Civil Engineering Reference
In-Depth Information
Figure 18.9
Variation of water content and undrained strength with depth in the ground for
normally consolidated and overconsolidated soils.
have been modified by the ground conditions as indicated by the broken lines: the
normally consolidated soil has dried due to evaporation and vegetation while the
overconsolidated soil has wetted and swelled due to rainwater penetrating cracks.
Figure 18.9(b) shows the variations of undrained strength
s
u
with depth cor-
responding to the water contents in Fig. 18.9(a). The undrained strength in the
overconsolidated soil is about 170 kPa corresponding to the water content close to
the plastic limit except near the surface. The undrained strength in the normally con-
solidated soil increases linearly with depth and is about 170 kPa at a depth of about
80 m, with an increase near the surface.
18.10 Accounting for variability
If you make a number of separate measurements of the same parameter, for example
water content of a lorry-load of soil, you will obtain a range of results. Your results
will be different because of errors in measurement of small weights of wet and dry soil
and because of the variation of true water content throughout the lorry-load. There
will be variations in any soil parameter you measure due to experimental errors and
due to the natural variation of soil in the ground. In soils the parameter may be a state
dependent parameter in which case its true value will vary with state, but we have seen
how to normalize test data to take account of state.
There are essentially three main ways in which engineers select design values from a
set of test results all measuring the same material parameter. Figure 18.10 shows a typ-
ical distribution of results as the number of observations plotted against the observed
value and this is a common plot in statistical analysis. (For simplicity I have shown a
symmetric distribution but for soil test data it could well be skewed.)
There is a mean value for which there are approximately as many larger values
as there are smaller values. There is a worst credible value and if you have mea-
sured any values smaller than this you have discarded the results for one reason or
another. There is a value called the moderately conservative value which is somewhere
