Page 309 - Numerical Analysis and Modelling in Geomechanics

290 ANALYSIS AND DESIGN OF PILE GROUPS
decreased (Feature No. 2). The latter aspect cannot be reproduced in the
interaction factor method and it can only be accounted for by a “complete”
approach.
3 Load-deformation coupling
Pile-soil interaction is a three-dimensional problem, and each of the load
components has deformation-coupling effects, i.e. there is an interaction
between the axial and lateral response of the piles. Modelling of this aspect
becomes important when a pile group is subjected to a combination of
vertical and horizontal loads. In this case, only a proper consideration of the
interaction between the axial and lateral response will lead to a realistic
estimate of the loads acting on the piles, which will be increased for the
piles in the leading rows and decreased for those in the trailing rows of the
group. However, in current design practice, such interaction effects are not
properly accounted for, and the axial and lateral responses of the piles are
treated separately.
4 Soil non-linearity
A fundamental limitation of the linear elastic methods is that they result in
a considerable overestimation of the load concentration at the outer piles of
the group, and this may lead to an overconservative design. Indeed, it has
long been recognised that consideration of soil non-linearity results in a
reduction of the stiffness of the piles, the reduction being greater for piles at
a greater load level, i.e. for the corner piles. Consequently, as the total
applied load increases, the share of the load carried by the corner piles
progressively decreases. This results in a redistribution of the loads in the
individual piles, leading to a more uniform distribution than that predicted
by linear models. Ideally, for an axially loaded pile group, all piles will carry
the same load as the total applied load approaches the ultimate load capacity
of the group.

Table 10.2 summarises the above-mentioned features and their effect on the
prediction of load at group corners. The table also shows the ability of the
computer programs discussed above to model such aspects of group behaviour.
It is worth noting that all the features mentioned above may be modelled using
the PGROUPN analysis, whereas the other programs can only model some of
these aspects, thereby neglecting important features of group behaviour. There is
thus a number of compelling arguments for adopting a design methodology
which deals with group effects on a more fundamental basis.

PGROUPN method of analysis
The PGROUPN analysis is based on a complete non-linear BEM formulation,
extending an idea first proposed by Butterfield and Banerjee (1971) and
incorporated into a number of computer programs, including PGROUP (Banerjee
and Driscoll, 1976), GAPFIX (Poulos and Hewitt, 1986) and that developed by