You are on page 1of 14

1378

Reliability approach for the side resistance of piles


by means of the total stress analysis (a Method)
C. Cherubini and G. Vessia

Abstract: The evaluation of the pile–soil adhesion plays a fundamental role in the estimation of the side resistance for to-
tal stress analysis. Over the years, researchers have presented proposals for adhesion factor formulations even though only
a few of them have shown a certain agreement in numerical and (or) methodological terms. Hence, several real-size exper-
imental analyses have improved the understanding of the pile–soil adhesion phenomenon and mechanism. Nevertheless,
the undrained shear strength (cu) values depend on the experimental technique employed. Such results force engineers to
make a difficult choice among various formulations. A reliability analysis is performed in this paper to take into considera-
tion the variations in formulations and values of the side resistance of bored piles in clayey soils. This study involves piles
having different lengths and diameters, which are supposed to be bored in Matera clays. Such soil is characterized by
means of laboratory investigation campaign, and its mechanical and stochastic main features are reported here. Values of
reliability index  are calculated by means of the first-order reliability method.
Key words: bored pile design, side resistance, matera clay, reliability-based design, undrained shear strength, adhesion factor.
Résumé : L’évaluation de l’adhésion pieu – sol joue un rôle fondamental dans l’estimation de la résistance des fûts pour
l’analyse en contrainte totale. Au cours des ans, les chercheurs ont présentés des propositions pour les formulations du fac-
teur d’adhésion même si seulement quelques-uns d’entre eux ont montré une certaine concordance en termes numériques
et (ou) méthodologiques. Ainsi, plusieurs analyses expérimentales à pleine échelle ont amélioré la compréhension du phé-
nomène et du mécanisme d’adhésion pieu – sol. Néanmoins, les valeurs de résistance au cisaillement non drainé cu dépend
de la technique expérimentale utilisée. De tels résultats obligent les ingénieurs à faire un choix difficile parmi les diverses
formulations. Dans cet article, on fait une analyse de fiabilité pour prendre en considération les variations dans les formu-
lations et les valeurs de résistance des fûts des pieux forés dans les sols argileux. L’étude implique des pieux ayant diffé-
rentes longueurs et diamètres, qui sont censés être forés dans les argiles de Matera. Un tel sol est caractérisé au moyen
d’une campagne d’essais en laboratoire et on fait rapport ici de ses caractéristiques mécaniques et stochastiques. Les va-
leurs de l’indice de fiabilité  sont calculées au moyen de la méthode de fiabilité du premier ordre.
Mots-clés : conception de pieux forés, résistance des fûts, argile de Matera, conception basée sur la fiabilité, résistance au
cisaillement non drainé, facteur d’adhésion.
[Traduit par la Rédaction]

Introduction 0.60 m in diameter. The piles are designed to be bored in


Matera clay. A reliability-based design is carried out accord-
The evaluation of adhesion between pile and soil, indi- ing to a complete characterization of the random structure of
cated as ca, plays a prominent role in the calculation of the the undrained shear strength (cu) presented here.
side resistance of piles in clays carried out in terms of total
stresses (referred as ‘‘a method’’).
Many national design provisions and recommendations Classic methods for evaluation of adhesion
suggest estimating ca by means of the undrained cohesion (cu) factor
reduced by a coefficient, namely , which represents the per- The side resistance of piles (Qs), in terms of total stresses, is
centage of cu mobilized by the pile–soil adhesion mechanism. commonly calculated using the following expression:
This paper contains a brief review of the most widely used
methods for the evaluation of adhesion by means of similar ½1 Qs ¼ cu DL
formulations that take into account the relationship between where
undrained cohesion and effective vertical stress. Moreover,  is the adhesion coefficient
the calculations of the reliability index  through the first- cu is the undrained shear strength
and second-order reliability method (FORM and SORM) are D is the pile diameter
carried out for piles of 15 and 20 m in length and 0.40 and L is the pile length

Received 15 September 2005. Accepted 22 June 2007. Published on the NRC Research Press Web site at cgj.nrc.ca on 18 January 2008.
C. Cherubini1 and G. Vessia. Technical University of Bari – Via Orabona, 4 – 70125 Bari – Italy.
1Corresponding author (e-mail: c.cherubini@poliba.it).

Can. Geotech. J. 44: 1378–1390 (2007) doi:10.1139/T07-061 # 2007 NRC Canada


Cherubini and Vessia 1379

The evaluation of  is carried out using series of equations, Fig. 1. Trends of the adhesion factor recommended by various
usually functions of cu, which have been proposed over the authors: K = Kerisel 1965, T = Tomlinson 1957, W = Woodward
years by several authors. The trends more often used are plot- and Boitano 1961, P = Peck 1958, D = Dennis and Olson 1983,
ted in Fig. 1 (Sladen 1992). A = API 1974, S = Sowers and Sowers 1970, M = McCarthy 1977
Furthermore, it is worth remembering that eq. [2], pro- (adapted from Sladen 1992).
posed by Kulhawy and Jackson (1989), is based on 106
bored piles and 41 of which were tested in compression and
65 in tension (Fig. 2)

pa cu
½2  ¼ 0:21 þ 0:26 for   1 3
cu pa

where pa represents the atmospheric pressure in the same


measurement units as cu.
Figures 1 and 2 show that for low values of cu or cu/pa, 
is assumed to be 1 whereas for higher cu values  magni-
tude can be taken from the various equations proposed in
the literature ranging from about 0.5 to 0.2. However, the
significant dispersion of the various curves should be taken
into consideration as the authors do further on.
Coduto (1994) suggests different relationships between 
and cu for the two cases of bored and driven piles. His study
provides analytical rules for the variations of , based on a
large number of data, given in Fig. 3 and Table 1. Fig. 2. Correlation between  and cu/pa for bored piles (adapted
Hence, it may be inferred that, on one hand, there is a no- from Kulhawy and Jackson 1989).
table dispersion of data values, which are at times well be-
low the interpolating trend, and, on the other hand, for low
values of cu, in the case of driven piles, the experimental
data show values of  even higher than 1.
Figures 4 and 5 show both  values in function of cu for
glacial till and boulder clay (Weltman and Healy 1978) and
the variation of  with respect to cu for some typical stratig-
raphies (Battaglio and Lancellotta 1985; Tomlinson 1977).
Moreover, Skempton (1959) found that  values for bored
piles in London clay vary from 0.3 to 0.6 according to cu
variation with a mean value of 0.45. The upper bound value
of cu considered is 96 kPa. The role of softened material at
the pile–soil interface appears to be fundamental in the
development of such adhesion (Lancellotta and Calavera
1999).
According to Christoulas (from Broms 1981) a reasonable
value of adhesion factor  must be commonly assumed as
0.45 for piles bored in stiff clayey soils.
Recently, the approaches directly adopted in many Euro-
pean countries for the evaluation of  or ca have been sum-
marized in the Report ERTC3 (1999).  0 0:45
In all the cases, the diagram in Fig. 6 shows how the ulti- 
½3  ¼ 0:5 vo
mate side resistance calculated according to the Italian na- cu
tional regulations is notably variable.
This equation is derived without taking into account the
way in which piles are put on site (bored or driven). However,
Variation of  values with respect to the
the same author stated that eq. [3] should be preferably used
effective stress level to substitute the coefficient  in the case of driven piles.
As early as 1972, Vijayvergiya and Focht observed that Furthermore, it should be highlighted that eq. [3]
the adhesion factor  increases as the stress level grows pro- (Sladen 1992) has given good results as compared with
vided that all other conditions are equal. measured side resistance in the case of bored concrete piles
Sladen (1992) attempted to make a relationship between in overconsolidated blue clay (Baldassarre et al. 1993).
0
the adhesion factor and the ratio of vo =cu according to the Recently, Cherubini (1999) has formulated the following
following expression: expression:
# 2007 NRC Canada
1380 Can. Geotech. J. Vol. 44, 2007

Fig. 3. Undrained shear strength versus  for (a) bored and (b) dri- Table 1. Analytical relationships between  and cu for
ven piles (adapted from Coduto 1994). bored and driven piles (Coduto 1994).
Bored piles  = 1 for cu £ 51 kPa
Bored piles  = 0.32 + 250cu–1.5 for cu > 51 kPa
Driven piles  = 1 for cu £ 32 kPa
Driven piles  = 0.35 + 170cu–1.6 for cu > 32 kPa

Fig. 4. Variation of  with cu for glacial till and boulder clay


(adapted from Weltman and Healy 1978).

 0:2  0 0:3
D vo
½7  ¼ 0:5
L cu

where D (diameter) and L (length) measure the slenderness


of the pile, which is assumed to influence the  values (see
Kraft et al. 1981; Semple and Rigden 1984).
Moreover, Fig. 7 shows the variation of  in function of
0
cu =vo according to American Petroleum Institute (1993) for-
mulations together with  values carried out by Nowacki et
al. (1992) who makes  depend on the index of plasticity.
Particularly relevant is the circumstance that the lower the
IP value the bigger the reduction rate of .

From a to the side-resistance determination


 0
0:36 According to the previous discussion, a method seems to
vo
½4  ¼ 0:5 be simple in theory but troublesome in practical application
cu because of the uncertainties related to the evaluation of .
which resembles eq. [3] (Sladen 1992) and the expressions Indeed, the evaluation of cu (both on site and in laboratory
(Lacasse and Boisard 1994) proposed by American Petroleum tests) reveals high dispersion (Kulhawy and Jackson 1989).
Institute (API 1993), presented as Accordingly, Chen and Kulhawy (1993), on the basis of nu-
 0 0:5 merous experimental data, studied the differences in magni-
vo 0 tude amongst cu values provided by different compression
½5  ¼ 0:5 for cu =vo  1:0 tests performed on triaxial device (unconsolidated undrained
cu
(UU), consolidated undrained (CU), and consolidated iso-
 0:25 tropically undrained compression (CIUC)). The analysis was
0
vo 0 performed on soil samples from four clay types: normally con-
½6  ¼ 0:5 for cu =vo > 1:0
cu solidated (NC), lightly overconsolidated (LOC), moderately
overconsolidated (MOC), and highly overconsolidated (HOC).
In Table 2,  values are estimated according to Sladen, Some results are briefly given in terms of average values
API, and Cherubini formulations and for few assigned values (see Table 3). Experimental outcomes point out how the use
0
of the ratio of vo =cu . It is found that such values are very of less reliable tests (UU, CU) than the CIUC test may no-
close indeed. ticeably underestimate cu. It is more evident in the case of
Kolk and Van der Velde (1996), starting from the API normally consolidated clays. On the other hand, for MOC
equation, suggested the use of the formula and HOC the cu values are, on average, constant.
# 2007 NRC Canada
Cherubini and Vessia 1381

Fig. 5. Variation of  with cu for typical stratigraphies (adapted from Tomlinson 1977).

Hence, over the years some authors have proposed differ- Other authors (Kezdi and Rethati 1988) suggested the in-
ent correcting factors to be introduced into side-resistance troduction of another factor S to take account of the side
formulation to take account of the cu value sensitivity to the conditions during the perforation
experimental technique.
As a matter of fact, Dennis and Olson (1983) considered ½9 qs ¼ Scu
84 load tests performed on tubular piles installed off-shore. where S = 1 for vertical plain shafts and S = 1.2 for tapered
The authors proposed to evaluate the factor  on the basis shafts according to Balasubramaniam et al. (1981).
of the cu values coming from UU tests performed on high On the basis of these results, the bearing capacity of
quality samples. These values are suggested to be taken as bored piles reveals an extremely variable activity as with re-
a reference. Thus, the side-resistance formulation is modi- spect to the determination of the side contribution given by
fied according to the following expression: adhesion. A way to overcome this point is proposed here for
the case of bored piles in Matera clay. The tool involved is
½8 qs ¼ cu Fc FL the reliability analysis, which takes the variability in  and
cu values into account.
where Fc is a factor for converting the value of resistance de-
termined by other tests into the reference values. Generally,
Fc is assumed to be Reliability-based design of bored pile side
(1) 1.1 when cu value is determined from CU tests performed resistance
on samples collected by thin-walled sampler embedded Owing to the previous considerations, the design of bored
only with pressure; piles in clay soils should be developed by means of a
(2) 1.8 for the cu value determined from nonconfined sam- reliability-based approach because it deals with loads and re-
ples collected by dynamically embedded sampler; sistances regarded as random variables.
(3) 0.7 for the cu value determined from on site vane tests. Hence, variability and uncertainty represent another de-
FL is a correction factor that takes into account the length gree of freedom in design activity. Such a new ‘‘dimension’’
of the pile. It is assumed to be equal to 1 for length up to in engineering design transforms the evaluation of ‘‘safety’’
30 m and 1.8 for length >53 m. in the estimation of ‘‘probability of failure’’ by means of the
# 2007 NRC Canada
1382 Can. Geotech. J. Vol. 44, 2007

Fig. 6. Evaluation of the ultimate side bearing capacity in terms of total stress according to the various National Regulations in use
throughout Europe (adapted from ERTC3 1999).

Table 2. Adhesion coefficient values for dif- Thus, for the reliability-based design the main point is the
ferent formulations and different 0vo =cu values. calculation of , that is, as eq. [11] shows, the calculation of

the mean and the standard deviation of SM.
0 Several methods have been developed to accomplish such
vo =cu Sladen API Cherubini
statistical moments of SM, such as the first-order reliability
0.25 0.28 0.35 0.33 method (FORM), which was geometrically interpreted by
0.5 0.37 0.42 0.43 Hasofer and Lind (1974) and the second-order reliability
1.0 0.50 0.50 0.55 method (SORM) by Breitung (1984) among others. Accord-
2.0 0.68 0.71 0.71 ing to such techniques, the mean and the standard deviation
3.0 0.82 0.87 0.82 of the performance function of n random variables are calcu-
lated by means of Taylor series stopped at the first- and the
reliability index . The function to be considered is called second-order terms, respectively. The first- and the second-
performance function (PF) or safety margin (SM). It is given order reliability methods are employed in this paper and re-
by the difference between the resistance R and the load Q sults are compared to estimate the need of SORM precision
in the cases studied.
½10 SM ¼ R  Q

Accordingly, the reliability index  can be defined as the Reliability index calculation
distance of the mean value of the performance function from Results from loading tests reported by Cherubini and
its critical value, which is 0: Vessia (2005) are now recalled to explain the reason for the
SM choice concerning the use of  formulation. Such tests are
½11 ¼ performed on bored piles in Matera clays and the investigated
SM piles are 0.60 m in diameter and 9, 10, and 11 m in length. In
The failure probability can be related to the reliability in- that paper, Coduto’s formulation of the adhesion factor  was
dex  by the following relationship (Leporati 1979): found to work better than others and gave a mean  value of
0.428 for a mean value of undrained shear strength cu equal to
½12 Pf ¼ ðÞ 175 kPa. Such  value falls within the value suggested for
moderately consolidated clays by Christoulas, that is, 0.45,
where () is the cumulative distribution function of a stan- and the value proposed by AGI for the bored piles in moder-
dard normal distribution. More recently, Withiam et al. ately overconsolidated clays, that is, 0.40. Results from
(1997) presented failure probability values (Pf) for normal Cherubini and Vessia (2005) suggest hereinafter employ
distribution and lognormal distribution starting from the Coduto’s expression of  for piles bored in Matera clays.
most common  values (Table 4). Nevertheless, this choice does not eliminate the model errors,
# 2007 NRC Canada
Cherubini and Vessia 1383

0
Fig. 7. Pile soil adhesion versus plasticity index and cu =vo ratio (adapted from Nowacki et al. 1992) together with recommendations from
American Petroleum Institute (1993).

Table 3. Laboratory mean values of cu(UU)/cu(CIUC) Table 4. Probability of failure values for normal
ratio and cu(CU)/cu(CIUC) ratio for natural clay samples and lognormal distribution related to some com-
characterized by different OCR values. mon  values (after Baecher and Christian 2003).

cu(UU)/cu(CIUC) cu(CU)/cu(CIUC) Probability of failure, Pf


Type of clay (average value) (average value) Reliability Normal Lognormal
NC 0.61 0.51 index,  distribution distribution
LOC 0.76 0.71 3.0 1.3510–3 1.1510–3
MOC 0.91 0.88 3.5 2.3310–4 1.3410–4
HOC 1.01 1.01 4.0 3.1710–5 1.5610–5
4.5 3.4010–6 1.8210–6
which are represented by the  variability. Moreover, Phoon 5.0 2.8710–7 2.1210–7
and Kulhawy (2005), among others, suggested the evaluation 5.5 2.9010–8 2.4610–8
of the  variability as model error by means of axial loading
tests compared with the theoretical model of the side resist- hawy and Jackson (1989) made some compression loading
ance. They discussed the opportunity of verifying the statisti- tests on bored piles in clayey soils. Table 6 shows the values
cal independence of the resultant  variability, concluding of correlation coefficients from 41 tests that vary from –0.4
that such procedure cannot substitute the long lasting method to –0.8. In this study, two cases are investigated (cu = –0.5
of expressing  as a function of cu. According to Kay (1993) and 0) to estimate the influence of the correlation, which is
model bias varies generally between 0% and 25%. Further- a troublesome parameter to be quantified, on the reliability
more, Kulhawy and Jackson (1989) made an interesting sur- index of the pile side resistance.
vey on the model uncertainty  in the estimation of pile side
resistance. They predicted the mean value of side resistance Reliability analyses are carried out by means of STRUREL
by means of eq. [2] and divided them for measured side re- (1996). The SORM method is chosen to calculate the reli-
sistance for compression loading. Results are reported in ability index () of the performance function (SM) below
Table 5. From these data, the total value of coefficient of var- ½13 PF ¼ cu DL  Ql
iation of model uncertainty can be drawn, which is 28%.
Thus, the coefficient of variation of  is assumed to be 20% where Ql is the side load acting on the shaft of the pile whose
in this paper to take account of the model bias. range is 350 kN 7 1550 kN for pile diameter of 0.40 m
Moreover, due to the negative correlation between  and and 500 kN 7 2300 kN for pile diameter of 0.60 m.  is the
cu, as shown before, a correlation coefficient  should be adhesion factor. With regards to the reliability analyses,
taken into account in the reliability analysis. A few studies  = 0.428 is taken as the mean value of the adhesion fac-
for correlation quantification of  and cu are available. Kul- tor; its coefficient of variation (indicated as COV) is equal
# 2007 NRC Canada
1384 Can. Geotech. J. Vol. 44, 2007

Table 5. Statistics for  correlation as predictor (after Kulhawy Table 6. Correlation coefficient for  versus su/pa (after Kulhawy
and Jackson (1989)). and Jackson (1989)).

Values of Qs(predicted)/Qs(measured) Data set


Statistic Group 1 Group 2 Group 3 Total Statistic Group 1 Group 2 Group 3 Total
Number of points 9 24 7 40 Number of points 10 24 7 41
Mean 0.85a 0.92 0.83 0.90b Correlation –0.835 –0.717 –0.373 –0.606
COV 0.21a 0.28 0.33 0.28b coefficient
a
mean = 0.96 and COV = 0.43 when one aberrant point is included.
b
mean = 0.92 and COV = 0.33 when one aberrant point is included. Moreover, the variance reduction function for cu is calcu-
lated by Vanmarcke’s (1983) simplified expression:
to 20% and its distribution function is taken to be normal
and lognormal distribution alternatively. cu is the undrained 
½15 2 ¼
shear strength with a mean value equal to 175 kPa, stan- L
dard deviation cu equal to 60 kPa, and the distribution
function is assumed to follow the normal and the lognor- where  is the scale of fluctuation and L is the length of the
mal distribution alternatively. These values are drawn from pile investigated.
the spatial variability characterization of Matera clays as il- The performance function for the reliability analysis is
lustrated by Cherubini et al. (2005). Measurements collected compared to a large range of side load values (expressed in
for the variability study were performed until 24 m of depth. kN) to establish a reasonable value of the reliability index .
The standard deviation value reported here (60 kPa) is ap- Results from the analyses are shown in Figs. 8–9. Here, only
proximately 20% reduced from the standard deviation mea- reliability index values from FORM analyses are reported,
sured value of 76 kPa so that the uncertainties related to the provided that the SORM results (not reported here) are quite
investigation procedures and measurement errors (Rethati coincident with the FORM one in these cases.
1988) are not considered. As a matter of fact, according to Each graph points out the differences in results between
Rethati’s work, the standard deviation of natural variability lognormal distribution and normal distribution of both cu
of soil, h, results from the following expression: and  for a given pile length, diameter, and a correlation co-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi efficient. Such differences increase as  increases. This is
½14 h ¼ 2  2rp more evident for lognormal distribution than for the normal
one. In the case of lognormal distribution, the increase in 
where  is the standard deviation of any physical character- values depends on the scale of fluctuation and on the corre-
istic derived by experimental investigation, and rp is the lation coefficient value. Such behaviour is common for piles
component of the standard deviation characterizing the mea- of different length and diameter, therefore only the results
surement errors. Rethati (1988) reported numerical values of from Fig. 8 are now discussed.
these two components of standard deviation of few soil phy- As seen from the four graphs, under the same geometrical
sical characteristics of Szeged soil. Such data suggest rp and side loading conditions, the lognormal distribution for cu
vary in a range of about 12% to 25% of . and  variables gives higher  values than the normal distri-
Phoon and Kulhawy (1999) summarized the total meas- bution. This occurrence could be attributed to the asymmet-
urement errors in terms of coefficient of variation coming ric shape of the lognormal distribution with respect to the
from a variety of common laboratory tests for cu measure- normal one. Furthermore, the gap in  values among the
ment among other properties (Hammit 1966; Johnston 1969; three values of the cu scale of fluctuation is negligible just
Sherwood 1970; Singh and Lee 1970; Minty et al. 1979). for  equals to 0.5 and 1 m when cu = –0.5 and cu and 
They found a 20% mean value of measurement error related are normally distributed.
to cu evaluation. Here, the cu values are prevalently meas- Moreover, the influence of the correlation coefficient
ured by UU tests, which are less precise than CIUC tests. seems to heavily affect  values. As Figs. 8a and 8b show,
Accordingly, the 20% reduction in standard deviation is for a 854 kN side load, provided that cu and  are normally
thought to be a correct assumption. distributed and cu scale of fluctuation is equal to 1 m,  =
To save the general purpose of this study, authors do not 4.6 for cu = –0.5 whereas  = 3.5 for cu = 0.0. These
deal with measurement errors. outcomes suggest pay attention on the values given to the
D is the diameter of piles considered equal to 0.40 and correlation coefficient of  and cu. For different clayey soils,
0.60 m, respectively. The dimension of piles is chosen from such parameter has different magnitude. When it is not eval-
those commonly used in deep foundation design in Matera uated for the case studied, it should be wiser and safer not to
clay. L is the length of pile assumed to be 15 and 20 m. consider cu = 0.0.
As far as the vertical scale of fluctuation is concerned, the The results also show that, as the scale of fluctuation in-
magnitude of a few metre values is suggested by the litera- creases,  values decrease. Such behaviour is due to the in-
ture. For the case studied (Matera clays), measures of cu crease in the variance reduction function with the increase in
scale of fluctuation () are not available. Nevertheless, three the scale of fluctuation. Furthermore, differences between 
possible values of the scale of fluctuation are considered: values for normal and lognormal distributions increase when
0.5, 1.0, and 2.0 m, according to evaluations performed on the scale of fluctuation decreases according to the four cases
similar clays (Cherubini 1997; Cafaro and Cherubini 2002). illustrated by Figs. 8 and 9.
# 2007 NRC Canada
Fig. 8. Reliability index versus side loads for four pile geometries: (a) and (b) L = 15 m and D = 0.4; (c) and (d) L = 15 m and D = 0.6.
Cherubini and Vessia

#
2007 NRC Canada
1385
1386

Fig. 9. Reliability index versus side loads for four pile geometries: (a) and (b) L = 20 m and D = 0.4; (c) and (d) L = 20 m and D = 0.6.

#
2007 NRC Canada
Can. Geotech. J. Vol. 44, 2007
Cherubini and Vessia 1387

At this stage, it can be interesting to compare the results Table 7. Side bearing capacity of bored piles in Matera
from Figs. 8 and 9 with the ones from the partial safety factor clay calculated by means of the a method through the
approach proposed by Eurocode 7, assuming as the ‘‘charac- Eurocode 7 approach.
teristic value’’ of undrained shear resistance, cuk, the expres-
Side bearing capacity, P (kN)
sion suggested by Schneider (1997) and Orr and Cherubini
(1999), that is: Diameter Length
  (m) (m) DA1 DA2 DA3
COV 0.4 15 793 1010 854
½16 cuk ¼ c u 1 
2 0.4 20 1058 1346 1139
0.6 15 1190 1515 1282
where c u is the mean value of cu measurements and COV is 0.6 20 1587 2020 1709
expressed in decimal notation.
The Eurocode 7 gives three possible approaches for geo-
technical design of piles. The designer shall choose one of of fluctuation () equal to 2.0 m. On the contrary, in the case
them. In this paper, the authors considered all the three ap- of cu = 0.0, for cu and  normally distributed, none of the
proaches to verify how much the  value can change. The cases are properly acceptable but those related to the first
three possible combinations of partial safety factors are the two scales of fluctuation for lognormal distribution.
following:
Those results cannot therefore be predicted in advance.
Design approach 1, DA1, where ‘‘M1’’+ ‘‘R4’’ partial
Thus, without a detailed statistical characterization of the
safety factors are combined;
design variables, partial safety factor approaches do not al-
Design approach 2, DA2, where ‘‘M1’’+ ‘‘R2’’ partial ways give a satisfactory reliability level. The design ap-
safety factors are combined; proach DA1 seems to be the most reliable in pile design,
Design approach 3, DA3, where ‘‘M2’’+ ‘‘R3’’ partial although it is not adequate for large scales of fluctuation, as-
safety factors are combined. suming normally distributed cu and  variables.
In the above three possible combinations, the partial safety Besides, analyses are performed according to the deter-
factor for soil resistance M corresponds to M1 = 1.0 and ministic approach suggested by the Italian Geotechnical Pro-
M2 = 1.4 while the partial safety factor related to the entire visions (D.M. LL.PP. 11/3/1988). They impose a minimum
side resistance of bored piles R corresponds to R2 = 1.1, value of global safety factor of 2.5 for designing according
R3 = 1.0, and R4 = 1.3. to the theoretical model, that is:
In all the cases, Eurocode 7 does not suggest a partial
safety factor for the model variable  but it uses the R factor ½17 Pl ¼ cu DL
to take into account model uncertainties.
The results from partial safety factor analyses are given in The geometrical characteristics of the investigated piles
Table 7, and they are obtained from eq. [13]. are the same as in the partial safety factor analysis, and the
mean value of c u = 175 kPa and  = 0.428 are taken into
The graphs in Figs. 8–9 show the reliability index values
account. The results are summarized in Table 11. As can be
related to side-resistance values according to the three de-
seen global safety factor of 2.5 provides much lower values
sign approaches suggested by Eurocode 7. These values of
of allowable side loads than those accomplished by means
 are also summarized in Tables 8–10.
of the partial safety factor. Table 11 shows a difference of
The UNI EN 1990 – Annex B (2002), for the purpose of at least 30% with respect to the values of DA1 in Table 7.
reliability-based design, recommends the minimum  values In such cases, the Italian Geotechnical Provisions are highly
corresponding to three consequences classes: CC1, CC2, and conservative according to the global safety factor method,
CC3. Three reliability classes: RC1, RC2, and RC3 are asso- which by no means takes into account the variability of
ciated with the three consequences classes. For class RC2 soil-resistance parameters.
structural members (such as piles) and for ultimate limit
state design for 50 year reference period, the target reliabil-
ity index shall be ‡3.8. When class RC1 is considered,  = Conclusions
4.3 shall be taken. This paper reports and discusses the most significant con-
The results from the three approaches show the DA1 cor- tributions to the estimation of the side resistance of piles in
responds to  ‡ 3.8 except for the scale of fluctuation equal terms of total stresses by means of the a method. As for the
to 2.0 m related to normal distribution of cu and . This is not adhesion factor , there are a lot of available theories and
true for the last two approaches, which are DA2 and DA3. experimental relationships, and, despite of similar trends,
For DA2, all cases show  < 3.8. This means that for the differences in resulting  values are at times significant.
M1 = 1.0 and R2 = 1.1 the safety margin reduces too This is obviously because the relationships drawn from ex-
much, thus resulting in an unsatisfactory reliability level. perimental data are often connected to particular types of
For DA3, the correlation coefficient cu = –0.5 only guar- clay and to particular testing apparatus while theoretical re-
antees the respect of the minimum  value of 3.8 for the lationships often suffer from excessive simplifications.
three scales of fluctuation and the two probability distribu- Therefore, it would be necessary, in the opinion of the au-
tions. In this case, the  values are much higher than 3.8, thors, to create a database of experimental data to evaluate
which vary from 4.4 to 5.7 according to different scales of the side resistance to carry out statistical analyses based on
fluctuation for normal and lognormal distributions. As in the a significant number of data and relative to the widest possi-
previous cases the lower  values are registered for the scale ble spectrum of different types of clay.
# 2007 NRC Canada
1388 Can. Geotech. J. Vol. 44, 2007

Table 8. Reliability index of side bearing capacity according to the limit state design DA1 of bored piles in Matera
clay for two cases of cu normally and lognormally distributed.

DA1 P (kN)  Norm  Lognorm


Case1: cu = 0  = 0.5 m  = 1.0 m  = 2.0 m  = 0.5 m  = 1.0 m  = 2.0 m
Pl (0.4 m; 15 m) 793 4.1 3.7 3.1 4.9 4.3 3.6
Pl (0.4 m; 20 m) 1058 4.1 3.7 3.1 4.9 4.3 3.6
Pl (0.6 m; 15 m) 1190 4.1 3.7 3.1 4.9 4.3 3.6
Pl (0.6 m; 20 m) 1587 4.1 3.7 3.1 4.9 4.3 3.5
Case2: cu = –0.5
P2 (0.4 m; 15 m) 793 5.0 5.0 4.1 6.5 6.0 5.0
P2 (0.4 m; 20 m) 1058 5.0 5.0 4.1 6.5 6.0 5.0
P2 (0.6 m; 15 m) 1190 5.0 5.0 4.1 6.5 6.0 5.0
P2 (0.6 m; 20 m) 1587 5.1 5.0 4.1 6.5 6.0 5.0

Table 9. Reliability index of side bearing capacity according to the limit state design DA2 of bored piles in Matera
clay for two cases of cu normally and lognormally distributed.

DA2 P (kN)  Norm  Lognorm


Case1: cu = 0  = 0.5 m  = 1.0 m  = 2.0 m  = 0.5 m  = 1.0 m  = 2.0 m
Pl (0.4 m; 15 m) 1010 2.6 2.3 1.9 2.7 2.4 2.1
Pl (0.4 m; 20 m) 1346 2.6 2.3 1.9 2.8 2.5 2.0
Pl (0.6 m; 15 m) 1515 2.6 2.4 1.9 2.8 2.5 2.0
Pl (0.6 m; 20 m) 2020 2.6 2.4 1.9 2.8 2.5 2.0
Case2: cu = –0.5
P2 (0.4 m; 15 m) 1010 3.3 3.3 2.8 3.9 3.5 2.9
P2 (0.4 m; 20 m) 1346 3.3 3.2 2.7 3.8 3.5 2.8
P2 (0.6 m; 15 m) 1515 3.3 3.2 2.6 3.7 3.4 2.6
P2 (0.6 m; 20 m) 2020 3.3 3.2 2.7 3.8 3.5 2.8

Table 10. Reliability index of side bearing capacity according to the limit state design DA3 of bored piles in Matera
clay for two cases of cu normally and lognormally distributed.

DA3 P (kN)  Norm  Lognorm


Case1: cu = 0  = 0.5 m  = 1.0 m  = 2.0 m  = 0.5 m  = 1.0 m  = 2.0 m
Pl (0.4 m; 15 m) 854 3.7 3.4 2.8 4.3 3.8 3.1
Pl (0.4 m; 20 m) 1139 3.7 3.3 2.7 4.3 3.7 3.1
Pl (0.6 m; 15 m) 1282 3.7 3.4 2.8 4.3 3.7 3.1
Pl (0.6 m; 20 m) 1709 3.7 3.3 2.7 4.2 3.7 3.1
Case2: cu = –0.5
P2 (0.4 m; 15 m) 854 4.5 4.5 3.7 5.7 5.2 4.4
P2 (0.4 m; 20 m) 1139 4.5 4.5 3.7 5.7 5.2 4.4
P2 (0.6 m; 15 m) 1282 4.5 4.5 3.7 5.7 5.2 4.4
P2 (0.6 m; 20 m) 1709 4.5 4.5 3.7 5.7 5.3 4.4

Table 11. Allowable side bearing capacity of With regards to the reliability analyses developed here, a
bored piles in Matera clay calculated according large range of possible side loads and related reliability in-
to the a method divided by the global safety dexes were investigated according to the criteria proposed
factor from Geotechnical Italian Provisions. by Eurocode 7. The partial safety factor approach with the
reliability-based design approach was compared in terms of
P (kN) divided by FS = 2.5 reliability index  from the analyses performed on bored
P (0.4 m; 15 m) 565 piles in Matera clay. The values of  were provided accord-
P (0.4 m; 20 m) 753 ing to different probability distributions, correlation coeffi-
P (0.6 m; 15 m) 847 cients, and scales of fluctuation for cu and . Differences in
P (0.6 m; 20 m) 1130 the reliability level can be drawn whenever cu and  are
considered as normal or lognormal distributions. Those dif-
ferences are lower for lower  values and in the case of log-
normal distribution. Furthermore, the level of reliability is
reduced when the scale of fluctuation increases.
# 2007 NRC Canada
Cherubini and Vessia 1389

The comparison between reliability and the partial safety tions. In Proceedings of the 9th International Conference on
factor approaches in pile design suggests that one be careful Structural Safety and Reliability, Rome, Italy, 19–23 June 2005.
when poor statistical details are given on design variables. Mill Press, Rotterdam, the Netherlands. pp. 931–936.
This warning comes from the evidence that  values for Coduto, D.P. 1994. Foundation design, principles and practices.
side resistance are influenced by the scale of fluctuation, Prentice Hall Inc., Englewood Cliffs, N.Y.
the type of distribution, and the correlation coefficient of cu Dennis, N.D., and Olson, R.E. 1983. Axial capacity of steel pipe
and  variables, which cannot be taken into account when piles in clay. In Proceedings of Geotechnical Practice in Off-
characteristic values are calculated. shore Engineering, Austin, Texas. Edited by Stephen Wright.
American Society of Civil Engineers, New York. pp. 370–388.
Moreover, the Italian Geotechnical Provisions, based on
Decreto Ministero dei Lavori Pubblici 11 marzo. 1988 (D.M.
the global safety factor approach, was compared with the re- LL.PP.). Norme tecniche riguardanti le indagini sui terreni e
liability study and the partial safety factor approaches. This sulle rocce, la stabilità dei pendii naturali e delle scarpate, i cri-
approach has been found to be highly conservative with re- teri generali e le prescrizioni per la progettazione, l’esecuzione e
spect to all other approaches. il collaudo delle opere di sostegno delle terre e delle opere di
fondazione. Istruzioni per l’applicazione’’, GU 1.6.1988, n. 127
suppl.
References ERTC3 1999. European Regional Technical Committee « Piles »
American Petroleum Institute (API). 1974. Recommended practice Survey report on the present-day design methods for axially
for planning, designing and constructing fixed offshore platforms, loaded piles. European practice, Published at the occasion of
API RP2A, 5th ed. American Petroleum Institute, Dallas, Tex. the XII ECSMFE, Amsterdam, the Netherlands.
American Petroleum Institute (API). 1993. Recommended practice Eurocode 1 UNI EN 1990:2002. Basic of structural design. UNI,
for planning, designing and constructing fixed offshore plat- Italy.
forms, 20th ed. American Petroleum Institute, Dallas, Tex. Eurocode 7 UNI ENV 1997:2004. Geotechnical design – Part 1:
Baecher, G.B., and Christian, J.T. 2003. Reliability and Statistics in general rules. UNI, Italy.
Geotechnical Engineering. John Wiley & Sons Ltd., UK. Hammit, G.M. 1966. Statistical analysis of data from comparative
Balasubramaniam, A.S., Photo-Yannvat, C., Gaseshanathan, R., and laboratory test program sponsored by ACIL. U.S. Army Engi-
Lee, K.K. 1981. Performance of friction piles in Bangkok subsoils. neer Waterways Experiment Station, Vicksburg, Miss., Miscella-
In Proceedings of the 10th International Conference on Soil Me- nous Paper 4-785.
chanics and Foundation Engineering, Stockholm, Sweden, June Hasofer, A.M., and Lind, N.C. 1974. An Exact and invariant first-
1981. pp. 605–610. order reliability format. Journal of the Engineering Mechanics
Baldassarre, G., Cherubini, C., and Monterisi, L. 1993. The behaviour Division, 100 (EM1): 111–121.
of bored piles subjected to load testing as compared to theoretical Johnston, M.M. 1969. Laboratory comparison tests using com-
estimates. In Proceedings of the 2nd International Geotechnical pacted fine-grained soils. In Proceedings of the 7th International
Senior on deep foundations and auger piles, Ghent, Belgium, June Conference on Soil Mechanics and Foundation Engineering,
1993. A.A. Balkema, the Netherlands. pp. 223–226. Mexico City, Mexico, 25–29 August 1969. Vol. 1, pp. 197–202.
Battaglio, M., and Lancellotta, R. 1985. Pali di fondazione nei ter- Kay, J.N. 1993. Probabilistic design of foundations and earth struc-
reni coesivi. In Atti delle conferenze di Geotecnica di Torino, tures. In Proceedings of the Conference on Probabilistic Meth-
XII Ciclo. ods in Geotechnical Engineering, Canberra, Australia, February
Breitung, K. 1984. Asymptotic approximations for multinormal in- 1993. Edited by K.S. Li and S.-C.R. Lo. A.A. Balkema, Rotter-
tegrals. Journal of Engineering Mechanics, 110: 357–366. dam, the Netherlands. pp. 49–62.
Broms, B. 1981. Pile foundations. General report session 8. In Pro- Kerisel, J. 1965. Vertical and horizontal bearing capacity of deep
ceedings of the 10th International Conference on Soil Mechanics foundation in clay. In Proceeding of Symposium on Bearing Ca-
and Foundation Engineering, Stockholm, Sweden, June 1981. pacity of Settlement of Foundations, Duke University, Durham,
pp. 113–125. N.C. pp. 45–52.
Cafaro, F., and Cherubini, C. 2002. Large sample spacing in eva- Kezdi, A., and Rethati, L. 1988. Handbook of soil mechanics.
luation of vertical strength variability of clayey soil. Journal of Vol. 3. Elsevier, Amsterdam, the Netherlands.
Geotechnical and Geoenvironmental Engineering, 128: 558–568. Kolk, J., and Van der Velde, E. 1996. A reliable method to de-
Chen, Y.J., and Kulhawy, F.H. 1993. Undrained strength interrela- termine friction capacity of piles driven into clays. OTC,
tionships among CIUC, UU and CU tests. Journal of Geotechni- Houston, Tex.
cal Engineering, 119: 1732–1749. doi:10.1061/(ASCE)0733- Kraft, L.M., Focht, J.A., and Amerasinghe, S.F. 1981. Friction ca-
9410(1993)119:11(1732). pacity of piles driven into clay. Journal of the Geotechnical En-
Cherubini, C. 1997. Data and considerations on the variability of gineering Division, 107: 1521–1541.
geotechnical properties of soils. In. Proceedings of the Conference Kulhawy, F.H., and Jackson, C.S. 1989. Some observations on un-
on Advances in Safety and Reliability, ESREL, Lisbon, Portugal, drained side resistance of drilled shafts. Foundation Engineering:
17–20 June 1997. Pergamon, Elsevier Science Ltd., Oxford, UK. current principles and practices Congress Evanston Illinois, 2:
pp. 1583–1591. 1011–1025.
Cherubini, C. 1999. Discussion on the paper of Y. Robert 1997. A Lacasse, S., and Boisard, P. 1994. Consequence of the new API
few comments on pile design. Canadian Geotechnical Journal, RP2A guideline for piles in soft clays. In Proceedings of the
35: 905. 13th International Conference on Soil Mechanics and Founda-
Cherubini, C., Giasi, C.I., and Lupo, M. 2005. Interpretation of tion Engineering, New Delhi, India, 5–10 January 1994. Edited
load tests on bored piles in the city of Matera. Journal of Geolo- by S. Saxena. Oxford & IHB Publishing Co. Ptv. Ltd., New
gical and Geotechnical Engineering, 23: 349–364. Delhi, India. pp. 527–530.
Cherubini, C., and Vessia, G. 2005. The bearing capacity of piles Lancellotta, R., and Calavera, J. 1999. Fondazioni. McGraw-Hill,
evaluated by means of load tests according to reliability calcula- Libri, Italy.
# 2007 NRC Canada
1390 Can. Geotech. J. Vol. 44, 2007

Leporati, E. 1979. The assessment of structural safety. Research Conference Houston, Houston, Tex. Paper OTC 1718. pp. 865–
Studies Press, Oregon, Ore. 874.
McCarthy, D.R. 1977. Essentials of soil mechanics and founda- Weltman, A.J., and Healy, P.R. 1978. Piling in boulder clay and
tions. Reston Publishing Company, Inc., Reston, Va. other glacial tills. Construction Industry Research and Informa-
Minty, E.J., Smith, R.B., and Pratt, D.N. 1979. Interlaboratory test- tion Association, Report PG5.
ing variability assessed for a wide range of N.S.W. soil types. In Withiam, J.L., Vojtko, E.P., Baker, R.M., Duncan, J.M., Kelly,
Proceedings of the 3rd International Conference on Applications B.C., Musser, S.C., and Elias, V. 1997. Load and resistance fac-
of Statistics and Probability in Soil and Structural Engineering, tor design (LRFD) for highway bridge substructures (participant
Sydney, Australia, 29 January – 2 February 1979. Edited by workbook). Federal Highway Administration, Washington, D.C.
D.G. Ingles. Pink Panter Printery, Sydney, Australia. Vol. 1, Woodward, R., and Boitano, J. 1961. Pile loading tests in stiff
pp. 221–235. clays. In Proceedings of the 5th International Conference on
Nowacki, F., Kalsrud, K., and Sparrevik, P. 1992. Comparison of Soils Mechanics and Foundation Engineering, Paris, France,
recent tests on OC clay and implications for design. In Proceed- Vol. 2, pp. 177–189.
ings of Recent large scale full instrumented pile tests in clay.
Thomas Telford Ltd., London, UK. Norwegian Geotechnical In- List of Symbols and Abbreviations
stitute, Paper 22, pp. 511–538.
Orr, T.L.L., and Cherubini, C. 1999. Considerations on the applic- AGI associazione Geotecnica Italiana
ability of semi-probabilistic Bayesian methods to geotechnical de- ca adhesion
sign. In XX Convegno Nazionale di Geotecnica, Parma, Italy, 22– cu undrained shear strength
25 September 1992. Pàtron, Bologna, Italy. pp. 421–426. cuk characteristic value of undrained shear strength
cu mean value of undrained shear strength
Peck, R.B. 1958. A study of the comparative behaviour of friction
CIUC consolidated isotropically undrained compression
piles. National Research Council, Washington, D.C. Highway
COV coefficient of variation
Research Board, Special Report 36. CU consolidated undrained
Phoon, K.K., and Kulhawy, F.H. 1999. Characterization of geo- D diameter of pile
technical variability. Canadian Geotechnical Journal, 36: 612– Fc converting factor
624. doi:10.1139/cgj-36-4-612. FL correct factor
Phoon, K.K., and Kulhawy, F.H. 2005. Characterization of model FORM first-order reliability method
uncertainties for drilled shafts under undrained axial loading. FS global safety factor
GSP 131 Contemporary Issues in Foundation Engineering, IP plasticity index
ASCE. L length of pile
Rethati, L. 1988. Probabilistic solutions in geotechnics. In Develop- pa atmospheric pressure
ments in Geotechnical Engineering, Elsevier, Vol. 46. Pf probability of failure
Schneider, H.R. 1997. Definition and determination of characteris- PF Performance function
tic soil properties. Contribution to Discussion, In Proceedings of Pl side resistance of pile
the 14th International Conference on Soil Mechanics and Foun- qs side resistance
dation Engineering, Hamburg, Germany. A.A. Balkema, the Q load
Netherlands. pp. 2271–2274. Ql side load
Qs total side resistance
Semple, J., and Rigden, M. 1984. Shaft capacity of driven pipe
R resistance
piles in clay. In Symposium of analysis and design of pile foun-
S side condition factors
dation, ASCE, San Francisco, Calif. SM safety margin
Sherwood, P.T. 1970. Reproducibility of results of soil classifica- SORM second-order reliability method
tion and compaction tests. Road Research Laboratory, STRUREL structural reliability analysis code
Crowthorne, UK. Report LR 339. UU unconsolidated undrained
Singh, A., and Lee, K.L. 1970. Variability in soil parameters. In  adhesion factor
Proceedings of the 8th Annual Engineering Geology and Soils  mean value of adhesion factor
Engineering Symposium, Idaho. pp. 159–185.  reliability index
Skempton, A.W. 1959. Cast in situ bored piles in London Clay.  Lognorm reliability index for lognormal distribution
Géotechnique, 9: 153–173.  Norm reliability index for normal distribution
Sladen, J.A. 1992. The adhesion factor: applications and limita- R partial safety factor for pile resistance
tions. Canadian Geotechnical Journal, 29: 322–326. M1 partial safety factor for soil resistance for M1 con-
Sowers, G.B., and Sowers, G.F. 1970. Introductory soil mechanics dition
and foundations. The MacMillan Company, New York. M2 partial safety factor for soil resistance for M2 con-
STRUREL. 1996. A structural reliability analysis program: theore- dition
tical manual. RCP GmbH, Munchen, Germany. 2 variance reduction function
 scale of fluctuation
Tomlinson, M.J. 1957. The adhesion of piles driven in clay soils.
SM mean value of safety margin
In Proceedings of the 4th International Conference on Soils Me-  correlation coefficient
chanics and Foundation Engineering. Thomas Telford Ltd., Lon- cu correlation coefficient between  and cu
don, UK. Vol. 2, pp. 66–71.  total standard deviation
Tomlinson, M.J. 1977. Pile design and construction practice. View- h standard deviation of natural variability of soil
point Publication, Australia. rp standard deviation of measurement errors
Vanmarcke, E.H. 1983. Random fields: analysis and synthesis. MIT SM standard deviation of safety margin
Press, Cambridge, UK. 0
vo effective vertical stress
Vijayvergiya, V.N., and Focht, J.A. 1972. A new way to predict  cumulative distribution function of normal distribu-
capacity of piles in clay. In Proceeding Offshore Technology tion

# 2007 NRC Canada

You might also like