410
ALONSO, GENS AND JOSA
(5) is to assume (Fig. 2(a))
s +
Pat
A~(p”)l,~ = N(0) - N(s) = K, In -
(6)
P.1
In other words,
p’
is the net mean stress at which
one may reach the saturated virgin state, starting
at a partially saturated condition, through a
wetting path which involves only (elastic) swell-
ing.
If Equation (6) is introduced into Equation (5),
the following relationship is obtained
@) =
(&_qo)-Kl/wKl
This Equation defines the set of yield
p.
values
for each associated suction (it can be considered
therefore as a family of yield curves in a (p, s)
space). In order to isolate a single yield curve it is
necessary to specify the preconsolidation net
mean stress for saturated conditions
(po*)
which
may be viewed as the hardening parameter in
Equation (7). This Equation plays a central role
in the model developed and explains, as discussed
at length later, not only the apparent increase in
preconsolidation stress associated with increasing
suction,
but also the collapse phenomena
observed in wetting paths. For this reason, the
yield curves (7) will be named LC yield curves
(after loading-collapse). An LC yield curve
through points 1 and 3 has been sketched in Fig.
2(b). Note that when
po* =
pc
the LC yield curve
becomes a straight line
(p. = p’).
In this case,
changes in s do not result in plastic deformations.
Only the elastic component, as given by Equation
(4), is maintained.
Existing information on the volumetric behav-
iour of partially saturated soils, as revealed by
suction controlled tests, may be used to support
some of the model assumptions. The results of a
number of isotropic and oedometer tests of this
type were compiled by Lloret & Alonso (1985)
and reported in terms of state surfaces (the void
ratio was expressed as a function of net vertical
or net confining stress and suction). Surface fitting
analysis was performed in order to find the most
appropriate analytical expression for the analysed
data. The tests considered involved different soil
types (low plasticity clays, silty clay and clayey
sand) and suction values ranging from zero to 100
kPa in most of the cases.
Most of the reported state surfaces have been
obtained through stress paths which involve
reductions in suction from an initial value and a
subsequent increment in net mean (or vertical)
stress
p.
A path of this kind is shown in Fig. 3.
For this particular path, the final specific volume
for the (generic) stress state (p, s) may be written
:
St__--------_
~.~~~~~__~~~~~~
t
PC
P
P
Fig. 3. Wetting and loading stress path in (p, s) stress
plane
ufp,
s) =
v(p’, si) + Au(p”) ’ + l(s)
In s
(8)
si
where the p-loading was supposed to take place
along virgin states. This expression may be com-
pared with the following empirical equation
which was found by Lloret & Alonso (1985) to fit
well experimental results
v=a+bln$+clns+dln4lns
P
= a + c In s + (b + d In s) In !-
PC
(9)
where
a, b, c
and
d
are constants.
Identifying Equations (8) and (9) suggests that
V(P”, SJ + Av(p”) ’ = a + c In s
.W
(10)
which supports the logarithmic variation of v
with suction adopted in Equation (6).
In addition Equation (9) provides an expres-
sion describing the increase in soil stiffness with
suction:
a(s) =
b +
d In s
(11)
This expression may adequately predict an
increase in soil stiffness with suction if
d < 0.
It
predicts, however,
an unlimited increase in
stiffness for increasing s and this is a limitation
for wide ranges of suction changes.
An asymptotic maximum stiffness, which is
probably closer to real behaviour, is predicted by
the alternative equation
n(s) = d(O)[(l - r) exp (-Bs) +
r]
(12)
where
r
is a constant related to the maximum
stiffness of the soil (for an infinite suction),
r = l(s + co)/l(O),
and B is a parameter which