Elsevier

Engineering Structures

Volume 161, 15 April 2018, Pages 207-222
Engineering Structures

SMART 2013: Lessons learned from the international benchmark about the seismic margin assessment of nuclear RC buildings

https://doi.org/10.1016/j.engstruct.2018.02.023Get rights and content

Highlights

  • An international benchmark tocompare seismic assessment methods started in 2011.

  • A synthesis of the work carried outby 42 participating teams is presented.

  • The main lessons and conclusions from the SMART 2013 project are exposed.

Abstract

In this paper, the main findings and conclusions drawn from the second international benchmark named SMART 2013 and jointly organized by the French Sustainable Energies and Atomic Energy Commission (CEA) and Electricité De France (EDF) within the framework of a wide research program entitled “Seismic design and best-estimate Methods Assessment for Reinforced concrete buildings subjected to Torsion and nonlinear effects” (SMART) are presented. A 1:4-scaled reinforced concrete (RC) specimen, representing a part of a nuclear auxiliary building and designed according to French guidelines for a PGA level equal to 0.2 g, was subjected to shaking table tests; results of this experimental campaign are used as reference data for this benchmark. The input ground motions considered in the seismic loading sequence are mainly natural bi-axial signals (main shock and aftershock) recorded during the Northridge earthquake that took place in California, USA in 1994 and have a PGA (Peak Ground Acceleration) about 1.8 g. These high-intensity seismic loadings allow assessing the relevancy of nonlinear numerical models when they have to deal with strong nonlinearities due to concrete cracking. The results produced by the 42 teams which participated in the international benchmark show that (i) the dynamic behavior of the specimen is well captured when dealing with the design level, (ii) the displacement are underestimated when dealing with the beyond design behavior, (iii) the peak frequency shifts are well captured and (iv) the damaging effect of the Northridge aftershock is almost null. Last, seismic safety margins of the specimen are quantified by two mechanical indicators; the results confirm the fact that the RC specimen which was designed according to the codes applicable in the French nuclear industry, exhibits noticeable good performance level regarding collapse prevention.

Introduction

When dealing with reinforced concrete (RC) structures for which the main function is to ensure the energy production, a specific attention is paid to assess (new buildings) or reassess (existing buildings) their safety level. The case of nuclear power plants (NPPs) is highly monitored in order to anticipate, manage and, ideally, cancel the consequences and the effects that would be caused by an accidental event. When designing the structure of such specific buildings, one can observe a trend which lies in considering a lateral force resisting system (LFRS) based on the combination of shear walls and beams. The reason for this choice is mainly due to the fact that this design strategy tries to combine the dissipative and ductility properties of frame structure with the stiff character of purely wall based structures. The story-drifts can be controlled in the lower levels of the structure because of the stiff nature of RC shear walls. On the contrary, the frames increase the dissipative capability of the whole building that leads to an increase of the displacement response of the structure. When this type of structures is regular or even slightly irregular, a consensus on the confidence level related to the assessment methodologies is nowadays accepted in the international earthquake engineering community. However, the case of highly irregular frame-wall structures needs to be investigated, especially in the nonlinear behavior range. Indeed, geometric, mass, stiffness, and strength irregularities may lead to three-dimensional effects such as torsion coupled with bending, increasing with the eccentricity between the torsion center and the mass center.

The safety quantification of such complex RC structures and related equipment regarding the seismic risk requires (i) to assess the seismic safety margins defined as the distance between a limit state expressed as a load bearing capacity and a structural response, (ii) to estimate with an acceptable confidence level the floor response spectra (FRS) useful for equipment seismic reassessment and (iii) to take into account uncertainties related to the input ground motions and to the input material parameters used to calibrate probabilistic structural models. Therefore, numerical analyses should be carried out to estimate as best as possible the responses of a given structure considering extreme seismic loadings for which the intensity measure (IM) overcomes the ones required at the design stage. In addition, because an accurate knowledge of the beyond-design dynamic behavior of the structure is needed to quantify the seismic safety margins, it is also needed to use nonlinear laws to describe the dissipative mechanisms related to the constitutive materials. These dynamic assessment methods are known as “best-estimate approaches” and aim at describing as accurately as possible the physics involved in the degradation process of the structure when an extreme seismic scenario occurs. Furthermore, the relevancy of advanced nonlinear models is not only related to the material parameters to be considered but also to the variability of the input ground motion used to perform the structural assessment. Material parameters and input ground motions are all subjected to uncertainties that should be taken into account. From the aforementioned discussion, it is obvious that improvements in the fields of nonlinear modeling as well as uncertainties modeling propagation are research fields of primary importance for the earthquake engineering community.

The past decades were marked up by major events that gathered the earthquake engineering community along the same path of improvements in the research fields of structural dynamics of low span RC shear walls and related assessment methodologies. The former Nuclear Power Engineering Corporation of Japan (NUPEC) organized a similar international benchmark as the one reported in this paper, under the auspices of the Nuclear Energy Agency (NEA) of the Organization of Economic Cooperation and Development (OECD), twenty years ago. The RC structure under consideration was regular and U-shaped with low span shear walls. The main conclusions were that advanced nonlinear dynamic methods still had to be improved, in particular when dealing with overdesign seismic ground motion leading to the structure working close to its ultimate limit state [1], [2], [3], [4], [5]. Some years later, between 1996 and 2002, an extensive experimental campaign was conducted in the scope of the CAMUS research program at the French Atomic Energy and Sustainable Energies Commission (CEA). A symmetric in plane five story RC wall 1/3th scaled mock-up [6], [7] was subjected to shaking table tests to improve the knowledge of this type of RC structures [8]. The acquired experimental data were used in two international benchmarks, held in 1998 and in 2003, to assess the predictive capabilities of existing methodologies. It appeared that the seismic safety margins were frequency dependent; this conclusion was confirmed by the related numerical simulations. In 2006, a blind prediction contest on the seismic response of a 7-story full-scale RC building with cantilever structural walls acting as the LFRS was launched by the Network for Earthquake Engineering Simulation (NEES), Portland Cement Association and University of California at San Diego. The objective of that research program was to check the seismic response of RC wall systems designed for lateral forces by means of a displacement-based design methodology to emphasize the interaction between the walls and the slabs [9], [10], [11], [12], [13], [14]. To address the complex issue of the safety margins quantification in case of strongly irregular RC structures regarding the seismic risk, a wide research project named Seismic Design and Best-Estimate Methods Assessment for Reinforced Concrete Building Subjected to Torsion and Nonlinear Effects (SMART) started in 2006 by the CEA and Electricité De France (EDF), under the auspices of the International Atomic Energy Agency (IAEA). The first part of this project, named SMART 2008, included both an international benchmark [15] and an experimental campaign [16], [17] based on seismic tests carried out on the AZALEE shaking table, as part of the TAMARIS experimental facility operated by the Nuclear Energy Division (NED) of the CEA. A mock-up representing a typical simplified part of an electrical nuclear RC building at the 1/4th scale was designed and built, according to the well-known Cauchy-Froude similitude law. The trapezoidal three-story specimen was composed of three walls with openings forming a U-shape and was designed according to the French current nuclear engineering practice, that means the use of response spectrum method and Eurocode 2 criteria for reinforcement design, at the design earthquake level. Seismic inputs of increasing intensity up to a maximum Peak Ground Acceleration (PGA) of 0.9 g were applied to the mock-up; these synthetic accelerograms were generated from the design spectrum. Research revealed the existence of seismic safety margins. As seismic safety margins quantification needs to consider structural response indicators, several indicators were analyzed [15]. Nevertheless, two key points were identified as potential sources of improvements. The first one was related to the choice of the input ground motions, which induced a progressive damage of the RC specimen that is not fully representative of a natural seismic scenario and the second one was related to the way of controlling the boundary conditions, in particular at the interface between the shaking table and the RC mock-up.

Faced with the success of the SMART 2008 project and given the interest of the earthquake engineering community in addressing this complex issue, CEA and EDF decided in 2011 to extend the SMART project by starting a second part, named SMART 2013, with the aim of improving the aforementioned aspects. A new experimental campaign including shaking table seismic tests was carried out in July 2013 with the AZALEE shaking table, on a RC specimen having a similar shape as the one tested during SMART 2008. Pictures of the SMART 2013 RC specimen are shown in Fig. 1.a and 1.b. More precisely, the SMART 2013 RC specimen is a 1/4th scaled mock-up representing a part of a nuclear auxiliary building. It was designed in 2010 according to the French practice and guidelines prevailing in 1998 [18] and 2006 [19]. The design PGA is equal to 0.2 g. The only difference between the SMART 2008 specimen with the one studied in SMART 2013 is related to the design of the foundation. This part of the specimen has been reinforced in order to improve the specimen/shaking table connection. A detailed description of the design assumptions can be found out from [20].

The seismic bi-axial loadings were based on an extreme seismic scenario composed of three main sequences: the design signal (synthetic – PGA = 0.2 g), a main shock (natural – PGA = 1.78 g) and the first aftershock (natural – PGA = 0.37 g), both recorded at the Tarzana Cedar Hill monitoring station, during the Northridge earthquake that occurred in California, USA, in 1994 [21]. Such high PGA levels were chosen to examine the beyond design behavior of the SMART 2013 RC specimen. In addition, the boundary conditions at the interface between the SMART 2013 RC mock-up and the shaking table were particularly well controlled and specific data related to the shaking table dynamic behavior itself were also acquired. The experimental measurements were used in a second international benchmark devoted to the assessment of the beyond design responses of the RC mock-up subjected to the aforementioned seismic scenario and to the quantification of its vulnerability within a probabilistic framework.

The SMART 2013 international benchmark was organized between February 2012 and September 2014 and was concluded by an international workshop which took place in Saclay, France from 25th to 27th November 2014 [22]. 42 participating teams from all over the world were registered. The list of the participants is provided in Appendix A, including the parts of the benchmark they were involved in.

The objectives of the benchmark were (i) to assess the capabilities of advanced best-estimate methods in predicting the seismic response of a complex RC specimen, subjected to beyond design dynamic loadings that may occur in case of extreme seismic events; in particular the capabilities of nonlinear numerical models to capture the structural damage from a natural seismic scenario consisting of a main shock and an aftershock in a satisfactory way for a given magnitude/distance couple, (ii) to improve the use of probabilistic methodologies addressing random and epistemic uncertainties to estimate the fragility curves, (iii) to share about seismic assessment methodologies and attempt to build a consensus within the international seismic engineering community. To reach these objectives, the results from the SMART 2013 experimental campaign were extensively used.

The SMART 2013 international benchmark was composed of four stages. Stage 1 was devoted to the characterization of the numerical models used by all the participating teams. Several data regarding the spatial/time discretization, the time integration algorithms used and the ways of taking the boundary conditions into account were asked to the participants. A description of the structural model was also required. In order to assess the relevancy of the assumptions considered in the constitutive laws formulations (concrete, steel and steel/concrete interface), a description of the effects taken into account was required. Therefore, each participant was asked to carry out basic static tests considering more or less complex (both monotonic and cyclic) loading paths on a representative volume element (RVE) of concrete, steel and RC. No dynamic loading was considered in stage 1. Stage 2 aimed at calibrating the numerical finite element (FE) structural models in the elastic range. In order to reach this objective, modal analyses considering various boundary conditions and transient analysis were required: modal properties and time history responses at various points. Only two low-intensity seismic loadings, with PGA equal to 0.1 g, were considered: a random signal (run #6) and a synthetic seismic signal (run #7) corresponding to 50% of the design seismic loading in terms of PGA. Both measured seismic inputs and outputs were given to the participants. In order to allow the participants to control the boundary conditions accurately, they were provided with displacements and accelerations time histories measured at the shaking table actuators. In addition, CEA also provided them with a numerical FE model of the AZALEE shaking table accounting for the exact position of the actuators to allow an accurate description of the whole dynamic system (RC specimen and shaking table). In stage 3, blind nonlinear dynamic computations for medium to high-intensity seismic loading sequences (7 successive seismic motions, with PGA ranging from 0.2 to 1.78 g) and corresponding time history responses at various points were asked to the participants. The nonlinear analysis of 7 seismic loadings, 2 being optional, was required. Only the seismic inputs were provided to the participants. The measured outputs were not available when stage 3 was ongoing, this strategy enabled to analyze the predictive capabilities of the assessment methodologies used by the participants. Finally, stage 4 was devoted to a numerical vulnerability analysis of the RC specimen within a probabilistic framework addressing random and epistemic uncertainties. The purpose of this stage was to assess the effect of the type of uncertainties on the fragility curves considering various failure criteria and engineering demand parameters. Two sub-stages were considered. In the first one, the numerical model was assumed to be linear elastic. Participants were free to use their own methodology to compute the fragility curves. In the second sub-stage, participants had to consider nonlinear constitutive laws to describe the energy dissipation and the failure mechanisms. The methodology to compute fragility curves was imposed, assuming a lognormal distribution of the random variables. For all sub-stages, the set of input ground motions was provided.

This paper aims at presenting the main conclusions and findings from the SMART 2013 international benchmark, emphasizing the assessment of the seismic safety margins of the RC specimen that was studied. To reach this aim, this paper is outlined as follows. In Section 2, the panel of numerical models developed by the participating teams is presented. In particular, the modeling assumptions related to the nonlinear constitutive laws used are presented. In addition, the results from the calibration stage are also summed up and discussed. Section 3 is devoted to the estimation of the seismic safety margins of the SMART 2013 RC structure. Two indicators are considered in order to describe the structural responses under seismic loadings. The results obtained by the participants are compared with the experimental ones and are positioned with respect to different damage thresholds in order to (i) quantify the seismic safety margins of the specimen and (ii) assess the capability of nonlinear models to corroborate the experimental observations under a blind contest. At the end of this section, the results obtained within the framework of stage 4, dedicated to the vulnerability analysis through fragility curves calculations, are presented and discussed. In Section 4, the conclusions from the SMART 2013 international benchmark are drawn regarding several aspects such as (i) the key assumptions to consider in order to ensure the relevancy of a nonlinear structural model when the assessment of the seismic safety margins is aimed, (ii) the capability of advanced nonlinear FE models to make predictive seismic assessments in case of a strongly irregular RC structure and (iii) the relevancy of the two indexes studied to assess the seismic safety margins.

Section snippets

Modeling assumptions, constitutive laws and initial structural calibration

In this section, the modeling assumptions made by the participants to deal with the required analyses are presented. The various modeling strategies used are briefly presented before focusing on some local results allowing the characterization of the nonlinear constitutive laws used to describe the material response of steel, concrete, or reinforced concrete. In addition, the ability of the numerical models to describe the seismic response of the specimen under low-level dynamic loading in a

Assessment of the seismic safety margins

In this section, the results from the structural robustness and probabilistic vulnerability analyses are presented. The results provided by the participants were post-processed in order to quantify the seismic safety margins by means of two mechanical indicators that are defined in the following. The analysis was conducted seismic sequence by seismic sequence. The analysis presented in this section allows (i) quantifying the seismic safety margins, (ii) studying the damaging effect of an

Main findings and lessons learnt from the international benchmark

In this section, the main key points highlighted during the benchmark that have been discussed with the benchmark’s participants are reported. The main findings reported herein have been shared with the benchmark’s participants during the final workshop jointly organized by CEA, EDF and partially endorsed by the IAEA at CEA center located in Saclay (France) in November 2014.

Concluding remarks

In this paper, the main findings from the international benchmark SMART 2013 carried out from February 2012 to November 2014 and jointly organized by CEA and EDF under the auspices of IAEA within the framework of the research project “Seismic design and best-estimate Methods Assessment for Reinforced concrete buildings subjected to Torsion and nonlinear effect” (SMART) have been presented. The objectives of the benchmark were (i) to quantify the ability of advanced nonlinear approaches to

Acknowledgements

The Authors would like to thank all the institutions which participated in the SMART 2013 project and in the International Benchmark for the huge work carried out. CEA and EDF are gratefully thanked for material and financial supports. The work reported in this paper has also been supported by the SEISM Institute (http://www.institut-seism.fr).

References (32)

  • N. Ile et al.

    Nonlinear analysis of reinforced concrete shear wall under earthquake loading

    J Earthquake Eng

    (2000)
  • Bisch Ph, Coin A. The CAMUS research programme. In: Proceedings of the 11th european conference on earthquake...
  • Queval JC, Combescure D, Sollogoub P, Coin A, Mazars J. CAMUS experimental program. In: Plane tests of 1/3rd scaled R/C...
  • M. Panagiotou et al.

    Shake-table test of a full-scale 7-story building slice. Phase I: Rectangular wall

    J Struct Eng

    (2011)
  • J.D. Waugh et al.

    Lessons learned from seismic analysis of a seven-story concrete test building

    J Earthquake Eng

    (2010)
  • P. Martinelli et al.

    Simulation of the shaking table test of a seven-story shear wall building

    Earthquake Eng Struct Dyn

    (2009)
  • Cited by (13)

    • Robust energy-based model updating framework for random processes in dynamics: Application to shaking-table experiments

      2022, Computers and Structures
      Citation Excerpt :

      The test sequence consists in an alternation of bi-axial gradually damaging seismic inputs of increasing level and of random ground motions with low acceleration level chosen such that the first eigenmodes of the experimental system are excited but without adding further damage to the RC specimen. Complementary information can be found in [3–5,69]. In spite of the huge numerical efforts made for modeling the experimental SMART2013 test-results, the recorded data itself has never been used for model updating purposes.

    • Evaluation of seismic torsional response of ductile RC buildings with soft first story

      2021, Structures
      Citation Excerpt :

      Experimental studies on a nuclear structural model (shear-wall-based plan asymmetric RC building with vertical irregularity) are carried out considering synthetic [49] and recorded [50] bi-directional ground motions. For a similar nuclear structure, considering uncertainties associated with the numerical methods and different modeling strategies (ranging from line model to 3D finite element models), numerical studies investigated such structure's seismic behavior for higher excitation levels [51,52]. The numerical models exhibited a high robustness level for both global and local damage indicators compared to the experimental results of nuclear structural model.

    • Damage development analysis of the whole nuclear power plant of AP1000 type under strong Main-aftershock sequences

      2021, Nuclear Engineering and Design
      Citation Excerpt :

      Wang et al. (2019a, 2019b) show that a shield building will enter a working state of plastic damage if subjected to beyond design basis main-aftershock sequence earthquakes. Many more studies in this field can also be found in the literature (Banci et al., 2018; Lo Frano and Forasassi, 2010; Medel-Vera and Ji, 2016; Richard et al., 2015, 2016b,a, 2018; Saouma and Hariri-Ardebili, 2019). The seismic performance of the whole NPP under single earthquakes has been studied in the literature (Bausys et al., 2008; Politopoulos et al., 2015; Kumar and Whittaker, 2017).

    • Influence of various parameters in the seismic soil-structure interaction response of a nuclear power plant

      2020, Engineering Structures
      Citation Excerpt :

      Dynamic analysis of NPP buildings is required for design and construction approval. Different approaches can be found in recent literature [2–6] in which the present nonlinear modelling strategies used for RC NPP are well summarized by [7]. On the other hand, the seismic soil structure interaction can be dealt with different approaches, initially categorized as (a) Substructure methods and (b) Direct methods [8,9].

    View all citing articles on Scopus
    View full text