better characterize and quantify seismic margins of existing NPP.
Figure 4. Chuetsu Oki 2007 EQ and KK NPP Location (Left). (Right) Table of maximal horizontal acceleration values recorded during
NCOE at the base mat level of the seven reactor buildings versus design values (data from the IAEA report, 2007).
The implementation on the KARISMA case-study will consist in nonlinear transient simulations of nuclear building
behaviour, taking into account the site effects, the interaction between soil, structure and large equipment. Finally
fragility curves and "HCLPF" [high confidence low probability of failure] by numerical methods of propagation of
uncertainties will be produced, allowing to compare several meta- models aiming at improving the efficiency of
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 113
computations and to evaluate several intensity measures chosen to parameterize the seismic ground motions. This
task is managed in 2 phases during the project (1) the initial stage where basic knowledge available at the beginning
of SINAPS@ and current best- estimate practices are used, (2) the final stage, where the major achievements of the
project SINAPS@ will be considered. Figure 18 illustrates geological data and structural Reactor Building model that
will be used as input in the initial phase. The identification of the sources (hypotheses, data, methods, uncertainties)
that impact the amount of seismic margins (or provisions) should be appreciated considering results coming the
initial and final stages.
Figure 18. (Left) General 2D stratigraphy of the KK site (SINAPS@ interpretation from IAEA Tecdoc 2013). (Right) Example of
structural model of Reactor Building Unit 7, as used during the 2010 KK benchmark, used in the intial phase of KK test case in SINAPS@.
(ii) to consolidate methods of probabilistic risk assessment, including algorithmic performance aspect for
practical studies, even if R&D activities in the disciplinary field of stochastic analysis and removal of associated
technological barriers have to be conducted in other collaborative projects (indeed, the necessary methods are
beyond the scope of the seismic hazard and concern other natural phenomena such as storms, waves...).
This KK demonstrative test case has a pivotal role in SINAPS@ due to a need of:
expertise on the seismic loading hypothesis coming from seismologists; in particular, we have to assess
the representativeness of the set of seismic ground motions which have to be selected or generated
by mathematical means, as input data for the fragility curves computation..
a numerical tool to compute non-linear soil and structure interaction, specific analysis on nonlinear site-
effects, and finally coupling SSI and site effects with structural seismic behavior numerical simulation,
nonlinear SSI and reinforced concrete structural predicting models, and expertise on methods used
to predict the seismic motion transferred from structure up to equipment.
4. CONCLUSION
Assessing seismic risk in the frame of nuclear plant safety requires a strong collaboration between researchers and
engineers covering a broad scientific domain. This article presents the state of practice in France for each step of
the seismic risk analysis, exhibits some limitations or difficulties in the application of current regulation, at least in
order to quantify the seismic margins. The SINAPS@ project focuses on a continuous analysis of completeness and
gaps in databases (all data types, from geology, seismology, site characterization and materials), of the reliability
or deficiency of models available to describe physical phenomena (prediction of seismic motion, site effects, SSI,
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 114
materials constitutive laws in nonlinear domain), and of the relevance or weakness of methodologies used to
performed seismic risk assessment. This critical analysis conducted confronting methods and available data to the
international state of the art systematically addresses the uncertainties issue. Some SINAPS@ preliminary results
are presented and discussed, among them some obtained thanks to the seismic instrumentation of the Argostoli
Greek test site: the weak motions recorded at the beginning of the SINAPS@ project allow study rheological and
geometrical site effects, the spatial seismic motion coherency: these data also help to constrain and validate
numerical simulation of seismic wave field in complex nonlinear medium. In addition to the importance of recording
high quality and qualified real seismic motions, we also show how SINAPS@ project tends to improve the prediction
of structural and equipment seismic behaviour combining the interpretation of data gathered in dedicated
experimental campaigns and numerical simulations. As complex physical phenomena interfere during the seismic
wave field travelling it is required to properly describe the full 3D nonlinear interactions on the whole path from
the fault up to the soil, the structure and equipment: in the frame of the seismic margins assessment there is also
a need to develop innovative and efficient methods validated against real data as proposed in SINAPS@ with the
KK case study. This work performed in SINAPS@ is essential to provide background information on regulatory
approaches currently applied in France to account for seismic risk in nuclear safety context (RFS 2001-01 to estimate
site specific seismic hazard level, and the Nuclear Safety Authority ASN Guide 2/01 specifying the provisions of
seismic design of civil works and equipment, and acceptable methods for estimating the seismic response of
structures interacting with the equipment). Products of the project should able formulating recommendations
about the potential applicability of new approaches accounting more realistically for the actual behavior of soil,
structures and materials under high seismic loadings, with explicit integration of uncertainties. SINAPS@ challenges
to provide the best data, improve, validate and disseminate new methodologies for seismic risk assessment,
quantifying their reliability, and finally enabling authorities to take consolidated decisions from an economical and
societal point of view. Then all the actors should take advantage of the SINAPS@ research results to better justify
the level of safety of the facilities and to define the appropriate arrangements to maintain over time this safety
level.
ACKNOWLEDGMENTS
The work carried out under the SINAPS@ project receives French funding managed by the National Research
Agency under the program “Future Investments” (SINAPS@ reference No. ANR-11-RSNR- 0022). SINAPS@ is a
SEISM Institute project (http://www.institut-seism.fr/en/). Authors thank the reviewers for providing interesting
comments allowing to clarify and improve the article.
REFERENCES
Anderson, J. and Hough, S., 1984. A model for the shape of the Fourier amplitude spectrum of acceleration at high
frequencies. Bulletin of the Seismological Society of America, Vol. 74, N°5, pp1969– 1993.
ASN Guide 2/01, Prise en compte du risque sismique à la conception des ouvrages de génie civil des installations nucléaires
de base, à l'exception des stockages à long terme des déchets, radioactifs”, 2006. Nuclear Authority Safety website
(http://www.asn.fr/index.php/Divers/Autres-RFS/Guide-ASN-Guide-2-01-ex- RFS-V.2.g).
ASN CSS report : Evaluations Complémentaires de Sûreté, rapport de l'Autorité de Sûreté , 2011. Nuclear Authority
Safety website (http://www.asn.fr/index.php/L-ASN-en-region/Division-de- Marseille/Actualites-de-votre-
region/Rapport-de-l-ASN-sur-les-evaluations-complementaires-de-surete-ECS).
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 115
Bard, P.-Y., A. Imtiaz, C. Cornou, E. Chaljub, G. Cultrera, A. Rovelli, P. Bordoni, F. Cara, G. Di Giulio, G. Milana, V. Pessina,
M. Pisciutta, M. Vassallo, G. Calderoni, N. Theodoulids, A. Savvaidis, K. Makra, D. Kementzetzidou, K. Pitilakis, E. Riga, F.
Gelagoti, D. Fäh, J. Burjánek, S. Parolai, T. Boxberger, P. Moczo, J. Kristek, F. Hollender, C. Guyonnet-Benaize and M.
Cushing, 2015. Comparison between data and numerical models, Deliverable D11.6, NERA Activity JRA1: Waveform
modelling and sitecoefficients for basin response and topography, EC FP7 project 262330, 213 pages.
Berge-Thierry C. Cotton F., Scotti O., Griot-Pommera D.A. and Fukushima Y., 2003. New empirical response spectral
attenuation laws for moderate European earthquakes. Journal of Earthquake Engineering, Vol. 7, N° 2, pp 193-222.
Berge-Thierry C., 2013. Seismic Hazard Assessment and Uncertainties Treatment: Discussion on the current French
regulation, practices and open issues, Proc. of the Probabilistic Safety Assessment (PSA) of Natural External Hazards
Including Earthquakes Workshop, Nuclear Regulation NEA/CNRA/R(2014)9, Prague, Czech Republic, pp
275-287 (https://www.oecd- nea.org/nsd/docs/2014/csni-r2014-9.pdf).
Berge-Thierry C., L. Bollinger and F. Hollender, 2015. Seismic hazard assessment for nuclear plants: current practices in
France, uncertainties and key scientific issues, Proc. of the workshop “Best Practices in Physics-based Fault Rupture Models
for Seismic Hazard Assessment of Nuclear Installations”, Vienna, Austria.
C. Berge-Thierry C., Wang F., Feau C., Zentner I., Voldoire F., Lopez-Caballero F., Le Maoult A., Nicolas M. and F. Ragueneau,
2016. The SINAPS@ French Research Project: first lessons of an integrated seismic risk assessment for nuclear plants
safety, submitted to the 16th World Conference on Earthquake, Santiago Chile, January 2017.
Bommer, J.J., K.J. Coppersmith, R.T. Coppersmith, K.L. Hanson, A. Mangongolo, J. Neveling, E.M. Rathje, A. Rodriguez-
Marek, F. Scherbaum, R. Shelembe, P.J. Stafford & F.O. Strasser, 2014. A SSHAC Level 3 probabilistic seismic hazard analysis
for a new-build nuclear site in South Africa. Earthquake Spectra, VOL. 31, N°2, pp 661-698, doi:
http://dx.doi.org/10.1193/060913EQS145M.
Chartier T., Clement C., Baumont D., Scotti O. and Baize S., 2014. Probabilistic Seismic Hazard Assessment In the Upper
Rhine Graben, Eastern France, Oral presentation at the American Geophysical Union, Abstract on
https://agu.confex.com/agu/fm14/webprogram/Paper28894.html.
Chaljub E., C. Cornou and P.Y. Bard, 2006. Numerical benchmark of 3d ground motion simulation in the valley of Grenoble,
french alps, Proc. of the Third International Symposium on the Effects of Surface Geology on Seismic Motion Grenoble,
France, 30 August - 1 September.
Code Aster, http://www.code-aster.org
Code MISS3D, http://www.mssmat.ecp.fr/mssmat/moyens/moyens_techniques_logiciels/miss%5d
Combescure C., Dumontet H. and Voldoire F., 2015. Dissipative Homogenised Reinforced Concrete (DHRC) constitutive
model dedicated to reinforced concrete plates under seismic loading, International Journal of Solids and Structures, Vol.
73-74, pp78-98, doi:10.1016/j.ijsolstr.2015.07.007
(http://www.sciencedirect.com/science/article/pii/S0020768315003030).
Cottereau R, Clouteau D, Soize C., 2007. Construction of a probabilistic model for impedance matrices, Computer Methods
in Applied Mechanics and Engineering; 196(17–20), pp 2252–2268, doi: 10.1016/j.cma.2006.12.001.
Cottereau R, Clouteau D, Soize C., 2008. Probabilistic impedance of foundation: Impact of the seismic design on uncertain
soils, Earthquake Engineering & Structural Dynamics, Vol. 37, N°6, pages 899–918, DOI : 10.1002/eqe.794.
Cultrera, G., Andreou, T., Bard, P.-Y., Boxberger, T., Cara, F., Cornou, C., Di Giulio, G., Hollender, F., Imtiaz, A.,
Kementzetzidou, D., Makra, K., Savvaidis, A., Theodoulidis, N. & The Argostoli Team (2014). The Argostoli (Cephalonia,
Greece) experiment. Proc. of the Second European Conference on Earthquake Engineering and Seismology (2ECEES),
Istanbul, Turkey, August 24-29, Paper #2265.
Cushing M.E., Hollender F., Guyonnet-Benaize C., Perron V., Imtiaz A., Svay A., Mariscal A., Bard, P.Y., Cottereau R., Lopez
Caballero F., Theodoulidis N., Moiriat D. and Gélis C. 2016. Close to the lair of Odysseus Cyclops : the SINAPS@ postseismic
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 116
campaign and accelerometric network installation on Kefalonia island – Site effect characterization experiment, Proc. of
the 7th International INQUA Meeting on Paleoseismology, Active Tectonics and Archeoseismology (PATA), 30 May to 3
June, Crestone, Colorado, USA,
https://www.researchgate.net/publication/303864531_Proceeding_of_the_7th_International_INQUA_Me
eting_on_Paleoseismology_Active_Tectonics_and_Archeoseismology).
Huguet M., Kotronis P., Erlicher S. and Voldoire F., 2016. Prediction of Cracking in RC Walls Under Cyclic Loading by Means
of a Novel Global Constitutive Model, Proc. of the 9th International Conference on Fracture Mechanics of Concrete and
Concrete Structures, doi: 10.21012/FC9.105.
IAEA 2007. Online Report entitled, “Preliminary findings and lessons learned from the 16 July 2007 earthquake at
kashiwazaki-Kariwa NPP”, International Atomic Energy Agency
(http://www.iaea.org/inis/collection/NCLCollectionStore/_Public/40/010/40010606.pdf).
IAEA 2010, Seismic Hazards in Site Evaluation for Nuclear Installations, 2010. International Atomic Energy Agency, Specific
Safety Guide SSG-9,
http://www-pub.iaea.org/MTCD/publications/PDF/Pub1448_web.pdf).
IAEA-Tecdoc-1722 2013, Review of Seismic Evaluation Methodologies for Nuclear Power Plants Based on a Benchmark
Exercise, 2013. International Atomic Energy Agency, IAEA-TECDOC-1722, (http://www-
pub.iaea.org/MTCD/publications/PDF/TE-1722_web.pdf).
Imtiaz, A., 2015. Seismic wave field, spatial variability and coherency of ground motion over short distances: near source
and alluvial valley effects. PhD Thesis, defended on January 6, 2015, University of Grenoble, 358 pages
(https://tel.archives-ouvertes.fr/tel-01148138/file/IMTIAZ_2015_archivage.pdf).
Kishta E., Richard B., Giry C. and F. Ragueneau, 2015. Plate formulation based on strong discontinuity method for
reinforced concrete components subjected to seismic loadings, Proc. of COMPLAS XIII,
(http://congress.cimne.com/complas2015/admin/files/fileabstract/a983.pdf).
KikNet Strong Motion Database, National Research Institute for Earth Science and Disaster Resilience, Japan,
http://www.kyoshin.bosai.go.jp/
Laurendeau A., F. Cotton, O.J. Ktenidou, L.F. Bonilla and F. Hollender, 2013. Stiff-soil/Rock Site Amplification: Dependency
on VS30 and kappa, Bulletin of the Seismological Society of America, Vol. 103, pp 3131-3148 ; doi :10.1785/0120130020.
Laurendeau A., Bard P.Y., Hollender F., Ktedinou O.J., Foundotos L., Hernandez B. and V. Perron, 2016. Prediction of
reference motions (1000<Vs<3000 m/s) from corrected Kik-Net records of the local site effects, Proc. of the 5th IASPEI /
IAEE International Symposium: Effects of Surface Geology on Seismic Motion, August 15-17, 2016.
Luco, J.E., F. C. de Barros, 2004.Assessment of predictions of the response of the Hualien containment model during forced
vibration tests, Soil Dyn. Earth. Engng., Vol. 24, pp 1013–1035, doi:10.1016/j.soildyn.2003.12.006.
Mylonakis G. and Gazetas G., 2008. Seismic soil-structure interaction: beneficial or detrimental, J.
Earthquake Engineering, Vol. 4, N°3, pp 277-301, doi: 10.1080/13632460009350372.
Oliver-Leblond, C., Delaplace, A., Ragueneau, F. and Richard B., 2013. Non-intrusive global/local analysis for the study of
_fine cracking, International Journal for Numerical and Analytical Methods in Geomechanics, Vol. 37, N°8, pp 973–992 doi:
10.1002/nag.2155.
Paludo L., Bouvier V. and R. Cottereau, 2016. Scalable parallel scheme for sampling of Gaussian random fields over large
domains, submitted for publication in International Journal for Numerical Methods in Engineering.
Pitilakis D. and D. Clouteau, 2010. Equivalent linear substructure approximation of soil-foundation- structure interaction:
model presentation and validation, Bull. Earth. Engng. Vol. 8, N°2, pp 257–282, doi : 10.1007/s10518-009-9128-3.
PROCEEDINGS OF THE 2016 SINAPS@ PLENARY SESSION | 117
RFS2001-01, Règle fondamentale de sûreté n°2001-01 relatives aux installations nucléaires de base. Détermination du
risque sismique pour la sûreté des installations nucléaires de base, 2001. Nuclear Authority Safety website
(http://www.asn.fr/index.php/Divers/Autres-RFS/RFS-2001-01).
SSHAC (1997). Recommendations for probabilistic seismic hazard analysis: guidance on uncertainty and use of experts (No.
U.S. Nuclear Regulatory Commission Report, NUREG/CR-6372), U.S. Nuclear Regulatory Commision Report, NUREG/CR-
6372. Washington, D.C.
SSHAC(2009), Implementation of the SSHAC Guidelines for Level 3 and 4 PSHAs—Experience Gained from Actual
Applications By Thomas C. Hanks, Norm A. Abrahamson, David M. Boore, Kevin J. Coppersmith, and Nichole E. Knepprath
Open-File Report 2009-1093, USGS.
Shinozuka, M. and Deodatis, L., 1991. Simulations of stochastic process by spectral representations, Applied Mechanics
Reviews, Vol. 44, N° 4., pp 191-205, doi:10.1115/1.3119501.
Scotti O., Clement C. and D. Baumont, 2014. Seismic Hazard for design and verification of nuclear installations in France:
regulatory context, debates issues and ongoing developments, Bollitino di Gefisica Theorica ed Applicata, Vol. 55, N°1, DOI
10.4430/bgta0080.
SIGMA Project, Seismic Ground Motion Assessment, http://www.projet-sigma.com/index.html.
Svay A., 2015. Numerical assessment of spatial variability of seismic ground motions effects on SSI analyses. Ph.D. thesis
annual report. EDF – CentraleSupélec.
Vitse M., Neron D. and Boucard P.A., 2014. Virtual charts of solutions for parametrized nonlinear equations, Computational
Mechanics. Vol 54, N° 6, pp 1529-1539, doi:10.1007/s00466-014-1073-6.
Voldoire F., 2006. Computational methods applied to the seismic margin assessment of power plants and equipment, Proc.
of the First European Conference on Earthquake Engineering and Seismology, Geneva, Switzerland, paper number: 1038.
Zio Enrico, 2014. Integrated deterministic and probabilistic safety analysis: concepts, challenges, research directions,
Nuclear Engineering and Design, 280:413-419, pp1-7, doi: 10.1016/j.nucengdes.2014.09.004.
Proceedings of the 2015 SINAPS@ Plenary session
ANR-11-RSNR-022