Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
6, 519±531
Limit state design applied to geotechnical structures requires La meÂthode de l'eÂtat limit appliqueÂe aux structures geÂotech-
the selection of characteristic values of ground parameters niques a besoin des valeurs caracteÂristiques des parameÁtres
and the use of partial safety factors. These factors, and the du massif et de l'utilisation des facteurs partiels de seÂcuriteÂ.
way that characteristic values are selected, will govern the Ces facteurs et la facËon comme les valeurs caracteÂristiques
calculated probability of failure of the structure. The paper sont seÂlectioneÂes deÂtermineront la probabilite de rupture de
presents a methodology for estimating the probability of la structure. Cet article preÂsente une meÂthodologie pour
failure provided by the application of Eurocode 7 to the estimer la probabilite de rupture fournie par l'application
design of spread foundations and embankments. A para- du Eurocode 7 au projet des fondations super®cielles. Un
metric study is presented enabling the identi®cation of the eÂtude parameÂtrique est presente permettant l'identi®cation
reduction factors by which the mean value of soil resistance des facteurs de reÂduction que doivent multiplier la valeur
parameters should be multiplied to obtain the characteristic moyenne des parameÁtres reÂsistants du sol pour obtenir la
value that will ensure a target probability of failure. valeur caracteÂristique assurant une probabilite ®xeÂe de rup-
ture.
KEYWORDS: bearing capacity; embankments; footings; limit state
design.
Table 1. Probability of failure and reliability index (adapted from Meyerhof (1993, 1995))
Structure Probability of failure Reliability index, â
Geotechnical works:
offshore foundations 1 3 10ÿ2 ±4 3 10ÿ3 2:3±2:7
earthworks 4 3 10ÿ3 ±1 3 10ÿ3 2:7±3:1
retaining structures 1 3 10ÿ3 ±4 3 10ÿ4 3:1±3:4
foundations 4 3 10ÿ4 ±1 3 10ÿ4 3:4±3:7
Reinforced concrete structures 5 3 10ÿ4 ±1 3 10ÿ5 3:3±4:3
Steel structures , 1 3 10ÿ4 . 3:7
Table 2. Partial factorsÐultimate limit states in persistent and transient situations (ENV 1997-1, 1994)
Case Actions Ground properties
Permanent Variable
VQ VQ VQ VQ
0´2 0´4 0´6 0´2 0´4 0´6 0´2 0´4 0´6 0´2 0´4 0´6
0´0 0´03 0´07 0´10 0´07 0´13 0´20 0´10 0´20 0´30 0´12 0´24 0´36
(1´06) (1´07) (1´09) (1´12) (1´14) (1´15) (1´17) (1´19) (1´20) (1´20) (1´21) (1´22)
0´05 0´05 0´08 0´11 0´07 0´14 0´20 0´10 0´20 0´30 0´12 0´24 0´36
(1´06) (1´07) (1´08) (1´11) (1´13) (1´14) (1´17) (1´18) (1´19) (1´19) (1´21) (1´22)
0´1 0´09 0´11 0´13 0´09 0´15 0´21 0´11 0´21 0´30 0´13 0´24 0´36
(1´06) (1´07) (1´08) (1´11) (1´12) (1´14) (1´16) (1´18) (1´19) (1´19) (1´20) (1´22)
0´15 0´13 0´14 0´16 0´12 0´17 0´22 0´13 0´21 0´31 0´13 0´25 0´36
(1´05) (1´06) (1´07) (1´10) (1´12) (1´13) (1´15) (1´17) (1´18) (1´18) (1´20) (1´21)
GROUND PARAMETERS AND FAILURE PROBABILITY IN DESIGN TO EC 7 523
relation to the mean value (high variability), and a high value Evaluation of the mean value and coef®cient of variation of
represents a small variation of the property (low variability). R. With regard to the resistance, the mean value, R m , and the
The evaluation of the scale of ¯uctuation requires a large standard deviation, ó R , are involved in the equations de®ning the
number of tests performed at several points in the soil mass. reliability index because V R ó R =R m . In the next few lines, a
Only thus is it possible to de®ne the variance function from presentation is made of some methodologies for evaluating those
which that parameter is determined (Wickremesinghe & Campa- parameters based on the corresponding parameters for the shear
nella, 1993). In the case of the evaluation of the undrained strength properties on which R depends.
shear strength a large number of tests are normally carried out. According to the classical equations for the bearing capacity
However, the same does not usually occur when the effective of shallow foundations, the resistance may result from various
shear strength parameters are to be evaluated. random parameters, such as geometrical data, unit weight, and
Yu & Mostyn (1993) indicate values of 0´32±0´67, 0´61, 1´02 strength properties of the ground. For the sake of simplicity, in
and 1´56 m for the scale of ¯uctuation of the undrained shear EC 7 the uncertainties associated with the geometrical data and
strength of four clay deposits and a value of 2´1 m for the with unit weight are normally not considered directly but are
drained shear strength of a clay. Thorne & Quine (1993) included, indirectly, in the selection of the characteristic values
mention that the scale of ¯uctuation very often ranges from 1 to of the properties of materials and actions.
3 m, in spite of being quite different from deposit to deposit, Therefore it can reasonably be assumed that the main source
and that it is typically lower in the vertical direction than in the of uncertainty of R results from the variability of the parameters
horizontal one. concerning the shear strength. Thus, considering geometrical
Let us assume that, for a speci®c problem, it is possible to data and unit weight as deterministic variables, the resistance of
de®ne the dimensions of the zone of the ground involved in the the foundation becomes a function of a single random variable,
structural behaviour, and that MX is the random variable which is the mean value of the governing strength parameter in
representing the mean value of the values of X in that zone. the zone of soil that affects the structural behaviour, as de®ned
Naturally MX m X m but ó MX depends not only on ó X but in the previous section. In this case the probability density
also on the dimension of the zone under consideration: the function of R, f R (r), is related to that of MX , f MX (x), as
larger this dimension, the smaller ó MX is. Various forms for follows (Benjamin & Cornell, 1970):
ó MX ó MX (ó X , dimensions) have been proposed, which have R g(X ) ) f R (r) dr f MX (x) dx (17)
been deduced from different theories of autocorrelation. Never-
theless, for most applications it can be assumed that (Thorne & This relation means that the probability of R taking a value
Quine, 1993) in the interval of width dr centred on r is equal to the
ó MX ó X (ð i ä Xi =Li )1=2 (15) probability of X being situated in the interval centred on the
corresponding value x g ÿ1 (r), but with width equal to
where the zone involved is de®ned by the dimensions Li , and dx d g ÿ1 (r). By using this relation it is possible to de®ne the
ä X i is the scale of ¯uctuation in the corresponding directions. function f R when f MX is known, as is illustrated in Fig. 1.
The dimensions of the soil involved in the previous equations Subsequently, it is simple to calculate the parameters referred to
have to be carefully selected. By taking into account the above.
considerations presented by Bauduin (1998) on this issue, two Another possibility consists of determining directly the mean
types of situation can be considered: value and the standard deviation of R, using a Taylor-series
expansion about the mean value MX m to approach the function
(a) The limit state is governed by the mean value (or overall
g(x):
mean) of all the ground affected by the structure if, on the
X g (i) (MX m )(x ÿ MX m )i
one hand, the volume of the ground affected is so great that g(x) ' (18)
the in¯uence of weak spots is compensated by stiffer zones i
i!
and if, on the other hand, the structure has a capacity to
in which g (i) (MX m ) represents the value of the ith derivative
redistribute loads from the weak spots to the stiffer zones.
when x MX m . By using only the ®rst three terms of Taylor-
(b) The limit state is governed by a low value (or local mean)
series expansion, one obtains
when the limit state may be governed by weaker spots or
the structural failure may occur before load redistribution g (2) (MX m )ó MX 2
has taken place. R m g(MX m ) (19)
2
Assuming, for spread footings, that the dimensions of the ó2R [ g (1) (MX m )ó MX ]2 g (1) (MX m ) g (2) (MX m )ì MX (3)
zone that controls the potential structural collapse depend on 2
the dimensions, B and L, of the foundation and not on the g(2) (MX m )
dimensions of all the ground affected by the structure, which ( ì MX (4) ÿ ó MX 4 ) (20)
2
represents a cautious option, then it is reasonable to consider
Lvertical Nv B (16a) where ì MX ( n) being the nth central moment.
This approach usually leads to good results when the coef®-
Ltransverse Nt B (16b) cient of variation V MX is small. For large values of this coef®-
Llongitudinal L (16c) cient, the applicability of the procedure depends on the form of
the function g(x), and when the latter is signi®cantly non-linear,
in which Nv and Nt depend on the angle of shearing resistance it depends on the number of terms of the Taylor-series expan-
(Meyerhof, 1948). For the 2D and 3D situations, Table 4 shows sion. Note also that when many terms of that expansion are
equations that make it possible to evaluate ó MX =ó X for various used, it becomes necessary to know the high level moments and
cases, by assuming that the scales of ¯uctuation are equal for central moments of the random variable MX , which is not
all the directions and are de®ned by ä X . always possible to achieve.
r = g(x)
x = g –1(r)
dr
P [r ≤ R ≤ n 1 dr ] = fR(r)dr
dr
=
1
dx dx
dr
Equal areas
dx
dr = dx
dr
fR(r)
fMX(x)
fR(r) x
Fig. 1. Determination of f R (r) when f MX (x) and r g(x) are known (Benjamin & Cornell, 1970)
Determination of the reliability index linearly on cu , the values of R m and ó R may be calculated
Bearing in mind the previous developments, the proposed exactly from the ®rst terms of equations (19) and (20):
approach to obtain the reliability index can ®nally be sum- R m g(Mcum ) (r 1)sc (2 ð) BL Mcum (23)
marised. Considering the values of V G and ãG for permanent
(1)
actions, those of V Qi and ãQ for variable actions, and the values ó R g (Mcum ) sc (2 ð) BLó Mcu (24)
of X mi , V X i , ãX i and á X i for the shear strength properties, the V R ó R =R m V Mc u =(r 1) (25)
calculation sequence is as follows:
where Mcum , ó Mcu and V Mcu are, respectively, the mean value,
(a) Determine V E from equation (13), using V G and V Qi , and
the standard deviation and the coef®cient of variation of the
compute ãE from equation (14), using ãG and ãQ .
variable Mcu , which represents the average of the values of cu
(b) Calculate á E using one of equations (9) as a function of
in the zone relevant for design, and
VE.
(c) Select MX mi and V MX i , considering the values of X mi and q rsc (2 ð)Mcum (26)
V X i and the observations presented above.
From equation (3), the design value of the resistance is given
(d ) Determine R d from MX mi , ãX i and á X i , using equation (3).
by
(e) Evaluate Em from R d , ãE and á E, using equation (4). In
principle, æ 1. R d (r á c u =ãc u )sc (2 ð) BL Mcum (27)
( f ) Determine R m and V R from MX mi and V MX i , using the
Therefore, introducing equations (23) and (27) in equation (4),
methods discussed above.
one obtains
( g) Finally, compute the reliability index, â, from equations (8),
using Em , V E , R m and V R . r á c u =ãc u
Em =R m (28)
æãE á E (r 1)
By assuming that R and E follow a normal distribution, equa-
EXAMPLES OF APPLICATION tion (9a) yields á E and, lastly, equation (8a) leads to the
Total stress stability analyses reliability index. In the case of a log-normal distribution, equa-
Governing equations. In this paragraph the proposed tions (9b) and (8b), respectively, should be used instead.
methodology is applied to the stability analysis of embank-
ments and spread foundations on clayey ground using total Application to embankments on soft soils. In the case of
stresses. The bearing equation for those cases is embankments on soft soils, q is usually zero (and so r 0), and
R [q sc (2 ð)cu ] BL (21) since the permanent loads signi®cantly exceed the live loads, it
can be assumed, in a simpli®ed way, that Qi 0; consequently,
where V E V G and ãE ãG (equations (13) and (14) respectively). On
sc 1 0:2 B=L (22) the other hand, the coef®cient of variation of the action effect,
which is signi®cantly controlled by the dead load of the ®ll, can
and q is the total vertical stress at the level of the foundation be assumed to be close to zero (V E 0), from which, assuming
base. a normal distribution and using equation (9), it follows that
If the geometrical data and unit weight of the ground are á E 1. Therefore, from equation (28), it can be deduced that
assumed to be deterministic parameters, the undrained shear
strength, cu , remains the sole random variable in the function Em ácu
(29)
expressing the resistance. On the other hand, since R depends R m æãG ãc u
GROUND PARAMETERS AND FAILURE PROBABILITY IN DESIGN TO EC 7 525
For normal and log-normal distributions, the reliability index is 1·0
then obtained, respectively, as
1 ÿ á c u =(æãG ãc u )
â (30a) 0·9
V Mcu
reliability index should be higher than 3´4±3´7 (see Table 1). log-normal distributions and signi®cantly lower, usually
From Table 6 the following conclusions for spread foundations under 0´5, for normal distributions.
on clayey soils can be extracted: (e) Values of V Mcu higher than 0´2 are likely to prevent the
adoption of spread footings, unless in soils with high shear
(a) The in¯uence of the depth of the foundation base on the strength; fortunately, in very stiff or hard soils such high
reliability index, represented by r, is negligible. values of the coef®cient of variation are usually not found.
(b) The type of distribution adopted has a signi®cant in¯uence
on the reliability index; in the case of a log-normal Using the values of Table 6, the reduction factors were
distribution, the value of â is less sensitive to the coef®cient evaluated assuming â 3:55 and r 0:05. In Fig. 3 these
of variation of the undrained shear strength, V Mcu , than in values are compared with Schneider's proposal (equation (31)).
the case of a normal distribution. It appears that this proposal varies from slightly conservative to
(c) As regards the variability of the action effect, represented quite conservative for log-normal distributions and for normal
by V E, it can be veri®ed that the reliability index is affected distributions when V c u is less than 0´4 (V Mcu , 0:12). For the
by the value of V Mcu : the higher the value of V E , the less it latter distribution, and for V c u higher than 0´4, the proposal
is affected. This conclusion is supported by the fact that seems to be unsafe in some cases.
higher partial safety factors of the action effect, ãE , are
associated with higher values of V E .
Effective stress (drained) stability analyses
From a practical point of view, it is important to emphasise Governing equations. Assuming a cohesionless soil as the
the following conclusions deduced from the analysis of Table 6: bearing ground, the resistance is expressed by
R (q9N q s q 0:5Bã9Nãsã) BL (32)
(a) If V Mcu is about 0´05 a characteristic value of undrained
shear strength equal to the mean value obtained in the tests where
can be used in the design.
(b) When V Mcu is about 0´1 the mean value obtained in the tests N q e ð tan ö9 tan2 (458 ö9=2) (33a)
should be multiplied by 0´9 so as to obtain the characteristic Nã 2(N q ÿ 1) tan ö9 (33b)
value.
s q 1 sin ö9(B=L) (34a)
(c) If V Mcu is about 0´15, the mean value from the test results
should be multiplied by 0´7±0´8 in order to obtain the sã 1 ÿ 0:3(B=L) (34b)
characteristic value in structures in which permanent loads
prevail, and by 0´8±0´9 in structures subjected to signi®cant In the following calculations the variables B, q9 and ã9 are
variable actions. assumed to be deterministic. On the other hand, the effective
(d ) When the coef®cient of variation obtained from test results, stress on the horizontal plane at foundation level is taken as
V cu , is particularly high and/or the dimension B is low, q9 rBã9sã M tan ö9m (35)
which leads to values of V Mcu of about 0´20 or higher, there
is a large in¯uence of the type of statistical distribution on where M tan ö9m is the mean value of the variable M tan ö9,
the resistance and on the action effect; in fact, for which represents the average of the values of tan ö9 in the zone
V Mcu 0:20 the coef®cient of conversion of the mean of the soil mass affecting the foundation design. Introducing
value into the characteristic value should be 0´6±0´7, for equations (35) and (33b) in equation (32), we obtain
GROUND PARAMETERS AND FAILURE PROBABILITY IN DESIGN TO EC 7 527
1·0
VE γE
0·0/1·0
0·9
Maximum reduction factor, αcu
0·1/1·0
0·8
0·2/1·1
0·7
0·3/1·2
0·6
Schneider
0·5
(a)
1·0 VE γE
0·0/1·0
0·9
Maximum reduction factor, αcu
0·1/1·0
0·8
0·2/1·1
0·7
0·3/1·2
0·6
Schneider
0·5
0 0·1 0·2 0·3 0·4 0·5 0·6 0·7
Coefficient of variation of the test results, Vcu
(b)
Fig. 3. Spread foundations on clayey soils: reduction factors (V Mcu 0:3 V cu ) for: (a) normal distribution;
(b) log-normal distribution
(N q ÿ 1)M tan ö9 function R will exhibit a distribution not far from a log-normal
R rs q N q à (36)
M tan ö9m one. Then, in the following calculations, normal and log-normal
distributions are assumed for M tan ö9 and R respectively. The
in which latter assumption requires that the log-normal distribution is
à B2 Lã9sã M tan ö9m (37) adjusted to the actual distribution of R obtained from the
(normal) distribution of M tan ö9.
Taking account of equation (3), the design value of R is As the distribution of R is known, it is possible to determine
expressed by R m , ó R and V R, which are used for obtaining the reliability
R d [rs qd N qd (N qd ÿ 1)átan ö9 =ãtan ö9 ] Ã (38) index. This procedure naturally implies that a log-normal dis-
tribution of the action effect is also assumed.
where N qd and s qd are calculated on the basis of equations
(33a) and (34a) respectively, by using
ö9d a tan(átan ö9 M tan ö9m =ãtan ö9 ) (39) Application to spread footings. Taking account of equation
(35), r 1 corresponds to a footing embedded between 0´4B
Then, taking into consideration equation (4), we obtain and 0´6B for ö9 308, and between 0´7B and 1´0B for ö9 458.
Em Rd In the calculations presented below r is taken equal to 0 or 1.
(40) For the coef®cient of variation, V E , and the partial safety factor
Rm æã E á E R m of the effect of the actions, ãE , the same pairs of values adopted
Owing to the assumption that R and E have log-normal above have been considered.
distributions, á E is calculated by equation (9b) and, ®nally, â is According to Table 10 in Appendix 1, it is reasonable to
obtained from equation (8b). expect the coef®cient of variation of tan ö9 to be lower than
Since the function R g(M tan ö9) is highly non-linear, in 0´20 and 0´15 for ö9 equal to 308 and 458 respectively. Consid-
contrast to what has been assumed above, the determination of ering square footings with B 2:0 m and the ö9 equal to 308
R m and ó R from equations (19) and (20) is not feasible for the and 458 respectively. Considering square footings with
present case. Then, as discussed below, a methodology based on B 2:0 m and the scale of ¯uctuation of the ground also equal
the procedure illustrated by Fig. 1 has to be applied. to 2:0 m, the coef®cient of variation of the mean value,
Owing to the equation that de®nes the factor N q, the function V M tan ö9 , in the zone affecting the behaviour of the foundation
R g(M tan ö9) is controlled by the exponential term eð tan ö9 . is then (see Table 4) lower than or equal to 0´1 and 0´05 for ö9
Therefore, if the variable M tan ö9 has a normal distribution, the equal to 308 and 458 respectively. The values of these para-
528 CARDOSO AND FERNANDES
meters will increase for footings of smaller dimensions and for (VMtan ö9 , 0:1), and seems to be optimistic when Vtan ö9 is
higher scales of ¯uctuation, according to Table 4. higher than 0´2.
Table 7 includes the values of the coef®cient of reliability, â,
obtained for the ranges of ö9 and V M tan ö9 mentioned above and
for the reduction factor, á M tan ö9 , in the range 1´0±0´5.
CONCLUSIONS
For spread footings the target value of the reliability index
A methodology has been presented for the evaluation of the
should be above 3´4±3´7 (see Table 1). Analysis of this Table 7
probability of failure provided by the application of Eurocode 7.
suggests the following comments regarding the reliability of
If a given target probability of failure, corresponding to
spread foundations under drained conditions:
socially acceptable risk, is assumed for each type of geotechni-
(a) The in¯uence of the depth of the base of the foundation is cal structure, then the safety provided by Eurocodes must
not relevant. comply with these values. This compliance depends not only on
(b) The greater V E , the smaller is the in¯uence of V M tan ö9 on the values of the partial safety factors for the actions and
the reliability index. This general conclusion is corroborated ground properties but also on the de®nition of characteristic
by the fact that greater values of V E are associated with values of the resistant parameters of the soil.
values of the partial safety factor on the action effects, ãE , On the basis of a parametric study for embankments on soft
that are also high. ground and spread foundations on cohesive or non-cohesive
(c) For any pair of values of the coef®cients of variation soils, it was possible to evaluate the reduction factors by which
(V M tan ö9 and V E ), the values of the reliability index are the mean value of strength parameters (cu or tan ö9) should be
similar, almost independent of the value of M tan ö9. multiplied to obtain characteristic values that will ensure that
the target probability of failure is achieved.
This ®nding leads to the following practical conclusions, which For embankments on soft ground it was possible to conclude
are independent of the value of the angle of shearing resistance that, for common situations, the reduction factor can be taken
of the soil: as equal to 1´0: that is, the characteristic value can be taken as
equal to the mean value. Furthermore, the reduction factor will
(a) If V M tan ö9 is of the order of 0´05, a characteristic value of be lower than the range 0´8±0´9 only in quite uncommon cases.
tan ö9 equal to the mean value obtained from tests can be As expected, for spread foundations on clayey soils reduction
used in the design. factors are highly dependent on the coef®cient of variation of
(b) If V M tan ö9 is around 0´1, the characteristic value should be cu , especially when this is larger than 0´3, and, to a certain
obtained by multiplying the mean value by 0´8±0´9. extent, on the relative importance of the permanent and variable
(c) When VMtan ö9 is about 0´15, which, in principle, will occur loads.
only in cases of low angles of shearing resistance, the For spread foundations on typical non-cohesive soils
characteristic values should be adopted by multiplying the (Vtan ö9 , 0:15) the range 0´95±1´00 seems to be appropriate for
mean value by 0´6, for the case of structures where the reduction factor. When Vtan ö9 exceeds 0´15 the reduction
permanent loads are predominant, or by 0´7 when there are factor becomes dependent on the value of the angle of shearing
signi®cant live loads. resistance, decreasing as and when this angle reduces.
Concerning the simpli®ed equation proposed by Schneider
Reduction factors were evaluated through Table 7 assuming (1997) for evaluation of the reduction factor, it can be con-
â 3:55 and r 0:05. In Fig. 4 the values obtained are cluded that:
compared with Schneider's proposal. It appears that this propo-
sal is quite conservative for ö9 equal to 458; for ö9 equal to (a) it is generally quite conservative for embankments on soft
308, it is slightly conservative if Vtan ö9 is less than 0´2 ground
1·0
VE γE
30° 0·0/1·0
Maximum reduction factor, αtanφ
0·9
30° 0·1/1·0
30° 0·2/1·1
0·8
30° 0·3/1·2
45° 0·0/1·0
0·7
45° 0·1/1·0
45° 0·2/1·1
0·6
45° 0·3/1·2
Schneider
0·5
0 0·1 0·2 0·3 0·4
Coefficient of variation of the test results, Vtanφ′
Fig. 4. Spread foundations on non-cohesive soils: reduction factors (V M tan ö9 =Vtan ö9 0:5 and 0´25 for 308 and 458
respectively) for log-normal distribution
(b) it supplies estimates that vary from slightly to quite Table 9. Coef®cients of variation of live loads in buildings (Calgaro,
conservative for the most common cases of spread found- 1996)
ations Live loads in buildings VQ
(c) for this type of structure it may lead to some unsafe results
when the coef®cient of variation is high and the variables Houses 0´25
exhibit a normal distribution. Of®ces basements and ground-¯oor 0´37±0´54
other ¯oors 0´26±0´39
Furthermore, this study leads to the conclusion that the
in¯uence of the two types of statistical distribution considered, Shops pharmacies 0´16
which have quite different tails, is signi®cantly relevant only drugstores 0´22
megastores 0´32
when the coef®cient of variation of the ground resistance is
bookshops 0´38
high to very high. groceries 0´44
Finally, it should be mentioned that with the introduction of
minor changes the approach presented can be applied to any Warehouses pharmacies 0´18±0´21
other sets of partial safety factors for actions and ground drugstores 0´25±0´30
megastores 0´36±0´43
resistance properties. bookshops 0´43±0´51
groceries 0´50±0´60
Parking of cars 0´03±0´05
APPENDIX 1: COEFFICIENTS OF VARIATION OF ACTIONS AND
OF SHEAR STRENGTH PROPERTIES OF THE GROUND The ®rst value refers to large areas and the second one to areas of
COLLECTED FROM THE LITERATURE about 10 m2
In this appendix a presentation is made of the values of the coef®cients
of variation of the actions and resistant properties of the ground collected
from bibliography. EQi effect of variable action Qi
Tables 8 and 9 show, respectively, the values of the coef®cients of EQki , EQmi characteristic, mean value of EQi
variation of actions due to the weight of the ground and structure and of FU cumulative distribution function
surcharges (live loads) in buildings. G permanent action
Table 10 shows a summary of the values of the coef®cients of Gm mean value of G
variation of the strength properties of the soils, determined from L length of the foundation
laboratory and in situ tests. Lvertical , dimensions of the zone of ground governing the limit
Ltransverse , state
Llongitudinal
NOTATION MX i , Mcu , random variable representing the mean value of
B width of the foundation M tanö9 X i , cu , tan ö9 in the zone governing the limit state
E random variable representing the action effect MX mi , Mcum , mean value of MX i , Mcu , M tan ö9
Ed , Ek , Em design, characteristic, mean value of E M tan ö9m
Ed,dst , Ed,stb design value of the effect of destabilising and stabilising Nt , N v ratio between Ltransverse , Lvertical and B
actions N q , Nã bearing capacity factors
EG effect of permanent actions N qd bearing capacity factor calculated by ö9d
EGk , EGm characteristic, mean value of EG Pc reliability or con®dence
All the ranges 0´12±0´85 Hammitt (1966); Grolimund & Recordon (1972); Alonso (1976);
Fredlund & Dahlman (1977); Cassan (1979); Baghery (1980); Kulhawy (1992);
Cherubini et al. (1993); Meyerhof (1993, 1995)
(0´34) Becker (1996b)
ö9 , 308 0´03±0´15 Kulhawy et al. (1991); Cherubini (1992)
30±408 0´10±0´22
All the ranges 0´05±0´25 Schultze (1972); Singh (1972); Harr (1977); Baghery (1980); Kulhawy (1992);
Manoliu & Marcu (1993); Cherubini et al. (1993); Meyerhof (1993, 1995)
(0´13) Becker (1996b)
tan ö9 0´07±0´15
Lumb (1966); Schultze (1972); Alonso (1976); Kulhawy et al. (1991)
The mean value of V is indicated in brackets when available
Xi
Pf probability of collapse Limit States Design for Foundations. Part I. An overview of the
Qi variable action foundation design process. Can. Geotech. J. 33, 956±983.
Qmi mean value of Qi Becker, D. E. (1996b). Eighteenth Canadian Geotechnical Colloquium:
R random variable representing the resistance Limit States Design for Foundations. Part II. Development for the
R d , Rk , R m design, characteristic, mean value of R National Building Code of Canada. Can. Geotech. J. 33, 984±1007.
V E , VG , coef®cients of variation of E, G, X i , MX i , Mcu , Benjamin, J. R. & Cornell, C. A. (1970). Probability, statistics, and
V X i , V MX i , M tan ö9, Qi , R, cu decision for civil engineers. McGraw-Hill, New York.
V Mcu , V M tan ö9 , Bolton, M. D. (1993). What are partial safety factors for? Proceedings
V Qi , V R , V c u of the international symposium on limit state design in geotechnical
Xi random variable on which the resistance depends engineering, Copenhagen, vol. 3, pp. 565±583.
X di , X ki , X mi design, characteristic, mean value of X i Calgaro, J. A. (1996). Introduction aux Eurocodes. SeÂcurite des constru-
a G , a Qi parameter or function modelling the effect of G, Qi tions et bases de la theÂorie de la ®abiliteÂ. Paris: Presses de l'EÂcole
cu undrained shear strength Nationale des Ponts et ChausseÂes.
f E , f R , f MX probability density function of the random variables E, Cassan, M. (1979). Determination probabiliste des caracteÂristiques meÂ-
R, MX caniques des sols. ConfeÂrence prononceÂe aÁ l'EÂcole Centrale des Arts
f Ej R joint probability density function of the random variables et Manufactures. (Referred in Magnan (1982)).
E and R Cherubini, C. (1992). Collecting coef®cients of variation of some
q, q9 total and effective vertical stress at the level of the geotechnical properties. Unpublished report referred in Cherubini et
footing base al. (1993).
s c , s q , sã shape factors of the bearing capacity equation Cherubini, C., Giasi, C. I. & Rethati, L. (1993). The coef®cients of
Ëki , Ëmi ratio between EQi and EG de®ned by the respective variation of some geotechnical parameters. In Probabilistic methods
characteristic or mean values in geotechnical engineering (eds K. S. Li and S. C. R. Lo), pp.
á E , áG , coef®cients of conversion of E, G, Qi , X i 179±184 A. A. Balkema, Rotterdam.
á Qi , á X i ENV 1997-1 (1994). Eurocode 7: Geotechnical design. Part 1: General
â reliability index rules. Brussels: CEN, European Committee for Standardization.
äX i scale of ¯uctuation of X i ENV 1991-1 (1999). Eurocode 1: Basis of design and actions on
ö9 effective angle of shearing resistance structures. Part 1: Basis of design. Brussels: CEN, European Com-
ö9d design value of ö9 mittee for Standardization.
ãE , ãG , ãQ , partial safety factors for E, G, Q, X i , cu , tan ö9 Fredlund, D. G. & Dahlman, A. E. (1977). Statistical geotechnical
ãX i , ãcu , properties of glacial Lake Edmonton sediments. Proc. 1st Int. Conf.
ãtan ö9 Applications of Statistics and Probability in Soil and Structural
ã9 effective soil unit weight Engineering, Hong Kong, 203±228.
ì MX central moment Grolimund, J. P. & Recordon, E. (1972). EÂtude statistique sur les
r factor proportional to the vertical (total or effective) reÂsultats d'essais de laboratoire effectueÂs sur un sol consideÂreÂ
stress at the level of footing base comme homogeÁne. Laboratoire de GeÂotechnique de l'EÂcole Polytech-
ó R, ó X , standard deviation of R, X , MX , cu , Mcu , tan ö9 nique FeÂdeÂrale de Lausanne.
ó MX , ó cu , Hammitt, G. M. (1966). Statistical analysis of data from a comparative
ó Mcu , ótan ö9 laboratory test program sponsored by ACIL. Vicksburg: US Army
Ø ci coef®cient of combination of action effects Engineers Waterways Experiment Station.
æ ratio between R d and Ed Harr, M. E. (1977). Mechanics of particulate mediaÐa probabilistic
approach. McGraw-Hill, New York.
Kulhawy, F. H. (1992). On evaluation of static soil properties. Proceed-
ings of the ASCE specialty symp. stability and performance of slopes
REFERENCES and embankments, New York, vol. 2, pp. 95±115.
Alonso, E. (1976). Risk analysis of slopes and its application to slopes Kulhawy, F. H., Roth, N. J. S. & Grigoriu, N. B. (1991). Some statistical
in Canadian sensitive clays. GeÂotechnique 26, No. 3, 453±472. evaluations of geotechnical properties. Proc. 6th Int. Conf. Applica-
Baghery, S. (1980). ProbabiliteÂs et statistiques en meÂcanique des sols. tions of Statistics and Probability in Soil and Structural Engineering,
Analyse probabiliste de la stabilite et des tassements de remblais sur Mexico City, 705±712.
sols compressibles (Site expeÂrimental de Cubzac-les-Ponts). TheÁse de Lo, S.-C. R. & Li, K. S. (1993). Issues in reliability-based design in
docteur-ingeÂnieur, EÂcole Nationale des Ponts et ChausseÂes, Paris. geotechnical engineeringÐa discussion. Proceedings of the interna-
Bauduin, C. (1998). Eurocode 7: Background and introduction to tional symposium on limit state design in geotechnical engineering,
practical applications. Notes from the Short Course on EC 7, Copenhagen, vol. 3, pp. 659±663.
University of Porto, October 1998. Lumb, P. (1966). Variability of natural soils. Can. Geotech. J. 3, 74±97.
Becker, D. E. (1996a). Eighteenth Canadian Geotechnical Colloquium: Magnan, J. P. (1982). Les meÂthodes statistiques et probabilistes en
GROUND PARAMETERS AND FAILURE PROBABILITY IN DESIGN TO EC 7 531
meÂcanique des sols. Presses de l'Ecole Nationale des Ponts et ability in Soil and Structural Engineering, Hong Kong, 371±388.
ChausseÂes, Paris. Simpson, B. & Driscoll, R. (1998). Eurocode 7: a commentary. London:
Manoliu, I. & Marcu, A. (1993). 25 years of utilisation of the limit state ARUP/BRE, Construction Research Communications Ltd.
concept in the Romanian Code for geotechnical design. Proceedings Singh, A. (1972). How reliable is the factor of safety in foundation
of the international symposium on limit state design in geotechnical engineering? Proc. 1st Int. Conf. Applications of Statistics and
engineering, Copenhagen, vol. 2, pp. 533±542. Probability in Soil and Structural Engineering, Hong Kong, 389±424.
Meyerhof, G. G. (1948). An investigation of the bearing capacity of Terzaghi, K. (1943). Theoretical soil mechanics. John Wiley & Sons,
shallow footings on dry sand. Proc. 2nd Int. Conf. Soil Mech. New York.
Found. Engng, Rotterdam 1, 237±243. Thorne, C. P. & Quine, M. P. (1993). How reliable are reliability
Meyerhof, G. G. (1993). Development of geotechnical limit state design. estimates and why soils engineers rarely use them. In Probabilistic
Proceedings of the international symposium on limit state design in methods in geotechnical engineering (eds K. S. Li and S. C. R. Lo),
geotechnical engineering, Copenhagen, vol. 1, pp. 1±12. pp. 325±332. A. A. Balkema, Rotterdam.
Meyerhof, G. G. (1995). Development of geotechnical limit state design. Yu, Y. F. & Mostyn, G. R. (1993). Spatial correlation of rock joints. In
Can. Geotech. J. 32, 128±136. Probabilistic methods in geotechnical engineering (eds K. S. Li and
Orr, T. L. & Farrell, E. R. (1999). Geotechnical design to Eurocode 7. S. C. R. Lo), pp. 241±255. A. A. Balkema, Rotterdam.
London: Springer-Verlag. Wickremesinghe, D. & Campanella, R. G. (1993). Scale of ¯uctuation
Schneider, H. (1997). De®nition and determination of characteristic soil as a descriptor of soil variability. In Probabilistic methods in
properties. Proc. 14th Int. Conf. Soil Mech. Found. Engng, Hamburg geotechnical engineering (eds K. S. Li and S. C. R. Lo), pp. 233±
4, 2271±2274. 239 A. A. Balkema, Rotterdam.
Schultze, E. (1972). Frequency distributions and correlation of soil Whitman, R. V. (1984). Evaluating calculated risk in geotechnical
properties. Proc. 1st Int. Conf. Applications of Statistics and Prob- engineering. J. Geotech. Engng 110, No. 2, 145±188.