Professional Documents
Culture Documents
INTRODUCTION
1141
depth resulting from pounder penetration and the depth and level of soil
improvement beneath the impact location is proposed. Case histories of two
dynamic compaction projects are analyzed to demonstrate the viability of
the proposed model.
Pounder
ft /^^ /V$s
°'i -iT
s %W
Soil springs j Soil Surrounding
and dashpots ^ 1 1 column soil
777779
n-1
K ftnV*n Simulation of
field problem
One-dimensional
finite element model
1142
equation of motion of the pounder is similar to (1), except that the density
of the soil is replaced by the density of the pounder material (p), and the
axial stiffness is characterized by the Young's modulus of the pounder ma-
terial (£); the frictional resistance at the side of the pounder during pen-
etration, after the crater has been formed, is assumed to be small (i.e., ks
= cs = 0).
Axial strain, E
1143
da (e^Y
M = — = M0 , "' .,- (virgin loading) (3)
cte (E* — e) z
cr = M 0 (unloading) (5)
(e* - e)
where i| = ratio of residual strain in the unloaded state at b (eb) to peak
strain at state a (e 0 ) (r\ = e t /e„ is a measure of the plastic behavior); and
E^ = corresponding strain on the virgin loading curve at the same stress level
as the strain on the unloading branch (see Fig. 2). Making e the subject of
the relation, (4) now becomes
E - T|S0
-'(T^i) "••(6)
The tangent-constrained modulus during unloading can then be determined
as follows:
,. da (da\/de\ MQ (e*) 2 . , ,. ,
M (unloadmg) (7)
= Ts = U j {Te) = ( T ^ o~=^
A linear stress-strain response is assumed during reloading until the stress
level given by the virgin-loading curve is reached, and thereafter, t h e stress-
strain behavior follows the virgin-loading curve. The linear behavior during
reloading is characterized by the tangent constrained modulus at state b
(see Fig. 2), regardless of whether t h e soil has been completely unloaded.
This phenomenological soil model is characterized by three parameters,
namely, M () , E*, and j \ . T h e parameters M 0 , e*, and r\ can be determined
using Ginsburg's model [(2), (4), and (5)] to fit the measured axial stress-
strain response from dynamic uniaxial strain tests o n sand samples. O n e
such test setup for fairly uniform loose sand of density 1.45 t/m 3 was reported
by Heierli (1962). Two typical axial stress-strain curves of the sand measured
from these tests are shown in Figs. 3(A) and 3(b), corresponding to single
and multiple loading-unloading responses, respectively. The computed axial
stress-strain response using (2), (4), and (5) with parameters M0 = 9.0 MN/
m 2 , E* = 0.45, and t) = 0.8 are also shown in Figs. 3(a) and 3(b). T h e
phenomenological soil model characterized by (2), (4), a n d (5) can be used
to obtain a reasonable fit of the measured response, not only during virgin
loading, but also during unloading as well as reloading.
D u e to the difficulty of obtaining undisturbed samples from t h e site, it
is perhaps preferable to estimate the initial tangent constrained modulus,
M 0 , based on correlation with in situ tests results. O n e such correlation with
the cone resistance of the soil reported by L u n n e and Christoffersen (1983)
is shown in Fig. 4. Sands before dynamic compaction m a y perhaps be as-
sumed to be normally consolidated. N o such correlations are available for
e* and y\. It is, however, of interest to note that analysis of further test
1144
t 8 12
(a) Loading and unloading curve (b) Loading,unloading and reloading curve
•S 300
•5L
//
//
J 250
Ov;r-Consolidated/
| 200 Sands /
•i ' 50
/ //
o
/ //
•E 100
a>
en
t= // ^-^^s
d
z^^Normally Consolidated Sands
~ 50
— o
0 10 20 30 tO 50 60 70
Cone resistance, qc I MN/m')
^ = ( T 3 - ^ •••• "(i°)
C (11)
" = T^T
where v, = soil Poisson ratio. This approximation is somewhat crude, but
it is not critical in the numerical simulation since the base of the soil column
lies beyond the depth of influence.
Vesic et al. (1965) observed experimentally that very rapidly loaded foot-
ings show a punch-through type of failure mode, characterized by near-
vertical slopes of the sides of the depression with little heave of the adjacent
soil. This mode of failure differs from general shear failure observed in
statically loaded footings. This punch-through failure mode is also observed
in DC projects in the field. This effect is simulated in the numerical model
by the penetration of the soil column into the ground through the failure
of the soil springs when they exceed their static resistances. This assumes
the soil springs behave in an elastic-perfectly plastic manner. For granular
soils, the static resistances of the adjacent soil springs can be determined
from
Qs = K,d'v tan ef>Av (12)
Qb = Ny,Ab (13)
where Qs = shaft static resistance at the node; Qh = base resistance; As
= shaft area associated with the node; A,, = base area of soil column; cr',
= vertical effective stress at the appropriate depth; <$> = friction angle of
sand; Ks = coefficient of lateral pressure; and Nq = bearing capacity factor.
NCI is assumed to be similar to that used for pile design and can be obtained
from standard design charts [e.g., Meyerhof (1976)]. When the soil spring
fails, the dashpot is disconnected in the numerical solution, and radiation
damping ceases across the column-soil interface. Radiation damping is re-
established when the spring force becomes less than the static resistance.
In the finite element approach, both the pounder and soil column are
discretized into a number of one-dimensional elements. Each element has
two degrees of freedom, one at each node. Assuming a linear variation in
axial deformation along the element and applying the weighted residual
method [see e.g., Smith and Griffiths (1988)] to (1) gives the following
element matrix equation for the soil column:
32w „ 5w
Mc —r
—2 ++ C — + Kcw + Ksw = 0 (14)
at dt
where Mc = element mass matrix of soil column; C = damping matrix of
dashpot; Kc = element stiffness matrix of soil column; Ks = stiffness matrix
of soil spring; and w = vertical displacement vector. The expressions for
1146
element matrices are assembled in the usual manner leading to the following
global matrix equation for the pounder-soil interaction problem.
M*w + C*w + K*w = 0 (15)
where M*, C*, and K* = the assembled global mass, damping, and stiffness
matrices, respectively; and w, w, and w = the nodal vertical acceleration,
velocity, and displacement vectors, respectively. The global matrix (15) is
solved using a Newmark implicit direct step-by-step integration scheme. The
analysis is performed by prescribing the impact velocity of the pounder to
the associated nodes as the initial condition. The impact velocity can be
obtained from the expression
v = V2g#£ (16)
where v = impact velocity of pounder; g = acceleration due to gravity; H
= height of fall; and £ = efficiency of the drop (£ = 1 for free fall). Some
energy loss can be expected if the pounder is allowed to drop by means of
cable and winch.
In the present work, the tangent stiffness approach is used to compute
the nonlinear response of the soil. A small time step is adopted to allow
the nonlinear stress-strain behavior of the soil column elements as well as
the surrounding soil springs to be followed closely, and the response of the
system is computed incrementally. At the end of each drop, the maximum
incremental strains in these elements are about 8%; for these strain levels,
Teng (1981) showed that for one-dimensional compression analysis of a soil
column, small-strain theory, which is used in the present work, is adequate.
The element sizes in the soil column are updated at the end of each drop
to ensure that small-strain theory is stilf relevant. The complexity of the
present problem, particularly that of the soil behavior, does not merit the
refinement of large-strain theory.
PREDICTIVE PROCEDURE
The initial Ks can be estimated from the cj> angle from Ks — 1 — sin cj>. The
parameter Nq is assumed to be similar to that used for pile design and can
be obtained from standard design chart [e.g., Meyerhof (1976)]. The pa-
rameters e* and r\, based on the work reported earlier, typically range from
0.4 to 0.5 and 0.8 to 0.9, respectively.
For each drop of the pounder, wave equation analysis is performed by
prescribing the impact velocity of the pounder as given by (16). At the end
of each drop, the analysis gives the pounder penetration and residual axial
stresses and strains in the soil column elements. The element sizes and soil
densities of the soil column elements are then updated for the next drop.
The estimation of the improved soil properties of the soil column after
impact cannot be directly obtained from the simple phenomenological soil
model used. It is, however, plausible to assume the relative density of soil
after impact (£>,.) is related to the residual strains in the soil column elements
(er) by the following relationship
D,. ~ Dri, , , ^r n„ ,
D,,» — £>,.,,/ \e* —
where D,^ = initial relative density of soil; D,., = limiting relative density
of soil (typically 90% to 100%); E0 = initial axial strain (= 0); and e* =
limiting strain (typically 0.4-0.5). The initial relative density of soil can be
determined in the laboratory or estimated based on Meyerhofs (1976) cor-
relation with the friction angle, estimated from cone resistance using (18),
i.e.
cf> = 28° + 15°D, (20)
Schmertmann's (1976) correlation of relative density with measured cone
resistance, which is commonly used, is strictly applicable to normally con-
solidated sands only. If the sand is overconsolidated, which is relevant to
sand after DC, the measured cone resistance must first be transformed to
that of an equivalent normally consolidated sand before the correlation can
be used. However, the coefficient of lateral pressure, which is the essential
parameter of this transformation, is difficult to determine, and the results
are sensitive to this parameter. The improved friction angle of sand after
DC can be estimated from (20) using the improved Dr from (19). The
improved soil density (ps) can be estimated using
Pmax(A 9min
Dr = ~ \ (21)
PsVrmax Pmin/
where pmax and pmin = maximum and minimum dry soil densities, respec-
tively; these values can be determined in the laboratory. However, pmax and
pmin may change as a result of crushing of the sand grains during the com-
paction process. Nevertheless, the numerical results were not sensitive to
this approximation, since ps lies within a narrow range.
The impact of the pounder will improve not only the properties of the
soil directly beneath it, but the surrounding soil as well, although to a
somewhat lesser extent, and this effect reduces with distance from the point
1148
adjacent to the soil column would be quite similar to those of the soil column,
the improvement of the surrounding soil may be approximated by updating
the soil parameters of the soil springs and dashpots before the analysis for
the next drop. The improved cone resistances are first estimated from the
updated friction angles based on (18). The constrained moduli are then
estimated from the curve for overconsolidated sands in Fig. 4, and there-
after, the soil shear moduli of the soil springs and dashpots are determined
using (17). Because of the difficulty quantifying the effect of compaction
on the coefficient of lateral pressure (Ks), this parameter after impact is
estimated based on a similar relationship given in (19), except that the
relative densities of soil are replaced by the corresponding coefficients of
lateral pressure. There is some evidence to suggest that Ks of overconsoli-
dated sands lies in the range of 1.2 to 2.0 [see Baldi et al. (1982)], and
hence, the limiting coefficient of lateral pressure can be taken as 2.0. The
parameter Nq for the base of the soil column that is beyond the depth of
influence is generally unchanged. It is, however, noted that the computed
pounder penetration and degree and depth of improvement of the soil,
which are essentially governed by the nonlinear behavior of the soil column,
are found to be insensitive to the updating of the soil parameters of the soil
springs and dashpots.
For the next drop, the soil elements in the soil column take on the residual
axial stresses and strains from the previous drop. The degree and depth of
improvement of the soil in the soil column in terms of relative density or
friction angle can then be estimated by comparing these values after the
last drop in the pass with that before DC.
If more than one pass is executed, the infilling of the crater using loose
sand can be simulated in the model by adding new soil elements to the top
of the existing soil column until its new level is reached. The soil properties
of these new elements can be estimated from typical values corresponding
to loose sand. These new soil elements and their adjacent soil springs will
have zero initial stress and strain, unlike the existing soil elements. The
analysis can then be repeated.
This predictive procedure is summarized in a step-by-step manner as
follows.
1. Discretize the pounder and the soil column into one-dimensional elements.
2. Determine the input parameters.
A. Pounder: E and p.
B. Soil column elements: M0, e*, T|, and ps.
C. Soil springs and dashpots: Gs, p„ vs, (j>, Ks, and Nq.
3. For each drop in the pass:
A. Perform wave equation analysis using estimated impact velocity of the
pounder. The analysis gives the pounder penetration and residual axial
stresses and strains in the soil column elements.
B. Update the soil column element sizes, the densities of the soil column
elements, and the parameters of the soil springs and dashpots for the
next drop.
4. At the end of the pass, obtain the cumulative pounder penetration, relative
density, and friction angle of the improved soil in the soil column elements.
1149
CASE STUDIES
The one-dimensional wave equation model was validated for small strain
levels in the soil with the results of a related theoretical problem of an
impulsively loaded circular footing resting on a linear elastic half-space and
a laboratory impact test on a circular surface footing on sand (Chow et al.
1991). The numerical solutions agreed closely with the analytical solutions
and the laboratory test results. Available data from two case histories of
DC are analyzed herein to assess the viability of the proposed wave equation
model in predicting the improvement of soil due to high-energy impact in
the field.
4-
r'L re ...,„•
i
!i #f + CPI CI
Cent of print !i CPTC2
I I I ____J I
1.22 m 0.30 m
FIG. 5. Indianapolis Site: Drop Pattern and Location of CPT at Preliminary Trial
Sand
Clay
sandy clay at the base of the soil column were assumed to be as follows:
Poisson ratio of 0.4, density of 1.8 t/m3, undrained shear strength (c„) of
120 kN/m2 (estimated from cone resistance), shear modulus of 150 c„, and
ultimate bearing pressure of 9 c„.
The computed crater depth after each drop in the preliminary trial is
shown in Fig. 7, and the results agree reasonably well with the measured
values at drop point 1. No crater depth measurements at footing H-l were
reported. The friction angle and relative density profiles before and after
DC for the trial, estimated from correlations with measured cone resistance
[see Fig. 6(A)] using (18) and (20), are shown in Fig. 8(a). Before compac-
tion, the friction angle was typically 31°, corresponding to a relative density
of about 30%. After compaction, the friction angle of sand in the top 3 m
of the soil column was improved to about 33.5°, corresponding to a relative
density of about 40%. The computed relative density and friction angle
profiles, shown in Fig. 8(a) are slightly greater than those estimated from
the correlations with measured cone resistances throughout the entire depth
of the soil column. However, the computed results for the relative density
and friction angle profiles after DC at the location of footing H-l during
the main works, shown in Fig. 8(b), agree reasonably well with those es-
timated from correlations with the measured cone resistance.
Before
OC " H
- - A f t e r OC
- Estimated
from q c
- /—*'"" Computed
20 U 60 BO 100 20 to
Relative density,0r |%| Relative density, Dr (%)
_ U _ Water tcKe
loose to
medium
sands
silty
clay
loose
sands
limestone
Cone penetration test results carried out at the impact point (F6) before
compaction as well as after each pass are shown in Fig. 9.
In the analysis, the cone penetration resistance profile measured before
DC was used to estimate the soil properties of the 14-m-deep soil column,
as well as the surrounding soil springs and dashpots. In the top 10 m and
between 12- and 14-m depth, which comprised mostly loose sand, the initial
tangent-constrained modulus, friction angle, and relative density were es-
timated using the curve for normally consolidated sands in Fig. 4, (18), and
(20), respectively. The friction angle of this sand ranged from 33° to 37°,
whereas its relative density ranged typically from 30% to 60%. The param-
eter Ks [~l — sin(4>)] was about 0.4-0.46. The Poisson ratio of the sand
was assumed to be 0.33, and the soil density was estimated from initial
relative density using (21); the densities pmax and pmin were assumed to be
1.85 t/m3 and 1.35 t/m3, respectively [see Lambe and Whitman (1979)]. The
remaining soil parameters for the sand assumed in the analysis were: Dr,
= 90%, e* = 0.45, and T\ = 0.85. The silty clay found between 10 to 12
m is below the anticipated depth of improvement, and hence, it is reasonable
to assume its behavior to be linear elastic. The soil parameters of the silty
clay were assumed as follows: Poisson ratio of 0.4, density of 1.8 t/m3,
undrained shear strength (c„) of 115 kN/m2 (estimated from the measured
cone resistance), and shear modulus of 150 c„. The density, Poisson ratio,
and shear modulus of the underlying limestone, located 14 m below the
surface, were assumed to be 1.8 t/m3, 0.25, and 60 MN/m2, respectively. It
is noted that the underlying limestone that was beyond the depth of im-
provement has little influence on the response of the soil column.
The computed crater depth (cumulative pounder penetration) after each
drop in the first pass is shown in Fig. 10(a). The computed results show
reasonably good agreement with the field measurements at the impact point
(F6) and another impact point just next to it. As shown in Fig. 10(b), the
computed relative density and friction angle profiles after the first pass,
following the procedure described in the earlier section, also agree reason-
ably well with those estimated using Meyerhofs (1976) correlations, given
in (18) and (20). The friction angle and relative density profiles before DC,
estimated from the measured cone resistance values at the impact point
1153
0 20 1.0 SO 10 100
Relative density, Dr(%)
FIG. 10. Kampung Pakar Site: Computed and Measured Results after First Pass
Mtosuied IFS1
p — Computed
inn .
FIG. 1 1 . Kampung Pakar Site: Computed and Measured Results after Second
Pass
(F6), are also included to show the improvement achieved after the first
pass.
For the second pass, the infilling of the crater depth from the first pass
was simulated by adding new soil elements to the existing soil column and
the analysis repeated as in the first pass. The computed crater depth after
each drop in the second pass is shown in Fig. 11(a), and the results compare
favorably with the field measurements at the impact point (F6). As shown
in Fig. 11(b), the computed relative density and friction angle profiles after
the second pass also agree reasonably well with those estimated from cor-
relations with the measured cone resistances. The influence of ground water
was not considered in the wave equation model. Nevertheless, it appears
1154
CONCLUSIONS
ACKNOWLEDGMENTS
The work upon which the paper is based is funded by the National Science
and Technology Board of Singapore under the R D A S grant number ST/86/
05.
APPENDIX I. REFERENCES
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M., and Pasqualini, E. (1982).
"Design parameters for sands from CPT." Proc. 2nd Eur. Symp. on Penetration
Testing, International Society for Soil Mechanics and Foundation Engineering,
425-432.
Chow, Y. K., Wong, K. Y., Karunaratne, G. P., and Lee, S. L. (1988). "Wave
equation analysis of piles—A rational theoretical approach." Proc. 3rd Int. Conf.
Application of Stress—Wave Theory on Piles, Canadian Geotechnical Society, 208-
218.
Chow, Y. K., Yong, D. M., Yong, K. Y., and Lee, S. L. (1991). "Numerical
modelling of dynamic compaction." Proc. 7th Int. Conf. on Computer Methods
and Advances in Geomechanics, International Association for Computer Methods
and Advances in Geomechanics, 237-242.
Ginsburg, T. (1964). "Propagation of shock waves in the ground." /. Soil Mech.
Found. Div., ASCE, 90(1), 125-163.
Heierli, W. (1962). "Inelastic wave propagation in soil columns." /. Soil. Mech.
Found. Div., ASCE, 88(6), 33-63.
Holeyman, A. (1985). "Unidimensional modellization of dynamic footing behavior."
Proc. 11th Int. Conf. on Soil Mech. and Found. Engrg., International Society for
Soil Mechanics and Foundation Engineering, 761-764.
Lambe, T. W., and Whitman, R. V. (1979). Soil mechanics. John Wiley and Sons,
New York, N.Y.
Lee, S. L., Chow, Y. K., Karunaratne, G. P., and Wong, K. Y. (1988). "Rational
wave equation model for pile driving analysis." /. Geotech. Engrg., ASCE, 114(3),
306-325.
1155
1156
1157