Professional Documents
Culture Documents
W. D. L. FINN
F m r l ofApplied
~ Science, Unive,:si~~
of British Colrrmbirr, Vrrncower, B.C., Cnnrra'n V6T 1 W5
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by SAVANNAHRIVNATLABBF on 11/23/14
AND
P. M. BYRNE
Departtnetzt o f Civil Engitzeeritzg, Uniwrsity of Britisl~Columbia, Vmcoiwer, B.C., Catzndn V6T 1 W 5
Received October 10, 1975
Accepted May 26, 1976
An analytical method for predicting the settlement of a horizontal stratum of dry sand subjected
to earthquake excitation is presented. An equivalent linear viscoelastic dynamic analysis is used to
determine the time-history of shear strains at various levels within the stratum. The vertical
strains caused by these shear strains are calculated by an empirical relationship shown to be in
close agreement with laboratory test data over a wide range of relative densities. This allows the
settlement distribution with depth to be calculated as a function of time.
The method was applied to a 50ft (15 m) stratum of sand. A base acceleration corresponding to
0-10 s of the N-S component of El Centro earthquake was used. Settlement distribution 11s.depth,
as well as surface settlements for various relative densities and earthquake scaling factors are
shown. The effect of surcharge loading on settlements is investigated in an approximate manner.
The results may be used to estimate differential settlement of a structure due to an earthquake.
The results show that while the compaction of the upper layers may be sufficient to reduce the
For personal use only.
settlements due to static loads to tolerable amounts, deep-seated settlements will still occur
during an earthquake.
On presente une mithode analytique pour predire le tassement d'une couche horizontale de
sable sec sujette B une excitation seismique. Une analyse equivalente dynamique en
visco-Clasticite lineaire est utilisee pour determiner I'historique des contraintes de cisaillement B
divers niveaux au sein de la couche. Les deformations verticales causies par ces contraintes de
cisaillement sont calculkes par une relation empirique dont on montre l'accord etroit avec les
donnees d'essais de laboratoire dans une large gamme de densites relatives. Ceci permet le calcul
en fonction du temps de la repartition du tassement avec la profondeur.
La methode a ete appliquie a une couche de sable de 50 pieds (15 m) d'epaisseur. On a utiiise
une acceleration de base correspondant a 0-10 s de la composante N-S du tremblement de terre
d'EI Centro. On montre la repartition du tassement avec la profondeur, ainsi que les tassements
de surface pour diverses densites relatives et differents coefficients d'echelle seismique. L'effet
de la surcharge additionnelle sur les tassements est ktudiee d'une maniere approximative.
Les resultats peuvent 2tre utilises pourestimer le tassement differentiel d'une structure d t i un
tremblement de terre. Les r6st1ltatsmontrent qtle, si le compactage des couches superieures peut
6tre suffisant pour reduire les tassements dus aux charges statiques B des quantites tolerables, des
tassements profonds se produiront malgre tout au cours d'un tremblement de terre.
[Traduit par la revue]
Can. Geotech. J . , 13, 355 (1976)
repeat the above process for each layer of the acceleration. If this is known, an estimate of
sand stratum and integrate the vertical strains the probable maximum settlements will be
to obtain the surface settlement. obtained by multiplying it by a factor of
The second step involves an approximation. about 2.5.
It has recently been shown by Martin et al. A detailed study of volume changes in sand
(1975) that the effect of an irregular shear caused by cyclic shear strains has been pre-
strain history depends not only on the mag- sented by Martin et al. (1975). Their results
nitude of the pulses in the record but the order show that the increment in volumetric strain, y ,
in which they are applied. The procedure for depends on the magnitude of y and the total
determining equivalent cycles of uniform shear volumetric strain, cVd, accumulated during the
strain does not take this fact into account. The previous cycles of shear strain. The accumu-
number of simple shear tests required to cover lated volumetric shear strain is an index of the
the relative density range in a site such as amount of previous straining or the strain
Niigata (Seed and Idriss 1967) could make history of the sand. The volumetric strain
step (iii) time-consuming and expensive. increment may be expressed analytically by
If the relationship between shear strain and
volume change were expressed analytically
then steps (ii), (iii), and (iv) could be elimi- The sand used in the study was Crystal
nated and the distribution of settlements could Silica No. 20 having a Dlo = 0.5 mm and a
be computed directly in conjunction with the uniformity coefficient of 1.5. For this sand at
dynamic analysis in step (i). This type of com- a relative density, D, = 45%, the constants
bined analysis has been made possible by the are C1 = 0.80, C2 = 0.79, C3 = 0.45, and
development of a general analytical expression C4 = 0.73 when the strains are expressed in
between shear strains and vertical strain by percentages. Volumetric strains computed using
Martin et al. (1975). The method of analysis [:I.] were in remarkably close agreement with
which allows the direct computation of the measured strains over a wide range of strain
settlements in the sand due to a given earth- histories at D, = 45%.
quake loading is presented here. It is applicable Procedures for determining the constants C1
to dry or unsaturated sands only. through Ci are given by Martin et al. (1975).
F I N N AND BYRNE 357
These constants vary with sand type and rela- provided the shear strain histories at various
tive density. However, we have found they levels in the deposit can be computed. In
need be determined only at one relative den- horizontal layers of sand the volumetric strains
sity Drl as the volumetric strain increment at are equal to the vertical strains and from the
another relative density Dr2 is given by latter the settlements may be computed.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by SAVANNAHRIVNATLABBF on 11/23/14
Calculation of Settlements
in which R is a parameter that varies with A stratum of unsaturated sand resting on
relative density. When the reference relative horizontal bedrock is shown in Fig. 2. The
density, D,., = 45% and the accumulated stratum is of thickness, H, and the properties
strain E,.~, in [ I ] is computed also at the same of the sand are allowed to vary in the vertical
relative density the following expression for R direction only. Shaking is due to shear waves
was established to agree with the experimental propagating upwards from the rock base. Un-
data presented by Silver and Seed ( 1971) : der these assumptions, the dynamic response
becomes that of a one-dimensional shear beam.
Since the layer properties may vary vertically
in a random manner, the shear beam can be
approximated by the discrete mass model
The computed and measured values of accu- shown in Fig. 2. The masses are connected by
mulated volumetric strains after 10 cycles of appropriate springs and dampers. The stress-
various magnitudes of shear strain for D , = 60 strain characteristics of sands are non-linear,
and 80% are shown in Fig. 1. The agreement hysteretic, and strain dependent. These charac-
For personal use only.
between measured and computed values of teristics are introduced into the model by
accumulated volumetric strain appears ade- using the equivalent linear strain dependent
quate for practical purposes. shear moduli and viscous damping ratios de-
During an earthquake a sand deposit is fined by Seed and Idriss (1969) and shown in
subjected to a train of non-uniform shear Fig. 3. The elemental spring stiffnesses, kt, are
strains. The procedure outlined above for cal- based on the current moduli, Gi, at the middle
culating the volumetric strains has been shown of each layer of thickness, hi;ki = Gi/hi.
by Martin et nl. (1975) to be applicable to The shear modulus G at any time is given
such a sequence of non-uniform strain cycles. by
Therefore the distribution of volumetric
strains in a sand deposit can be determined [41 G = 1000 K,(a,,,')'/'
in which is the mean effective normal stress
and K, is a parameter which varies with strain
and relative density as shown in Fig. 3.
Meosured (Seed and Silver 1972)
i8t Predicted
"
0.01 0.1 1.0
Sheor S t r o i n in percent
:
.- 12
An iterative procedure is used for solving
Q (a) [ICI]. Initial values for shear modulus and
: 8 I/ damping ratio are assumed to be those for a
strain, = The response of the system
to the prescribed base motion is then computed
"lo-4 lo-' 10- 10 -' I over a short time interval At such as 0.5 s
S h e a r Strain in%
80 I I I I I assuming viscoelastic behaviour. The distribu-
tion of average shear strains within the soil
profile is determined. If these strains differ by
more than an arbitrarily prescribed amount
from the strains assumed at the beginning of
the analysis, new values for the moduli and
damping are selected and the system of equa-
tions again solved. The procedure is repeated
until two consecutive sets of strains are ob-
tained which agree within the prescribed toler-
ances. At this stage strain compatible stiffnesses
For personal use only.
in which hl = A~ = A,,
25
N
,
:
c
.-
:: 0
~151 [CI = [MI[+:I[A:I[ + I ~ [ M I
"
L"
summed, the distribution of settlements FIG. 5. Dynamic stresses and associated strains.
throughout the depth of the layer are obtained
as a function of time. 45, 60, and 8 0 % . The dependence of settle-
The distribution of total settlements under ment on relative density is obvious. It should
shaking by the N-S component of El Centro be noted that even very dense sands (D, =
(1940) with maximum acceleration of 0.32 g 80% ) can undergo measurable settlement
is shown in Fig. 6 for relative densities D, = under strong shaking. In the example shown,
360 CAN. GEOTECH. J. VOL. 13, 1976
0 0.1
S e t t l e m e n t in f e e t
0.2 -
0 0.1 0.2 0.3 0.4 0.5
For personal use only.
M a x . B a s e A c c e l e r a t i o n in gravity units
FIG.6. Settlement vs. depth for various densities.
FIG. 7. Surface settlements vs. maximum base
the settlement was of the order of 0.25 in acceleration.
(0.64 cm) when the relative density D, =
building has an opposite effect. During an
80%.
earthquake, the base shear generated in the
The total settlement in a given sand stratum
building by inertia forces generates additional
depends on the frequency content and magni-
tude of the base input motion, because these shear stresses and strains in the sand and tends
factors have a predominant effect on the dy- to increase settlements.
namic response of a given sand layer. Since A very approximate model of the action of
the motion is fed in at bedrock or an equiva- a structure was used to investigate these effects.
lent stiff layer, the frequency content should This model took into account only the in-
be rather high. This kind of frequency content creased effective pressures between the sand
is represented reasonably well by the El Centro grains due to the weight of the structure and
record. If necessary the record may be scaled the increased shear stresses generated by the
to higher frequencies. mass of the structure. The flexibility of the
The effect of variations of the maximum structure and such vibration modes as rocking
accelerations are shown as functions of the were not taken into account as they would be
relative density in Fig. 7. The non-linear nature incompatible with the simple one-dimensional
of the response of the sand is clearly evident. analysis used.
The effect of a structure on settlements
Eflect of a Structure was modelled by adding an extra slice to the
The settlement of a sand will be affected by discrete mass system with a very high stiffness
the existence of a structure founded on the and a mass sufficient to cause foundation pres-
sand. The weight of the structure increases the sures ranging from 2000 to 8000 psf (1 to 4
mean normal effective stress in the sand and so kg/cm2). The distribution of settlements with
increases the shear modulus G, in accordance depth caused by the first 10 s of El Centro
with [4]. The increased values of G lead to in- (1940) are shown in Fig. 8 for both the free
creased resistance to shearing strain and hence field case of no surcharge ( a = 0 ) and for a
to volume change. This effect tends to reduce building causing a surcharge of a = 4000 psf
the total settlement. However, the mass of the (2 kg/cm3). The presence of the surcharge
FINN AND BYRNE 361
I0 -
/ the f u l l depth o f
undrained state are almost identical until the
-
/ the L a y e r
- pore-water pressures in the undrained sands
-
m
Z 20-
/'
/
D,=SO% -
rise above 30% of the overburden pressures.
The settlement of the saturated sands would
be delayed until dissipation of these excess
.-
a
pore pressures occurred. For higher pore-
water pressures, settlements can also be com-
3 0
puted, but at such higher water pressures there
is more concern about a possible foundation
40
failure or complete liquefaction than with
settlements.
Conclusions
An analytical method has been presented
Surface Settlement in feet for computing estimates of the settlements of
FIG. 10. Settlement distribution in non-uniform strata of unsaturated sand during an earth-
For personal use only.
show that the relative density of the sand, the 1969. Influence of soil conditions on ground mo-
time history of earthquake accelerations, and tions during earthquakes. ASCE, J . Soil Mech. Found.
Div. 95(SM1), pp. 99-137.
the surface masses and pressures of structures 1970. Soil moduli and damping factors for dynamic
are the primary factors affecting the settle- response analyses. Rep. No. EERC 70-10, Univ. Calif.
ments. E a r t h q ~ ~ a kEng.
e Res. Cent.. Berkeley, Calif.
The procedure and results described above SEED, H. B. and S I L V E RM. , L. 1972. Settlement of dry
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by SAVANNAHRIVNATLABBF on 11/23/14