Page 303 - Numerical Analysis and Modelling in Geomechanics

284 ANALYSIS AND DESIGN OF PILE GROUPS
based methods of analysis. Numerical techniques for pile group analysis may be
broadly classified into the following two categories:

(a) continuum-based approaches;
(b) load-transfer (or subgrade reaction) approaches.

The latter category, based on Winkler spring idealisation of the soil, employs
load-transfer functions to represent the relationship between the load at any point
along the pile and the associated soil deformation at that point. Such a semi-
empirical method is widely adopted for the analysis and design of single piles,
especially where non-linear soil behaviour has to be considered and/or soil
stratification is complicated (e.g. the “t-z” or “p-y” curve methods of analysis).
The computer programs PILGP1 (O’Neill et al., 1977),FLPIER(Hoit et al.,
1996) and GROUP (Reese et al., 2000) are included in this category. The main
limitations associated with this approach are as follows:

1 The modulus of subgrade reaction is not an intrinsic soil property but instead
gives the overall effect of the soil continuum as seen by the pile at a specific
depth, and hence its value will depend not only on the soil properties but
also on the pile properties and loading conditions. Thus, no direct tests can
be conducted to establish force-displacement relationships for that particular
pile and soil type, and hence these curves have to be derived from the data
obtained by conducting a field test on an instrumented pile. However, due to
the high costs, such a test is rarely justifiable for onshore applications and
hence standard load-transfer curves are usually adopted in practice. This
implies that a significant amount of engineering judgement is needed when
formulating these curves for site conditions which differ markedly from the
recorded field tests. Murchison and O’Neill (1984) have compared four
commonly adopted procedures for selecting p-y curves with data from field
tests, and their results show that errors in pile-head deflection predictions
could be as large as 75%. Huang et al. (2001) employed several sets of p-y
curves derived from DMT data for the analysis of laterally loaded piles, and
none of the p-y curves yielded reasonable predictions of the measured pile
deflections.
2 The load-deformation relationship along the pile is modelled using discrete
independent springs and no information is available from the analysis
regarding the deformation pattern around the pile. Disregarding continuity
through the soil makes it impossible to find a rational way to quantify the
interaction effects between piles in a group. Thus, in evaluating group
effects, recourse is made to an entirely empirical procedure in which the
single pile load-transfer curves are modified on the basis of small-scale and
full-scale experiments performed on pile groups in different types of soil.
Although Reese and Van Impe (2001) reported some successful analyses of
this kind for pile groups under lateral loading, the uncertainties on the