where (cu/ ov')nc has been assumed to be 0.23. This relationship assumes the removal of overburden pressure due to geological processes: for the line labelled 'minimum', 5m of overburden is removed and, for the one labelled 'maximum', 70m. There are seven results in London Clay (solid squares) and five results in Lambeth Clay (open squares) that are below the minimum — these may be considered physically impossible values for the in situ clay. There are also four results in London Clay that are above the maximum — these too may be physically unreasonable values (although they might be believable if more than 70m of overburden was removed).

We asked the same one hundred engineers as before to select a characteristic line (or lines) through this data based on a 'cautious estimate'. The thin solid lines on Figure 5.11 show the outcome.

The bandwidth of the interpretations in the London Clay increases from perhaps 50kPa at +10m OD to about 100kPa at -20m OD. Rather alarmingly, the ratio of the most optimistic to the most pessimistic assessment of cu is as much as three-fold. The seven results below the minimum line and the four above the maximum were (almost always) ignored.

In the Lambeth Clay, estimates of cu ranged from 50kPa to 300kPa — the lower value is clearly influenced by the five physically impossible values that fall below the minimum line.

5.4.2 Singapore marine clay

Figure 5.12 shows the results of about forty field vane tests performed at a site in Singapore, where bored piles and barrettes were installed for a residential development.21 Ground conditions typically comprise up to 35m of soft to firm marine CLAY, overlying silty SAND/sandy SILT, overlying SANDSTONE.

The marine clay has low to medium sensitivity (St = 2- 7). Laboratory tests suggest it is under-consolidated (over-consolidation ratio, OCR < 1) but this may be in error owing to sample disturbance - independent studies suggests that Singapore marine clay is lightly over-consolidated (OCR = 1.5- 2.5) and this is supported by the soil's liquid limit exceeding its natural water content.

The data shown on Figure 5.12 indicates that the marine clay's undrained strength cu increases with depth at a steady rate and lies slightly above the following relationship with vertical effective stress o'v: cu = 0.22<

(shown by the dashed line), which is commonly used to estimate the undrained strength of normally consolidated clays.

Figure 5.12. Results of field vane tests in Singapore marine clay, with engineers' interpretation of the clay's characteristic undrained strength

As for the Holborn case study, we asked more than one hundred engineers and engineering geologists to select a characteristic line (or lines) through this data, on the basis of Eurocode 7's definition of the characteristic value as a cautious estimate. The lines on Figure 5.12 show the outcome of this study.

A majority of engineers drew a straight line through the data points; others preferred a stepped relationship. The bandwidth of the interpretations is about 10kPa and remains fairly constant with depth.

Figure 5.13 shows the results of more than eighty undrained triaxial compression tests performed in the laboratory on U100 samples taken from eight different boreholes from the same site in Singapore. The measured shear strength is highly scattered, caused in large part by sample disturbance prior to laboratory testing.

Figure 5.13. Results of undrained triaxial compression tests in Singapore marine clay, with engineers' interpretation of the clay's characteristic undrained strength

Also shown on Figure 5.13 is the relationship between undrained strength and vertical effective stress given above for normally consolidated clays. The results are generally below the line, confirming the assessment based on the vane tests that sample disturbance occurred during sampling.

We asked the same one hundred engineers to select a characteristic line (or lines) through this data based on a 'cautious estimate'. Figure 5.13 shows the outcome.

The interpretations are far more variable than for the field vane results, which reflects the greater variability in the laboratory measurements. The engineers were particularly conservative in their estimates of cu deeper than 20m, where a value of 35kPa was popular.

It is interesting to note that the dashed line — which represents a typical cu/ o'v relationship for normally consolidated clay — has a better agreement with the triaxial than the vane test results. However, the laboratory data has considerable scatter, indicating that sample disturbance (both physical and stress-related) was significant. Disturbance is less likely for the in situ vane tests. If a correction factor22 of 0.8 is applied to the vane test results (based on the soil's plasticity index of 40-60%), better agreement would be obtained with the cu/ o'v relationship.

5.4.3 Thames Gravels at Gravesend

Figure 5.14 shows the results of about forty standard penetration tests (SPTs) performed at a site near Gravesend, in Kent, where about one thousand driven-cast-in-situ piles were to be installed.23 Two site investigations ('A' and 'B') were undertaken at the site and gave markedly different results.

In Investigation A, the SPT blow counts obtained in the Thames Gravels between 14m and 22m below ground level show very little variation with depth, as can been seen from the open white symbols on Figure 5.14. By contrast, in Investigation B, the blow counts obtained are highly variable and appear to increase slightly with depth — witness the closed black symbols.

As for the Holborn and Singapore studies, we asked more than forty civil and geotechnical engineers to select a characteristic line (or lines) through this data, based on a 'cautious estimate'. The engineers were given the data for Investigations A and B separately.

Many of the engineers questioned whether the data from Investigation A is credible, since it shows very little scatter in a stratum that is expected to be highly variable. We have therefore ignored Investigation A in this case study.

The interpretations obtained for Investigation B are shown by the lines on Figure 5.14.

Figure 5.14. Results of standard penetration tests in Thames Gravels, with engineers' interpretation of the characteristic value for Investigation B (black symbols). Interpretation for Investigation A (white symbols) is not shown

It is difficult to understand how, when asked for a cautious estimate of the characteristic value of blow count N, the engineers could produce such a wide spread of interpretations. The data presented is fairly typical of the quality and variability of data obtained in most site investigations. This exercise therefore poses serious questions about how the industry selects design parameters.

5.4.4 Conclusions from the case studies

Our conclusion from these studies is that engineers are not particularly good at selecting a cautious estimate of the characteristic value, particularly when the available data is scattered. Statistical treatment of large data sets (such as these) may help to guide engineers in this task (see Section 5.5).

0 0

Post a comment