Evaluating reliability in geotechnical engineering - Risk and Reliability in Geotechnical Engineering - page 150

Table 3.6 Values of coefficient of variation (COV) for geotechnical properties and in situ tests
Property or in situ test result

Coefficient of variation— COV (%) Source

3%-7% Harr (1984), Kulhawy (1992)

Unit weight ( γ )

0%-10% This chapter, Lacasse and Nadim

(1997)

Buoyant unit weight ( γ b )

Effective stress friction angle
( φ ')

2%-13% Harr (1984), Kulhawy (1992)

Undrained shear strength (S u ) 13%-40% This chapter, Kulhawy (1992), Harr

(1984), Lacasse and Nadim
(1997)

Undrained shear strength (S u ) Clay - UU Triaxial (10%-30%) Phoon and Kulhawy (1999)

Clay - UC Triaxial (20%-55%) Phoon and Kulhawy (1999)

Undrained strength ratio (S u /
σ v ')

5%-15% This chapter, Lacasse and Nadim

(1997)

Compression index (C c ) 10%-37% This chapter, Kulhawy (1992), Harr

(1984)

Preconsolidation pressure (p p ) 10%-35% This chapter, Harr (1984), Lacasse

and Nadim (1997)

Coefficient of permeability of
saturated clay (k)

68%-90% This chapter, Harr (1984)

Coefficient of permeability of
partly saturated clay (k)

130%-240% Harr (1984), Benson et al. (1999)

Coefficient of consolidation (c v ) 33%-68% This chapter.

Standard Penetration Test blow
count (N)

15%-45% Harr (1984), Kulhawy (1992)

Electric Cone Penetration Test
(q c )

5%-15% Kulhawy (1992)

Mechanical Cone Penetration
Test (q c )

15%-37% Harr (1984), Kulhawy (1992)

Dilatometer Test tip resistance
(q DMT )

5%-15% Kulhawy (1992)

Vane shear test undrained
strength (S v )

10%-20% Kulhawy (1992)

Plastic Limit 6%-30% Phoon and Kulhawy (1999)
Liquid Limit 6%-30% Phoon and Kulhawy (1999)
Source: Duncan, J.M., 2000, Factors of safety and reliability in geotechnical engineering, Journal of Geotechnical Engineering ,
Vol. 126, No. 4, p. 307-316. Used with Permission from ASCE.

However, the accuracy of the three-sigma rule depends on the accuracy with which HCV
and LCV can be estimated. Estimating these very extreme values has been found to be more
difficult than it might at first appear (Christian and Baecher, 1999). Because the ±3σ range
covers 99.73% of all values, only 0.27% of all values should lie outside the ±3σ range. This
corresponds to only one instance in 370, a very rare occurrence.
Studies have shown that there is a strong tendency for engineers to estimate ranges of val-
ues between HCV and LCV that are too small. One such study, described by Folayan et al.
(1970), involved asking a number of geotechnical engineers to estimate the possible range
of values of C c /(1 + e) for San Francisco Bay mud, with which they all had experience. The
results of this exercise are summarized in Table 3.7. On average, these experienced engineers
were able to estimate the average value of C c /(1 + e) within about 15%, but they underesti-
mated the COV by about 67% as compared with the results of 45 laboratory tests.