SEISMIC SSI ANALYSIS COMPARISON BETWEEN DETAILED ... - …

Transactions, SMiRT-24

BEXCO, Busan, Korea - August 20-25, 2017

Division X

SEISMIC SSI ANALYSIS COMPARISON BETWEEN DETAILED

AND DISCRETIZED MODELING OF AN AUXILIARY CONTROL

BUILDING

Samer El-Bahey1, PhD., P.E., Stephane Damolini2, Todd Radford1, Ph.D., Cory Figliolini2

1 Senior Engineer, JENSEN HUGHES, USA 2 Principal Engineer, JENSEN HUGHES, USA

ABSTRACT

In recent years, the nuclear industry and the Nuclear Regulatory Commission (NRC) have made a

tremendous effort to assess the safety of nuclear power plants (NPPs) as advances in seismology have led

to the perception that the potential earthquake hazard in the United States may be higher than originally

assumed. The Seismic Probabilistic Risk Assessment (S-PRA) is a systematic approach used in NPPs in

the U.S. to realistically quantify the seismic risk by identifying the dominant contributors to seismic risk

and core damage. One major aspect of realistically quantifying the seismic demands is by performing

Seismic soil-structure interaction (SSI) analysis.

SSI analysis of NPPs has been historically performed in the frequency domain using a lumped-

mass stick model of the structure to capture the global dynamic response of the system required to

calculate realistic in-structure response spectra (ISRS). With the recent advances in computer software

and hardware technology, it is now possible to perform SSI analysis of detailed finite element structural

models in the frequency domain.

This paper presents the results of a seismic SSI analysis of an Auxiliary/Control Building (ACB)

using both lumped mass stick and detailed 3-dimension Finite Element (FE) representations of the

structure. Because of the complexity of the structure, the stick model consists of multiple interconnected

sticks developed using GT-STRUDL and calibrated against a detailed FE model of the structure with a

fixed base developed using ANSYS before introducing the SSI effects. SSI effects were analyzed using

EKSSI for the stick model and ACS/SASSI for the detailed FE model.

The results of the detailed FE model in terms of response spectra, are calculated and compared

against those of the lumped-mass stick model.

24th Conference on Structural Mechanics in Reactor Technology


Division IX (include assigned division number from I to X)

INTRODUCTION

Seismic Probabilistic Risk Assessment (S-PRA) studies have been performed in many of the US

Nuclear Power Plants over the last two decades. The S-PRAs were initially performed to answer safety

concerns in heavily populated areas, then evolved to satisfy the NRC’s request for information regarding

severe accident vulnerabilities in Generic Letter 88-20, Supplement 4 USNRC., (1991). The NRC

encourages the use of PRA for making risk informed decisions and has developed a Risk-Informed

Regulation Implementation Plan USNRC., (2000) and associated regulatory guides. Most of the initial S-

PRAs performed in the US in the 1980s, contained a level of uncertainty arising from the seismic hazard

and uncertainty in the fragilities of structure, systems and components (SSCs) which resulted in the

spread of the level of uncertainty in the calculated Core Damage Frequency (CDF).

Following the March 2011 Great Tahoku Earthquake and its catastrophic consequences on the

Fukushima Daiichi NPP, it was clear that relying on uncertainties in the design could lead to catastrophic

consequences. From which, the Nuclear Regulatory Commission (NRC) established a Near Term Task

Force (NTTF) to conduct a systematic review of NRC processes and regulations and to determine if the

agency should make additional improvements to its regulatory system. The NTTF developed a set of

recommendations intended to clarify and strengthen the regulatory framework for protection against

natural phenomena. Subsequently, the NRC issued a 50.54(f) letter on March 12, 2012 requesting

information to assure that these recommendations are addressed by all U.S. nuclear power plants. The

50.54(f) letter requests that licensees and holders of construction permits under 10 CFR Part 50 re-

evaluate the seismic hazards at their sites against present-day NRC requirements and guidance.

Advances in characterizing earthquake source, travel path, and local site effects have led to the

perception that the potential free field earthquake hazard in the United States may be higher than

originally assumed. The effect of SSI is yet still a major uncertainty in the seismic design of nuclear

power plants.

SOIL-STRUCTURE INTERACTION OVERVIEW

The ground motion observed by any structure is different than the free field motion due to the

following interactions:

• Inertial Interaction: Inertia developed in the structure due to its own vibrations gives rise to

base shear and moment, which generates displacements and rotations of the foundation relative to the

free-field due to the flexibility of the soil-foundation system. This added flexibility affects the building

frequency by shifting it towards the flexible range. The system overall damping is also affected by the

added displacements as energy dissipation via radiation damping and hysteretic soil damping rises

affecting the overall system damping.

• Kinematic Interaction: The presence of stiff foundation elements at or below the ground surface

cause foundation motions to deviate from free-field motions as a result of ground motion incoherence,

wave inclination, or foundation embedment.

Commonly used methods for capturing the SSI effects are either:

Direct Analysis: where the soil and superstructure are included in the same finite element model

and analyzed as one system. This could be performed using multiple SSI software like FLUSH by

representing the soil as a continuum along with foundation elements. The direct analysis method is rarely

used in practice due to the computational complexity.




Substructure Approach: where the structure is initially analyzed having a fixed-base, from which

the dynamic characteristics of the structure are calculated including the modal frequencies, Eigen vectors,

and Eigen values. The kinematic effects are then addressed using frequency dependent transfer functions

relating the free-field motion to the foundation input motion (FIM) taking into account the soil column

properties. The inertial interactions are then addressed by calculating frequency dependent impedance

functions to represent the stiffness and damping of the soil-foundation interface depending on the soil

column properties. The superposition inherent in a substructure approach requires an assumption of linear

soil and structure behavior, although in practice this requirement is often followed only in an equivalent-

linear sense.

AUXILIARY AND CONTROL BUILDINGS

The chosen Auxiliary Building (AB) is a multi-story, structural steel and reinforced concrete

structure which houses the safety injection system, residual heat removal system, CVCS monitoring

system, auxiliary feedwater pumps, steam and feedwater isolation and relief valves, heat exchangers,

other pumps, tanks, filters, and demineralizers, and heating and ventilating equipment.

The chosen Control Building (CB) is a rectangular structural steel and reinforced concrete

structure which houses the access control areas, control room, upper and lower cable spreading rooms,

electrical and mechanical equipment rooms, and locker rooms. The intermediate floors and roof are

reinforced concrete slabs supported by structural steel beams and girders. The floor and roof framing are

supported by exterior reinforced concrete bearing walls and interior steel columns.

The AB shares a common base mat and wall with the CB. The building interior is enclosed on

one side by the reactor building wall.

SITE CONDITIONS AND SEISMISITY

The geologic column underlying the chosen site consists of soil deposits over a sequence of older,

competent sedimentary lithified rock formations. The soil deposits are related to glacial and postglacial

processes during the Quaternary Period and consist of loess, clay, and till. In the area of the Nuclear

Island, these deposits are removed, and as needed, replaced by engineered fill. The uppermost competent

rock layer is the Graydon Chert Conglomerate. Underlying the Graydon Chert Conglomerate are the

Burlington and Bushberg formations of Mississippian age. The Burlington Formation consists of

limestone. The Devonian-age Snyder Creek Formation underlies the Bushberg Formation throughout the

site area.

No faults have been identified within 12 miles of the site, and no historic earthquake epicenters

have been reported within approximately 30 miles. The closest earthquake of any size was a md = 2.3

event at a distance greater than 20 miles from the site.

SITE-SPECIFIC GROUND MOTION AND SOIL PROPERTIES

In accordance with the 50.54(f) letter and following the guidance in the SPID EPRI., (2013A), a

probabilistic seismic hazard analysis (PSHA) was completed in a separate effort using the recently

developed Central and Eastern United States Seismic Source Characterization (CEUS-SSC) for Nuclear




Facilities (CEUS-SSC, 2012) together with the updated EPRI Ground-Motion Model (GMM) for the

CEUS EPRI., (2013B). For the PSHA, a lower-bound moment magnitude of 5.0 was used, as specified in

the 50.54(t) letter. Information pertaining to the Hazard Consistent Strain-Compatible Properties for upper

bound, UB, best estimate, BE, and lower bound, LB, soil cases are obtained from the PSHA and used

herein.

The site-specific ground motion considered herein is based on the new Ground Motion Response

Spectrum (GMRS) developed as part of the PSHA effort along with its associated Foundation Input

Response Spectrum (FIRS) and the Soil properties for the Lower-Bound (LB), Best Estimate (BE) and

Upper-Bound (UB) cases as shown in Figure 1. The properties include hazard-consistent strain-

compatible properties (HCSCPs) for the full column profile for FIRS hazard level.

Artificial time histories corresponding to the FIRS are generated herein based on the fitting and

enveloping requirements of NUREG. (1987), Section 3.7.1 Option 1 Approach 2 at 5% damping.

A convolution analysis is then performed using the SOIL module of ACS/SASSI for each

combination of earthquake direction and soil profile. The developed artificial FIRS hazard level

foundation input time histories are input as outcrop motions in soil column convolution analyses, to

generate surface time histories. Time histories are convolved to the surface of the soil profile for the X, Y,

and Z directions using the appropriate hazard consistent LB, BE, and UB soil properties as shown in

Figure 2.

Surface time histories are then converted to 5% damped Response Spectra and compared against

the Review Level Earthquake Surface Response Spectra (RLESRS) that corresponds to the GMRS as

shown in Figure 3. It can be seen from the plots that the envelope of the RS from the 3 time histories for

each horizontal direction is a good match to the horizontal RLESRS (RLESRS_H). The envelope of the

RS from the 3 time histories for the vertical direction is a good match to the vertical RLESRS

(RLESRS_V). The conservatism in higher frequencies results from additional soil column amplification

above the outcrop and it will not significantly affect the structure response. The comparisons verify that

there is no significant deficiency in the envelope of the time histories when compared to RLESRS.




Figure 1. Full Column Profile for shear-wave velocity (Vs), compression-wave velocity (Vp), and

shear




Figure 2. Artificial Time Histories Corresponding to the FIRS (X, Y, and Z)




Figure 3. Surface Time Histories Response Spectra compared against the RLESRS at 5% Damping




STRUCTURAL MODEL DESCRITIZATION

To include eccentricities in a lumped mass stick model between centers of gravity and centers of

stiffness as well as to adequately capture local vertical modes in the Control Building where floor systems

may be relatively flexible, the response of the Auxiliary/Control building was captured through

decoupling the horizontal and the vertical responses through two structural fixed base GT-STRUDL

models.

The horizontal model developed is a three-dimensional stick model and shown in Figure 4. The

mass eccentricities were accounted for at each major floor elevation by lumping the floor masses at their

respective centers of mass and connecting them through rigid members to the vertical sticks located at the

centers of rigidity. The model is used to calculate the natural frequencies of the structure and its

eigenvectors.

No Auxiliary/Control horizontal model concrete stick elements were anticipated to be

significantly cracked at the Ground Motion Response Spectrum (GMRS) review level earthquake (RLE).

Therefore, no reduction in stiffness parameters for these elements are necessary per the guidance of

ASCE 4-13 Table 3-2 ASCE., (2013).

Figure 4. Schematic Diagram for Enhanced Horizontal Model of Auxiliary/Control Building

Although there are no major structural irregularities in the ACB, the floor system in the Control

Building is relatively flexible. The aforementioned features made it necessary to model the ACB in the

vertical direction considering two sticks connected through horizontal beams. The intent of the vertical

stick model is to adequately capture the global wall and column modes in addition to the local Control

Building floor modes. One vertical stick represents the Auxiliary Building and the outer structural walls

of the Control Building and the other stick represents the Control building supporting steel columns.

Vertical floor flexibility is represented by the horizontal members. A schematic of the vertical model is

shown in Figure 5.




The 1984’ elevation of the Control Building was found to have vertical frequencies greatly

exceeding 20 Hz and therefore modeled integrally with the ACB at that elevation. The length of the

horizontal elements representing the floor system was taken to be very small (0.1 ft) to eliminate the

unrealistic rocking effect due to the eccentricity of the model from the chosen coordinate system. Floor

inertias were adjusted accordingly. Given the framing of the slabs into perimeter walls, the composite

floor beams were considered fixed to the structural walls; whereas they were modeled pinned to the

columns since the spans on opposite sides of the columns may feasibly oscillate out of phase. The

horizontal translational degrees of freedom were restrained for all nodes. Effective inertias of the

perimeter walls were also calculated in order to account for their additional vertical floor flexibility.

Figure 5. Schematic Diagram for the Auxiliary/Control Vertical model

LUMPED MASS STICK MODEL SSI ANALYSIS

The EKSSI computer programs used herein for SSI analysis were developed by Professor

Eduardo Kausel of the Massachusetts Institute of Technology (MIT), and verified by Stevenson and

Associates (S&A). The EKSSI software package includes multiple modules. The following two modules

were used for the current analysis. The SUPELM program module computes the frequency-dependent

dynamic impedance of the foundation. The foundation is assumed to be rigid and cylindrical in shape,

which is reasonable. SUPELM can also compute transfer functions allowing for the determination of time

histories at the bottom of the foundation using the SUPELM KININT module. The EKSSI program

module provides the frequency domain solution, including SSI effects, to a dynamically-loaded structure

that is supported on compliant soil. The EKSSI program performs the SSI analysis by combining the

building model and the foundation impedance matrix, subjecting the combined model to input

acceleration time histories, and determining the response at required nodes.

Fixed-base modal properties for the ACB are calculated using GT-STRUDL software. The time

histories applicable to the free field surface are calculated using SPECTRASA software.

Impedance functions for the substrata are calculated using SUPELM. The transfer functions are

used by the KININT module to generate time histories at the foundation bottom.




The structural model and the foundation impedance functions are combined in EKSSI to form the

soil-structure interaction model. The models are then analyzed in EKSSI using the input time histories.

Resultant response time histories are calculated separately in the X, Y, and Z directions at all levels of

interest. Structural inherent damping was considered at 4% for un-cracked concrete structures.

DETAILED FINITE ELEMENT SEISMIC MODEL

A best estimate fixed base seismic model of the ACB was first created using the ANSYS finite

element analysis package as shown in Figure 6 and a modal analysis was performed. Structural walls and

floors were individually modeled with shell elements. Beam elements were used for columns and for floor

framing. Structural dead load mass and distributed equipment load were taken into account. The

maximum mesh size was set to 48 inches for all 2D elements. Composite action for floor framing was

credited where applicable. Material densities were used for dead loads and adjustments to mass were

made to account for modeling simplifications. Mass for fixed equipment and live load were also added.

Fixed boundary conditions at the base of the foundation elements (top of bedrock) were applied.

Figure 6. ACB ANSYS Model Overview

DETAILED FINITE ELEMENT MODEL SSI ANALYSIS

The ACS/SASSI program was used for the SSI analysis of the detailed FE model to account for

coupling between the structure and the soil, effects of flexibility in the underlying media, and effects of

flexibility in the structural basemat. Embedment of the ACB was considered and the flexible volume sub

structuring approach was used. A fixed-base version of the created ACS/SASSI model is compared to the

fixed-base ANSYS model for validation of the conversion. The transfer functions for a representative set

of nodes of the converted ACS/SASSI model were compared to the ones of the original ANSYS model.

The ACS-SASSI model was modified to be a fixed-based model similar to the ANSYS model. To

simulate a fixed base in the ACS/SASSI model, the volumic soil elements and the embedded soil layers

were removed and the ground elevation was set at the foundation of the building. All the remaining layers

of soil were turned into extremely stiff hard rock. The same properties were applied to the half-space. A




comparison between the transfer functions from ACS/SASSI and ANSYS, for the fixed-based model is

presented in Figure 7 at the Control Building Elevation 2090’ for the X, Y and Z transfer functions.

Figure 7. ACS-SASSI and ANSYS fixed-base models X, Y and Z transfer functions, CB El. 2090’




Preliminary runs were performed using the Fast-Flexible Volume method. For this method, the

interaction nodes were comprised of the excavated volume boundary nodes, along with user-determined

internal node layers within excavation volume. This method is a good compromise between

computational resources and modelling refinement.

For each soil case, for each direction, results were inspected for reasonableness. Out of the 9

analyses (3 soil cases, 3 directions each), the Lower Bound X-direction transfer functions and response

spectra were not satisfactory, as the transfer functions had high, unexpected spikes, at around 8.2Hz. This

is attributed to local soil modes, which can result from a Fast-Flexible Volume analysis, as not all the

excavated volume nodes were defined as interaction nodes.

A full Flexible Volume analysis was then performed instead of the Fast-Flexible Volume for the

Lower Bound, X-direction. Results were inspected and no unexpected spikes in the transfer functions

were observed.

SEISMIC RESPONSE ANALYSIS COMPARISON

Average floor In-Structure response spectra for the detailed FE Model were generated by creating

a nodal envelope of the LB, BE, and UB response and then averaging all nodal envelopes across each

floor elevation for each of the three directions. The In-Structure response spectra for the lumped mass

stick model was generated by enveloping the LB, BE, and UB responses at each node representing a floor

elevation for each of the three directions. Comparison between results for select floor elevations are

shown in Figure 8 and Figure 9. The maximum observed accelerations for select floors are presented in

Table 1. The comparisons indicate a generally close agreement between the calculated maximum

accelerations at major floor elevations specially in the horizontal direction. However, in terms of the

calculated in-structure response spectra, the results of the two models slightly differ despite an overall

similarity in spectral shapes.

Table 1. Comparison of Maximum Accelerations

Location

LMSM FEM

Maximu

m Sa (g)

Maximu

m Sa (g)

X Y Z X Y Z

C

ontrol

1

974’

0

.79

0

.83

2

.82

0

.72

0

.84

0

.82

1

984’

1

.03

1

.08

2

.56

0

.88

0

.88

1

.87

2

000’

1

.31

1

.25

2

.38

1

.1

1

.33

2

.93

A

uxiliary

2

047’

2

.47

2

.15

1

.41

2

.15

2

.21

1

.98




Figure 8. ISRS Comparison between LMSM and FEM for the Auxiliary Building




Figure 9. ISRS Comparison between LMSM and FEM for the Control Building

CONCLUSIONS

A 3-D detailed FE model of an Auxiliary Control Building was developed and analyzed to

evaluate the seismic SSI response using ACS/SASSI. The FE model consisted mainly of solid/shell

elements representing concrete floors, walls and basemat. The model considers spatial distributions of

mass and stiffness in the structure, including the flexibility of the basemat, walls and floors. In addition,

the model is made sufficiently refined to capture the out-of-plane flexural response of floors and walls.

The results of the detailed model at key locations in the structure have been computed and compared with

those of a comparable lumped-mass stick model typically used in SSI analysis. The comparisons indicate

a generally close agreement between the calculated global responses, such as maximum accelerations at

major floor elevations. However, in terms of the calculated in-structure response spectra, the results of the

two models slightly differ despite an overall similarity in spectral shapes. The differences in the spectral

responses are more pronounced in the vertical response of the structure. This may be attributed to the

flexibility of the floors, walls and basemat; the spatial distribution of mass in the structure and the limited

number of modes that can be represented by the stick model.




REFERENCES

ASCE., (2013), “Seismic Analysis of Safety-Related Nuclear Structures and Commentary, ASCE 4-13”

EPRI., (2013A), “Seismic Evaluation Guidance: Screening, Prioritization and Implementation Details

(SPID) for the Resolution of Fukushima Near-Term Task Force Recommendation 2.1: Seismic.”

EPRI 1025287, Palo Alto, CA

EPRI., (2013B), “Ground-Motion Model (GMM) Review Project.” EPRI 3002000717

NUREG. (1987), “Commission, N.R., NUREG-0800,". Standard Review Plan for the Review of Safety

Analysis Reports for Nuclear Power Plants."

USNRC., (1991), “Supplement 4, Individual Plant Examination of External Events (IPEEE) for Severe

Accident Vulnerabilities—10CFR 50.54 (f)(Generic Letter No. 88-20)”, Rockville.

USNRC., (2000), “Risk-Informed Regulation Implementation Plan”, Washington, DC: Document US

NRC SECY-00-213

Documents

SEISMIC SSI ANALYSIS COMPARISON BETWEEN DETAILED ... - …