Verification of geotechnical reliability using load tests and integrity tests - Risk and Reliability in Geotechnical Engineering - page 537

statistics of a number of commonly used methods for design and construction of driven piles
(Zhang et al. 2001). In this table, failure of piles is defined by the Davisson criterion. For
each design method in the table, the driven pile cases are put together regardless of ground
conditions or types of pile response (i.e., end-bearing or frictional). Ideally, the cases should
be organized into several subsets according to their ground conditions and types of pile
response. The statistics in the table are intended to be used only to illustrate the proposed
methodology. The ASD approach results in designs with levels of safety that are rather
uneven from one method to another (i.e., β = 1.74-3.11). If an FOS of 2.0 is used for all these
methods and each design analysis is verified by two positive proof tests conducted to twice
the design load (i.e., x T > 1.0), the statistics of these methods can be updated as shown in
Table 14.6 . The updated bias factors are greater, but the updated COV R values are smaller
than those in Table 14.5 . The updated β values of all these methods fall into a narrow range
(i.e., β = 2.22-2.89) with a mean of approximately 2.5.
Now consider the case when an FOS of 2.0 is applied to all these methods in Table 14.5
and the design is verified by two proof tests conducted to failure at an average load of
twice the design load (i.e., x T = 1.0). The corresponding updated statistics using the Bayesian
sampling theory ( Equations 14.12 through 14.14 ) are also shown in Table 14.6 . Both the
updated bias factors (λ R = 0.97-1.12) and the updated COV R values (COV R = 0.21-0.24)
fall into narrow bands. Accordingly, the updated β values also fall into a narrow band (i.e.,
β = 1.78-2.31) with a mean of approximately 2.0.
The results in Table 14.6 indicate that the safety level of a design verified by proof tests is
less influenced by the accuracy of the design method. This is logical in the context of Bayesian
statistical theory in which the information of the empirical distribution will play a smaller
role when more measured data at the site become available. This is consistent with foundation
engineering practice. In the past, reliable designs were achieved by subjecting design analyses
of varying accuracies to proof tests and other quality control measures (Hannigan et al. 1997;
O'Neill and Reese 1999). This also shows the effectiveness of the observational method (Peck
1969), with which uncertainties can be managed and acceptable safety levels can be main-
tained by acquiring additional information during construction. However, the importance
of the accuracy of a design method in sizing the pile should be emphasized. A more accurate
design method has a smaller COV R and utilizes a larger percentage of the actual pile capacity
(McVay et al. 2000). Hence, the required safety level can be achieved more economically.

14.6 relIabIlItY oF PIleS VerIFIeD bY IntegrItY teStS

14.6.1 Worked example
The impact of interface coring on the reliability of large-diameter bored piles with toe debris
is illustrated in this worked example. The same procedure can be applied to study the impact
of other integrity tests on the reliability of piles with various imperfections. If a pile contains
toe debris, the pile capacity will be adversely affected. A pile capacity reduction factor, R F ,
is used to measure the effect of toe debris, which is defined as the ratio of the capacity of the
pile with toe debris to the capacity of the pile without toe debris.
Strictly speaking, R F should be treated as a random variable. Owing to lack of sufficient
statistical data for R F , R F is treated as a deterministic quantity in this study, but its value will
depend on the debris thickness, and the length and diameter of the pile. Once R F is obtained,
the bias factor of the capacity of the pile with toe debris, λ RD , is given by

λ = R λ (14.27)

RD FR