Vous êtes sur la page 1sur 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254519379

Statistical Description of Characteristic Soil Properties

Article · January 2007


DOI: 10.4043/19117-MS

CITATIONS READS

7 884

4 authors, including:

Suzanne Lacasse Farrokh Nadim


Norwegian Geotechnical Institute Norwegian Geotechnical Institute
130 PUBLICATIONS   2,119 CITATIONS    201 PUBLICATIONS   3,218 CITATIONS   

SEE PROFILE SEE PROFILE

Amir Rahim
Norwegian Geotechnical Institute
5 PUBLICATIONS   24 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NGI avalanche warning View project

Investigation of the Earthquake Culture in Iran View project

All content following this page was uploaded by Farrokh Nadim on 01 September 2014.

The user has requested enhancement of the downloaded file.


OTC 19117

Statistical Description of Characteristic Soil Properties


Suzanne Lacasse and Farrokh Nadim, Intl. Centre for Geohazards, Norwegian Geotechnical Inst. (NGI);
Amir Rahim, NGI; and Tom R. Guttormsen, Norsk Hydro Production

Copyright 2007, Offshore Technology Conference


to point within a soil volume, and imperfect interpretation
This paper was prepared for presentation at the 2007 Offshore Technology Conference held in models, measurement errors and other sources.
Houston, Texas, U.S.A., 30 April–3 May 2007.
The selection of soil properties for use in geotechnical
This paper was selected for presentation by an OTC Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
assessment is often based on subjective judgment and
presented, have not been reviewed by the Offshore Technology Conference and are subject to accumulated experience. The uncertainties in the soil
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Offshore Technology Conference, its officers, or members. Papers presented at properties are only indirectly accounted for when the
OTC are subject to publication review by Sponsor Society Committees of the Offshore
Technology Conference. Electronic reproduction, distribution, or storage of any part of this
characteristic (design) value(s) are chosen.
paper for commercial purposes without the written consent of the Offshore Technology Statistics and probability are useful tools for the
Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not
more than 300 words; illustrations may not be copied. The abstract must contain conspicuous quantification of the mean (most probable, expected) value
acknowledgment of where and by whom the paper was presented. Write Librarian, OTC, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
and the possible range of values of a parameter. Statistical and
probabilistic methods can quantify the uncertainties and make
Abstract it possible to account for them in a rational and consistent
The paper discusses expressions such as "characteristic value", manner. They are however rarely used in practice to establish
"best estimate", "upper bound" and "lower bound" used to the design soil parameters. The reason for this is unclear, but
describe a design soil profile in practice through the reanalysis perhaps it has become a habit that no one questions, or the
of case studies. Characteristic (design) values of undrained restricted use of statistical methods may be a reflection that
shear strength were compared with the results of unbiased often there are not enough data available to actually
statistical analyses. It would seem that when one relies on implement statistical methods with confidence. DNV (2006)
laboratory test results to establish the characteristic (design) prepared a guidance note on the statistical representation of
strength, the experienced engineers tend to lie much lower soil data. The tools are explained in detail. The profession now
than the average, perhaps one standard deviation below the needs to make a recommendation of which values to use in
mean. On the other hand, when one relies on the results of in design.
situ piezocone tests, the characteristic strength can be much In this paper, examples of design soil parameters recom-
closer to the mean of the interpreted measurements in situ. The mended by NGI for offshore sites over the past two decades
authors suggest that the profession adopts a univocal definition were reevaluated using statistical methods, and comparisons of
for the term "Best Estimate" and for parameter variability. characteristic value with best estimates and variance are made.
This suggestion is proposed as a subject for debate in the
coming months so that agreement can be established for use in Background for study
practice. The authors also suggest that the geotechnical When describing the design soil profile, expressions such as
profession should use more extensively than before statistical "characteristic", "best estimate", "upper bound" and "lower
analysis when establishing design values, look beyond using bound" values are used in practice. There does not exist a
exclusively statistical analyses and start addressing the common understanding or a standard criterion for either the
variability and uncertainty in soil parameters explicitly by selection of "best estimate", "upper bound" and "lower bound"
implementing probability theory and reliability analyses. values or the uncertainty that is covered by these expressions.
The paper attempts to quantify some of these concepts through
Introduction the reanalysis of case studies in the North Sea and elsewhere,
A detailed geohazards assessment at an offshore site often where characteristic (design) values had been selected by
involves the assessment of the stability of submarine slopes experienced engineers. The characteristic values were
under static and dynamic (ocean wave and seismic) loading. compared with the results of unbiased statistical analyses. To
Even the simplest geotechnical calculation models for slope ensure neutrality for the statistical analysis, the cases were
stability and seismic response require mechanical soil analyzed by an engineer not having had any prior involvement
properties. These can never be established with complete in the original selection of the characteristic parameters.
certainty. Soils are naturally variable because of the way they Six analyses are presented. Three of the case studies were
are formed and the continuous processes of the environment for geohazards assessment while the other three were for
that alter them. The uncertainty in the mechanical properties of foundation design for offshore installations. Several parame-
offshore soils is due to both the natural variability from point ters are key to the such calculations. Important parameters
needed for slope stability evaluation, seismic response analysis
2 OTC 19117

and the evaluation of the load capacity of anchored


foundations include undrained shear strength, shear modulus Statistical description of soil properties
at small strains, unit weight of soil, pore pressure generation, A quantitative geotechnical variability investigation must rely
etc. Focus is placed on the undrained strength of the soil in the on sets (samples in statistical terms) of measured data which
paper. The characteristic (design) undrained shear strength in
the case studies was based on laboratory tests under different
stress conditions, the interpretation of in situ piezocone
penetration test or correlations for normalized undrained shear
strength as a function of overconsolidation ratio.

Uncertainties in soil properties


The selection of characteristic values of soil properties for use
in geotechnical design is often based on subjective judgment
and accumulated experience, and the uncertainties which are
involved with the soil properties are only to a limited extent
brought into the picture when the characteristic values are
chosen and reported to the client.
In the engineering literature - and geotechnical engineering
is no exception - the terms variability and uncertainty are often
employed interchangeably. While the two terms refer to
concepts which are significantly related, a clarification in
terminology would positively contribute to reducing the
confusion and ambiguities associated with the statistical
characterization of mechanical soil parameters for
geotechnical assessment.
Technically, variability can be defined as an observable
manifestation of one or more physical processes. In principle,
a variable property could be described, for instance, if a
sufficient number of measurements were available and if the Fig. 1 Descriptive and inferential analysis for the probabilistic
modeling of a random variable.
quality of the measurements themselves was sufficient to
ensure confident evaluation of the observations. Hence, the
are limited in size and quality. Hence, it is necessary to refer to
observation of variability implicitly provides a more or less
sample statistics. Sample statistics are imperfect estimators of
detailed assessment of the level of knowledge on a
the “real” population parameters. Hence, they are never
phenomenon of interest and of the capability to measure and
completely representative of the real distribution of the data,
model the phenomenon itself. Variability, which represents the
and are biased to some degree. “Sample statistics” refers to
natural randomness of the property, is also referred to as
any mathematical function for a sample. For most engineering
aleatory (or inherent) uncertainty.
purposes, sample statistics are more useful than the
Uncertainty can also be epistemic. Epistemic uncertainty
comprehensive frequency distribution (as given by a
refers to the lack of knowledge for a given soil property.
histogram, for instance). An infinite number of sample
Given the inherent nature of a physical process and the
statistics may be calculated from any given data set. For most
available level of knowledge and/or technology related to the
geotechnical engineering applications, however, it is usually
process itself, it is impossible in many situations to model the
sufficient to calculate the first two statistical moments of a
variability rigorously. In other situations, it may be practical to
sample, i.e. the mean and standard deviation. Higher moments
assume that a phenomenon is indeterminate (uncertain) when a
are unreliable when estimated from usual sample sizes
detailed description of variability is expected to be
available for offshore sites.
uneconomic or redundant. Figure 1 (Uzielli et al. 2006a)
shows the integrated descriptive and inferential analysis for
Parameter estimation for independent soil variables
probabilistic modeling of a random variable.
Let X be a soil parameter, for example the unit weight, for
The main sources of uncertainty in estimated mechanical
which n observations, x1... xn exist. The mean value of X is µ
soil properties within a nominally uniform layer are:
and the standard deviation is σ. Any dependency on depth or
• Limited geo-exploration
on any other conditioning quantity is disregarded in this
• Measurement errors example, i.e. X is a so-called independent variable.
• Limited parameter evaluation The mean value µ is estimated by the sample mean deno-
• Spatial variability of mechanical properties ted x and the standard deviation σ is estimated by the sample
These sources, except for the inherent spatial variability, are standard deviation, where n is the number of data points:
epistemic uncertainties. This type of uncertainty can be
reduced by, for example, by increasing the number of tests or
1 n
improving the measurement methods. x= ∑ xi
n i =1
OTC 19117 3

( )
1 n 2 for stability analysis of offshore installations. In sites A and F,
s= ∑ xi − x the characteristic undrained shear strength was based on
n − 1 i =1
triaxial and/or direct simple shear laboratory tests. At site, B,
D and E, CPTU profiles were used to determine undrained
The sample estimates x and s are central estimates of µ and σ.
The coefficient of variation is the ratio of the standard
deviation to the mean (CoV = σ/µ). ao
Y
Parameter estimation for dependent soil variables ε
Let Y be a soil variable, for example the shear strength, whose
variation with depth z can be reasonably well modeled as a li- a1
near function: Y is then a so-called dependent variable, where-
as z is independent. Assuming that n observations of pairs (zi, 1
yi) where i = 1, 2, 3,... n are available from a soil investigation,
the variation of Y with depth can be expressed as:

Y = ao + a1 z + ε

where the term a0 + a1z represents the linear mean variation


with depth, the term ε the natural variability of Y about the z
mean, and z the depth below the soil surface. The variable Y
vs. z is shown in Fig. 2. The coefficients a0 and a1 represent Fig. 2 Variability of a depth-dependent soil parameter.
the surface intercept and the depth gradient, respectively, of
the mean of Y. The variability term ε has zero mean and shear strength, while at site C, several series of different labo-
standard deviation σ. The standard deviation σ is assumed to ratory and in situ shear strength tests were used.
be constant, i.e. it is assumed to be independent of depth z. Although the geotechnical specialists who established the
The coefficients a0 and a1 in the linear expression can be characteristic (design) undrained shear strength profiles were
estimated from the n observed data pairs (z, y) as aware of the variability of the measurements at these three
sites and the uncertainty in the estimation of the undrained
n
(
∑ zi − z yi − y
i =1
)( ) shear strength, no attempt was made at the time to quantify the
variability or uncertainties.
a1 = n
(
∑ zi − z
i =1
)
2
su from laboratory measurements
When undrained shear strength is measured in the laboratory,
and the shear strength profile vs. depth consists of discrete
measurements, and the characteristic (design) undrained shear
ao = x − a1 z strength is often selected as a conservative estimate
representative of the available data. For the statistical study,
in which the mean and standard deviations were obtained from the
parameter estimation method for dependent variables as
1 n 1 n described above. Only the triaxial compression data were
y= ∑ yi and z= ∑ zi used, as they represent the most reliable test results.
n i =1 n i =1
su from interpreted from CPTU results
The standard deviation σ of the natural variability term is When the characteristic undrained shear strength profile is es-
estimated by the sample standard deviation: tablished on the basis of piezocone penetration (CPTU) tests,
the results can be interpreted from theoretical or empirical cor-
1 n relations. The empirical approach was used in this study. To
∑ ( yi − (ao + a1zi ))
2
s=
n − 2 i =1 interpret undrained shear strength from CPTU test results,
three main categories can be used:
When the standard deviation or variance of ε is not 1. su interpretation using “total” cone resistance
constant with depth, the derivation above is no longer valid. 2. su interpretation using “effective” cone resistance
DNV (2006) provides the equations for a0 and a1 for the 3. su interpretation using excess pore pressure
situation where the standard deviation of ε increases with
depth, and the increase is proportional with depth z. In most of the cases presented in this paper, the estimation
of su from CPTU tests used the corrected total cone resistance.
Undrained shear strength, su The interpretation was done with the following equation:
The three geohazards assessment sites are designated as A, B qt − σ vo
and C. The soil profiles at Sites D, E and F were established su =
N kt
4 OTC 19117

where qt is the corrected cone resistance (measured cone If one should interpret the undrained shear strength from
resistance plus measured pore pressure behind the cone), σvo is CPTU with one of the other two interpretation methods
the total overburden stress and Nkt is the cone factor. mentioned above (effective cone resistance or excess pore
There are two main sources of uncertainty in the pressure), the derivation of the mean and standard deviation of
evaluation of su from the above equation: the corrected cone su using the effective cone resistance or the excess pore
resistance qt and the cone factor Nkt. The uncertainty in the pressures can be done in a manner similar to the procedure
total overburden stress is considered as small as σvo is outlined above.
calculated from the unit weight of the clay times the depth. For Site E, the estimation of su from CPTU tests used two
Data in the literature show little variability in the unit weight alternative approaches; (1) correlation between corrected cone
of clay within a layer. Therefore, in the derivation below, the resistance Nkt and the pore pressure factor Bq (Lunne et al
uncertainty in the overburden stress is ignored. 1997); and (2) site-specific correlation between the pore
The mean and standard deviation of the undrained shear pressure factor Bq and the overconsolidation ratio (OCR)
strength su were evaluated using the first-order, second- followed by calculation of the su from the OCR using the
moment (FOSM) approach (Ang and Tang, 1984). The FOSM SHANSEP approach (Ladd et al 1977). The pore pressure
approach provides an effective means of investigating the factor Bq is defined as:
propagation of second-moment uncertainties. It provides an
approximate estimate of the central tendency parameter (e.g. u − uo ∆u
mean) and the dispersion parameter (e.g. standard deviation) Bq = =
q t − σ vo q t − σ vo
of a random variable which is a function of other random
variable. The statistical parameters for the undrained shear
strength were evaluated as follows: where qt is the corrected cone resistance, σvo the total overbur-
den stress and ∆u the measured excess pore pressure.
qt − σ vo
su =
N kt Characteristic soil profiles used in design vs.
statistical soil parameters
with mean at any depth as: The statistical results are shown as the mean of the undrained

su =
(q t − σ vo ) shear strength µsu, and the standard deviation represented by
the µsu ± σsu. The dispersion σsu represents both the spatial
N kt variability of the undrained shear strength in the soil layer in
Ignoring the uncertainty which may be present in the question (aleatory uncertainty) and the uncertainty in the
estimation of the overburden stress, the variance of the estimation of the in situ values (epistemic uncertainty).
undrained shear strength su, accounting for the uncertainties
for both qt and Nkt, becomes: Site A
At Site A, anisotropically consolidated triaxial compression
σ s2u ≈
σ q2t
+
(q t − σ vo )σ
2
2 tests were used to establish the characteristic undrained shear
2
N kt ⎛N ⎞
4 Nk
strength under triaxial conditions. Twenty-two tests were
⎜ kt ⎟ available between depths of 45 and 95 m. Figure 3 presents the
⎝ ⎠
results for Site A. The statistical analysis is shown as the
where the mean and standard deviation of the cone resistance means and ± one standard deviation (±1 std. dev.). The
are given by: characteristic (design) undrained shear strength profile was
n

∑q ti
taken at the lower bound of the measured CAUC-values. The
qt = i =1 characteristic suC selected was even lower than the mean minus
n one standard deviation below depth 65 m.

σ qt =
1 n
∑ qti − qt
n − 1 i=1
( )
2
Site B
At Site B, six fairly uniform CPTU profiles were analyzed
The soil design parameter reports that were used in this statistically. Figures 4 and 5 present the available CPTU data
study often provide upper bound and lower bound estimates of and the results of the statistical analysis at Site B compared
the cone factor. If the average cone factor was not given, the with the characteristic values. The undrained shear strength
statistical analyses assumed that the cone factor had a normal was interpreted from the CPTU's to correspond to loading
probability distribution and that the mean and standard under triaxial compression conditions, suC. The statistical mean
deviation of Nkt were: and the recommended characteristic value compare quite well,
and are in agreement at depth 1 to 2 m and at about 8 m depth.
N ktupper bound + N ktLower bound Otherwise, the statistical mean is only a few percentage points
N kt = (up to 5%) below the characteristic value. The coefficient of
2
variation was higher, approximately 16%.
N ktupper bound − N ktLower bound
σ N kt ≈
4 Site C
At Site C (Fig. 6), strength index tests (torvane, pocket pene-
trometer, lab vane), in situ vane tests, unconsolidated undrai-
OTC 19117 5

ned (UU) tests were available. More recent anisotropically and results illustrate the uncertainty in the interpretation method.
isotropically undrained triaxial compression tests (CIUC and For Site E1, the two interpretations gave very different results.
CAUC) were run at depths 28 m and 69 m. These are the most The characteristic profile was selected as close to the lower of
reliable results. The characteristic undrained shear strength in the two interpretations. For Site E2, the two interpretation
triaxial compression, suC, was obtained from a correlation with methods gave similar results. The characteristic strength was
in situ overburden stress (p'o) and overconsolidation ratio, again selected close to the lower of the two interpretations.
where the undrained shear strength was taken as 0.31·p'o. C
su (kPa)
Figure 7 presents the statistical analysis. Using the method 0 100 200 300 400 500 600 700 800

described above, the coefficient of variation about the means 0


CAUC test
was between 10 and 15%. Since there were only 4 data points Characteristic
available, the standard deviation was verified using the mean
Snedecor and Cochran (1964) approximation, where the range 20
mean +1 std.dev.
of values at a given depth is multiplied by a weighting factor mean -1 std.dev.

(see Lacasse and Nadim, 1996). This led to a coefficient of


variation between 6 and 14%. For a more complete study, one 40

should also include the statistics for the field vane tests. The
field vane showed a shear strength profile close to the triaxial

Depth (m)
compression results, and is considered to be reliable. 60

The strength index and UU tests show linearly increasing


shear strength with depth down to a depth of 70 m, and a de-
crease in strength down to 90 m. The data below 70 m are 80

probably affected by sampling disturbance. The triaxial data at


a depth of 70 m may also have been affected by sample
disturbance, though probably to a smaller degree than the 100

strength index tests.


Figure 7 compares the characteristic undrained shear
strength with the statistical analysis results. The mean is 10-
120

15% higher that the characteristic undrained shear strength. Fig. 3 Characteristic undrained shear strength in triaxial
The characteristic shear strength is close to the mean minus compression at Site A (1984) compared with result of
one standard deviation. The characteristic shear strength below statistical analysis.
a depth of 50 m tends to be higher than the mean minus one qt (MPa)
standard deviation. The relative position of the characteristic 0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4

profile and triaxial data at 69 m can confirm that the triaxial


0

data may have also been affected by sample disturbance at 70


m, if there is no change in the layering at that depth. 1
Mean_qt
DW50
Site D 2 DW51

At Site D, four highly variable CPTU profiles were available. DW51A


DW51B
Figure 8 reproduces the corrected cone resistance and Fig. 9 3
DW37
presents the characteristic lower and upper bound interpreted DW38
Depth (m)

suC (corresponding to triaxial compression) in the design 4

report. The strength recommendations were made for


calculations associated with skirt penetration. Figure 9 also 5

shows the statistical mean ± one standard deviation based on


the four CPTU results. The characteristic upper bound agrees 6
well the mean + one standard deviation in the top 3 m, but is
much higher than that below 3m. On the other hand, the 7
characteristic lower bound is about 2 standard deviations
lower than the mean in the top 3m, and one standard deviation
8
below the mean between 3 and 6 m.
Fig. 4 Results of 6 CPTU tests at Site C (2003).
Site E
At Site E, two CPTU profiles were available in two locations Site F
that had very different soil profiles. Figures 10 and 11 present Site F had a very uniform soft clay profile where both labora-
the results of the analyses (sites E1 and E2). For this site, the tory tests and CPTU tests were very uniform. Figure 12a pre-
effect of using two different methods of interpretation for the sents the laboratory data set analyzed statistically (CAUC and
CPTU results was studied, and the undrained shear strength DSS tests) and Fig. 12b the mean ± one standard deviation in
profiles form the two methods was compared. The two the two soil units for the DSS undrained shear strength. The
interpretations were used to obtain the undrained shear statistical analysis (Uzielli et al. 2006b) included not only the
strength for a direct simple shear state of stress suDSS. The statistical uncertainty, represented by the standard deviation of
6 OTC 19117

the data about the mean trend, but also the uncertainty in the not reflect a consistent safety margin and because of this
measurements and the uncertainty due to the limited number appears to be somewhat arbitrary, while remaining conserva-
of measurements. The resulting standard deviation represents tive. Slopes with nominally the same factor of safety could
therefore the total uncertainty. The coefficient of variation was have significantly different safety margins because of the
22% for the CAUC in soil units I and II, and 29% in unit I and uncertainties and how they are dealt with.
23% in unit II for the DSS tests. Statistical analysis was also C
su (kPa)
done for the CPTU results, and the CoV was lower. The 0 50 100 150 200 250

characteristic strength profile was only slightly lower than the 0

statistical mean of the laboratory data. CAUC/CIUC

10 Characteristic
C mean
su (kPa)
mean +1 std.dev.
0,0 5,0 10,0 15,0 20,0 25,0 20 mean -1 std.dev.
0
Characteristic
mean 30
1
mean +1 std.dev.

Depth (m)
mean -1 std.dev.
2 40

3
50
Depth (m)

4
60
5

70
6

7 80
Fig. 7 Characteristic triaxial compression undrained shear
8 strength at Site C compared with result of statistical
analysis.
Fig. 5 Characteristic triaxial compression undrained shear
strength compared with result of statistical analysis at qt (kPa)
Site B. 0 100 200 300 400 500 600 700
0

suC (kPa)
0 50 100 150 200 250
0 1
in-situ vane

10 UU
Pocket penetrometer
20 Torvane 2
Lab vane
30
CIUC/CAUC

40
3
Depth (m)

Depth (m)

50 qt KS01
qt KR02
60 qt-KP01
4 qt KN01
70 mean_qt

80
5

90

100
6

Fig. 6 Undrained shear strength tests at Site C (2003).


7
Discussion
In a deterministic framework, the stability situation for a slope Fig. 8 Results of 4 CPTUs at Site D (2001).
is expressed by the factor of safety. The factor of safety is
defined as the ratio of the characteristic resisting force to the The characteristic shear strength profile for sea floor
characteristic load (driving force). The conventional approach stability evaluation recommended by the experienced engineer
does not address the uncertainty in load and resistance in a seems to fit well with the statistically-determined mean value
consistent manner. The ambiguous definition of “characte- minus one-half to one standard deviation.
ristic” value requires that the engineer implicitly account for The case studies suggest that in the early 80's, the
uncertainties by choosing conservative values of load (high) characteristic undrained shear strength was selected close to
and resistance parameters (low). The choice, however, does the mean value minus one standard deviation. Characteristic
OTC 19117 7

undrained shear strengths from more recent case studies lie skirts are of interest, this is believed to be too conservative
close to the mean value minus one-half standard deviation, because of spatial averaging effects.
although there are exceptions.
suDSS (kPa)

100

120

140
suDSS

20

40

60

80
(kPa)
Su (kPa)

0
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 0
0
Based
Based on on
Nkt N
&Bq & Bq
kt

calculated
Mean Nmean (Nkt&Bq)
kt & B q
1 suC
SuC_recommended_Upper
recom. Upper Bound
5 Based
Based on on Bq &
Bq&OCR OCR
suC
SuC_recommended_Lower
recom. Lower Bound
suC
SuC_mean
mean Mean Bmean
calculated q & OCR
(Bq&OCR)
2
mean +1 std.dev.
+std Characteristic
Design, Capacity
mean -1 std.dev.
-std
3 10

Depth (m)
Depth (m)

15
5

20
7

8
25

Fig. 9 Characteristic upper and lower bound undrained shear


Fig. 11 Characteristic direct simple shear undrained shear
strength from CPTUs at Site D compared with result of
strength from CPTUs interpreted with 2 methods at Site
statistical analysis.
E2 (2003).
SuDSS
suDSS (kPa)(kPa)
100

120

140
20

40

60

80
0

5
Basedon
Based onNkt
Nkt&Bq
& Bq
Mean Nkt &
calculated Bq (Nkt&Bq)
mean
Basedon
Based onBq&OCR
Bq & OCR
Mean Bq &mean
calculated OCR(Bq&OCR)
10 Characteristic
Design
Depth (m)

Characteristic
15

20

(a) (b)
25

Fig. 12 (a) Undrained shear strength from laboratory tests at


Fig. 10 Characteristic direct simple shear undrained shear Site F (1989), (b) Characteristic undrained shear strength
strength from CPTUs interpreted with 2 methods at Site in direct simple shear at Site F compared with result of
E1 (2003). statistical analysis.

When the engineer is faced with two sets of data for the The statistical data also provided useful additional
shear strength profile, he tends to choose the lowest value of information, for example confirming the effects of sample
the two as the characteristic profile. It may be more logical to disturbance on the specimens tested in the laboratory.
combine the two sets of data through e.g. a Bayesian model. Although some patterns seem to emerge on the relationship
A characteristic strength equal to the mean minus 1 to 2 between the characteristic design soil parameters and the
standard deviations would probably fit well with the statistically-derived ones through the cases considered in this
characteristic profiles for situations where much of the study, the geotechnical engineers should look beyond using
available data are strength index tests and tests of unknown exclusively simple statistical analyses, and start addressing the
quality. This is also rational if the point values are of concern. variability and uncertainty in soil parameters explicitly.
If the undrained shear strength averaged over, for example, a Probability theory and reliability analyses provide a ration-
critical sliding surface or the length and periphery of long al framework for dealing with uncertainties and decision-
8 OTC 19117

making under uncertainty. Depending on the level of sophisti- When determining the characteristic undrained shear
cation, the analyses provide the following output: probability strength, using ±0.5 or ±1 standard deviations in the analyses
of failure or unsatisfactory performance, reliability index, the will depend on what the parameters will be used for (for
most probable combination of parameters leading to failure example, stability analysis versus skirt penetration analysis). A
and sensitivity of result to any change in parameters wider bound would be logical if one is mainly concerned with
the punctual values of the undrained shear strength. If the
Conclusions and recommendations undrained shear strength is averaged over a larger depth and/or
The study aims at setting the premises for a discussion in the volume of soil, the wide range is however probably too
profession on the consistency in the selection of design conservative because of spatial averaging effects.
parameters. The profession should aim at providing consistent The authors recommend the following when the parame-
safety margins in design, as was the intention with LRFD ters are based on high quality laboratory or in situ tests: for
codes. Yes the study shows that there can be large and highly design over large volumes of soil, the characteristic undrained
variable uncertainties in the soil parameters. Probability ap- shear strength could be set equal to the mean minus 0.5
proaches or some simplified version of this is probably the standard deviation; for design of installations over smaller
only rigorous way to approach a consistent safety margin. areas or volumes (e.g. penetration of short skirts), one should
Ideally characteristic parameters should be estimated to be use a characteristic value equal to the mean minus one
above/below a threshold within a distribution in which the standard deviation. For design based on less reliable test data,
user has a reasonable level of confidence. Threshold the range should be one to two standard deviations, depending
recommendations are made below. Such an approach would on the volume of interest. Furthermore for the stability
also give recognition to the necessity and cost-effectiveness of assessment of natural submarine slopes, the mean undrained
high quality and reliable soil sampling and testing. shear strength should be used. These definitions and suggest-
From the case studies, it would seem that when one relies ions are set forth as a subject for debate in the coming months
on laboratory test results to establish the characteristic (design) so that the issue can be discussed and agreement can be
strength, the experienced engineers tend to lie much lower reached for use in practice.
than the mean, perhaps at about one standard deviation below The authors believe that the geotechnical profession should
the mean. On the other hand, when one relies on the results of use more extensively and more routinely statistical analysis of
in situ piezocone tests, the characteristic strength can be much the parameters. Patterns emerge on the relationship between
closer to the mean of the interpreted measurements in situ. characteristic soil parameters and statistically-derived ones.
This may be due to the fact that one feels that for CPTU's in The authors recommend however that the profession should
clays, the selection of the cone resistance values is already look beyond simple statistical analyses, and address the
based on the data that lie towards the lower bound of the variability and uncertainty in soil parameters explicitly by
measured data. implementing probability theory and reliability analyses.
A study of the case study data bank suggests that the
experienced engineer was more conservative in the early to References
mid-80's than in the past 10-15 years. Ang, A.H.S. and W.H. Tang: Probability concepts in engineering
There is a need to unify the vocabulary used in the planning and design, New York. Wiley & Sons, 409 p, 1975.
description of soil parameters used in design. "Characteristic" DNV/Norsk Hydro Technical Report No. 2006-1437: “Guidance
Notes Statistical Representation of Soil Data”, 30 August 2006.
strength is probably fairly well understood by most
Lacasse, S., and F. Nadim: “Uncertainty in characterizing soil
practitioners to be the recommended value for design. The properties”, In Shackleford, C.D., Nelson, P.P. & Roth,
authors recommend that "best estimate" be used to designate M.J.S.(eds) Uncertainty in the Geologic Environment: From
the statistical mean obtained by methods similar to those Theory and Practice, ASCE SSP 58, New York, 49-75, 1996.
described herein. The use of the expressions "Lower Bound" Ladd, C.C., R. Foott, K. Ishihara, F. Schlosser, F., and H.G. Poulos,
and "Upper Bound" should be discouraged, but one should “Stress deformation and strength characteristics”, SOA Report.
express the "uncertainties" (or the "variability") as the Proc. ICSMFE Tokyo, Vol. 2, 421-494, 1977.
standard deviation or the coefficient of variation. Lunne, T., P.K. Robertson and J.J.M. Powell: Cone penetration
The profession should encourage the use of the testing in geotechnical practice, London Spon Press, 312 p,
1997.
characteristic value for design. There should be a series of
Nadim, F., G. Biscontin and A.M. Kaynia: “Seismic triggering of
characteristic profiles for different design cases, e.g. skirt submarine slides”, 2007 Offshore Technology Conference,
penetration, stability of skirted gravity foundations, seismic Paper 18911, Houston, Texas, U.S.A., 30 April–3 May 2007.
response (Nadim et al 2007). The characteristic profile could Snedecor, G.W., and W.G. Cochran: Statistical methods, Univ of
be more sophisticated, and include a range or a standard Iowa Press, 140 p, 1964.
deviation to account for uncertainty and specific modifications Uzielli, M., S. Lacasse, F. Nadim and K.K. Phoon: “Soil variability
to account for shape or depth effects. The standard deviation analysis for geotechnical practice”, In T.S. Tan, K.K. Phoon,
considered should represent the total uncertainty, and not only D.W. Hight & S. Leroueil (eds.), Proc. 2nd Intern. Workshop
the statistical uncertainty represented by the standard deviation "Characterization and Engineering Properties of Natural Soils",
Singapore, Taylor & Francis NL., Vol. 3, 1653-1752, 2006a.
of the data about the mean trend. This means that the
Uzielli, M., S. Lacasse, F. Nadim and T. Lunne: “Uncertainty-based
uncertainty in the measurements and the uncertainty due to the characteristics of Troll marine clay”, In T.S. Tan, K.K. Phoon,
limited number of measurements should also be evaluated and D.W. Hight & S. Leroueil (eds.), Proc. 2nd Intern. Workshop
represented in the standard deviation. "Characterization and Engineering Properties of Natural Soils",
Singapore, Taylor & Francis NL., Vol. 4. 2753-1782, 2006b.

View publication stats

Vous aimerez peut-être aussi