V, Earthquake Engineering, Tenth World Conference © 1992 Baikema, Rotterdam.

ISBN 90 5410 060 5

Earthquake induced displacements of soil~structures systems

Peter M. Byme & Hendra Jitno

Department ofCivil Engineering, University ofBritish Columbia, Vancouver, B. C., Canada
Francisco Salgado
Vancouver, B.C.. Canada

ABSTRACT: An analysis procedure is presented for predicting the earthquake induced displacements of
earth dams. The procedure extends the simple Newmark method from a single-degree-of-freedom rigid
plastic to a multi-degree-of-freedom flexible system using energy concepts. The method is applied to the
San Fernando dams. The lower dam suffered a flow slide on its upstream side while the crest of the upper
dam moved 2 meters downstream during the 1971 San Fernando earthquake. The predicted and observed
displacements are in good agreement in terms of both the magnitude of displacements as well as their
pattern.

1 INTRODUCTION under the gravity loads and inertia forces. For

many structures such as dams, these deformations
A number of earth dams have failed or undergone may be acceptable. It is important, therefore, to
large displacements due to seismically induced develop simple reliable methods for predicting
liquefaction, for example, two Chilean mine such displacements, and this is the object of this
tailings dams (Dobry and Alvarez (1976)), and the paper.
Mochikoshi tailings dam in Japan (Marcuson The deformation analysis proposed here is a
(1978); Ishihara (1984)). A number of other dams pseudo-dynamic finite element procedure which
have undergone large deformations but have not allows both the ine tia forces from the earthquake
actually failed in as much as the reservoir was not as well as the softgairig effect of the liquefied soil
released. The classical example of this -was the to be considered. The method is essentially an
near failure of the lower San Fernando dam in extension of Newmark’s procedure from a rigid~
which a liquefaction induced flow slide occurred on plastic single-degree-of-freedom system to a
the upstream side removing the crest of the dam. flexible multi-degree-of-freedom model. The
Of perhaps more interest from the analysis point of procedure together with the results are presented
view was the behaviour of the Upper San in this paper.
Fernando dam. The crest of this dam moved about
2 m downstream during the San Fernando
earthquake of 1971. Both of these dams have been 2 ANALYSES PROCEDURE
extensively studied in the past and their
movements and soil properties have been Prediction of earthquake induced movements of
established. These dams have therefore been used earth structures is a difficult problem. Complex
to verify the proposed deformation analysis effective stress dynamic analyses procedures have
procedure. been proposed (Finn et al. (1986); Prevost (1981))
The dynamic response of soil-structure systems but are essentially research tools and not-generally
involving soils whose properties change markedly appropriate for analysis of most dam structures.
with cyclic loading is a difficult problem. The The simplest analysis procedure is that
difficulty mainly arises from the complexity of the proposed by Newmark (1965) in which a potential
stress-strain relations of the soil - particularly when slide block is modelled as a single-degree-of
pore pressure rise and liquefaction occurs. The freedom rigid plastic system. Any prescribed time
strains to trigger liquefaction are generally small history of acceleration can then be applied at the
(<1%). However, once liquefaction is triggered base and the resulting displacements computed by
large but limited deformations occur as the soil numerical integration. Newmark also found that
strain hardens and regains stiffness and strength the maximum displacement at the end of the

1407
shaking period could be estimated from simple since Newmark assumed the material to be rigid
formulae by considering the earthquake to be plastic, the internal force or resistance is constant
approximated by a number of pulses. with displacement as shown in Figure lb. The
There are two concerns when applying work done is the area beneath the resistance line.
Newmark’s simple procedure to an earth stmcture The external force is the gravity driving force, Mg
such as a dam: (1) the soil, particularly in zones Sllla, and in this case is constant with displacement
where liquefaction is triggered is not rigid plastic; as shown in Figure 1b. WEXT is therefore the area
and (2) the single-degree-of-freedom model does beneath the driving force line. The net work done
not allow the pattem of displacements to be is the difference between the two areas, namely the
computed. Byrne (1990) and Byrne et al. (1991) shaded area and this must equal 1/2 MV2.
discusses this and show a way of allowing for a Now WINT = (s.L)D, where s = the shear
general stress-strain relation as well as extending strength of the soil and L is the length of the slide
Newmark’s approach to a multi-degree-of-freedom block, and WEXT = (Mg Si11cr)D. Thus equation 2
system. Basically a pseudo-dynamic finite element reduces to
procedure is used in which earthquake induced
displacements which satisfy energy considerations
are achieved by use of a horizontal seismic D(s.L - Mg sina) = 1/2 MV2 (3)
coefficient. The appropriate seismic coefficient is
the one which satisfies the work-energy equation 01'
and is found by trial-and-error as described by D = 1/2 MV2/(s.L - Mg sine)
Byme et al. (1991). This approach is described
here. It is first applied to the Newmark problem = 1/2 V2/gN (3a)
and then extended to a general stress-strain and
multi-degree-of-freedom system.
Newmark’s simplified model is that of a block where the yield acceleration, N, is given by
of mass M resting on an inclined plane of slope cm,
and subjected to a velocity pulse, V, relative to the
base (Fig. 1a). The resulting displacement is given N = (s.L - Mg sine)/Mg (4)
by
D = 6V2/2gN (1) Equation 3a is for a single velocity pulse and
when 6 pulses are considered the result is identical
to Newmark’s equation 1.
where D = maximum displacement, V = the Soil when triggered to liquefy will not behave in
velocity pulse which Newmark took as the a rigid plastic manner and this is examined herein.
maximum ground velocity, N = the yield The triggering of liquefaction of loose saturated
acceleration, i.e., the acceleration as a fraction of sandy soils by earthquake loading is a small strain
"g" required to initiate yield and sliding, and g = phenomenon (Byrne (1990)). Upon liquefaction
the acceleration of gravity. The number 6 in his the stress in the soil drops from A to B. Its
formula comes from considering 6 pulses of resistance then increases with strain to a residual
velocity V which Newmark found gave agreement value, s,., shown in Fig. 1c. The driving force from
with the integrated records when the ratio the ground slope remains constant, however, so
N/A<0.13 as it usually is for practical cases of that the system accelerates and deforms. When
concern. the strain reaches point C the material has
His model will now be developed in terms of hardened so that the stress developed is now
work-energy and this will allow its extension to a sufficient to balance the driving stress as shown in
general formulation. Fig. 1d. However, the system has a velocity at this
The work-energy theorem states that the work point and the stress continues to increase until
done by the internal forces or stresses minus the point D is reached where the net energy (Wm-I ~
work done by the external forces must equal the WEXT) is zero. If the system also had an initial
change in kinetic energy of the system, namely, velocity at the time liquefaction was triggered it
would carry on to point E. If the driving stress
exceeds the residual strength, a flow slide will
WINT ' WEXT = 1/2 M V2 (2) occur.
Comparing the rigid plastic Newmark approach
with the proposed extension to a general stress-
The work done by the internal forces depends strain relation (Figs. 1b and 1d) it may be seen that
on the stress-strain relations of the material and Newmark is missing the displacement from A to D.

1 408
 V = velocity energy equation 2 and this is described in detail in
M = Mass of the block (7)-
D = dlsplacement V For a multi-degree~of-freedom system a finite
element approach can be used. The displacements
are computed from the solution of
‘ \
/\\
E 3.
<5 [K]{A} = {F + AF} (5)
Force or Stress
sou RESISTANCE where [K] is the global stiffness matrix of the
(b)D / E system, {A} is the vector of nodal displacements,
{F} is the static load vector acting on the system
3:'T:?i=it iiiiiii Y-*3. 1:llllVz'ri:i-1~1l'<ii'E' *:~:'§?i (gravity plus boundary loads), and {AF} is an
A additional load applied to satisfy the energy
omvmo FORCE balance of equation 2. If {AF} = 0, then for the
single-degree-of-freedom, a displacement corres-
Displacement or StTaln ponding to C would be predicted. An additional
force is required to balance the energy and predict
Stress Pre-cycllc/ Pre-llquefaction points D or E. This additional force can be
considered as a seismic coefficient. However, its
value is not related to the peak ground
(c) .................. acceleration but is selected by trial and error so as
Sr to balance the energy in accordance with equation
5 l 2.
fycyclo <_ Post-liquetactlon For the multi-degree-of-freedom system Wm-I
equals the work done by the element stresses and
strains, and WEXT equals the work done by the
B Strain’ static load vector = {F}'{A}T. The additional force
Stress {AF} is not included as it is merely an artiface to
obtain the appropriate displacements.
(d) RESISTANCE D E The procedure has been incorporated into the
finite element computer code SOILSTRESS
§3:':1§I§3r5:?:"'§1
121' - ' -:-:1 (Byrne and Janzen (1981)) and found to give an
exact agreement with Newmark when the
‘onlvime F nos assumptions made correspond to a single-degree
of-freedom rigid plastic system. It gives good
Post-liquefaction
agreement with liquefaction induced field
B Straln observations reported in Hamada et al.. (1987).
The procedure predicts a flow slide failure of the
Fig. 1: (a) Block on inclined plane subjected to Lower San Fernando dam on its upstream side as
velocity pulse V; (b) Work-energy, was observed. This occurred because the residual
Newmark; (c) Characteristic of pre- and strength after liquefaction was triggered was
post-liquefaction monotonic stress-strain insufficient to resist the gravity driving stresses,
curves; and (d) Work-energy, extended consequently the upstream slope slumped coming
Newmark. to rest when its final geometry was consistent with
residual strength together with energy
considerations. Due to space limitation the lower
This could be a very considerable displacement San Femando dam predictions are not shown.
since strains of 20 to 50% are commonly required This paper concentrates on predicting the seismic
to mobilize the residual strength, s,. However, in displacements of the upper San Fernando dam.
carrying out analyses where liquefaction is
triggered only one pulse is considered appropriate,
whereas Newmark considered a range of pulses up 3 UPPER SAN FERNANDO DAM
to 6 depending on the ratio N/A. So there may be
compensating factors here. The San Fernando dams were located
For a single-degree-of-freedom system, the approximately 14 km from the epicentre of the
displacement can be computed directly from the M6.6 Richter Magnitude earthquake which

1409
 occurred on February 9, 1971. The Lower dam The post-quake values for unliquefied zones 1, 2
suffered a major flow slide on its upstream side, and 4 were kept the same as the pre-quake values.
while the crest of the upper dam moved For the liquefied sandy zone. 5, the modulus
approximately 2 m downstream. number, KG, was drastically reduced to model the
The Upper San Fernando dam was constructed very large reduction in stiffness that occurs upon
commencing in 1921 to provide reservoir storage liquefaction, and the strength was dropped to its
capacity for the Los Angeles Aqueduct. The main residual value. These reduced values were based
body of the dam was constructed of hydraulic fill on the average normalized standard penetration
resting upon an alluvium foundation soil. The top value, (N1)50 = 15 together with post-liquefaction
dyke section was constructed of rolled fill. A cross- strengths and strains proposed by Seed and Harder
section of the dam showing the various material (1990), Byrne (1990), and Seed et al. (1984).
types, the water table, and the predicted zones of The clayey core, material 3, although not
liquefaction is shown in Fig. 2. considered to liquefy was modelled as undergoing
For the proposed analysis procedure, the zones a marked reduction in stiffness. Based on data
in which liquefaction is triggered must first be presented by Byrne et al. (1984), the initial
identified. The assumed zones of liquefaction modulus, and hence KG, was reduced by a factor of
were based on the analysis carried out by Seed et 3.3 as shown in Table 1 due to pore pressure rise.
al. (1973). Their study involved a comparison of A large pore pressure rise in the core was
the cyclic resistance of the soil and the dynamic recorded. The residual strength of the clay was
stresses caused by the earthquake. Liquefaction based on Torvane shear tests reported by Seed et
was assumed to be triggered in the zones where the al. (1973). The authors suggested an s,/wvo =
dynamic stresses exceeded the cyclic resistance. 0.24 and based on an average 0' ,,,, = 110 kPa, this
The cyclic resistance was based on laboratory leads to s,. = 26 kPa as shown on the table.
testing of samples and the dynamic stresses were The zones of liquefaction would drain
evaluated from a dynamic analysis using a sometime after the earthquake, causing additional
modified form of the Pacoima record scale to a volumetric strains and settlements. Potential
peak acceleration of 0.6g. volumetric strains of 2% were assumed based on
The triggering of liquefaction is a small strain Tokimatsu and Seed (1987) and an average (N1)60
phenomenon, with strains generally less than 1%. = 15, and were accounted for in the analysis by
Such strains are generally not of concern for earth applying appropriate loads to each liquefied
dams and are not considered herein. It is the much elements in a ma.nner similar to that commonly
larger strains that arise when the gravity forces used in thermal elasticity.
together with inertia forces act on the very much
softer post-liquefaction stress-strain curves that are
of concern. These strains can be in the range 10- 4 RESULTS
50%. It is these strains and the resulting
displacements that are computed in the analysis The predicted deformed finite element mesh is
proposed herein. shown in Fig. 3, and the predicted and observed
The soil properties required for the analysis are displacements at locations A to F on the figure, are
the pre- and post-quake values and these are listed listed in Table 2. The observed values were taken
in Table 1. The general procedure for obtaining from Serff et al. (1976). It may be seen that the
these is described in Byrne (1990) and Byme et al. predicted and observed horizontal displacements
(1991). The pre-quake values for all 5 soil types are in excellent agreement at points A to E. At
were based on data reported by Seed et al. (1973). location F the predicted displacement is lower than

Liquefied S011
_.. Z _ _ _ _ _ _ _ _ _ 1!-§~
‘ -§u.-I!___
-IIlm===5-- I
IIIIIIIEIIIIIHIIIJI
|II I.I II
I:'_:::IlIIHII:::t:::;;_:I
i=;i=__F?% I ---%§§E€?-i?ééi-
" -_i 1:11‘- lI!I 'l nlyIIIIIlll III ll!|wllllll I II‘=II=I

Illl
I I I|I \—Il l\-iIIII\ —Il l I\ I I; Il —-l ln—-nun.nun -—II{ iIIIl. -III. _-I- -II -—II IIIII l mnll—n —Il I—II :::::IIIIIIIIIIIIIIIII I IIIII - I I I IIIIIII I IIIII ‘- IIIIIII I 1— IIIII IIIII IIIII III
I I I Ir II“II.I
Fig. 2: Upper San Fernando Dam: Soil Types, Zones of Liquefaction and Water Table.

1410
 Table 1. Soil properties used in the analysis. Table 2. The observed and predicted displace-
ments of Upper San Femando Dam.
Soil '1 ¢<=K(; 11KB m Rr
Type k.N/ma (kPa)
Displacements (meter)
1
Allu- 20 37 0 117 .802000 .40 .66 Horizontal Vertical
vium
Location‘ Observed Fred. Observed Pred.
2
Hydr. 19 37 0 175 522000 .26 .78
Fill A 1.49 1.59 -0.76 -0.90

3 B 1.49 1.72 -0.58 -0.53

Clay 19 37 0 175 52 2000 .26 .18
Core (16) (50) (5) (1000) (15) (.18) C 1.95 1.91 -0.06 -0.01
4
Rolled 22 25114125 .76 2000 .38 .90 D 2.19 2.07 ~O.43 +0.05
Fill
E 1.77 1.79 -0.52 -0.13
s
Liqf. 19 37 0 175 .52 2000 .26 .18 F 1.10 034 -0.06 “-0.02
Soil (24) (0.7) (0) (2000) (.26) (.01)
*Refer to Fig. 3.
Egg: - Numbers in brackets indicate the proper- "Negative value signifies settlement.
ties after earthquake.
- Pre- and post-earthquake properties of soil
1, 2 and 4 are assumed identical. 5 SUMMARY

An analysis procedure is presented for predicting

the observed by a factor of 3. The predicted and earthquake induced displacements of dams. The
observed vertical displacements at the crest points procedure is an extension of the simple Newmark
A and B are in good agreement. However, at method from a single to a multi-degree-of-freedom
locations D, E and F, the predicted vertical system taking into account the softened stress-
displacements are lower than observed. strain response of liquefied soils.
The lower predicted displacements, both The method is based on the assumption that
vertical and horizontal, in the downstream toe area the strains to trigger liquefaction are small and can
suggest that the zone of liquefaction extended be neglected compared to the subsequent strains
further towards the toe than assumed. Observed caused by the combined gravity and inertia forces
sand boils in the toe area also indicate that the acting on the softened post-liquefaction stress-
zone of liquefaction likely extended to the toe. strain curves. It is these curves that control
However, in the analysis presented here, the zones liquefaction induced strains and displacements.
of liquefaction were assumed to be those computed The method was applied to the San Fernando
by Seed et al. ( 1973). dams. The procedure predicts that the lower San

Original Mesh Final Mash

" """"".¢-u|€:E::|||lnnnu|IIII
" 2451!!I=n- II=5’inin1IIIIIIIIII0 E (‘I

4dIIIIIIIIrJ1;;IIIIIIIIIIIIs:r - F
--IIIIIlllnmvqvlllnlllllll I~,'1' v
||llll
\
I.’
—-IIIIIIllllllll-IQIIIIIIIIIIII
Q--I-I--Inunnnnnnnmsnnnnnnnn

1-": IIII1-"1_ll
ill‘; —IllZIII—IlI ZIII1-"1-" I-II —I|l
Fig. 3: Upper San Fernando Dam Computed Deformed Shape. Displacement
Magnification Factor = 2.

1411
Fernando dam would suffer a flow slide on its Induced by Seismic Liquefaction", Computers
upstream slope in agreement with the field and Gcotechnics 4, 1987, pp. 197-220.
observation. The procedure also predicts that the Ishihara, K., "Post-Earthquake Failure of a Tailings
crest of the upper San Fernando dam will move Dam due to Liquefaction of the Pond Deposit“,
downstream about 2 m in agreement with the field Proc. International Conf. on Case Histories in
observations. In addition, the predicted pattern of Geotechnical Eng., St. Louis, Vol. 3, 1984, pp.
deformations is generally in good agreement with 1129-1143.
the field observations. Marcuson, W.F., "Liquefaction Failure of Tailings
Dams Resulting fiom the Near Izu Oshima
Earthquake, 14 and 15 January 1978", Proc. 6th
6 ACKNOWLEDGEMENTS Pan American Conference on SMFE, Vol. 2,
1979, pp. 69-80.
The authors are grateful to NSERC for its financial Newmark, N.M., "Effects of Earthquake on Dams
support and to Ms. Kelly Lamb for her typing and and Embankments", Geotechnique, Vol. 15, No.
presentation of the paper. 2, 1965, pp. 139-160.
Prevost, J.H., "DYNA-FLOW: A Nonlinear
Transient Finite Element Analysis Program",
7 REFERENCES Rpt. No. 81-SM-1, Dept. of Civil Engineering,
Princeton University, Princeton, NJ, 1981.
Byrne, P.M., "A Model for Predicting Liquefaction Seed, H.B., Lee, KL, Idriss, I.M. and Makdisi, F.,
Induced Displacements", Soil Mech. Series No. "Analysis of the Slides in the San Fernando
147, Department of Civil Engng., University of Dams During the Earthquake of February 9,
B.C., Vancouver, 1990. Also, 2nd International 1971", Report No. UCB/EERC-73/2, University
Conference on Recent Advances in of California, Berkeley, June 1973.
Geotechnical Earthquake Engineering and Soil Seed, R.B. and Harder, LF., "SPT-Based Analysis
Dynamics, St. Louis, Paper 7.14, March, 1991. of Cyclic Pore Pressure Generation and
Byrne, P.M. and Janzen, W., "SOILSTRESS: A Undrained Residual Strength", I-I. Bolton Seed
Computer Program for Nonlinear analysis of Memorial Symposium Proceedings, Vol. 2, 1990.
Stresses and Deformations in Soil", Soil Seed, H.B., Tokimatsu, K., Harder, L. and Chung,
Mechanics Series No. 52, Dept. of Civil R., "The Influence of SPT Procedures in Soil
Engineering, University of B.C., December Liquefaction Resistance Evaluations", Report
1981, updated January 1989. No. UCB/EERC-84/15, College of Eng., Univ.
Byme, P.M., Jitno, H., Garner, S., Lee, M. and of California, Berkeley, California, 1984.
Lou, J.K., "Analysis of Earthquake Induced Serff, N., Seed, H.B., Makdisi, F.I. and Chang,
Displacements of the Intake Structures, John C.Y., "Earthquake Induced Deformations of
Hart Dam", Proc. Canadian Dam Safety Earth Dams", Report No. UCB/EERC-76/4,
Conference, Whistler, B.C., 1991, pp. 97-116. College of Engineering, University of
Byrne, P.M., Morris, D.V. and Caldwell, J., California, Berkeley, 1976.
"Seismic Stability of a Tailings Impoundment on Tokimatsu, K.AM. and Seed, H.B., "Evaluation of
Soft Clayey Silt Deposits", 8th World Settlement in Sands Due to Earthquake
Conference on Earthquake Engineering, Vol. 3, Shaking", Journal of Geol. Engineering, ASCE,
pp. 381-388, San Francisco, 1984. Vol. 113, No. 8, pp. 861-878, 1987.
Dobry, R. and Alvarez, L., "Seismic Failures of
Chilean Tailings Dams", Joumal of the Soil
Mechanics and Foundations Division, ASCE,
Vol. 93, No. SM6, Nov. 1967, pp. 237-260.
Duncan, J.M., Byrne, P.M., Wong, K.S. and Mabry,
P., "Strength, Stress-Strain and Bulk Modulus
Parameters for Finite Element Analyses:
Stresses and Movements in Soil Masses", Report
No. UCB/BT/80-01, 1980.
Finn, W.D. Liam, Yogendrakumar, M., Yoshida,
N. and Yoshida, I-1., “TARA-3: A Program to
Compute the Response of 2-D Embankments
and Soil-Structure Interaction Systems to
Seismic Loadings", Dept. of Civil Engineering,
University of B.C., Vancouver, B.C., 1986.
Hamada, M., Towhata, I., Yasuda, S. and Isoyama,
R., "Study of Permanent Ground Displacement

1412