ML20147H512
| ML20147H512 | |
| Person / Time | |
|---|---|
| Site: | Oyster Creek |
| Issue date: | 02/26/1988 |
| From: | Berger E, Fierz H, Kluge D BASLER & HOFMANN, SWITZERLAND |
| To: | |
| Shared Package | |
| ML20147H481 | List: |
| References | |
| NUDOCS 8803080508 | |
| Download: ML20147H512 (25) | |
Text
_ _ _ _ _ _ _ _
, kb
(
PREDICTIVE RESPONSE COMPUTATIONS FOR VIBRATION TESTS AND EARTHOUAKE OF MAY 20, 1986 USING AN AXISYMMETRIC -
FINITE ELEMENT FORMULATION BASED ON THE COMPLEX RESPONSE METHOD AND COMPARISON WITH MEASUREMENTS - A SWISS CONTRIBUTION j E. Berger*, H. Fierz* and D. Kluge**
I ,
INTRODUCTION l
i Background l
^
A research project on the validation of current soil-structure interaction (SSI) methodologies - jointly sponsored by the U.S. Nuclear Regulatory Comission (USNRC) and the Electric Power Research Institute (EPRI) - has been under way sirce the be-ginning of 1986. Argonne National Laboratory (liNL) is coordinating the project on behalf cf the USNRC and issued a Statement of work (Ref. 1) to participating univer-sities and national laboratories in April 1986.
Basler & Hofmann (B&H) as consultant in civil safety matters to the Swiss Federal Office of Energy, Nuclear Safety Department (HSK) became aware of this EPRI/USNRC l Validation Project during the "Workshop on Soil-Structure Interaction" in Bethesda, j Maryland (Ref. 2). Because the firm has made extensive use in t.ie p6st 15 years of l its in-house SSI computer codes in connection with seismic reviews of several nu- !
clear installations in Switzerland, the e was a great interest in this project to validate the firm's own SSI methodology. Therefore, the USNRC was asked whether a )
Swiss participation in the project would be possible.
The USNRC welcomed such a participation, so that EPRI/USNRC Validation Project could benefit from the experience of other countries with other methodologies. Subsequent-ly, B&H prepared a proposal to HSK (Ref. 3) to fund an HSK Validation Project and to participate on the regulatory side in the overall project. This led to a coopera-
- Technical Staff, Basler & Hofmann, 8029 Zurich, Switzerland
- Project Manager, Federal Office of Energy, Nuclear Safety Department, 5303 Vdrenlingen, Switzerland gga 188n 888886 p PDR
i axisymmetvic finitt element computer program BHLUSH/BHNOUND (Ref 5). The pretteted response was reported in Progress Report for Phase 1 (Ref. 6) to ANL (overall pro-ject coordinator), HSK and USNRC', Wgk on Phase 1 continued from December 1986 to June 1987.
Phase 2. The measured vibration test data were received from ANL and the predicted response f om Phase 1 was compared with the measurerrents. The differences were interpreted'with the hid of' partnetric studies using the computer model established inphase 1.BasedoniherctultsofthesestudiesarefinedmodelsetCwasdeve-loped. Phase 2 to:k place dukit.g July 1987. <
Phase 3. The model set C from Phase 2 was then used to predict the response of the soil-structure system (containment model and surrounding soil) subjected to the ea-thquake motions recorded at a point in the free-field during the event of May 20, 1986. The same ;;omputer program as in Phase 1 was used. The results from Phases 2 and 3, i.e. the comparison of measured and predicted forced vibration responses, the results of the parametric studies, the development of model set C and the prediction of the earthquako respond were reported in Progress Report for Phases 2 & 3 (Ref 7) to ANL,4SK and USNRC. Phase 3 took place during August 1987, and the pre-i dicted earthquake response was sen l to ANL on magnetic tape on August 31, 1987.
Phase 4. The measured earthquake response data were received from ANL at the begin-ning of October 1987 and compared with the predicted response from Pht.se 3. One ad-ditional parameter variation was investigated to interprete some of the differences observed. The c'e,sults of the comparison and the parameter variation togtther with the major tir fings from Phases 1 through 3 were sumarized in a Final Report for the HSK Validatio'n Project (Ref. 8) to ANL, HSK and USNRC. This work together with the preparation of this papet wn., completed by the end of November 1987.
} ' l Q Ff.ofPaper [ ,
In this paper are pre unted the major findingrj ad results form the four phases of this project. The ag. lysis methods and the model sets used in the predictive re-sponst,;omputat)onsteedescribedinthenexttwosections.Theyarefollowedbya sect'on Nach on the comparisun of predictions with measurements f or the forced vi-brathn iests and for thiearthquake of May 20, 1986. Finally, a section on conclu-
)
sions 20mpletes this paper' ,
j.
f i j 1 !
The variation of the ground motions both horizontally and vertically in the discre-tized region of the site is b6sically considered by means of the discretized finite element representation of the soil-structure system.
As far as the earthquake generated wave pattern in the ground is concerned, the mo-tions in the discretized soil-structure system depend on the definition of the earthquake grourd motion in the free-field, that is at the boundary of the discre-tized soil-structure systen. The SSI-methodolgy used in this project can account for vertically propagating plane waves (shear- and compression waves) in the free-field only. A horizontal variation of the ground motion in the free-field is not possible.
Hence, each point on the surface in the free-field has the same motion, which may be an acceptable approximation in the immediate region around a structure, l The three dimensional representation of the 1/4-scale containment model is given by I the axysimmetric formulation of the structural model. Energy transmitting boundaries i at the vertical edges of the discretized soil-cylinder assure the proper dissipation of energy created by the soil-structure interaction effects, i.e. radiation damping j in a horizontal direction is considered. The bottom boundary of the finite element mesh consists of a rigid base. Hence, to account for proper radiation damping in the vertical direction, this boundary has to be chosen deep enough, so that the influ-ence of reflecting waves on the response of the structure can be neglected.
In the following, a brief description of the free-field and the SSI-analysis methods used in :s projut are given.
Free-Field Analysis The free-field analysis used in this project assumes vertically propagating plane waves in a horizontally layered soil deposit limited in depth by a rigid base (Ref. l
- 12, 13). Laterally the layers extend to infinity. The computaties were made with l
the computer code FLUSH.F77 (Ref. 9, 14). The objectives of the free-field analysis were threefold:
. Determine the strain-compatible dynamic soil properties in the free-field, so that the effects of the non-linear soil behaviour (primary non-linearity) can be accounted for in the SSI-analysis.
. Show the variation of the earthquake ground motion with depth in the free-field.
4 . Deconvolve the given earthquake acceleration time-histories in the free-field from the surface to the depth of the rigid base of the finite element mesh used in the SSI-analysis.
DESCRfPTION OF MODELS General An overview of the SSI model sett used in this project is given in Table 1, which contains for each model set the number of subsets used and their identification, the data packages available and explanatory comments.
Table 1 OVERVIEW OF SSI MODEL SETS USED IN HSK VALIDATION PROJECT No. of Subset Data Model Subsets in Identifi- Packages Set Each Set cation Used* Comments A - - -
Not Considered in <
HSK Validation Project B 4 B)
(a),(b),(c) Prediction of Response for (d),(e),(f) Forced Vibration Tests, Radial Excitation on Roof of Structure B
2 (a),(b),(c) Dyn. Soil Properties as in (d),(e),(f) Subset B3 , but Change of Soil and (g) Properties in Backfill B
3 Dyn. Soil Properties as in Sub-set B 2with "No Embedment" B
4 Dyn. Soil Properties as in Subset B2 with "Realistic Embedment" C 3 C hi (a),(b),(c) Prediction of Response for Horiz.
(d),(e),(f) Earthquake Excitation with "Re- l and alistic Embedment" !
(g),(h),(i) 2 C
VI C
VI Prediction of Response for Vert. l Earthquake Excitation with "Re- !
alistic Ertedment '
C h2 (a),(b),(c) Dyn. Soil Properties as in Sub-(d),(e),(f) set C but Reduction of Shear Wave (g),(h),(1) VelocNy Profile by 10 % l and (j)
- See List of References 1
The material damping of reinforced concrete at the low strains expected during the forced vibration tests was estimated to be 0,8 percent (Ref. 25). This value was used in the predictive response computations for the forced vibration experiment.
For the earthquake of May 20, 1986, a material damping value of 3,0 percent was used due to the 1:rger strains expected during the earthquake. This value is somewhat lower than that proposed for concrete under OBE loading conditions in the USNRC Reg.
Guide 1.61 (Ref. 26).
Soil Model used in Model Set B The soil model established for model set B was documented in detail in Ref. 6.
Hence, only the major features are summarized here. The data used was supplied in Data Packages (a), (b), (e) and (f) (Ref. 15, 16, 19 and 20).
Basic Model Recuirements. To investigate soil-structure interaction effects with the computer program BHLUSH/BHB0VND, the soil model has to satisfy certain requirements.
Firstly, the "ground surface" of the soil model needs to be level. Information on surface features in the area of the containment model indicated that a level ground surface may be assumed.
Secondly, the "soil deposit" needs to be horizontally layered. Hence, each layer has to be constant in thickness and extend to infinity in a horizontal direction. The seismic refraction survey (Ref.16) and the boring information from the subsurface investigations showed that the site is layered fairly horizontally so that the as- l sumption of horizontal layers is justified. l Thirdly, the method of analysis requires a finite "depth" of the soil model. For very deep soil profiles, it is therefore necessary to specify a rigid base deep enough in the soil profile, so that its influence on the response of the structure can be neglected. Because the rock level at the site appears to be at a depth of about 400 m below the ground surface, the soil profile was chosen 62 m deep. It could be shown, that a soil profile with this depth subdivided into 22 layers as shown in Figure 1 was adequate for the forced vibration analysis and that the rigid base had no influence on the response of the structure. l 1
l l
Typical Soil Profile. From the data of the subsurface investigations, a typical pro ,
file was established at the site, which consists of three soil layers:
. The top layer - a recent alluvial deposit - is ebout 15 m thick at the site and consists largely of grey silty fine sands (SM) occa-sionally inter 5edded with grey fine sandy silts and clays of low plasticity (ML'c. Typical N-values are in the order of 5 to 15.
. The second layer - also a recent alluvial deposit is about 20 m thick at the site and consists largely of grey fine to medium sands and some gravels (SP, GP). Typical N-values are in the order 25 to 50.
. The third layer - of Pleistocene age - consists largely of grey silts turning to grey silty clays (ML, CL). It reaches a depth of about 400 m under the containment model, where it is underlain by
%e Miocene basement rock. Typical N-values up to a depth of 60 m are in the order of 20 to 30.
Physical Soil Properties. The soil properties selected for each of the 22 soil layers, i.e. total unit weight, maximum shear modulus to represent elastic soil stiffness at very low strains, minimum damping ratio to represent material damping at very low strains and Poisson's ratio are given in Table 2.
The choice of the total unit weight values was based on results obtained in labora-tory tests and on the classification of soils reported in the boring logs. Table 2 3
shows that a weight of 19'000 N/m was selected for the more silty and clayey layers ,
3 and 19'500 N/m for the vore sandy and gravelly deposit. Maximum shear moduli were computed from the total unit weight and from the shear wave velocity profile shown in Figure 2, which represents the average of five cross-hole surveys conducted at j the site of the 1/4-scale containment model. The average shear wave velocity begins at about 100 m/s at the surface and reache3 a value of about 300 m/s at 60 m depth.
In the proposed soil model for model set B however, the computed variation with depth of the maximum shear moduli was smoothet: out and represented by the following
]
two straight lines:
for 0 m < z < 30 m 4 2 Gmax (2) = 3 + 11*(z/30) (10 kN/m )
for 30 m
- z = 60 m 4
G , (z) = 14 + 4*( ) (10 kN/m2)
Shear Wme Velocity m/s 0 WO 200 300 M)0 500 000 0 ' I I 'I ' I ' I '
C unee sur Forcel D 10
. ve asan see EA -
Sss - w g d 08 - - - - -- - - - - - - - - - - - - - .
e ,,
.m- . 2 y -
~ f 06 -- - - - - - - - - - - - - - - - -- - -- --
1 4 -
i i
3 r-
, y .,
a4 _ _ _ _ _ _ _ _ _ . _ _ . _ . - _ _ _ _ _ _ . _ . . __._ ..
, ', 558 Aaasyte ,
I o2 - - - - - - - - - - - --
.s - y 8;
j .-' e' e'
ur' io j -
E
- 1 -
Shear Strain , percent r 30 - -
3 f -
e 3 g 20 - - - - - - - --
g a
7, g - - - - . - .
-w -
)
w -
E m - . - - . . . . - .__ . _
ra g 5 __
(
o o
. . e-' e-' e> e-' e Shear Strain, percent J'
0 . . . . .
- v. . . .
Figure 2. Average Shear Wave Velocity Profiles Figure 3. Strain Dependence of Dynamic Soil Properties for Site and Model Sets B and C
1 Table 3 SOIL PROPERTIES USED IN FRRE-FIELD OF MODEL SET C
! Layer Layer Unit Shear Modulus Strain Compat. Damping Poisson's Ratio
- No. Thickn, Weight at Low Strain Shear Modulus Ratio Horiz. Ve rt .
2 2 (m) (N/m ) (kN/m ) (kN/m ) (%) ~
1 0,66 18'500 14'195 12'130 2,9 0,480 0,483 2 1,00 18'500 19'549 12'960 5,6 0,483 0,489 3 1,00 18'500 26'018 14'737 7,1 0,486 0.492 4 1,00 18'500 32'447 16'841 8,0 0,488 0,494 5 1,06 18'500 39'067 19'085 8,7 0,489 0,495 ,
6 1,28 18'500 46'519 21'712- 9,2 0,491 0.496 7 1,70 18'500 55'100 23'361 9,9 0,492 0,497 8 1,00 18'500 61'500 23'995 10,8 0,493 0,497 9 1,80 18'500 68'100 24'767 11,4 0,492 0,497 10 1,80 18'500 74'700 25'992 11,8 0,491 0.497 4
11 1,90 18'500 81'200 27'201 12,1 0,491 0.497 12 2,30 20'000 133'400 62'222 8,9 0.486 0,493
$ 13 2,30 20'000 141'900 64'161 9,2 0,485 0,493 l 14 2,40 20'000 150'500 66'349 9,5 0.484 0.493 15 1,60 20'000 113'600 ;7'653 12,4 0,488 0.496 !
16 1,60 20'000 119'500 3P 551 12,2 0,488 ),496 17 1,60 20'000 125'300 4I'217 12,0 0.487 0.496 18 1,60 20'0G 131'200 44'536 12,0 0.485 0,495 19 1,60 20'000 137'100 46'608 12,0 0.485 0.495
- 20 2,50 20'000 202'400 93'365 9,0 0.478 0,490 i
! 21 2,50 20'000 211'700 96'800 9,1 0,477 0,490 ;
! 22 6,00 19'000 151'200 59'330 10,0 - 0 ,v,3 l 23 6,00 19'000 158'850 63'158 10,0 -
C.453 !
> 24 6,00 19'000 166'500 67'000 10,0 - 0,492 l 25 7,00 19'000 174'640 71'150 10,0 - 0,492 4
e 1
l
l
)
tests reported in Data Packages (e) and (h) (Ref. 19, 22) as well as from other pu-blished data (Ref.11).
i Finally, the backfilled excavation pit was not taken into account in model set C.
COMPARISON OF PREDICTIONS WITH MEASUREMENTS FOR VIBRATION TESTS Predictive Comoutations The model subset B1 shown in Figure I was used to, predict the response of the struc-ture to forced vibrations. The response was calculated for an excitation on the roof in the radial direction only because the computer code BHLUSH/BHBOUND can not handle tangental excitations. The driving point was nodal point 13 (see Figure 1). Contrary to the frequency dependent force amplitude during the tests, the predictive computa-tions were made for a constant force amplitude of 98 kN over the frequency range between 0 and 30 Hz. This is consistent with the results of the vibration tests scaled to a force of 98 kN (Ref.18, 21). No iteration was performed to obtain ;
strain-compatible soil properties. ,
The response of the structure was obtained in form of displacement transfer func-tions at nodal points 16, 1, 67, 117 and 25 (see Figure 1) for the specified output directions normalized to a force of 98 kN. The nodal points with their associated output directions and the corresponding recording channels are given in Table 4.
The displacement transfer functions (displacement and phase) predicted for channels 2 and 3 at the top edge of the structure and for channels 12 and 13 on the basemat I are shown in Figures 4 through 7. Channels 2 and 12 recorded in the radial direc-tion, channels 3 and 13 in the vertical direction. The frequency range shown in F1-gures 4 through 7 was from 0 to 18 Hz only, because the measurements did not show any significant response of the structure above 18 Hz. The maximum displacements predicted at all relevant recording channels are given in Table 4. I The analysis method used to predict the displacement transfer functions did not give modal parameters directly. From the transfer functions however, the most significant mode of vibration can be identified by the marked resonant peak as a rigid body rocking mode. For this mode, the resonant frequency and the modal damping ratio have been determined as 4,64 Hz and 23 %, respectiv21y. The corresponding mode shape is displayed in Figure 8.
m._._m__.._m .. ,__,___ _, .
.D
/
e
?.
= .;
/ - -
J s . . . . . . . . . .. . . . . , . ,
- =
a a ..
e is .. .*
5- ..
. \
s -
'Qg w % :
N- ~
.......<.s u . . . .. . . , u n. s Figure 4. Nomalized Displacewnts and Phases at figure 5. Normalized Displacements and Phases at Channel 2-from Field Measurements ( - ) and from Channel 3 from Field Measurements ( - ) and from Models By (aaa) and C (+++) Models By (aaa) and C h1 I "')
Q ...
l .w CA i
/ f_
g C'__m+2- . _ r cr-I \ !
\. .
\- n \
\! !
\;
t,-
\ - ,ti s
\! :
\.i,: :
\i,. \ ,i,
\\ \'
\! \, r au-
\\ Y
\! _ . ,:::.-=7-Q 1
l l e m-Figure 8. Measured and Predicted Mode Shapes for Rocking Response The comparison of measured and predicted transfer functions shows that the measured ;
overall behaviour of the soil-structure system could be reproduced in a qualitative way by the predictive computations. The only resonance peak over the frequency range l of 0 to 18 Hz - indicating a rigid body rocking of the 1/4-scale containment model -
was predicted by the analysis, although there are considerable differences as far as the resonant frequency and the displacement amplitude at resonance are concerned.
j From a quantitative comparison of the measured and predicted results, the following differences can be identified:
The predicted resonant frequency of 4,64 Hz overestimates the measured value of 3,77 Hz (Ref. 21) by about 23 %, indicating a model behaviour which is too stiff. !
The modal damping ratio estimated from the predicted transfer func-tions is about 23 %. This compares with a measured value of about 10
% (Ref. 21). Taking into consideration, that material damping is not i very important at the low strain levels generated during the vibra-tion tests, it is concluded that radiation damping in the model is 1
considerably overestimated.
1
f ace. However, soils by nature have a much larger compressive than tensile stiff-ness. Therefore, the above desribed modelling of the embedded structura seems to tie the structural model much more rigidly into the ground than is the case in reality.
Consequently, the computed rocking response is considerably smaller than the l measured one (see Figure 8). l l
To study the effects of embedment, model subset B2was changed by making the four ,
soil elements 115, 116, 117 and 118 immediately adjacent to the structure (see Figure 1) very weak with a shear modulus of 100 kN/m 2. Hence, the embedment of the structure was relaxed "completely" (see Table 1, model subset 8 3). The computed re- I sponse of this model showed a resonant frequency of 3,4 Hz which falls below the l measured value of 3,77 Hz and a model damping ratio of 5 % which is less than the measured 10 %. The computed displacement amplitude at resonance is now about 3 times larger than the measured value.
i The effectiveness of this change showed the importance of a proper modelling of the embedment. Therefore, a final attempt to model the embedment of the structure as realistically as possible was made (see Table 1, model subset 4B ). The top soil ele- I ment 115 adjacent to the structure (see Figure 1) was left weak as in model subset !
2 B3 , i.e. 100 kN/m . The shear moduli of the underlying three elements 116, 117 and 118 were set equal to 1/4,1/2 and 3/4 of the values in the free field. The reaso-ning behind this assumption is, that a rocking structure will most likely develop l
small gaps between soil and structure near the ground surface, because the soil- l structure interface can not carry tension there. Further down in the soil, the over-burden pressure is responsible for a soil-structure interface prestressed in com-pression. Hence, the development of a gap due to the rocking structure will be less likely with increasing depth, until the soil-structure interface behaves as if soil and structure elements were connected to each other. At this point "full" embedment becomes active and the stiffness of the real system is represented by the model.
With this model subset B 4the computed resonant frequency, modal damping ratio and maximum displacement amplitude on the roof at resonance now have values of 4,25 Hz, 8,8 % and 1,63 mm, respectively. These values agree quite well with the measured ones (see Table 4),
l Dynamic Soil Properties at Site. As mentioned previously, the dynamic soil proper-ties in the free field for model set C (see Table 3) were slightly changed from l their original values in model set B to reflect the additional knowledge from a re-view of the shear wave velocities used for model set B and from a review of addi-l l
j
~
8..
1 -
i .
q 1
.. e-1 t .
4
.l .
4 l
l
, i. l 4
4 4
a..
s 5 ,
=
4
, J
, = w . .. =
Figure 9. Acceleration Response Spectra of Earthquake Input at Station FAl-5 (Top: i j N; Middle: E; Bottom: V; Damping: 5 %) l
I Table 5 MEASURED AND PREDICTED MAXIMUM ACCELERATIONS 1
Station Nodal Maximum Accelerations in Direction Code Pt.No. North East Vertical C
n C
y A, A p A, A p A, A p
2 2 Code (cm/sec ) (cm/sec ) ge,f,,e 2 ) ge,j,,e 2) 2 (c:n/sec ) (cm/sec2 )
F4UN 21 21 210 200.2 200 181,3 - 64 45,1 F4UE 21 21 212 200,0 204 181,4 41 40,7 F4US 21 21 211 200,2 200 181,3 69 - 52,8 F4UW 21 21 209 200,0 202 181,4 84 47,0 F4LN 113 129 ? 170,4 7 154,4 ? - 43,9 F4LE 113 129 127 170,5 147 154,4 40 39,8 F4LS 113 129 126 170,4 148 154,4 69 - 50,7 F4LW 113 129 124 170,5 148 154,4 82 46,7 FAl-1 130 150 130 188,2 157 161,4 47 37,6 FA2-1 130 150 163 207,4 158 156,9 - 44 38,9 FA3-1 130 150 134 203,9 162 161,2 58 44,6 FAl-2 174 202 172 197.9 157 159.2 40 38,3 FA2-2 174 202 207 214,9 160 157,2 69 39,3 FA3-2 174 202 155 205.5 146 1 61 ,0 51 42,4 FAl-3 237 277 224 213,3 159 154,8 44 39,7 FA2-3 237 277 211 217,0 166 155,9 46 40 FA3-3 237 277 185 209,9 7 157,7 47 40,8 FAl-4 279 327 200 212,5 156 154,9 43 40,5
, FA2-4 279 327 -
214,1 -
155,9 - 40,6 FA3-4 279 327 203 216,0 164 154,8 44 39,6 DHA6 180 208 110 156,3 151 148,3 43 37,7 !
DHAll 183 211 93 128,1 130 128,0 36 38,4 i DHA17 186 214 78 110,1 110 101,4 34 39,1 DHA47 - -
86 -
83 -
33 -
Free Field layer No:
FAl-5 1 203 203,6 154 154,3 40 39,4 FA2-5 1 189 203,6 155 154,3 42 39,4 FA3-5 1 194 203,6 182 154,3 41 39,4 l DHB6 7 135 154,2 150 145,9 39 39,3 DHB11 10 107 125,1 114 129,2 39 39,2 i DH817 13 90 109,5 100 102.3 36 39,2 l DHB47 24 97 -
79 -
31 37,9 1
1 Note: Am: Measured Acceleration Values Ap: Predicted Acceleration Values l
were compu%ed for the entire profile from the strain-compatible shear wave veloci-ties to agree with the compression wave velocities described before. The variation with depth of the maximum vertical accelerations is also plotted in Figure 10 to a depth of 60 m.
Measured and predicted maximum accelerations in the free-field for all three direc-tiens (N, E and V) are given in Table 5. Measured and predicted acceleration re-sponse spectra at 5 % damping for stations DHS6 (N E and V) and DHB47 (only V) are shown in Figures 11 and 12, respectively.
It is concluded that the predicted responses in the free-field compare quite well with the field measurements. An exception to this general conclusions may be the response due to the horizontal N-component for which the maximum accelerations at depth are overestimated. The following may account for this: The soil profile used for the free-field analysis was strain-compatible with the E-component. As already mentioned, this profile is softer and has greater damping ratios as the strain-com-patible profile for the N-component. At greater depth this possibly leads to an overamplification of the higher frequencies of the N-component and results in too high predicted maximum accelerations.
Finally, it is noted that the analysis method used does not account for horizontal spatial variation of the free-field motion. Thus, the predicted responses in the free-field at the two surface stations FA2-5 and FA3-5 are identical to the earth-quake input motion at station FAl-5. The measured motions, however, shows a certain horizontal spatial variation (see Table 5).
SSI-Analysis The SSI-Analysis of the soil-structure system was conducted separately for each of the three components (E N and V) of the earthquake excitation specified at station FAl-5 in the free field. The cut-off frequency was 15 Hz as for the free-field ana-lysis.
Firstly, an initial solution of the soil-structure system represented by the model ,
subset Chl (see Figure 1) was obtained for the E-component of the earthquake input motion. The strain-compatible dynamic soil properties as determined in the corre- '
spending free-field analysis were used. From this solution, the strains in the soil accounting also for the interaction effects of the structure in addicion to the effects of the free-field motion, were calculated to a depth of 22 m below the
4
~
i: <
i'
};-
f -
e..o.,,rm' Figure 12. Measured (---) and Predicted (---) Acceleration Response Spectra in Free-Field at Station DHB47 (Vertical Direction; Damping: 5 %)
ground surface. Then, new strain-compatible shear moduli and damping ratios were evaluated for this porcion of the soil model. Using these new dynamic soil proper- l ties, a second solution was computed, resulting in the final response of the soil-structure system due to the E-component of the earthquake input motion.
Secondally, the same soil properties - including the effects of secondary non-line-arities described above - were used to compute the response of the soil-structure system (model subset Chl) due to the N-component of the earthquake input motion.
l Thirdly, the vertical solution of the soil-structure system (see Figure 1 model subset Cyj) was obtained for the V-component of the earthquake input motion.
l l
Because of the cylindrical coordinates used in the axisymmetric finite element pro-gram BHLUSH/BHBOUND, the solutions for each of the three SSI-analyses were obtained in terms of radial, tangental and vertical response components at a point of inte-rest. Hence, to obtain the response in terms of N , E- and V-components, as they are recorded by the strong motion accelerometers, the radial, tangental and vertical J
components computed for each of the three SSI-analyses at a specified recording
4
~
.1
- e ,
~
1 em
- h 1
, l
\
i
- l J* ,
~
s.
- .,
- y, i
......,t~,3 Figure 13. Measured (-) and Predicted (---) Acceleration Response Spectra on Structure at Station F4US (Top: N; Middle: E; Bottom: V; Damping: 5 %) )
1
O
_n
~
4
- i. f -
)
,l
~. - ,
4
- e ,
I
~ . . . .
r,......,to i l
4 Figure 15. Measured (-) and Predicted (---) Acceleration Response Spectra on j Structure at Station F4LS (Top: N; Middle: E; Bottom: V; Damping: 5 %) t i
1
r . .m_.. . . , _ _ . , , , , , ,
1 3
. ~~~ ~%rm4 4 h ,^ '"
p ,x g A i a
3 I
5' I, i fh
- g. wnrM .,:_ NA F
d s
~ A*:$*le:.':"*+ $ -
w 8 . ....... . . . ... . . . . . . . . . . . . . . . , .
.. e.i Figure 17. Measured ( ) and Predicted (---) Acceleration-Time Histories on Ground-Surface at Station FAl-1 (Top: N; Middle: E; Bottom: V)
~
A comparison of maximum accelerations in the structure shows (Table 5), that the agreement is better in the E-direction than in the N-direction. While the predicted accelerations at the top of the structure in both horizontal directions deviate be-tween 5 % and 10 % from the measured values, the predicted accelerations at the bottom of the structure in the N-direction are about 35 % larger than the measure-ments. The E-direction is showing good results. In the vertical direction, the ;
measured accelerations are unterestimated by the analysis both at the top and the bottom of the structure.
In the vicinity of the structure on the ground surface (Stations FAl-1. FA2-1, FA3-1) and imediately underneath the structure (Stations DHA6, DHAll, DHA17), the measured values in the N-direction are again smaller than the predicted ones. The E- ,
direction is showing good agreement. The erratic behaviour of the measured and pre- I dicted vertical maximum accelerations in the imediate vicinity of the structure is difficult to interpret.
Approaching the free field, the results at the ground surface improve, which is self-evident with the earthquake input motion specified in the free field. The va-riation of the maximum accelerations with depth was predicted more accurately in the E-direction and the vertical direction than in the N-direction.
The acceleration response spectra (Figures 13, 15 and 16) show the most obvious dif- ;
ferences between measured and predicted responses in the N-direction at the top of the structure (see Figure 13). The marked peak at about 2,8 Hz most likely results from the predicted rocking frequency of the structure. The larger response computed at this frequency can also be observed on other horizontal and vertical acceleration !
response spectra (see Figures 15 and 16). To show the existence of this rocking fre-quency in model subset C hl in a more transparent way, the predicted amplitude trans-fer function between the N-components o' Station FAl-5 in the free-field and Station F4US at the top of the structure is plotted in Figure 20. It shows quite well the i resonant frequency at 2,8 Hz.
The acceleration-time histories plotted in Figure 14 also shows that the computed oscillations at about 2,8 Hz following the main peak response are much stronger than in the measured acceleration time history. This indicates that rocking at this fre-
- quencys may not be as important in the field as predicted by the computations, or rocking in the field may take place at an other frecuency. Furthermore, the N-com-
i:
} . ,
2+ =
- o i . .
s s
s-
.. s 5.
ht- ,
t VJ~wmypnpu ?h 3
- . . . . . . .a .. .. ..
e.......,t.-
1 I
Figure 19. Normalized Displacements and Phases at Channel 2 from Field Measurements l
(-) and from Model C h2 IH')
l Compared to the measured behaviour, model subset C seems still somewhat too stiff.
h2 l From the measured transfer function the maximum amplification is estimated to be '
between about 2.0 and 2,2 Hz. The softer soil profile led however to a partial im- l provement of the predicted results as shown by the accelaration response spectra and acceleration time histories displayed in Figures 21 and 22.
l
't
s 4
~
s 1.
- O f \j\
- . ~ _ _
I: -
j y
- .i
/ ....
. 1
. 1 I
- \
i:
~
I l
J*,
^
y r.......,t<,3 Figure 21. Acceleration Response Spectra on Structure at Station F4US from Field Measurements (-) and from Models Cg (a u) and Ch2(+++) (Top: N; Middle: L; Bottom:
V; Damping 5 %)
4
gained through the project will certainly be helpful in the analysis of futur SSE-problems.
It is our oppinion, that the differences observed between measured and predicted ;
responses can be attributed mainly to uncertainties about the parameters used in the analysis. These uncertainties result mainly from the inability to determine correct- ,
ly the dynamic soil properties and to translate measured or estimated system proper- l ties into appropriate model parameters. To minimize the effects of these uncertain-ties on the predicted response, a good portion of experience is required.
The question about the usefulness of forced vibration tests to improve earthquake response predictions has been answered positively, based on the experience gained i during this project. The results of the forced vibration tests were very helpful in
)
refining model subsets B) through B4. Furthermore, as good a match as possible be-tween measured and pradicted results for forced vibration tests tends to improve the predicted earthquake response.
I Finally, some findings worth mentioning were made during the course of this project:
. The SSI-methodology used during this project reaches its limits for I very soft and deep soil profiles. To reproduce correctly the beha- !
viour of the soil, the layers in the model should be chosen very l thin. This is particularly important in deconvolving correctly the earthquake motion in the free-field. However, this requirement may be in conflict with the computer time required to solve the problem.
. The measured free-field response in the N-direction suggests that 1 the use of independent strain-compatible soil properties for each !
direction improves the prediction of the response in the free-field and therefore also the response of the soil-structure system. This ;
should be further investigated. l
. The comparison of measured and predicted responses in the structure indicates that the soil underneath the structure may be softer than l used in the analysis. A rocking frequency of the structure around 2 '
Hz could in our opinion further improve the prediction of the struc-tural response.
In conclusion, it is hoped that the participation of HSK in the USNRC Validation Project has resulted in some valuable contributions in regard to the validation of SSI-methodologies. Throught this participation, HSK has improved its understanding of SSI-methodologies used in the analysis of nuclear power plants.
- 15. Data Package (a). First Phase Soil-Boring Test Data.
- 16. Data Package (b). SMART-1 Array Geological Data.
- 17. Data Package (c). Structural and Geometric Details of the 1/4-Scale Containment Model and Instrument Locations.
- 18. Data Package (d). Vibration Test Input Data.
- 19. Data Package (e). Soil Test Data-Phase 2.
- 20. Data Package (f). Geological Data from Within Array of Model Structure.
- 21. Data Package (g). Response Data from Vibration Test of Completed Structure.
- 22. Data Package (h). Additional Information on Soil Testing.
- 23. Data Package (i). Free-Field Motion from Strong Motion Event of 5/20/86.
- 24. Data Package (j). Complete Set of Ground Motion and Structural Response Records from 5/20/86 Event.
- 25. H. Bachmann. "Zur SchwingungsdEmpfung teilweise vorgespannter Leichtbeton- und Betentragwerke." Schweizer Ingenieur und Architekt Nr. 11, 1986,
- 26. U.S. NRC. "Damping Values for Seismic Design of Nuclear Power Plants." Regula-tory Guide 1.61., October 1973.
- 27. V. Langer, S. Tinic, E. Berger, P. Zwicky and E.G. Prater. "Full Scale Vibration Test on Nuclear Power Plant Auxiliary Building: Part I." Transactions of the 9th International Conference on Structural Mechanics in Reactor Technology, Volume K1, pp. 261, Lausanne 1987.
- 28. D. Flade, H. Steinhilber and A. Dietz. "Full Scale Vibration Test on Nuclear l Power Plant Auxiliary Building: Part II." Transactions of the 9th International Conference on Structural Mechanics in Reactor Technology, Volume Kl. pp. 267 Lausanne 1987.
4 :
l l
1 l ;
l 1
REFERENCES ,
- 1. Argonne Nati0nc1 Lcboratory. "Validation of Soil-Stureture Interaction (SSI)
Methodolgy: EPRI/USRNC Seismic Experiment." Statement of Work, April 1986.
- 2. Brookhaven National Laboratory and U.S. Nuclear Regulatory Comission. "Procee-dings of the Workshop on Soil-Structure Interaction Bethesda, Maryland, Juni 16
- 18, 1986." NUREG/CP-0054, December 1986.
- 3. Basler & Hofmann. "Validierung von Analysemethoden zur Lusung von Boden-Struk- :
tur-Interaktionsproblemen bei Kernanlagen im Falle von Erdbeben (Validation of Analysis Methods to Solve Soil-Structure Interaction Problems in Connection with Nuclear Installations Subjected to Earthquakes).- Proposal submitted to Swiss Nuclear Safety Department (HSK), September 1986.
- 4. E. Berger. "Seismic Response of Axisymetric Soil Structure Systems." Doctoral Dissertation, University of California, Berkeley,1975.
- 5. Basler & Hofmann. "BHLUSH/BHBOUND: Computer Program for the Dynamic Analysis of Axisymetric Soil-Structure Interaction Problems." User's Manual, Rev. O. Decem-ber 1986.
- 6. Basler & Hofmann. "Predictive Response Computations for Forced Vibration Experi-ments on 1/4-Scale Containment Model in Lotung Taiwan." Progress Report for Phase 1 of HSK Validation Project for Swiss Nuclear Safety Department (HSK),
July 1987.
- 7. Basler & Hofmann. "Comparison of Measured and Predicted Forced Vibration Res-penses and Prediction of Earthquake Response for 1/4-Scale Containment Model in ,
Lotung, Taiwan." Progress Report for Phases 2 & 3 of HSK Validation Project for '
Swiss Nuclear Safety Department (HSK), September 1987, 1
- 8. Basler & Hofmann. "Comparison of Predictions with Measurements for Vibration l I
Tests and Earthquake of May 20,1986 at 1/4-Scale Containment Model in Lotung.
Taiwan", Final Report of HSK Validation Project for Swiss Nuclear Safety De-partment (HSK), October 1987.
- 9. J. Lysmer et al. "FLUSH: A Computer Program for Approximate 3-D Analysis of Soil-Structure Interaction Problems." Earthquake Engineering Research Center, Report No. EERC 75-30 University of Ca.lifornia, Berkeley, November 1975.
- 10. H.B. Seed and I.M. Idriss. "Soil Moduli and Damping Factors for Dynamic Response 3
Analyses". Earthquake Engineering Research Center, Report No. EERC 70-10 University of California, Berkeley, December 1970. j
- 11. H.B. Seed, R.T. Wong, I.M. Idriss and K. Tokimatsu. "Moduli and Damping Factors i
. for Dynamic Analysis of Cohesionless Soils." Journal of Geotechnical Engi-neering, ASCE, Vol. 112, No. 11, November 1986. !
1
- 12. J.M. Roesset and R.V. Whitmann. "Theoretical Background for Amplification Stu-J dies." Research Report R69 - 15 Massachusetts Institute of Technology, Ccmbridge, March 1969. l i
- 13. P.B. Schnabel et al . "SHAKE: A Cor@ uter Program for Earthquake Response Analysis i of Horizontally Layered Sites." Earthquake Engineering Research Center, Report No. EERC 72-12, University of California, Berkeley, December 1972. ;
- 14. Basler & Hofmann. "FLUSH.F77: Computer Program for the Dynamic Analysis of Soil-Structure Interaction Problems." User's Manual, Rev. O. December 1986.
.__ . .. .__