The case history of the deep excavation for the National Gallery extension
in London is presented in this paper. Comparisons with data from other site
s show that it is typical for that of similar projects in Central London. C
lass 1 predictions of the retaining wall behaviour (prior to construction)
using the Model London Clay constitutive relationship considerably overesti
mated wall and ground movements. Retaining wall and ground movements are al
so considerably over-predicted by analyses using a simple linear elastic/pe
rfectly plastic soil model, despite optimistic parameters being assumed for
the soils. Predictions made using the constitutive model BRICK are closer
to the measured deflected shape, but are also higher than measured values.
Parametric studies of the effect of various parameters suggest that a "best
estimate" of the wall movements are still well in excess of those measured
. It is concluded that these differences are due to three-dimensional effec
ts and deficiencies in the model. The "beam-spring" computer software for r
etaining walls FREW gives similar results to the analyses using the simple
model. Analyses of the same problem carried out by a different operator usi
ng another finite element code, but with the same constitutive model, yield
ed somewhat different results and highlight the need for careful interpreta
tion of finite element analyses. (C) 2001 Elsevier Science Ltd. All rights
reserved.