Page 304 - Numerical Analysis and Modelling in Geomechanics

F.BASILE 285
general use of the approach in routine design remain (Rollins et al., 1998;
Rollins et al., 2000; Huang et al., 2001).
3 It is uncertain how the p-y curves are influenced by pile-head fixity. To date,
this issue has hardly been addressed, although Reese et al. (1975)
showed that the p-y relationships are affected by pile-head fixity. The
relevance of this aspect is obvious if the p-y curves from single pile tests are
to be used for pile group predictions where the pile-heads are restrained by a
cap.

In conclusion, the load-transfer approach may be regarded as a link between the
interpretation of full-scale pile tests and the design of similar piles rather than a
general design tool for pile group predictions.
Several hybrid approaches which combine a load-transfer analysis for single
pile response and a continuum model to estimate pile-soil-pile interaction have
been proposed (Chow, 1986a, 1987; Mandolini and Viggiani, 1997). However,
such analyses do not overcome the main limitation of the load-transfer approach
that is the questionable assessment of the empirical constants which define the
non-linear relationship on the basis of intrinsic soil properties.
The above shortcomings may be removed by means of soil continuum based
solutions which are generally based on the finite element method (FEM)
(Ottaviani, 1975) or the boundary element method (BEM) (Butterfield and
Banerjee, 1971). These solutions provide an efficient means of retaining the
essential aspects of pile interaction through the soil continuum and hence a more
realistic representation of the problem. Further, the mechanical characteristics to
be introduced into the model now have a clear physical meaning and they can be
measured directly. Finite element analyses are valuable for clarifying the
mechanism of load transfer from the pile to the surrounding soil but, especially
for pile groups, are not readily applicable to practical problems. The considerable
effort of data preparation and the high computational cost (particularly if non-
linear soil behaviour is to be considered) preclude the routine use of such
techniques in design. Some idea of the computational resources required may be
obtained from the non-linear FEM analysis of a laterally loaded 9-pile group by
Kimura and Adachi (1996) who reported a CPU time of 85 hours on a SPARC II
work-station.
By contrast, BEM provides a complete problem solution in terms of boundary
values only, specifically at the pile-soil interface. This leads to a drastic
reduction in unknowns to be solved for, thereby resulting in substantial savings
in computing time and data preparation effort. This feature is particularly
important for three-dimensional problems such as pile groups.
The following computer programs may be included in this category. DEFPIG
(Poulos, 1990), based on a simplified BEM analysis and the use of interaction
factors, models soil non-linearity in an approximate manner by means of an
elastic-plastic interface model. Two main shortcomings are associated with this
model: (1) the non-linear features of stress-strain behaviour are not captured