You are on page 1of 10

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

net/publication/323967716

Calibrating shaft and base resistance factors for design of drilled shaft
foundations

Conference Paper · March 2013

CITATIONS READS

0 82

3 authors, including:

Duong Bach P.H.A.J.M. Van Gelder


National University of Civil Engineering, Hanoi Delft University of Technology
4 PUBLICATIONS   2 CITATIONS    372 PUBLICATIONS   3,553 CITATIONS   

SEE PROFILE SEE PROFILE

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

We are working on linear and nonlinear multivariate time series analysis in modeling hydro-meteorological variables. View project

Predicting floodplain velocities due to embankment dam failure of Mosul dam View project

All content following this page was uploaded by P.H.A.J.M. Van Gelder on 23 March 2018.

The user has requested enhancement of the downloaded file.


Safety, Reliability and Risk Analysis: Beyond the Horizon – Steenbergen et al. (Eds)
© 2014 Taylor & Francis Group, London, ISBN 978-1-138-00123-7

Calibrating shaft and base resistance factors for design of drilled shaft
foundations

D. Bach & P.H.A.J.M. van Gelder


Delft University of Technology, Delft, The Netherlands

M.B. Hudson
AMEC, Los Angeles, California, USA

ABSTRACT: In recent studies, calibrating resistance factors of drilled shafts were mainly performed
for a common resistance factor. Uncertainties of predicted shaft and base resistances were lumped into a
unique resistance factor. This approach has not fully taken advantage of the Load and Resistance Factor
Design (LRFD) method, as opposed to when shaft and base resistance factors are calibrated separately.
A dataset including 26 Osterberg cell-tested drilled shafts is analyzed and the differences between pre-
dicted and measured resistances for the shaft and base are statistically analyzed. A reliability-based analy-
sis technique using Monte Carlo Simulation (MCS) is used to calibrate the resistance factors separately in
the framework of the LRFD method with specified target reliability levels. Several benefits from the use
of separating resistance factors on shaft and base are presented.

1 INTRODUCTION and base resistance were lumped into a unique


resistance factor. In fact, uncertainty degrees of
In the last decades, taking the lead from the devel- predicted resistances are very different for the two
opment of structure design, the LRFD method responses, for example, the construction methods
has gradually superseded the Allowable Stress affect the shaft and the base differently. Therefore,
Design (ASD) method for foundation design. The the derived calibration results have not fully satis-
LRFD approach makes possible the application of fied the requirement for the design of pile founda-
reliability analysis to quantify uncertainties associ- tions following the LRFD method.
ated with various load and resistance components. In this study, a dataset comprised of results
The resistance factor calibration is performed for a from 26 drilled shafts tested by the Osterberg cell
set of load factors and probability distributions for (O-cell) method is used to calibrate resistance fac-
loads that were specified in structure codes. tors separately for the shaft and the base responses.
In pile foundations, several studies on the resist- The resistances measured by the O-cell and the
ance factor calibration were carried out based theoretically predicted resistances are compared
on reliability analysis methods. The calibration and the difference between them is statistically
studies determined resistance factors for numer- analyzed through bias factors. A reliability-based
ous resistance prediction methods with different analysis technique using MCS is used to calibrate
target reliability levels. According to the First shaft and base resistance factors separately in the
Order Second Moment (FOSM) method, there framework of the LRFD method with target reli-
were Barker et al. (1991); Yoon & O’Neill (1997); ability indices. Several benefits from the use of
Withiam et al. (1997); McVay et al. (2000, 2002) and separate resistance factors are presented.
Kuo et al. (2002). Based on the First Order Reli-
ability Method (FORM), there have been Honjo
et al. (2002) and Paikowsky et al. (2003, 2004). The 2 CALIBRATION UNDER FRAMEWORK
Monte Carlo Simulation (MCS) method was used OF THE LRFD METHOD
by Allen et al. (2005); Abu-Farsakh & Yu (2010)
and Bach et al. (2012). The LRFD method states that a factored resist-
The resistance factors calibrated in the studies ance of a structural component is larger than a lin-
mentioned above applied to the total resistance of ear combination of factored load effects as given
piles. This means that all uncertainties of the shaft by a following general format:

2497

Fullpaper-Book-03-[CH201-CH300].indb 2497 8/22/2013 7:17:31 PM


n For the case where the O-cell is placed at the
φR ∑ γ iQi (1) bottom of the shaft, as the cell is pressurized,
i =1 the bottom of the cell moves downward, caus-
ing a reaction force at the lower plate of the cell
where φ = resistance factor; R = theoretically pre- (base resistance, Rb), while the upper plate of the
dicted resistance; γi = load factor for the ith load cell moves upward, mobilizing shear forces along
component; and Qi = designed value for the ith the shaft (shaft resistance, Rs). They are measured
load component. These factors are calibrated separately from each other (Fig. 1d).
using reliability methods based on the probabilistic An equivalent load-displacement curve as
characteristics of load effects and the resistance, head-down compressive tests (Fig. 1b) can be
including statistics and modelling uncertainties. obtained by using test results from the O-cell
The factors are calibrated to meet specified target tests. At an arbitrary displacement, d, there is an
reliability levels that were selected based on assess- upward load, Qou, and a downward load Qod. An
ing previous designs or cost-benefit analyses. equivalent load at the top of shaft is the sum of
In case of separating the shaft and base resist- the upward load, Qou, and the downward load, Qod.
ances from a total resistance R. Equation 1 can be Note that the buoyant weight of shaft, W, must
rewritten as: be subtracted from the measured upward load in
order to obtain pile resistance values. In case the
n slenderness ratio of the shaft is larger than 20, the
φS RS φbRb ≥ ∑ γ iQi (2) displacement, d, needs to be added to the elastic
i =1 shortening corresponding to the axial load in the
pile (Russo et al. 2003). If the displacement, d,
where φS, φb = shaft and base resistance factors, being considered is beyond the measured maxi-
respectively; RS, Rb = predicted resistances at the mum displacement of any resistance component
shaft and base of piles, respectively. Hence, the aim (Dumax or Ddmax) that does not correspond to the
of this study is to calibrate separately the φS and φb ultimate resistance value, the corresponding load
of drilled shafts for a specified prediction method is conservatively assumed constant at the value of
associated with different target reliability indices. the maximum load applied (Qmax).

3 OSTERBERG CELL TEST (O-CELL TEST) 4 BIAS FACTOR FOR RESISTANCE


AND LOAD
The high capacity of large-diameter drilled shafts
in combination with the high cost and large size 4.1 Bias factor for resistance
required for reaction systems sometime make
head-down compressive tests too costly or imprac- Consider a dataset comprised of N drilled shafts,
tical (Fig. 1a). The O-cell test often provides high where the ith shaft has a measured nominal resist-
capacity testing at a more affordable cost, and ance, RMi, and a theoretically predicted resistance,
therefore has become an attractive alternative for RPi. The resistance bias factor, λRi, for an ith shaft
testing drilled shafts. The O-cell is a hydraulic jack is then defined as:
placed within the shaft (sometimes at or near the
shaft bottom) to induce an upward and downward RM
λ Ri = Mi
(3)
vertical load (Fig. 1c). RPPi

Figure 1. Comparison of head-down compressive test and O-cell test: (a) Arrangement of head-down compressive
test; (b) Typical Load-Displacement curve in a head-down compressive test; (c) Arrangement of an O-cell test; (d)
Typical Load-Displacement curves in O-cell test.

2498

Fullpaper-Book-03-[CH201-CH300].indb 2498 8/22/2013 7:17:31 PM


Table 1. Load factors and probabilistic characteristics for dead and live loads.

Structure Type of load Load factor, γ Bias factor, λ COV Distribution Reference

Buildings Dead load 1.20 1.05 0.10 Normal Ellingwood and


Live load 1.60 1.00 0.25 Type I Galambos
(1982)
Bridges Dead load 1.25 1.05 0.10 Normal Nowak (1999)
Live load 1.75 1.15 0.20 Normal

The mean of resistance bias factors for whole by Paikowsky et al. (2004), the resistance fac-
dataset is obtained as: tor calibration is performed for a set of load
factors already specified in the structural code.
∑ i =1 λRi
N
Thus, the load factors are fixed. Load factors,
λR = (4) statistical parameters and probability distribu-
N tions for dead load and live load bias factors are
summarized in Table 1 for the case where only
The standard deviation of resistance bias fac-
dead load, QD, and live load, QL, are considered
tors is determined as:
in Equation 2.

∑ i =1(λRi λR )
N 2

σR = (5) 5 CALIBRATION BASED ON MONTE


N −1
CARLO SIMULATION
Finally, the coefficient of variation of resistance
bias factors is given as: This study follows the calibration procedure
based on MCS as recommended by Allen et al.
σR (2005) and Bach et al. (2012) to calibrate shaft
VR =
COV (6)
λR and base resistance factors. Based on Equa-
tion 2, if just the dead and live loads are con-
The predicted resistance is calculated using sidered following the ultimate limit state of
theoretical methods based on specific ground con- strength I, then a limit state function can be
ditions. The measured nominal resistance consists written as:
of the measured nominal shaft resistance and the
measured nominal base resistance. The measured RMs + RMb QMD − QML (7)
nominal shaft resistance is defined as the load
corresponding to a displacement that is equal to where RMs, RMb, QMD & QML = measured nominal
20 mm for sand or about 5 to 15 mm for clays shaft resistance, the measured nominal base resist-
(Osterberg 2001) or the plunging load in the O-cell ance, the measured value of the dead load and
tests, whichever comes first. Similarly, the measured the live load, respectively. If all measured terms in
nominal base resistance is defined as the load cor- Equation 7 are converted to predicted terms using
responding to a displacement that is equal to 5% of the bias factors of shaft resistance λRs, base resist-
the diameter of shaft (O’Neill & Reese 1999, Reese ance λRb, dead load λQD, and live load λQL, Equa-
et al. 2006). Paikowsky et al. (2004) concluded that tion 7 can be rewritten as:
this criterion provides a reliable and simple failure
interpretation. In case the nominal displacement
mentioned above exceeds the measured maximum RS λ RS + Rb λ Rb QD λQD − QL λQ
QL (8)
displacement, the load-displacement curve of the
O-cell tests is extrapolated using the stability plot Combining Equation 2 with Equation 8, after
method developed by Chin & Vail (1973) and dis- transformation, the limit state function is now
cussed by Neely (1991). obtained:
Statistical parameters and probability distribu-
tions for resistance bias factors are estimated in ⎡ ⎤
detail in Section 6.
g

QL ⎢
( QD
L + QL D )( Rb
RS + RS Rb )−Q
D
λQD

QL ⎥
Q − λQ
⎢ Rb QL ⎥
4.2 Bias factor for load φS + φ
⎢⎣ RS b ⎥⎦
The way to determine bias factors for loads is
also similar to that for resistances. As specified (9)

2499

Fullpaper-Book-03-[CH201-CH300].indb 2499 8/22/2013 7:17:34 PM


where terms in Equation 9 are the same as those analyzed by Fellenius (2011) and Schmertmann &
mentioned in Section 4. Schmertmann (2012). The shafts had 1.32 m diam-
MCS is used to generate random numbers that eter and were 16.5 through 20.1 m long. They were
are needed to independently extrapolate the Cumu- drilled through 7.2 through 10.5 m of artificial fill
lative Distribution Function (CDF) value for each consisting of sand and silt and into the underly-
random variable in the calibration process. In this ing natural sand with gravel. A mineral slurry was
calibration, there are four random variables which used to keep the walls of the shaft stable during
are the bias factors of shaft resistance, base resist- drilling.
ance, dead load and live load.
The computation steps are as follows:
6.2 Bias factors for shaft and base resistances
• Step 1: assign a target reliability index, βT
As mentioned in Subsection 4.1, resistance bias
• Step 2: select a trial shaft resistance factor, φS
factors are determined by means of measured
• Step 3: select a trial base resistance factor, φb
nominal resistances and predicted shaft and base
• Step 4: generate random numbers for each set of
resistances, respectively.
bias factors λRs, λRb, λQD, and λQL
The values of measured nominal resistances are
• Step 5: define the limit state function, g, as
estimated from the O-cell test results following the
described in Equation 9. Find the number of
limit displacement criteria proposed by Osterberg
cases in which g ≤ 0. The probability of failure is
(2001) and O’Neill & Reese (1999). Amongst the
then computed as:
26 O-cell tests, only two shafts had a measured
maximum upward displacement exceeding 20 mm.
count ( g ≤ ) The measured maximum upward displacement was
Pf = (10)
NS 32.3 mm (shaft 10B) and the measured minimum
upward displacement was 1.1 mm (shaft 22B).
where NS = number of simulations. In this study, Plunging behavior in shaft friction only occurred in
the number of simulations is 5 × 104. The cor- shaft 10B at the upward displacement of 25.6 mm
responding calculated reliability index, β, is then (see Fig. 2a). Although the upward displacement
defined as: of shaft 10 A reached a displacement of 20.6 mm,
the tendency of plunging was not clearly observed,
β = −Φ −1 ( ) (11) due to the fact that the test stopped at this level of
displacement. Obviously, the upward displacement
criterion of 20 mm for friction behavior in sand as
Here Φ−1 = inverse CDF of the standard normal proposed by Osterberg (2001) is suitable based on
distribution. If the calculated reliability index, the test results. Therefore, the load-displacement
β, is different than the specified target reliability curves of 24 out of the total of 26 shafts have to be
index, then, βT, the trial base resistance factor, extrapolated with the goal of estimating the meas-
φb, in Step 3 should be changed and a new itera- ured nominal shaft resistance.
tion needs to be repeated until |β − βT| ≤ toler- Regarding the measured maximum downward
ance. Repeat step 3 to step 5 for the next set of displacement, there are 14 shafts with a measured
trial shaft resistance factors, φS′ S . maximum downward displacement beyond the
• Step 6: from step 2 to step 5, a series of cou- displacement criterion of 65 mm (i.e. 5% diam-
ples of value for φS and φb may be derived that eter) as proposed by O’Neill & Reese (1999).
satisfies a specified target reliability index βT. A However, no shaft in all of the 26 shafts showed
method of selecting the couple of values needs the tendency of plunging behavior for end bear-
to be defined to help obtain the best results in ing, even with respect to shafts which have a large
the calibration process; the choice of methodol- displacement in combination with a large load.
ogy will be discussed in Section 6. For example, the results of shaft 19A indicated
a downward displacement of 83.1 mm at a rela-
tively large load of 8,663 kN, and shaft 19B also
6 CASE STUDY
reached a displacement of 100.8 mm at a load of
7,982 kN (see Fig. 2b). Hence, the concept of ulti-
6.1 Dataset for calibration
mate resistance at the base of shaft (or the pile toe)
A dataset of 26 O-cell equipped drilled shafts does not seem to be supportable, as indicated by
was collected from reports compiled by LAW/ Fellenius (2011). Of 26 shafts, the load-displace-
CRANDALL, Inc. (1994a, b, c) for the reme- ment curves of 12 shafts have to be extrapolated
diation project for the Los Angeles Memorial in order to estimate the nominal base resistance.
Coliseum after the Northridge earthquake in Los Since the upper natural materials and artificial
Angeles in January 1994. The O-cell graphs were fill at the site of the Los Angeles Coliseum are

2500

Fullpaper-Book-03-[CH201-CH300].indb 2500 8/22/2013 7:17:36 PM


According to Reese et al. (2006), the recom-
mended value of unit base resistance is predicted
as:

rb 0.0575N60 ( ) (14)

and the base resistance is then:

Rb Arrb (15)

here, A = cross section area at the base of the


shaft.
Statistical parameters for the resistance bias fac-
tors are estimated through Equations 3 through 6
and their probability distributions are checked by
the Kolmogorov-Smirnov test (Fenton & Griffiths
2008). All results are summarized in Table 2.

6.3 Calibration results based


on Monte Carlo Simulation
The limit state function, g, in Equation 9 is used
to calibrate the shaft and base resistance factors
with a procedure including six steps as described
in Section 5. Target reliability indices are assigned
as βT = 2.5 and 3.0, which correspond to the tar-
get probabilities of failure as Pf = 0.00621 and
0.00135.
The statistical parameters for the dead and live
loads are taken from Table 1 for buildings; the
Figure 2. Example of O-cell load-displacement curves
from the tests at the Los Angeles Memorial Coliseum.
probability distributions for the dead load and
live load are normal and type 1, respectively. The
ratio of the dead load to the live load, QD/QL, is
predominantly sandy soils, the predicted shaft
selected as 2.5 based on the study by Ellingwood &
resistance is calculated using the β method (O’Neill
Galambos (1982).
& Reese 1999):
The statistical parameters for the shaft and base
z2 resistances are taken from Table 2; the probability

RS ∫z
1
β zddA (12) distribution for both is log-normal. The ratio of
the base resistance to the shaft resistance, Rb/Rs,
where RS = predicted shaft resistance; σz′ = vertical is 0.289.
effective stress in soil at depth z; dA = differential Figure 3 indicates the relation between the shaft
circumferential area along the shaft; z1 & z2 = depth resistance factor, φS, and the base resistance factor,
of the top and bottom of the soil layer considered; φb, for two target reliability indices as 2.5 and 3.0.
and β = parameter determined as follows: It can be seen that there are numerous couples of
value of φS and φb derived that satisfy specified tar-
• For the sand and silt layer (first layer): get reliability indices. An inverse relation between
φS and φb is observed, that is, φS increases when φb
β
N60
15
( ) β ≤ 1.20
Table 2. Probabilistic characteristics of resistance bias
(13a) factors.
• For the sand with gravel layer (second layer): Standard
deviation,
β 0 75
β ≤ 1.80 (13b) Location Mean, λR σR COVR Distribution

Shaft 1.138 0.299 0.263 Log-normal


where N60 = uncorrected SPT index, z = depth Base 3.029 1.210 0.399 Log-normal
considered.

2501

Fullpaper-Book-03-[CH201-CH300].indb 2501 8/22/2013 7:17:38 PM


resistance factor continues to be used. Note that
in AASHTO 2007, shaft and base resistance fac-
tors were used separately for different prediction
methods. However, the values of these factors are
primarily obtained from experts’ judgments and
experiences.
In order to find out the difference in calculat-
ing the total resistance, two approaches are con-
sidered: (i) using a common resistance factor and
(ii) using separate shaft and base resistance fac-
Figure 3. Relation between φS and φb.
tors. Therefore, in this study, all the O-cell test
results are transformed to the equivalent head-
down compressive test results using the meth-
decreases, and vice versa. One issue arising herein
odology shown in Section 3. For demonstration,
is the question of which couple should be selected
shaft 4A is selected as an example. The O-cell
to provide the best results for the calibration proc-
test result is depicted in Fig. 4a, the equivalent
ess. Based on a study result presented in Bach et al.
head-down compressive test result obtained from
(2012), the calibrated resistance factor depends on
the O-cell test after transformation is shown in
the ratio of the coefficient of variation to the mean
Figure 4b.
of resistance bias factor, that is, COVR/λR. This
In case of using the equivalent head-down
relation seems to be linear when the limit state
compressive tests, statistical parameters of the
function is linear; the larger the ratio COVR/λR,
total resistance bias factor are estimated as
the smaller the resistance factor. From that a cer-
λR = 1.233, σR = 0.251 and COVR = 0.203. The
tain couple of φS and φb will be selected when the
probability distribution of the bias factor is log-
ratio of φS to φb approximates a ratio, temporarily
normal under the Kolmogorov-Smirnov test.
called “Acceptance Correlation Ratio”—ACR as
follows:

COVRb
φS λ RS COVRb λ RS
≈ ACR = = (16)
φb COVRS COVR λ R
S b
λ RS

where λRs & λRb = mean of bias factor of the


shaft and base resistances, respectively; COVRs &
COVRb = coefficient of variation of bias factor of
the shaft and base resistances, respectively.
Based on Equation 16, the ACR derived in this
case is 0.57. The values of couple of φS and φb,
therefore, are 0.95 and 1.65 for βT = 2.5, and 0.84
and 1.46 for βT = 3.0, respectively.

6.4 Benefit when using separate


resistance factors
In fact, recent studies have mainly focused on
calibrating a unique resistance factor. There are
two reasons leading to this selection. First, data
of measured resistances are collected from dif-
ferent types of testing, i.e., from the O-cell tests
and/or from the head-down compressive tests.
Normally, one would transform all the O-cell test
results to those under the head-down compres-
sive tests. Second, until now there has not been
an approach which is able to harmonically solve
the relation between shaft and base resistance Figure 4. Load-Displacement curves by O-cell and
factors after calibration. As a result, a common equivalent head-down compressive tests for shaft 4A.

2502

Fullpaper-Book-03-[CH201-CH300].indb 2502 8/22/2013 7:17:42 PM


Based on the calibration procedure described are not large. Conversely, in this case study where
in Bach et al. (2012), calibrated common resist- 26 shafts were located at a site, and they were con-
ance factors are 0.92 and 0.83 corresponding to structed by the same method with the same size
βT = 2.5 and 3.0. in a relatively stable geological condition. The
For comparison, shaft 4A is taken as an exam- uncertainty degree in the gathered dataset is not
ple: shaft and base resistances calculated accord- too high and thus the calibrated resistance factors
ing to the β method (FHWA method) are 9,692 kN are relatively large. Here, we can see that the pre-
and 3,165 kN, respectively. If we use a common dicted base resistances by the β method are much
resistance factor, the factored total resistance underestimated compared to the measured nomi-
will be: (9,692 + 3,165) x 0.83 = 10,671 kN (with nal ones. So, the calibrated base resistance factors
βT = 3.0); meanwhile, if we use separate resist- are large, even greater than 1 (φb = 1.65 and 1.46,
ance factors, the factored total resistance is now: see more in Subsection 6.3).
9,692 × 0.84 + 3,165 × 1.46 = 12,762 kN (with In the United States, several Departments of
βT = 3.0). So, the calculated total resistance using Transportation (DOTs) approved and used cali-
separate resistance factors, in this case, is larger brated resistance factors in a certain region with
than that using a common resistance factor with the terminology as “regionally calibrated resist-
an amount of 2,091 kN. Hence, the use of sepa- ance factors”. Abdel Salam et al. (2010) showed
rate resistance factors may achieve a more eco- that the LRFD regionally calibrated resistance
nomical design. factors reported for sands and clays are either
equal to or greater than the AASHTO recom-
mended values. In sand, the resistance factors
6.5 Comparison with resistance
are as much as 50% above those recommended
factors in ASSHTO 2007
by AASHTO, while values are as much as 100%
In order to compare shaft and base resistance fac- above the recommended values used for clay.
tors indicated in ASSHTO 2007, statistical param- Such large increases in resistance factors will
eters for the dead load and live load are taken from likely reduce the overall cost of bridge deep
Table 1 with respect to bridges; probability dis- foundations.
tributions for the dead load and the live load are
normal (Nowak 1999). The ratio of the dead load
6.6 Sensitivity analysis
to the live load, QD/QL, is selected as 3.46 (Bach
et al. 2012). The prediction method for the resist- In order to independently evaluate the influence
ance considered herein is the β method (FHWA of ratios, QD/QL and Rb/RS, on the value of cali-
method). brated resistance factors, the sensitivity analysis is
The calibrated values of φS and φb with βT = 3.0, performed (see Eq. 9). For the first step, the ratio
therefore, are 0.83 and 1.44, respectively. While Rb/RS is fixed with the value of 0.289 and the ratio
in ASSHTO 2007, the values of φS and φb with QD/QL is allowed to vary with values of 1.0, 1.5, 2.0,
βT = 3.0 are 0.55 and 0.50, respectively. A question 2.5 and 3.0; the variation of φS and φb with the ratio
arises herein regarding why there is a such consid- QD/QL will be observed. For the second step, the
erable difference in value associated with resistance variation of resistance factors with the increase of
factors. This issue can be explained as follows. the ratio Rb/RS will be considered, the ratio Rb/RS
First, the resistance factors in ASSHTO 2007 is taken as 0.25, 0.289, 0.5, 0.75 and 1.0, while the
were developed using either statistical analysis of ratio QD/QL is fixed as 2.5. The analysis results are
shaft load tests combined with reliability theory described in Figure 5.
(Paikowsky et al. 2004), fitting to Allowable Stress In Figure 5a, when the ratio Rb/Rs is fixed as
Design (ASD), or both. These two approaches 0.289, the resistance factors, φS and φb, slightly
resulted in significantly different resistance fac- decrease with the increase of ratio QD/QL and reach
tors; engineering judgment was used to establish a stable value when this ratio lies in a range from
the final resistance factor, considering the qual- 2.5 to 3.0. In Figure 5b, the ratio QD/QL is fixed
ity and quantity of the available data used in the as 2.5, and φS and φb, with an inverse tendency,
calibration. slightly increase with the increase of the ratio Rb/Rs
Second, Paikowsky et al. (2004) used the data and also reach a stable value as this ratio lies in
that were collected from many sites, e.g. from dif- a range from 0.5 to 0.75. After that the resistance
ferent testing methods, different soil parameters factors have a tendency of slight decrease when the
and different construction methods. Further, the ratio Rb/Rs is larger than 0.75.
size of tested shafts also varied considerably from Based on the obtained results, it can be con-
site to site. The variability or uncertainty degree cluded that the shaft and base resistance factors
of the data was rather high, therefore, resulting in are less sensitive with respect to the ratios QD/QL
the calibrated resistance factors that whose values and Rb/Rs.

2503

Fullpaper-Book-03-[CH201-CH300].indb 2503 8/22/2013 7:17:44 PM


1. A limit state function was formulated and a cali-
bration procedure based on MCS was proposed
to separately calibrate shaft and base resistance
factors for drilled shafts under the framework
of the LRFD method.
2. A ratio, temporarily called “Acceptance Corre-
lation Ratio”—ACR, was proposed aiming to
determine a unique couple of values for the shaft
and base resistance factors from the outcomes
of the calibration process based on MCS.
3. The use of separate shaft and base resistance
factors instead of a common resistance factor
may achieve a more economical design.
4. Through the comparison of obtained resist-
ance factors in this study and those in
ASSHTO 2007, once again it is confirmed that
the use of regionally calibrated resistance fac-
tors will likely reduce the overall cost of deep
foundations.
5. The shaft and base resistance factors are less
sensitive with respect to the variation of either
the dead to live load ratio or the base to shaft
resistance ratio in the calibration process.

ACKNOWLEDGEMENTS

This study is mainly funded by Project 322 of the


Vietnam Ministry of Education and Training and
partly supported by CICAT and the Section of
Hydraulic Engineering, Delft University of Tech-
nology, the Netherlands. The writers would like
to thank Ms. Dominguez, Los Angeles coliseum
administrator, and Dr. Fellenius, for supporting
Figure 5. Sensitivity analysis for shaft and base resist- and providing the material used in this paper.
ance factors.

REFERENCES

7 CONCLUSIONS Abu-Farsakh, M.Y., & Yu, X. 2010. Interpretation crite-


ria to evaluate resistance factors for axial load capac-
This study has presented a procedure for calibrating ity of drilled shafts. Transportation Research Record:
shaft and base resistances of drilled shafts under Journal of the Transportation Research Board, 2202(1):
20–31.
the framework of the LRFD method. A dataset Abdel Salam, S.S., Sritharan, S., & Suleiman, M.T. 2010.
of 26 O-cell tested drilled shafts was collected and Current design and construction practices of bridge
used for the calibration. Based on the geological pile foundations with emphasis on implementation of
condition on the site (the Los Angeles Coliseum), LRFD. Journal of Bridge Engineering, 15(6): 749–758.
the β method (FHWA method) was chosen aim- Allen, T.M., Nowak, A.S., & Bathurst, R.J. 2005. Trans-
ing to estimate shaft and base resistance factors for portation research circular E-C079: Calibration to
this method. determine load and resistance factors for geotechni-
In the scope of this study, only the ultimate cal and structural design. Technical report, Trans-
limit state of strength I for axially loaded drilled portation Research Board, Foundations of Bridges
and Other Structures Committee, General Structures
shafts was used, other limit states were not men- Committee, Washington, D.C.
tioned herein. The shaft and base resistance ASSHTO. 2007. LRFD bridge design specifications.
factors were calibrated according to two tar- Washington, D.C.
get reliability indices as 2.5 and 3.0. Based on Bach, D., van Gelder, P.H.A.J.M., Bakker, K.J. & Vri-
obtained results, conclusions drawn from the jling, J.K. 2012. Resistance factor calibration of drilled
study are as follows: shafts for bridge foundations. Proceedings of the 10th

2504

Fullpaper-Book-03-[CH201-CH300].indb 2504 8/22/2013 7:17:44 PM


International Probabilistic Workshop. Stuttgart, Ger- Nowak, A.S. 1999. National cooperative highway
many, 165–187. research program—Report 368: Calibration of LRFD
Barker, R.M., Duncan, J.M., Rojiani, K.B., Ooi, P.S.K., bridge design code. Technical report, Transportation
Tan, C.K., & Kim, S.G. 1991. National cooperative Research Board, Washington, D.C.
highway research program—Report 343: Manuals for O’Neill, M.W., & Reese, L.C. 1999. Federal highway
the design of bridge foundations. Technical report, administration—Report No. FHWA-IF-99–025:
Transportation Research Board, Washington, D.C. Drilled shafts: Construction procedures and design
Chin, F.K., & Vail, A.J. 1973. Behavior of piles in allu- methods. Technical report, Federal Highway Admin-
vium. Proceedings of the 8th International Confer- istration, Washington, D.C.
ence on Soil Mechanics and Foundation Engineering. Osterberg, J.O. 2001. Load testing high capacity piles:
Vol 2.1, Moscow, 47–52. What have we learned?. Proceedings of the 5th Inter-
Ellingwood, B. & Galambos, T.V. 1982. Probability- national Conference on Deep Foundation Practice,
based criteria for structural design. Structural Safety, Singapore.
1: 15–26. Paikowsky, S.G., Baecher, G.B., & Christian, J.T. 2003.
Fellenius, B.H. 2011. Capacity versus deformation Statistical issues of LRFD calibration for deep foun-
analysis for design of footings and piled foundations. dations. 12th Pan-American Conference on Soil Sym-
Geotechnical Engineering Journal of the SEAGS & posium, Vol. 2, 2839–2844.
AGSSEA, 42(2): 70–77. Paikowsky, S.G., Birgisson, B., McVay, M., Nguyen, T.,
Fenton, G.A. & Griffiths, D.V. 2008. Risk Assessment in Kuo, C., Baecher, G.B., Ayyub, B.M., Stenersen, K.,
Geotechnical Engineering. John Wiley and Sons. O’Malley, K., Chernauskas, L., & O’Neill, M. 2004.
Honjo, Y., Suzuki, M., Shirato, M., & Fukui, J. 2002. National cooperative highway research program—Re-
Determination of partial factors for a vertically loaded port 507: Load and resistance factor design (LRFD)
pile based on reliability analysis. Soils and Founda- for deep foundations. Technical report, Transporta-
tions: Journal of the Japanese Geotechnical Society, tion Research Board, Washington, D.C.
42(5): 91–109. Reese, L.C., Isenhower, W.M., & Wang, S.T. 2006. Analy-
Kuo, C.L., McVay, M.C., & Birgisson, B. 2002. Calibra- sis and Design of Shallow and Deep Foundations. John
tion of load and resistance factor design—Resistance Wiley and Sons.
factors for drilled shaft design. Transportation Research Russo, G., Recinto, B., Viggiani, C., & de-Sanctis, L.
Record: Journal of the Transportation Research Board, 2003. A contribution to analysis of Osterberg’s cell
1808: 108–111. load test. Proceedings of the 4th International Geotech-
LAW/CRANDALL, Inc. 1994a. Final report—Geo- nical Seminar on Deep Foundation on Bored and Auger
technical inspection services—Los Angeles Memorial Piles, Ghent, Belgium, 331–338.
Coliseum Repair. Los Angeles, California. Schmertmann, J.H., & Schmertmann, C.P. 2012. Testing
LAW/CRANDALL, Inc. 1994b. Report of Foundation and remediation observational method for the design
Investigation, Proposed Los Angeles Memorial Coli- and construction of non-redundant pile foundations.
seum Repair. Los Angeles, California. GeoCongress 2012 Full-Scale Testing and Foundation
LAW/CRANDALL, Inc. 1994c. Report of Pile Load Design, 349–361.
Testing, Los Angeles Memorial Coliseum Repair. Los Withiam, J.L., Voytko, E.P., Barker, R.M., Duncan, J.M.,
Angeles, California. Kelly, B.C., Musser, S.C., & Elias, V. 1997. Federal
McVay, M.C., Birgisson, B., Zhang, L., Perez, A., & Highway Administration—Report No. DTFH61-94-
Putcha, S. 2000. Load and resistance factor design C-00098: Load and resistance factor design (LRFD)
(LRFD) for driven piles using dynamic methods—A for highway bridge substructures. Technical report,
Florida perspective. Geotechnical Testing Journal, Federal Highway Administration, US Department of
23(1): 55–66. Transportation, Washington, D.C.
McVay, M.C., Birgisson, B., Nguyen, T., & Kuo, C.L. Yang, L. 2006. Reliability-based design and quality con-
2002. Uncertainty in load and resistance factor design trol of driven piles. Ph.D. thesis, University of Akron,
Phi factors for driven prestressed concrete piles. Trans- United States of America.
portation Research Record: Journal of the Transporta- Yoon, G.L., & O’Neill, M.W. 1997. Resistance factors for
tion Research Board, 1808: 99–107. single piles from experiments. Transportation Research
Neely, W.J. 1991. Bearing capacity of auger-cast piles in Record: Journal of the Transportation Research
sands. Journal of Geotechnical Engineering, 117(2): Board,1569(6): 47–54.
3–16.

2505

Fullpaper-Book-03-[CH201-CH300].indb
View publication stats 2505 8/22/2013 7:17:45 PM

You might also like