ISGSR 2011 - Vogt, Schuppener, Straub & Bräu (eds) - © 2011 Bundesanstalt für Wasserbau ISBN 978-3-939230-01-4

A Comparative Study of Pile Design Using Eurocode 7 and RBDE

J. Wang, Y. Wang & Z. Cao
Department of Building and Construction, City University of Hong Kong, Hong Kong

ABSTRACT: A comparative study of pile design is presented using Eurocode 7 and an expanded reliabil-
ity-based design (RBDE) method that is recently developed by the authors. A design example that has
been used in the literature to illustrate Eurocode 7 is re-designed using RBDE. The RBDE method gives
designs that are consistent with the designs from Eurocode 7 or correspond to the target failure probabil-
ity (pT) adopted in EN 1990. The RBDE method allows design engineers to adjust the design pT easily to
accommodate the needs of a particular project without additional computational efforts. In addition, de-
sign engineers have the flexibility to make assumptions and/or simplifications deemed appropriate in de-
signs. Such flexibility is illustrated by exploring the effect of different probability distributions of soil ef-
fective friction angle on design.

Keywords: Pile, Eurocode 7, Reliability-Based Design, Monte Carlo Simulations

1 INTRODUCTION

Reliability-based design (RBD) of foundations has attracted increasing interest over the last two decades,
and several RBD methodologies have emerged, such as the partial factor design method in Eurocode 7,
the load and resistance factor design (LRFD) method for highway structure foundations (Barker et al.
1991, Paikowsky et al. 2004, Paikowsky et al. 2010), and the Multiple Resistance Factor Design (MRFD)
method for transmission line structure foundations (Phoon et al. 2003a&b). These RBD methods aim to
provide designs with appropriate degrees of reliability, which is usually expressed in probabilistic terms,
such as the target probability of failure pT = 7.2  10-5 or target reliability index T = 3.8 adopted in EN
1990 (European Committee for Standardization 2002). Through some calibration process, Eurocode 7
provides tabulated partial factors for actions (i.e., loads), material properties, and resistances. Design en-
gineers select appropriate partial factors from the table and carry out design calculations using a trial-and-
error approach. The calibration of partial factors in Eurocode 7 has been primarily based on deterministic
methods that calibrate to the long experience of traditional design with the aid of historical and empirical
methods (Orr and Breysse 2008). As the numerical values of the partial factors are obtained from deter-
ministic methods, it is of great interest to use full probabilistic methods (e.g., Monte Carlo Simulations
(MCS)) to investigate the performance of these partial factors in achieving the desired degrees of reliabil-
ity. In addition, the partial factors in Eurocode 7 aim for pT = 7.2  10-5 only, partial factors for other pT
values commonly are not available. This fact limits a designer’s flexibility to adjust the pT to accommo-
date specific needs of a particular project.
To address these limitations, an expanded reliability-based design (RBDE) method was recently devel-
oped that formulates the foundation design as an expanded reliability problem (Wang et al. 2011, Wang
2011). In this paper, a comparative study of pile design using Eurocode 7 and RBDE is described. After a
brief introduction of Eurocode 7 and RBDE, a pile foundation design example is described that has been
used to evaluate Eurocode 7 in literature (Orr 2005a). Then, the design example is re-designed using
RBDE and compared with the designs from Eurocode 7. In addition, the effect of the probability distribu-
tions of soil effective friction angle on designs is explored using RBDE.

481
2 PARTIAL FACTOR DESIGN METHOD IN EUROCODE 7

Eurocode 7 contains three Design Approaches (i.e., DA1 with Combination 1 (C1) or 2 (C2), DA2, and
DA3), and it aims to achieve that the probability of exceeding some limit states during a specified service
period of the structures is smaller than the pT valued adopted. Consider, for example, the ultimate limit
state (ULS), the pT = 7.2  10-5 is adopted in EN 1990. For the ULS design of piles under axial compres-
sion, the design equation is given as (e.g., Orr 2005b):
Fc,d  R c,d (1)
where Fc,d is the design action (load) and Rc,d is the design resistance of the pile. The design vertical ac-
tion, Fc,d is given as (e.g., Orr 2005b):
Fc,d   G G k   Q Q k (2)
where Gk is the characteristic permanent load, Qk is the characteristic variable load, G and Q are the rele-
vant partial load factors. The values of G and Q are given in Eurocode 7 and depend on the Design Ap-
proach being used. The design compressive resistance of piles is given by (Orr 2005b):
R c ,d  R b , d  R s ,d  R b , k  b  R s , k  s (3)
where Rb,d and Rs,d are the design base and shaft resistances, Rb,k and Rs,k are the characteristic base and
shaft resistances, and b and s are the relevant partial resistance factors.

3 EXPANDED RELIABILITY-BASED DESIGN (RBDE) METHOD

An expanded reliability problem herein refers to a reliability analysis of a system in which a set of system
design parameters are artificially considered as uncertain with probability distributions specified by the
user for design exploration purposes (Wang et al. 2011, Wang 2011). For example, consider the pile with
the pile length L as the design parameter. The design process is one of finding an L value that satisfies
both the ULS and SLS requirements and achieve the design target pT or T. In the context of RBDE, the L
of the pile is considered as independent discrete random variables with uniformly distributed probability
mass function p(L). The pile design process is re-formulated as a process of finding failure probabilities
corresponding to designs with various L values [i.e., conditional probability p(Failure|L)] and comparing
them with pT. Failure refers to events in which the load exceeds resistance (i.e., F > R). Feasible designs
are those with p(Failure|L) ≤ pT. Note that the uniform probability mass function p(L) does not reflect the
uncertainty in L, because L represents design decisions and no uncertainty is to be associated with it. In-
stead, it is used to yield desired information about p(Failure|L). Using Bayes’ Theorem (e.g., Ang and
Tang 2007), the conditional probability p(Failure|L) is given by:
p(L Failure)
p(Failure L)  p(Failure) (4)
p( L)
in which p(L|Failure) = conditional probability of L given failure. Since L is independent discrete uni-
form random variables, p(L) in Equation (4) is expressed as:
1
p( L)  (5)
nL
in which nL = number of possible discrete values for L. Using a single run of MCS, p(Failure) and
p(L|Failure) can be estimated. Details of the RBDE and MCS are given by Wang et al. (2011) and Wang
(2011).
RBDE can result in a large number of feasible designs. The requirement of the economic optimization
limit state (EOLS) then is adopted to finalize the design as the one with the minimum construction cost
(Wang and Kulhawy 2008, Wang 2009). The construction cost of pile is estimated using published, annu-
ally-updated, unit cost data, such as Means Building Construction Cost Data (Means 2007). The construc-
tion costs for all feasible designs are calculated as the product of their unit costs and pile lengths, and the
final design is determined accordingly by comparing their construction costs.

482
4 PILE FOUNDATION DESIGN EXAMPLE

Orr (2005a&b) illustrated Eurocode 7 using a bored pile design example shown in Figure 1. The bored
pile with a diameter B = 0.6 m is installed in sand with a total unit weight  = 21 kN/m3, a characteristic
effective friction angle 'k = 35 (effective cohesion c' = 0), and SPT-N = 25. Groundwater level is at a
depth of 2 m below the ground surface. The pile is designed to support an axial compression load with a
characteristic permanent load Gk = 1200 kN and a characteristic variable load Qk = 200 kN. The unit
weight of concrete is 24 kN/m3. The only design parameter is pile length L.

Gk=1200kN
Qk=200kN

GWL 2.0m

L=?

Sand
'k=35º
=21kN/m


Figure 1. Pile foundation example (after Orr 2005a).

Table 1. Partial factors for the pile example

Design Approach Permanent Variable Base Resis- Shaft Resis- Model Material
Load Factor, Load Factor, tance Factor, tance Factor, Factor, Factor,
G Q b s R M
DA1, C1 1.35 1.5 1.25 1 1.5 —
DA1, C2 1 1.3 1.6 1.3 1.5 —
DA2 1.35 1.5 1.1 1.1 1.5 —
DA3 1.35 1.5 — — — 1.25

Table 2. Summary of pile designs using Eurocode 7

Design Ap- Pile Length, Overall Factor of Safety, Probability of Failure,
proach L (m) OFS p(Failure)
DA1. C1 14.9 2.4 1.9  10-3
DA1. C2 14.6 2.4 2.6  10-3
DA2 14.0 2.3 4.4  10-3
DA3 16.7 2.7 2.4  10-4

Using three DAs of Eurocode 7, Orr (2005b) provided a set of model solutions to this design example.
The characteristic resistances are obtained as Rb,k = Abqbk and Rs,k = Asqsk, where qbk and qsk are the char-
acteristic base resistance and shaft friction obtained from the soil parameters and Ab and As are the areas
of the pile base and pile shaft. Table 1 summarizes the partial factors used for different DAs in this exam-
ple. The values of the partial factors b and s are corrected by a model factor, R = 1.5. The design resis-
tance for DA3 is obtained by applying the partial material factor M = 1.25 to tan'k to obtain 'd which is
483
used to obtain the design resistances. The base bearing resistance is given by qbk = v0' Nq, where v0' is
the vertical effective stress at the pile base and Nq is a bearing capacity factor estimated using Nq versus
' relationship proposed by Berezantzev et al. (1961). The shaft resistance is given by qsk
=sin'v0' tan'.
Table 2 summarizes the L values obtained from three DAs. The L values vary from 14.0 m to 16.7 m,
and the overall factors of safety, OFS (defined as Rc,k/Fk), vary from 2.3 to 2.7. This design example is re-
designed in the next section using the RBDE method (Wang et al. 2011, Wang 2011).

5 RBDE DESIGN

The RBDE approach conceptually contains four basic steps: (1) establish deterministic calculation mod-
els, (2) model geotechnical-related uncertainties, (3) perform MCS and identify a pool of feasible designs,
and (4) select the final design based on economic evaluation. To enable a consistent comparison with the
design by Orr (2005b), the deterministic ULS calculation models in this section follow those adopted in
the previous section. In addition, the permanent load G is treated as constant and equal to the characteris-
tic permanent load Gk (i.e. G=1200kN). The total unit weight  = 21 kN/m3 of soil is also taken as deter-
ministic.

5.1 Uncertainty Modeling

Uncertainties in design loads and material properties in Eurocode 7 are reflected through their respective
characteristic values. The uncertain variables in this design example include the variable load, effective
friction angle of soil, and length of pile, as shown in Table 3. The characteristic value for a design load in
EN 1990 is defined as the load magnitude that corresponds to 5% or 2% probability of exceedance (i.e.,
an upper 95% or 98% fractile of its probability distribution) (European Committee for Standardization
2002). The variable load Q is considered as a lognormal random variable with a coefficient of variation
COVQ = 0.5, and its characteristic value Qk is taken as the upper 95% fractile of the probability distribu-
tion. Then, the mean value (i.e., Qm) of variable load is calculated as:
Qk
Qm  (6)
1  1.645COVQ 
Using Qk = 200 kN (Orr 2005a&b) and COVQ = 0.5, Qm is estimated as 110 kN.
The effective friction angle ' of soil is considered as a lognormal random variable with a coefficient
of variation COV' = 0.1. The mean value (i.e., 'm) of effective friction angle of soil is calculated as
(Schneider 1997):
 'k
 'm  (7)
1  0.5COV' 
Using 'k = 35º (Orr 2005a&b) and COV' = 0.1, 'm is estimated as 36.84º.
In addition, the pile design parameter L is treated as independent discrete uniform random variable.
The possible L values vary from 12 m to 21 m with an increment of 0.3 m.

Table 3. Uncertain modeling in RBDE design

Variables Variable Load, Q Effective friction angle of soil, ' Pile Length, L
Mean COV* Min* Max* Mean COV* Min* Max* Interval
Values 110 kN 0.5 24º 40º 36.84º 0.1 12 m 21 m 0.3 m
Distribu-
tion Type Lognormal Distribution Lognormal Distribution Discrete Uniform Distribution
* COV = Coefficient of Variation, Min = Minimum, Max = Maximum

5.2 Monte Carlo Simulation (MCS)

MCS is performed using the software package Matlab (Mathworks 2010), which is equipped with random
number generators for various probability distributions, such as “lognrnd” for lognormal variables,
“normrnd” for normal variables, and “rand” for uniform variables. Random samples of the lognormally
distributed variable load and effective friction angle of soil are generated by “lognrnd” with their respec-
484
tive means and standard deviations. Because the Nq versus ' relationship proposed by Berezantzev et al.
(1961) is only applicable for the ' values between 24 and 40, the ' values are taken as 24 and 40, re-
spectively, when the ' values generated from the random number generator are smaller than 24 or larger
than 40. Random samples of uniformly distributed L are generated by “rand”. For each set of random
samples, the loads and resistances of the pile are calculated. The pile is considered “failed” when the sum
of the permanent and variable loads exceeds the bearing resistance. A single run of MCS with a sample
size of 9,000,000 is performed for RBDE, and the p(L|Failure), p(Failure), and conditional failure prob-
ability p(Failure|L) are estimated from MCS using Equation (4) accordingly.

0.1
Expanded RBD M ethod
Eurocode 7
Probability of Failure, p(Failure)

0.01

14.0m
14.6m
14.9m -4
0.001  T = 3.2 or p T = 6.9 10

16.7m
-5
0.0001  T = 3.8 or p T = 7.2 10

0.00001
12 15 18 21
Pile Length, L (m)
Figure 2. Conditional probability of failure from Monte Carlo Simulation.

5.3 Results
Figure 2 shows the conditional failure probability p(Failure|L) obtained from a single run of MCS. Note
that p(Failure|L) is a variation of failure probability as a function of L. Failure probability p(Failure) de-
creases as L increases. Figure 2 also includes the pT = 7.2  10-5 adopted in EN 1990, and feasible designs
are those that fall below the pT, as shown in the figure. In this example, the feasible designs are the piles
with L ≥ 17.7 m. The economic requirement then is adopted to determine the final design (Wang and Kul-
hawy 2008, Wang 2009). Since the construction cost is the product of pile depths and unit costs, the eco-
nomic design for a given value of B is the one with the minimum L value. Therefore, among the pool of
feasible designs obtained from RBDE, the final design is a pile with L = 17.7 m.

6 RESULT COMPARISONS

Figure 2 also includes the failure probability p(Failure) for the design pile length obtained from Eurocode
7 (see Table 2). Each p(Failure) is obtained by performing a run of MCS with a sample size of 1,000,000,
and the exact values of p(Failure) are summarized in Table 2.
These p(Failure) values vary from 4.4  10-3 to 2.4  10-4, and they follow the p(Failure) versus L
curve obtained from the RBDE method. They are all however significantly larger than the pT = 7.2  10-5
adopted in EN 1990. This implies that using the partial factors recommended in Eurocode 7 does not
guarantee automatic fulfillment of its target reliability. It is interesting to note that, however, these
p(Failure) values are consistent with the empirical foundation failure rates of about 10-2 to 10-3 (Baecher
1987).
In addition, it is worth noting that, in RBDE, feasible designs for different pT values are inferred di-
rectly from Figure 2 without additional computational efforts, which allows designers to adjust easily the

485
design pT to accommodate the needs of a particular project. To illustrate such flexibility, Figure 2 also in-
cludes the pT = 6.9  10-4 (i.e., T = 3.2) that have been adopted in the reliability – based designs of foun-
dations for transmission line structures in North America (Phoon et al. 2003a&b). The corresponding de-
sign is a pile with L = 16.2 m, which falls among the range of pile length obtained from Eurocode 7.

7 EFFECT OF PROBABILITY DISTRIBUTIONS OF SOIL EFFECTIVE FRICTION ANGLE

The RBDE method allows designers to make assumptions and/or simplifications deemed appropriate in
designs. This section illustrates this flexibility by using different probability distributions of soil effective
friction angle ' in the design and exploring its effect. The ' is considered as a normal random variable
with the same mean value (i.e., 'm =36.84º) and coefficient of variation (i.e., COV' = 0.1). Random sam-
ples of the normally distributed effective friction angle of soil are generated by the Matlab function
“normrnd” with its mean and standard deviation. The ' values are taken as 24 and 40, respectively,
when the ' values generated from the random number generator are smaller than 24 or larger than 40.
A single run of MCS with a sample size of 9,000,000 is performed for RBDE, and the results are shown in
Figure 3.
Figure 3 compares the results for normal distribution with those for lognormal distribution. The
p(failure) values both decreases significantly as the pile length increases. The p(failure) versus L relation-
ship for the normal distribution moves towards the upper right corner of the plot, indicating a significant
increase of failure probability for the same L value or a significant increase of design L value for the
same pT value. For the pT = 7.2  10-5 adopted in EN 1990, the feasible designs are the pile with L ≥ 18.9
m, and therefore, the final design is the pile with L = 18.9. For pT = 6.9  10-4 (i.e. 3.2), the final de-
sign pile length is 17.4 m. These results indicate that probability distribution types have significant effect
on the design.

0.1
Lognormal distribution
Normal distribution
Probability of Failure, p(Failure)

0.01


0.001  T = 3.2 or p T = 6.9 


0.0001  T = 3.8 or p T = 7.2 

0.00001
12 15 18 21
Pile Length, L (m)
Figure 3. Effect of probability distributions of soil effective friction angle.

8 SUMMARY AND CONCLUSIONS

A comparative study was presented for pile design using Eurocode 7 and RBDE. An example that was
used to illustrate Eurocode 7 was re-designed using RBDE. The RBDE method gives designs that are con-
sistent with the designs from Eurocode 7 or correspond to the target failure probability pT adopted in EN
1990. It is also found that, using the partial factors recommended in Eurocode 7 does not guarantee auto-
matic fulfillment of its target reliability, although the resulting failure probabilities are consistent with the

486
empirical rates of foundation failure. In addition, it is worth noting that, in RBDE, feasible designs for dif-
ferent pT values are inferred directly without additional computational efforts, which allows designers to
adjust easily the design pT to accommodate the needs of a particular project. The RBDE method also gives
designers the flexibility to make assumptions and/or simplifications deemed appropriate in designs. This
flexibility was illustrated by using different probability distributions for soil effective friction angle in the
design and exploring its effect. It was found that probability distribution types have significant effect on
the design.

ACKNOWLEDGEMENTS

The work described herein was supported by a grant from the Research Grants Council of the Hong Kong
Special Administrative Region, China [Project No. 9041484 (CityU 110109)] and a grant from City Uni-
versity of Hong Kong (Project No. 7002568). The financial supports are gratefully acknowledged.

REFERENCES

Ang, A. H.-S. and Tang, W. H. 2007. Probability concepts in engineering: Emphasis on applications to civil and environ-
mental engineering. Wiley, New York.
Baecher, G. B. 1987. Geotechnical risk analysis user’s guide. Report No. FHWA/RD-87-011, Federal Highway Administra-
tion, McLean, Va., USA.
Barker, R. M., Duncan, J. M., Rojiani, K. B., Ooi, P. S. K., Tan, C. K., and Kim, S. G. 1991. Manuals for the design of bridge
foundations. NCHRP Report 343, Transportation Research Board, Washington, DC.
Berezanysev, V.G., kristoforov, V.S., and Golubkov, V.N. 1961. Load bearing capacity and deformation of piled foundations.
Proceedings Ⅴ International Conference on Soil Mechanics and Foundation Engineering, Paris, France, 2, 11-15
European Committee for Standardization 2002. EN 1990: Eurocode – Basis of Structural Design, Brussels.
Mathworks, Inc. 2010. MATLAB – the language of technical computing, <http://www.mathworks.com/products/matlab/>
(March 9, 2010).
Means. 2007. 2008 RS means building construction cost data. R.S. Means Co., Kingston, MA.
Orr, T. L. L. 2005a. Design examples for the Eurocode 7 Workshop. Proceedings of the International Workshop on the
Evaluation of Eurocode 7, Trinity College, Dublin, 67-74.
Orr, T. L.L. 2005b. Model solution for example 3 - Pile foundation designed from soil parameter values. Proceedings of the
International Workshop on the Evaluation of Eurocode 7, Trinity College, Dublin, 85-86.
Orr, T. L. and Breysse D. 2008. Eurocode 7 and reliability-based design. Reliability-Based Design in Geotechnical Engineer-
ing: Computations and Applications, Chapter 8: 298-343, Edited by Phoon, Taylor & Francis.
Paikowsky, S. G., Birgisson, B., McVay, M., Nguyen, T., Kuo, C., Baecher, G., Ayyub, B., Stenersen, K., O’Malley, K.,
Chernauskas, L., and O’Neill, M. 2004. Load and resistance factor design (LRFD) for deep foundations. NCHRP Report
507, Transportation Research Board, Washington, DC.
Paikowsky, S. G., Canniff, M. C., Lesny, K., Kisse, A., Amatya, S., and Muganga, R. 2010. LRFD design and construction of
shallow foundations for highway bridge structures. NCHRP Report 651, Transportation Research Board, Washington, DC.
Phoon, K. K., Kulhawy, F. H., and Grigoriu, M. D. 2003a. Development of a reliability-based design framework for transmis-
sion line structure foundations. Journal of Geotechnical and Geoenvironmental Engineering, 129(9), 798-806.
Phoon, K. K., Kulhawy, F. H., and Grigoriu, M. D. 2003b. Multiple resistance factor design (MRFD) for shallow transmission
line structure foundations. Journal of Geotechnical and Geoenvironmental Engineering, 129(9), 807-818.
Schneider, H. R. 1997. Definition and determination of characteristic soil properties. Proceedings of the XII International Con-
ference on Soil Mechanics and Geotechnical Engineering, Hamburg, 2271-2274, Balkema, Rotterdam.
Wang, Y. 2009. Reliability-based economic design optimization of spread foundations. J. Geotech. and Geoenvir. Eng.,
135(7), 954-959.
Wang, Y. 2011. Reliability-based design of spread foundations by Monte Carlo Simulations. Geotechnique, doi:
10.1680/geot.10.P.016.
Wang, Y., Au, S. K., and Kulhawy, F. H. 2011. Expanded reliability-based design approach for drilled shafts. J. Geotech. and
Geoenvir. Eng., 137 (2), 140-149.
Wang, Y. and Kulhawy, F. H. 2008. Economic design optimization of foundations. J. Geotech. and Geoenvir. Eng.,134(8),
1097-1105.

487