You are on page 1of 15

LIPKIN,J., BENNETT,R. H. & MCTIGUE,D. F. (1986). Giorechnique 36, No. 1.

11-25

Consolidation under an isotropic total stress


increase : part II, experimental results
for marine clay

J. LIPKIN,* R. H. BENNETT? and D. F. McTIGUE$

Deformation and internal pressure changes have been Ceci s’accorde assez bien avec les mesures indbpen-
measured in a large sample (of the order of 1 m3) of dantes.
remoulded, reconsolidated marine sediment subjected to
a series of step-like, 6.9 MPa changes in hydrostatic KEYWORDS: clays; compressibility; consolidation;
pressure. Undrained surface displacements of 0.24.4 elasticity; pore pressures; soil properties.
mm were observed in response to each increment in
external pressure. The surface then slowly rebounded INTRODUCTION
with a characteristic time of about lo6 s. The excess An in situ heat transfer experiment (ISHTE) has
pore pressure induced by each step in pressure at the been under development for several years as part
boundary was of the order of from - 1 kPa to -2 kPa of the US Subseabed Disposal Program (SDP)
(suction) and likewise relaxed as external fluid diffused (Hollister, Anderson & Heath, 1981). It is
into the sample. Predictions based on a poroelastic intended that this experiment will be fielded over
material response model agree qualitatively with these
a one-year period at a water depth of approx-
observations when values of material properties are esti-
mated from independent measurements. The experimen- imately 6000 m in the central North Pacific
tal data can also be used to determine parameter values. Ocean. The experiment will involve the emplace-
Such a procedure yields very good fits to the data and ment of a heat source and associated instrumen-
indicates permeability, drained bulk modulus and solid tation in the sea floor to a depth of 1 m. The
bulk moduli of the order of lo-l6 m*, 10 MPa and 10 instrumentation will be capable of making in situ
GPa respectively, in reasonable agreement with inde- measurements of the sediment thermal, mechani-
pendent measurements. cal and chemical responses to this heat source.
The data acquired will be used to assess the valid-
Des changements de dbformation et de pression interne
ity of some of the numerical modelling techniques
ont et& measures dans un grand &chantillon (de I’ordre
de 1 m3) de sidiment marin remani& et reconsolidt
being used in the SDP feasibility study (Percival,
soumis g une s&ie de modification de pression hydrau- 1983).
lique par paliers de 6,9 MPa. Des d&placements non- A large-scale simulation experiment has been
drain&s superficiels compris entre 0,2 et 0,4 mm ont 8t& completed as a prelude to the fielding of ISHTE.
observks en rCponse B chaque augmentation de pression The details and purposes of this simulation
externe. Puis la surface s’est lentement restauree au experiment have been described by Percival
tours d’un temps caractkristique d’environ lo6 s. Les (1982); however, it is only necessary here to note
surpressions induites & la limite lors de chaque palier de that the experiment made use of a temperature-
pression etaient de I’ordre de - 1 kPa a -2 kPa
controlled pressure vessel to cool and pressurize a
(suction) et se relichaient de faGon analogue au fur et B
mesure que du fluide externe se diffusait dans
large volume (approximately 1 m3) of remoulded,
l’&chantillon. Des prkvisions bashes sur un modtle de reconsolidated marine sediment. Simulated deep
rkponse d’une mat&e porotlastique s’accordent de ocean conditions of 55.2 MPa hydrostatic press-
faGon qualitative lorsque les valeurs des propriCt&s des ure and 4°C were achieved in this way before
matitres sont estimCes B partir de mesures indkpen- carrying out the experiment. These conditions
dantes. Les donnies exp6rimentales peuvent s’employer were then maintained for 30 days while a heater
aussi pour dkterminer les valeurs des paramttres. Une experiment was conducted in the sediment, after
telle procbdure donne des rttsultats tris compatibles which the sediment was returned to atmospheric
avec les don&es et indique une perm&abilit&, un module
pressure.
de masse drain&e et des modules de masse solide de
l’ordre de lo-l6 m*, 10 MPa et 10 GPa respectivement.
In this Paper data obtained during the isother-
mal phases of the simulation experiment (i.e.
during pressurization and depressurization) are
presented and compared with the predictions of
Discussion on this Paper closes on 1 July 1986. For
further details see inside back cover. the poroelastic material response model presented
* Sandia National Laboratories, Livermore. by McTigue, Lipkin & Bennett (1986). The partic-
t Naval Ocean Research and Development Activity. ular data considered are time-resolved measure-
$ Sandia National Laboratories, Albuquerque. ments of
11
12 LIPKIN, BENNETT AND McTIGUE

(a) the displacement of the sediment-water inter- experiment demands careful treatment of the rela-
face tive compressibilities of the fluid and solid sedi-
(b) the sediment pore pressure at two interior ment constituents.
points in the sample volume. Sediment characteristics and sample prep-
aration are reviewed in the following section. The
A great deal of important information can be instrumentation used to obtain surface displace-
extracted from these data. Indeed, it will be ment and pore pressure data is described in the
shown that the simulation experiment provided a next section, and the data are presented and com-
unique opportunity to examine marine sediment pared with analytical predictions in the fourth
physical properties under simulated deep ocean section. Discussion and conclusions derived from
conditions. this study are given in the final section.
By fitting the predictions of the analytical
model to the experimental data for undrained dis- SEDIMENT CHARACTERISTICS AND
placement jumps, asymptotic (late time) displace- SAMPLE PREPARATION
ments, pore pressure jumps and relaxation rates, The sediment used in the ISHTE simulation
values of four sediment material constants can be experiment was dredged from the floor of the
estimated. Use of this procedure, however, also North Pacific Ocean about 1500 km north-west
requires independent knowledge of four addi- of Hawaii. The water depth in the area is about
tional material constants (McTigue et al., 1986). 5800 m. The sediment is approximately 65% clay
By way of review (from part I), the following eight and approximately 35% silt, with the clay frac-
parameters are regarded as fundamental in the tion dominated by illite. The in situ porosity is
material model about 0.75.
initial porosity Sample preparation was carried out by the
fluid bulk modulus University of Rhode Island Marine Geo-
elastic shear modulus mechanics Laboratory (Silva, Jordan & Cri-
first solid bulk modulus scenzo, 1984). The sediment was first sieved to
second solid bulk modulus remove manganese nodules. It was then reconsti-
drained bulk modulus tuted to a thick slurry by adding sea water and
permeability mixing. A reinforced steel tank was specially fab-
fluid viscosity ricated to contain the sediment sample during its
consolidation as well as during the simulation
&, p, K, and G are assumed to be known. experiment. This tank was 1 m in diameter with a
However, values for the solid bulk moduli K,’ fixed height of 1 m plus a detachable sleeve that
and K,” are difficult to obtain directly, and only a increased the height to 1.5 m to accommodate the
few have been reported in the literature. There- additional sediment volume needed before con-
fore K,’ and K,” are regarded as material con- solidation. A highly permeable drainage fabric
stants whose values are to be determined from and an adjacent layer of filter material were used
the displacement and pore pressure data. In addi- to line the tank before it was filled with the sedi-
tion, the permeability k and drained bulk ment slurry. In the filling process, the sediment
modulus K are found. slurry was poured into the tank in layers approx-
It should be noted that sediment permeability imately 0.2 m thick. After each layer had been
is perhaps the single most important mechanical added, the tank was sealed and a 30 kPa vacuum
parameter that arises in the radionuclide trans- was applied for 1200 s to remove any air that
port and porewater convection analyses that are may have been trapped in the slurry during the
being done for generic sub-sea bed waste reposi- mixing process. This procedure was carried out
tories (Seabed Programs Division 6334, 1983). An continuously for 43.2 x lo3 s (12 h). After the full
independent determination of permeability made height of 1.5 m had been reached, the top of the
from in situ measurements at high pressure and tank was covered by filter material and drainage
low temperature is thus a valuable result that can fabric.
provide increased confidence in existing labor- Reconsolidation was directed towards re-
atory data. In addition, the data and analysis pre- turning the sediment to its in situ porosity. This
sented here provide a unique opportunity to was accomplished by loading the sample with
explore effects of fluid and solid compressibility steel plates placed on the top surface. The load
on the deformation of a saturated marine sedi- was increased in three increments over 60.5 x lo4
ment. These effects are commonly neglected in s (7 days) to a maximum stress of 12.9 kPa. After
classical soil mechanics. However, it will be an additional 51.8 x lo4 s (6 days), the entire
shown that obtaining an understanding of the tank was placed in a cold box at 5°C and con-
sediment response observed in the simulation solidation was allowed to proceed for 6.22 Ms (72
CONSOLIDATION UNDER STRESS INCREASE 13

days). At this time, the load was removed, the linear variable differential transducer (LVDT)
sediment surface rebounded and, after an addi- made by Shaevitz Engineering, Pennsauken, New
tional 259 x lo4 s (3 days), the final state was Jersey, which was modified so that its electri-
reached. The sample was then trimmed to the cal output would be insensitive to hydrostatic
fixed height of the tank (1 m). pressure.
The trimmed material was sampled radially for Space limitations on the instrumentation frame
water content. These measurements indicate the required remote positioning of the LVDT. A lever
degree of uniformity in the sediment sample arm-sediment follower arrangement was there-
before the simulation experiment. As anticipated, fore developed so that displacements could be
the water content was lower near the drained measured near the centre of the sediment tank.
boundaries of the tank, but the variation was Such positioning of the follower was important
small. The porosity at the centre was 0.76, and for obtaining data during the thermal part of the
that at the boundary was 0.73, a change of about experiment. However, the resulting closeness of
4%. Post-test water content analyses show varia- the follower to the heater implant arm suggests a
tions of the same order (Silva er al., 1984). possible source of error in the isothermal dis-
placement data. A discussion of such measure-
INSTRUMENTATION ment errors is deferred to the final section.
General An additional consideration associated with
All the instrumentation used in the ISHTE the use of an LVDT in a simulated deep ocean
simulation experiment was attached to a steel environment is the need to isolate it from the
frame developed by the Applied Physics Labor- potentially detrimental effects of extended expo-
atory of the University of Washington, Seattle, sure to sea water. In the present application, this
Washington (Miller, Miller & Olson, 1984). This isolation was accomplished by mounting the
frame was in turn mounted on the sediment tank LVDT in a plastic tube with a very flexible latex
so that the instruments could be positioned rela- bladder on one end to permit access of the lever
tive to the rigid structure. In addition to the arm. The tube and bladder were filled with
instrumentation discussed in this Paper, the mineral oil, and the required electrical connec-
experiment included the use of a resistance heater, tions were made through oil-filled lines to the
thermal sensors (both thermocouples and pressure vessel feed throughs. Excitation and
thermistors), a thermal conductivity probe and a signal voltages were transmitted through these
porewater sampler. Miller er al. (1984) present lines, and the signal voltages were recorded at
details related to the development and calibration preprogrammed intervals on an Esterline Angus
of the thermal instrumentation. (Indianapolis, Indiana) data logger. The same
data logging system was used to record pore
Sediment surface displacement pressure data during the experiment. The record-
Figure 1 is a schematic drawing of the device ing interval was varied depending on the fre-
used to obtain a time-resolved measurement of quency of the changes expected to occur in a
the displacement of the top surface of the sedi- particular phase of the one-month experiment.
ment. The sensing element in this device was a An interval of 9 x 10’ s was used for the
LVDT electrical leads
(in oil-fllled tube)
/

s!/ LVDT housing

LVDT body (flxed)


Heater Implant tube
LVDT core rod (movable)

Fig. 1. Schematic diagram of the sediment surface displacement measuring system


14 LIPKIN, BENNETT AND McTlGUE

Differential
Tubing to pressure
pressure
Porous transducer (porewater)
Cone angle = 5-3”

conducttng electronic signal


cable conditioner

+D

Fig. 2. Schematic diagram of the deep ocean piezometer probe

3.9 x lo5 s (4.5 days) pressurization phase. The stainless steel housing that is pressure compen-
time required for each individual pressure change sated to in situ hydrostatic pressure (Fig. 2). The
during this phase was approximately 2 x lo3 s, stainless steel pressure sensor housing is separat-
with hold times at constant pressure after each ed from the titanium probe components by high
step of the order of 2 x lo4 s; the 9 x 10’ s data dielectric polycarbonate material. The total
logging interval was therefore adequate to lengths of the piezometer probes can be changed
capture the essential features of the sediment depending on the experimental design objectives.
response during this phase of the experiment. Only one pore pressure measurement at a presel-
Before pressurization, the sediment and the salt ected depth below the sediment-water interface
water above it were cooled to 277 k 1 K. This (mud line) is possible with each piezometer probe.
temperature was maintained with excellent stabil- A variable reluctance differential pressure
ity throughout the simulation experiment. Bench transducer measures the excess porewater press-
testing of the LVDT revealed that its voltage ure (differential above hydrostatic) directly. A 5
output at a given displacement depended on kHz sine wave is supplied to the differential trans-
ambient temperature. It was therefore essential to ducer by a carrier oscillator in the signal condi-
calibrate this device at a temperature equal to tioner unit, producing an alternating current
that used in the experiment. Such a calibration output from the Wien bridge-type transducer cir-
was carried out following the simulation experi- cuitry whose amplitude is proportional to the
ment using a controlled temperature environment transducer imbalance. The alternating current
chamber. The calibration factor obtained in this signal is amplified, demodulated and filtered by
way, 0.333 V/mm, was used to convert LVDT the signal conditioning unit, producing a k5 V
voltage output to sediment surface displacement. direct current output level corresponding to the
full-scale range of the transducer (f68.9 kPa).
Pore pressure (piezometer) probe Solid state signal conditioning electronics are
The piezometer probe consists of an 8 mm dia. enclosed (at atmospheric pressure) in a stainless
titanium tube attached .to a tip having a cone steel capsule located directly above the pressure
angle of approximately 5.3”. Details regarding the sensor capsule. Data were recorded with both
choice of probe tip design can be found in analog strip charts and a data acquisition system
Bennett & Faris (1979). A porous stone, which with a hard copy printer for the duration of the
allows porewater pressure to be transmitted to experiment.
the pressure sensor, is fastened between the tita-
nium tube and the probe tip (Fig. 2). Pore press- Testing and calibration of pressure transducers
ure is transmitted through the porous stone to an The pressure transducers were tested at high
internal tube attached to the pressure sensor. The hydrostatic pressure (68.9 MPa) over a period of
differential pressure sensor is pressure balanced 2.7 Ms (31 days) before the ISHTE simulation
by a similar internal tube that runs from the experiment to determine sensor characteristics
pressure sensor to the top of the porous stone and long-term stability (Bennett, Burns &
retainer. The pressure sensor is enclosed in a Lambert, 1982). The pressure sensors exhibit a
CONSOLIDATION UNDER STRESS INCREASE 15

Displacement gauge
Far field
piezomeler probe

piezometer probe

-Steel tank

Porous stone

cm
fT3.* 35.9cm y,ITcm
j
Fig. 3. Positions of piezometer probes in the sediment tank

zero shift during pressurization but display excel- imately 1.2 x lo3 s for the near field probe and
lent long-term stability under high pressure. A 3.7 x lo3 for the far field probe. The induced
correction can be made for the zero shift and excess pore pressure at the far field probe has a
applied to the pore pressure measurements time delay of approximately 36 s, whereas the
observed during pressurization. The two pi- near field probe has a nearly instantaneous decay
ezometer probes were calibrated at laboratory of pressure following insertion. Furthermore, the
ambient conditions before use in the simulation maximum pore pressures generated differ signifi-
experiment. Immediately following the experi- cantly: 6.6 kPa and 12.9 kPa for the near and far
ment, the pressure sensors were again checked field probes respectively. These discrepancies
and calibrated (McTigue, Lipkin & Bennett, suggest that the sediment near the heater may
1985). have been altered or disturbed by insertion of the
heater and various other probes before the simu-
Piezometer insertion data lation experiment. The effect of introducing a
Pore pressures were monitored with two short drainage path near the heater is treated
piezometer probes placed at different positions in approximately in the analysis in part I, and is
the sediment sample. The near field piezometer discussed extensively by McTigue et nl. (1985).
monitored pore pressures 0.015 m from the The radial variation in porosity described in
heater (skin to skin) and 0.17 m below the mud the foregoing also may have contributed to the
line. The far field piezometer measurements were difference in response at the far field probe. Post-
monitored 0.26 m below the mud line and 0.34 m test analyses for water content show a porosity of
from the heater (skin to skin), as depicted in Fig. 0.74 in the vicinity of the near field probe and
3. Each probe was inserted separately. 0.73 in the area of the far field probe. Extensive
During probe insertion, soil deformation correlations of undrained shear strength and
occurs and excess porewater pressures are gener- water content (Williams, 1982) indicate that this
ated. The maximum pressure occurs at the small decrease in porosity can increase the
probeesoil interface and decays at a characteristic undrained shear strength by as much as 50%.
time which depends on the probe size and sedi-
ment properties (Randolph & Wroth, 1979). Nor-
malizing the excess pressure using the initial and COMPARISON OF CALCULATED AND
final values obtained during decay and plotting MEASURED RESPONSES
the data as a function of time reveals significant A computer code was written to predict the
differences in the decay curves (Fig. 4). Applying surface displacement and the internal pore press-
the logarithmic fitting method from consolidation ure based on the analytical solutions given in part
theory (Lambe & Whitman, 1969), t,,, is approx- I. The cumulative response to a succession of
16 LIPKIN, BENNETT AND McTIGUE

I I I 1 4
0.1 1.0 10.0 100 1000
Term? ml”

Fig. 4. Normal&d dissipation of pore pressure induced by piezometer insertion

external pressure increases is calculated by super- K, can be written in terms of the fundamental
position. Early time approximations are used for parameters
each step until the elapsed time, normalized by
the relaxation time, exceeds 0.02. Because the K 1 - K,JK,”
1+&.3--K, 1 - K/K,’
relaxation time is approximately lo6 s and the
typical ramp time for each pressure increment is K, = KS’
K,’ 1 - K,JK,”
approximately lo3 s, the early time approx- ’ + ” z f 1 _ KJK s’
imation is employed well past the ‘corner’ at the
end of each external load step. The jumps in pore pressure difference scale with
A pressure-dependent fluid bulk modulus K, (1 - B)p, In this case, B is very close to unity,
was used in these calculations. Over the range of and approximately
pressure encountered in the ISHTE simulation
experiment, from 0 MPa to 55.2 MPa, K, for 1 _ B ~ 9. K 1 - WKs”
sea water at 4°C increases about 16%, from 2.14 K, 1 - KJK,”
GPa to 2.49 GPa (Riley & Skirow, 1975). Such
an increase affects the undrained volumetric Finally, the relaxation rate for both the displace-
behaviour of the sediment, i.e. the sediment ments and the pore pressure depends on the con-
stiffness increases during pressurization. The solidation coefficient c, which scales with kK/,u.
variation in K, is small between successive load The overall pattern of the computer simulations,
steps, however, and the effects of previous steps then, depends primarily on these three param-
decay exponentially with time. Thus, the solution eters.
for constant properties is used and K, is updated The first calculation of interest is to model the
to the current value to calculate the contributions experiment using the best available estimates of
of all previous load increments to the present material properties obtained from independent
deformation. This results in a very slight overesti- tests. These values are given in Table 1 along with
mate of K, when fitting the model to the data. the sources used to obtain them. The fluid
Further, it is assumed that the sediment always properties K, and p are well established and are
rests on the bottom of the tank, so that the mea- available in extensive tables (Riley & Skirow,
sured displacement represents twice the sym- 1975). The solid bulk moduli K,’ and K,” are
metric displacement. This assumption has been usually assumed to be equal Few measurements
examined in detail by McTigue et al. (1985). for clays are reported in the literature; that given
The displacement and pore pressure responses by Skempton (1961) is adopted. Representative
of the sediment are dominated by three factors, drained properties are given by Baladi & Akers
K, , B and c, defined in part I. The magnitude of (1981). Permeability and porosity measurements
the jumps in surface displacement scales with for the marine sediment used in the ISHTE simu-
p,, L/K,, where p. is the external pressure and L lation have been reported by Silva et al. (1984).
is half the sample length. The undrained modulus Calculations based on the set of parameters
CONSOLIDATION UNDER STRESS INCREASE 17

Table 1. Independent estimates of material properties

Parameter Symbol Value Source

Viscosity 1.66 x 10m3 Pa s Riley & Skirow (1975)


Fluid bulk modulus 2.14 x lo9 + 6.41~ Pa Riley & Skirow (1975)
Porosity 0.75 Silva et al. (1984)
Shear modulus 6.0 x lo5 Pa Baladi & Akers (1981)
Drained bulk modulus 2.0 x 10’ Pa Baladi & Akers (1981)
Solid bulk modulus K,‘, K,” 5.0 x 10” Pa Skempton (1961)
Permeability k (lG3.0) x 10-16m2 Silva et al. (1981)

given in Table 1 are shown in Figs 5-8 along with parameter sensitivity is high, a good fit should be
the data measured in the simulation experiment. attainable using reasonable values for the
In Fig. 5 the maximum calculated displacement is material properties.
seen to be 1.4 mm, while the observed value is 1.6 Fitting the model to the experimental data was
mm. However, it is quite evident that the com- carried out by trial and error, varying several
puted undrained modulus K, is too small and the parameters about their nominal, independently
relaxation rates are too large. In Fig. 6 results for measured values. This approach cannot guar-
the surface displacement during depressurization antee that a good fit yields a unique set of param-
show undrained displacements that are only eters. However, experience gained through
slightly too large, and relaxation rates that are numerous calculations strongly suggests that the
again too fast, particularly at late time when the data can be matched for only one, well-
cumulative relaxation from previous steps con- constrained set of material parameters.
tributes significantly. The large displacement K,, p, & and G are regarded as known, and k,
associated with the last depressurization step is K, K,’ and K,” are varied to find a good fit.
ascribed to expansion due to exsolution of air. Although the displacement history is dominated
The model does not account for this phenome- by only two parameters, K, and c, it was found
non. In Figs 7(b) and 8(b) the calculated pressures that the four primary parameters regarded as
are quite different from the data; 1 - B is clearly unknowns all act in concert in affecting the calcu-
far too large. lated displacements. The same is found for the
The model fails to represent the data using pore pressure trends, which again depend prin-
independent estimates of the material properties. cipally on 1 - B and c. Thus, each data set is
This suggests the possibility of seeking a set of used to determine values for all four primary vari-
parameters that yields a good fit, thus using the ables.
data to determine material properties. If the The best fits obtained for surface displacements

?
z V
X
E
i -a-

E
x
Z-,2- /
: /

-16-

I I

0 IO 20 30 40
T&me: s X 1 O4

Fig. 5. Comparison of sediment surface displacement data and


model predictions during the pressurization phase of the experi-
ment (magnitudes of the model parameters used are given in
Table 1)
LIPKIN, BENNETT AND McTIGUE

Time. s X 10“

Fig. 6. Comparison of sediment surface displacement data and


model predictions during the depressurization phase of the
experiment (magnitudes of the model parameters used are given
in Table 1)

are shown in Figs 9 and 10 for the pressurization solid modulus K,” is commonly assumed to be
and depressurization phases respectively. Simi- equal to K,‘, but the sensitivity to this parameter
larly, the best fit for the far field pressure data is is very high, and it was found that good fits could
shown in Fig. 11. The values of the material be obtained only with distinctly different values.
properties required for these three fits are given in This observation is usually ascribed to uncon-
Table 2. It is evident that the parameter values nected porosity in the soil (e.g. Biot & Willis,
obtained for these three cases are generally quite 1957). It should be noted that the required K,”
consistent among themselves, and they are all values are close to, but greater than, the fluid
within an order of magnitude of the independent modulus K,.
parameter estimates summarized in Table 1. The It is particularly interesting to note the appar-
drained bulk modulus K, solid modulus K,’ and ent stiffening exhibited by both the data and the
permeability k are all smaller than the indepen- calculations for surface displacement during
dent measurements by factors of 2-3. The second pressurization. The last eight external pressure

0
Far field

-10

m
5
m- 20

ii
a,
?z -30
0
Y
2
(0 -40
a,
a

- 50

- 71 I L I

0 10 20 30 20 30
Ttme: s x 1 O4 Time’ s X 1 O4
(a) (b)

Fig. 7. (a) Sediment pore pressure changes measured by the far field piezometer during the pressurization phase of the
experiment and (b) a comparison of far geld piezometer data and model predictions during the pressurisation phase of
the experiment (magnitudes of the model parameters used are given in Table 1)
CONSOLIDATION UNDER STRESS INCREASE 19

-60

I I . 1
0 10 20 30 10 20 30
Time: s X 1 O4 Time: s X 1 O4

(a) (b)

Fig. 8. (a) Sediment pore pressure changes measured by the near field piezometer during the pressurization phase of the
experiment and (b) a comparison of near field piezometer data and model predictions during the pressurization phase of
the experiment (magnitudes of the model parameters used are given in Table 1)

increments are all of the same magnitude (6.9 pressures were measured at single points, and
MPa), yet the undrained displacements associated thus may reflect material inhomogeneities and
with each become successively smaller (Fig. 5). other local effects. In contrast, the displacements
The model results show the same trend due to the that were measured result from strains integrated
pressure-dependent increase in K,, but the degree over the entire body, and thus they tend to be
of apparent stiffening seen in the data is under- very smooth.
estimated. Despite the apparent success in modelling
Model calculations for the far field pore press- surface displacement and far field pore pressure
ure capture the essential.features of the measured histories, a serious problem is revealed by calcu-
response, although the computed undrained lations for the near field pressure. Fig. 12 shows
jumps are too small for the earliest and latest model results for the near field using the param-
external increments. The pressure data are prob- eters obtained by fitting the far field data. There
ably ‘noisier’ than the displacement data. Pore is clearly a large discrepancy between the model

I‘- -_
Data 1
I
I

-16-

I I I

0 10 20 30 0 1 2 3 4 5 6

Time: s X 1 O4 Time. s X 1 O4

Fig. 9. Comparison of sediment surface displacement Fig. 10. Comparison of sediment surface displacement
data and beat-fit model predictions during the preasuriza- data and best-fit model predictions during the depres-
tion phase of the experiment (magnitudes of the model surization phase of the experiment (magnitudes of the
parameters are given in Table 2) model parameters are given in Table 2)
20 LIPKIN, BENNETT AND McTIGUE

-71 I I I I I 1
0 10 20 30 0 10 20 30
Tfme: s X 1 O4 Time: s X 1 O4

Fig. 11. Comparison of far field piezometer data and Fig. 12. Comparison of near field piezometer data and
best-fit model predictions during the pressurization phase model predictions based on the parameters determined by
of the experiment (magnitudes of the model parameters fitting data for the far field probe (Fig. ll), pressuriza-
are given in Table 2) tion phase of the experiment (magnitudes of the model
parameters are given in Table 2)

and the data. The model predicts a maximum pressure difference larger. The opposite is
cumulative pressure difference of about -8 kPa, observed.
while the test reached only -3 kPa. Further- A second possible influence on the far field
more, the measured cumulative pressure differ- measurements may have entered through the
ence in the far field (- 6 kPa) exceeds that in the radial variation in porosity discussed previously.
near field (- 3 kPa). This observation cannot be However, while a small change in porosity can
represented by the model. Since the excess pore have a significant effect on the shear strength, its
pressure relaxes by the diffusion-like process of effect on the elastic properties and permeability is
fluid flow, the characteristic time for the relax- expected to be small. Indeed, on the basis of
ation scales with 12/c, where 1is the drainage path typical variations in drained bulk modulus with
length. In the experimental configuration (Fig. 3), porosity (Hamilton, 1971), the modulus in the
the minimum drainage path from the far field denser far field region may have been as much as
probe is 0.14 m (to the side boundary), while that 15% greater than that in the near field. The per-
for the near field probe is 0.17 m (to the top meability may have been decreased as much as
boundary). Thus, the relaxation time for the near 25% relative to the near field (Silva & Calnan,
field is expected to be longer, and the cumulative 1981). These effects are partially offsetting and

Table 2. Material properties determined bv fitting models to data*


I _

T Cylinder model

Displacement Far field kpressurizati Displacement Far field Near field


pressure displacemeni pressure pressure

K(x 106Pa) 7.0 7.0 8.0 7.0 6.0 6.0


K,’ ( x 10’0 Pa) 1.7 1.7 1.2 1.5 1.7 1.7
K,” ( x 109 Pa) 3.0 2.7 4.0 3.2 2.8 2.7
k Cx IO-"II?) 8.3 8.3 10.0 8.3 8.3 8.3
6.28 7.60 4.06 5.47 7.07 7.60
9.13 12.23 5.09 7.52 10.90 12.23
7.03 5.09 13.00 8.12 4.96 4.36
3.55 1.61 9.06 4.65 1.97 1.38
2.60 2.60 3.52 2.60 2.27 2.27
0,458 0.458 0,463 0,458 0.452 0.452
0.49995 0.49996 0.49993 0.49994 0.49996 0.49996
‘u{: 1 i5.2 MPa 0.49997 0.49998 0.49994 0.49996 0.49997 0.49998

* p. K,, 4, and G are fixed to the values given in Table 1


CONSOLIDATION UNDER STRESS INCREASE 21

result in a maximum decrease in the consoli-


dation coefficient c of about 15% in the far field
and, consequently, a similar decrease in the relax-
ation time. Calculations of the type shown in Fig.
11, when repeated for the near field pore pressure,
indicate a relaxation time of the order of l/20
that observed in the far field. This can be ascribed
only to an unexpectedly small value of the length
scale 1.
These considerations strongly suggest the
possibility that a shorter drainage path existed in
the neighbourhood of the near field probe. In
part I, the model for an annular region was devel-
oped to address this problem. Results using such
a modified model are summarized in the follow-
ing section. -8I
0 10 20 30
Time, s X 10“
APPLICATION OF ANNULUS MODEL Fig. 14. Comparison of far field piezometer data and
Smaller cumulative pressure differences at the annulus model predictions during the pressurization phase
near field probe, ostensibly further from a drained of the experiment (magnitudes of the model parameters
boundary than the far field probe, raise the possi- are given in Table 2)
bility that free fluid may have penetrated the sedi-
ment near the centre of the tank. This seems consequently, can be expected to lead to rapid
likely since surface cracks were observed extend- relaxation of pore pressure changes in this area.
ing from the heater probe as it was inserted into Figures 13-15 show the best fits obtained for
the sediment. A solution for the response of an the annulus model. The solutions for step changes
annular porous body to external pressure changes in external pressure were used for these calcu-
was developed in part I to simulate the penetra- lations, and the early time approximations were
tion of free fluid near the centre of the sediment not employed. A reasonable representation of the
tank. Because the configuration of the cracks or displacement data can again be obtained cap-
gaps that may have been present in the sediment turing both the undrained step responses and the
is unknown, the inner radius is arbitrarily taken relaxation rates (Fig. 13). The far field pressure
to be equal to the radius of the heater probe. This data are also well represented (Fig. 14). The
results in a minimum drainage path to the near material properties derived from both sets of cal-
field pore pressure probe of 0.02 m (Fig. 3) and, culations differ little from those obtained from the

x -8-
E
-i
c
E
g-12-
s
::
0

-16-

-3.51 I I
0.0 IO.0 20.0 30.0
Time sXl@ Time. s X 1 O4

Fig. 13. Comparison of sediment surface displacement Fig. 15. Comparison of near field piezometer data and
data and annulus model best-fit predictions during the annulus model predictions during the pressurization phase
pressurization phase of the experiment (magnitudes of of the experiment (magnitudes of the model parameters
the model parameters are given in Table 2) are given in Table 2)
22 LIPKIN, BENNETT AND McTIGUE

solid cylinder model (Table 2). This is to be change in K, results in only a 44% increase in
expected, because the surface displacement inte- K,. It is possible that drag on the heater probe
grates over the entire sample and the far field may manifest itself more strongly as the surface
probe should sense little influence from boundary displacement increases, contributing to the appar-
conditions at small values of r/R,, where R, is ent stiffening.
the outer radius. A second factor that could give rise to a large
The principal difference in the annulus calcu- apparent undrained modulus K, is the assump-
lations is seen in the results for the near field pore tion that the sediment always rested on the
pressure. The model is able to represent the essen- bottom of the tank. The calculated surface dis-
tial features of the data (Fig. 15) including the placement represents the axial strain integrated
maximum pressure difference attained and the over the entire sample length. If, however, the
relaxation rates. The calculated pressure jumps sediment were able to deform without settling
are notably larger than those measured during over a time-scale that is comparable with or
the middle portion of the test. Most importantly, longer than the time of each pressure increment
however, it should be emphasized that the (about 1800 s), the measured displacements could
annulus model predicts smaller cumulative press- be as little as one-half of the value calculated.
ure differences at the near field probe, as Simplifying and conservative assumptions were
observed. Material constants determined from used by McTigue et al. (1985) to estimate the
this fit are shown in Table 2 along with those likelihood of such an occurrence. These results
from the other fits. suggest that a small gap between the bottom of
the sediment sample and the sample tank would
SUMMARIZING REMARKS close on a time-scale that is much less than that
Calculations based on independent measure- of the pressure build-up and therefore would not
ments or estimates of material properties exhibit contribute to the displacement of the top surface
marked departures from the experimental data of the sample.
(Figs 5-S). To some extent, such discrepancies are A third factor that results in a high apparent
related to parameter sensitivity in the poroelas- undrained bulk modulus is deformation of the
ticity model. There are several other factors, steel tank. The surface displacements were mea-
however, that may have contributed to the mea- sured relative to the tank, which was assumed to
sured response. These are discussed briefly in the be rigid. However, the tank itself was also fully
following paragraphs. immersed in the external fluid and deformed with
The sediment sample is idealized as a finite cir- each pressure increment. For a bulk modulus of
cular cylinder that deforms freely in response to K,,= 180GPa, the tank would shorten by about
external pressure changes. There are two possible 0.01 mm with each pressurization step. Thus, the
constraints, however, that may affect the actual measured undrained displacements are about 3%
experimental results. The first of these is due to low, making the material appear to be slightly
the heater probe, which was fixed to a rigid struc- stiffer than it actually is. This small correction is
ture and penetrated 0.33 m into the sample. If a neglected.
no-slip condition or any sort of frictional resis- Values for the modulus K, of clay particles
tance prevailed along the 12.7 mm radius probe, used in the initial model calculations are another
the surface displacement, which was measured at area of uncertainty. Slates are comprised of clays
a radius of 38 mm, may have been restricted. The and have almost no porosity, so that moduli
material would then appear to be stiffer than it obtained from measurements on slates should
actually is. It should be noted that the undrained provide a reasonable estimate for individual clay
bulk modulus determined by matching the model particles. Clark (1966) reports K, values for slate
and data is of the order of 6 GPa, while conven- in the range 20-50 GPa. Another estimate for an
tional measurements for clays usually yield values aluminium silicate clay can be obtained by adopt-
about half as large. ing the modulus for metallic aluminium, 69 GPa.
In addition, it was noted that the model under- These values bracket that used by Skempton
predicts the apparent stiffening observed in the (1961), 50 GPa, which was used in the prelimi-
undrained response. Assuming that no relaxation nary calculations.
takes place during each pressure increment, the The sensitivity of the calculations to changes in
undrained modulus K, inferred from the surface the input parameters has been examined by
displacement measurements increases from about varying each parameter about its nominal best-fit
5 GPa to 13 GPa during pressurization. The value while holding the other parameters con-
model incorporates a pressure-dependent fluid stant. Because &,, K, and p are well known, and
modulus K, that gives rise to an increase in K,. since the value of G has little influence on the
However, for the calculations shown in Fig. 5, the deformation, the bulk moduli K, K,'and K,", and
CONSOLIDATION UNDER STRESS INCREASE 23

3E
i; 1 o-6 -
E
a,
;”
5
8
s
fj 10-7
=
::
6
0
I
I
10~8 j I 1 8
0’1 1.0 10.0 100.0
Drained bulk modulus K. MPa

(a)

7- I

6-

Baladi -

I I
0’001 0 010 I 0.100
1 I 1
1.000 1
10~000

Shear modulus G, MPa

ib)

Fie. 16. Predicted dependence of the consolidation coefficient on (a) the drained
buik modulus and (h) the shear modulus

the permeability k, are regarded as unknowns. 18) varies roughly linearly with K,’ near the
The permeability affects only the consolidation nominal fit, and is very sensitive to K,“. These
coefficient c, which varies linearly with k. Thus, results are summarized qualitatively in Table 3.
the influence of variations in K, K,’ and K,” on In conclusion the analysis of the isothermal
the observable parameters c, 1 - B and K, are of mechanical response of the sediment in the
particular interest. Figs 1618 show the results of ISHTE simulation experiment shows that the
these sensitivity calculations. The consolidation dominant deformation processes are well under-
coefficient c varies only with changes in K, and c stood. Fitting calculated results using a poroelas-
and K are approximately proportional near the tic model to surface displacement and pore
nominal fit. The parameter c (Fig. 16) is quite pressure data determines a number of material
insensitive to K,, and thus it does not vary with properties. These results are of particular interest
pressure. The undrained pore pressure difference because they are derived from a very large sedi-
scales with 1 - B (Fig. 17), which varies linearly ment sample under simulated deep ocean condi-
with K and varies very strongiy with K,” as K,” tions. Some of the inferred properties can be
approaches K,. The undrained modulus K, (Fig. checked against independent laboratory measure-
24 LIPKIN, BENNETT AND McTIGUE

Sohd bulkmodulus K,":GPa Indeed, the mechanical response data from the
1 10 100 1000 simulation experiment indicate that the
10-z
undrained Poisson’s ratio of the sediment is v, z
0.49995. None the less, all the measured effects
during pressurization and depressurization are
due to the small difference in the compressibilities
of the water and clay particles. The model shows
10-z-
that, because the water is more compressible than
co
the solid, a change in external pressure results in
I a very small difference between the pressure in the
sediment pores and that at the sediment bound-
aries. The sediment undergoes an undrained iso-
10-b-
tropic deformation, which relaxes as external
water flows into the body. The material thus
expands slowly until only the strain due to com-
pression of the solid remains. This phenomen-
I I ology is clearly seen in the data for surface
10-s ' ,
0.1 1 10 100 displacement and pore pressure.
DraInedbulkmodulus K: MPa Calculations based on independently measured
Fig. 17. Predicted dependence of 1 - B on the drained or estimated material properties show order-of-
bulk modulus and the solid bulk modulus KS” magnitude agreement with the data but fail to
represent the detailed behaviour accurately. The
ments, while others are uniquely determined in parameter sensitivity is found to be quite high,
this analysis. however, and very good fits are obtained by
The sediment is modelled as a porous elastic varying the magnitude of the material properties
material with compressible constituents. Such an over reasonable ranges. The best fits are found for
analysis distinguishes this problem from many values of permeability, drained bulk modulus and
others in soil mechanics where the water and soil solid bulk modulus that are all within a factor of
particles are assumed to be incompressible. 2-3 of independent measurements.

Table 3. Sensitivity of derived parameters to changes in material properties

Parameter Sensitivity of parameter to the following

K KS’ K," G k
Moderate Zero Zero Low Moderate
If, Moderate Zero High Zero Zero
K.. Zero Moderate High Zero Zero

SolIdbulk modulus K,‘, K,“: GPa


Fig. 18. Predicted dependence of the undrained hulk modulus on the solid bulk moduli
CONSOLIDATION UNDER STRESS INCREASE 25

ACKNOWLEDGEMENTS ISHTE simulation experiment. Report SAND83-


The ISHTE simulation experiment was a col- 1847, Sandia National Laboratories, Albuquerque.
laborative effort of numerous individuals under (Available from National Technical Information
the direction of C. M. Percival (Sandia National Service, US Department of Commerce, Springfield,
Virginia.)
Laboratories). The successful execution of the
McTigue, D. F., Lipkin, J. & Bennett, R. H. (1986).
experiment was in large measure due to the tech-
Consolidation under an isotropic total stress
nical expertise of L. 0. Olson and the staff of the increase: part I, model analysis for compressible
University of Washington Applied Physics constituents. Gtotechnique 36, No. 1, l-9.
Laboratory. Sample preparation and geotechnical Miller, J. B., Miller, V. W. & Olson, L. 0. (1984).
analyses were carried out by the University of ISHTE simulation APL-UW engineering report. In
Rhode Island Marine Geotechnical Laboratory, 1982 Subseabed Disposal Program annual report:
co-ordinated by A. J. Silva. Technical assistance thermal response studies October 1981 through Sep-
for the piezometer instrumentation was provided tember 1982, pp. 81-168 (ed. C. M. Percival). Report
SAND82-2717, Sandia National Laboratories, Albu-
by J. T. Burns (Naval Ocean Research and Devel-
querque. (Available from National Technical Infor-
opment Activity). Technical assistance for the
mation Service, US Department of Commerce,
sediment surface displacement instrumentation Springfield, Virginia.)
was provided by E. Boespflug (Sandia National Percival, C. M. (1982). Laboratory simulation of deep
Laboratories). An anonymous referee pointed out ocean in situ heat transfer experiment. In Oceans ‘82
the potential influence of sample inhomogeneity Conference Record, pp. 679-684. Washington:
due to lower water content near the drainage. Marine Technology Society and IEEE Council on
This work was supported by the US Department Oceanic Engineering.
of Energy under contract DE-AC04-76DPOO789. Percival, C. M. (1983). The Subseabed Disposal Program
in situ heat transfer experiment (ISHTE). Report
SAND80-1202, Sandia National Laboratories, Albu-
REFERENCES querque. (Available from National Technical Infor-
Baladi, G. Y. & Ackers, S. A. (1981). Constitutive mation Service, US Department of Commerce,
properties and material model development for Springfield, Virginia.)
marine sediments in support of the subseabed dis- Randolph, M. F. & Wroth, C. P. (1979). An analytical
posal program. In Subseabed Disposal Program solution for the consolidation around a driven pile.
annual report January to December 1980, vol. II, part Int. J. Numer. Analyt. Meth. Geomech. 3, 217-229.
1, pp. 621-781 (ed. K. R. Hinga). Report SANDIl- Riley, J. P. & Skirow, G. (1975). Chemical oceanography.
1095/H, Sandia National Laboratories, Albu- New York: Academic Press.
querque. (Available from National Technical Seabed Programs Division 6334 (1983). The Subseabed
Information Service, US Department of Commerce, Disposal- Program: 1983 status report Report
Springfield, Virginia.) SAND83-1387, Sandia National Laboratories, Albu-
Bennett, R. H., Burns, J. T. & Lambert, D. N. (1982). querque. (Available from National Technical Infor-
Fabrication and testing of deep ocean piezometer mation Service, US Department of Commerce,
system and components. In Subseabed Disposal Springfield, Virginia.)
Program annual report January to September 1981, Silva, A. J. & Calnan, D. I. (1981). Geotechnical aspects
vol. II, part 1, pp. 641-645. Report SAND82-0664/ of subseabed disposal of high level radioactive
II, Sandia National Laboratories, Albuquerque. wastes. In Subseabed Disposal Program annual report
(Available from National Technical Information January-December 1979; vol. II, appendix, pp. j35-
Service, US Department of Commerce, Springfield, 744. Report SAND80-2577/H. Sandia National
Virginia.) Laboratdries, Albuquerque. (Available from Nation-
Bennett, R. H. & Faris, J. R. (1979). Ambient and al Technical Information Service, US Department of
dynamic pore pressures in fine-grained submarine Commerce, Springfield, Virginia.)
sediments: Mississippi Delta. Appl. Ocn Res. 1, 115- Silva, A. J., Jordan, S. A. & Criscenzo S. J. (1984). URI
123. technical report of the simulation experiment for in
Biot, M. A. & Willis, D. G. (1957). The elastic toe% situ heat transfer experiment project. In 1982 Sub-
cients of the theory of consolidation. J. Appl. Mech. seabed Disposal Program annual report: thermal
24594-601. response studies October 1981 through September
Clark, S. P. (1966). Handbook of physical constants. 1982, pp. 253-340 (ed. C. M. Percival). Report
Memoir 97, Geological Society of America, New SAND82-2717, Sandia National Laboratories, Albu-
York. querque. (Available from National Technical Infor-
Hamilton, E. L. (1971). Elastic properties of marine mation Service, US Department of Commerce,
sediments. J. Geophys. Res. 76, 579-604. Springfield, Virginia.)
Hollister, C. D., Anderson, D. R. & Heath, G. R. (1981). Skempton, A. W. (1961). Effective stress in soils, con-
Subseabed disposal of nuclear wastes. Science 213, crete, and rocks. In Pore pressure and suction in soils
1321-1326. pp. 4-16. London: Butterworths.
Lambe, T. W. & Whitman, R. V. (1969). Soil mechanics. Williams, N. D. (1982). The effects of elevated tem-
New York: Wiley. perature on the engineering-. properties of seapoor
McTigue, D. F., Lipkin, J. & Bennett, R. H. (1985). sediments. PhD thesis, University of California,
Isothermal mechanical response of sediments in the Berkeley.

You might also like