Você está na página 1de 21

Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

Contents lists available at ScienceDirect

Engineering Science and Technology,


an International Journal
journal homepage: www.elsevier.com/locate/jestch

Full Length Article

Probabilistic-based analysis of a shallow square footing using Monte


Carlo simulation
Ressol R. Shakir
Assistant Professor of Geotechnical Engineering, Civil Engineering Department, College of Engineering, University of Thi-Qar, Iraq

a r t i c l e i n f o a b s t r a c t

Article history: In this research, the bearing capacity (BC) of square shallow foundations at a depth of 0.5 m has been
Received 14 May 2018 computed with different width ranged from 1 m to 4 m utilizing direct approach based on cone tip resis-
Revised 24 July 2018 tance (qc). The qcs have been measured through performing the Cone Penetration Tests (CPTs) on soil at
Accepted 24 August 2018
four locations in the site of water intake project in Nasiriyah, Iraq as part of a soil site investigation pro-
Available online xxxx
gram. The suggested applied vertical loads were 50–550 kPa. The qcs below the foundation level were
considered uncertain and were simulated as random variables. Different probability distributions (PDs)
Keywords:
were used to simulate the qcs values: normal, logistic, lognormal, Gamma, Weibull, Inverse Gaussian
Bearing capacity (BC)
Reliability analysis
and Rayleigh distributions. Choosing the appropriate distribution was based on a goodness of fit (GOF)
Monte Carlo simulation (MCS) test. Monte Carlo simulation (MCS) was used to model the limit state function (LSF) based on a direct
Cone Penetration Test (CPT) solution. It was found that the bearing capacity is sensitive to both the coefficient of variation of qcs
Shallow foundation and applied vertical load. It also was found that different PDs for soil parameters may result in the dif-
ferent PD of BC. Varying the width of shallow foundations changes the values of qcs and their probability
distributions. According to this change, the reliability index of BC is slightly affected by the width of the
foundation. The reliability index shows a good value at stress 100 kN/m2 higher than the target reliability
and the reliability equal to zero at 550 kN/m2. A nonlinear behavior was observed between the probabil-
ity of bearing capacity failure and the coefficient of variation of the applied vertical load.
Ó 2018 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction Recently, the CPT has been used increasingly in Nasiriyah as an


essential field test in the subsurface exploration especially by Oil
Increasing the geotechnical structures constructed in Nasiriyah Company in Iraq due to vital projects intended to be constructed
which are located in southern Iraq through the last two decades in this city. CPT is a simple and fast test which gives a continuous
placed the engineers towards their responsibility to develop and and understandable reading [3–5] but it is expensive compared to
implement an efficient probability-based analysis and design. Shal- other field tests, in Iraq. The probabilistic analysis uses the contin-
low foundations are considered as the most geotechnical struc- uous data obtained through CPT efficiently since it needs much
tures encountered through analysis and design of buildings, a data. There are two approaches used to compute the BC of shallow
group of buildings and also projects of oil companies such as footings: direct approach and indirect approach. The indirect solu-
pipelines and water intake structures. A shallow foundation is tion involves obtaining the parameters of soil based on the results
most economical application commonly used to support small obtained from the CPT [6,7]. Then, they can be used to determine
and medium loading structures [1,2]. Different methods may be the soil BC using analytical and empirical equations. The direct
used to solve the bearing capacity (BC) of shallow foundations such solution is based on the qc which is measured through CPT at the
as analytical solution, practical solution based on parameters influence zone [8]. The BC of shallow foundations is calculated
obtained through in-situ testing methods, and full-scale loading using different empirical equations according to the type of soil
tests such as plate load test and presumed values presented by underneath the foundations. The traditional method used to com-
codes and handbooks. The field tests such as CPT are efficient pute the BC is called the deterministic solution which is based on
where good correlations are commonly used to identify the soil average soil parameters.
parameters. The uncertainties associated with predicting BC are implicitly
included in the deterministic solution in the global FOS. Despite a
E-mail address: rrshakir@eng.utq.edu.iq
long experience of using deterministic approach to solve the bearing

https://doi.org/10.1016/j.jestch.2018.08.011
2215-0986/Ó 2018 Karabuk University. Publishing services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
2 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

capacity (BC) of shallow foundations, it has not reflected the influ- MCS to solve the LSF of BC. This research will give a reliable design
ence of uncertainty of soil parameters accurately. The uncertainties for the foundation to support a water intake structure at an area
in the soil parameters can be included in the BC equations to imple- located at Nasiriyah city where very few studies have been pre-
ment the reliability analysis by taking the soil parameters as random sented on reliability analysis.
variables. Probabilistic-based design of shallow foundations is the
best method commonly used practically nowadays that reflect the 2. Project site
effect of the uncertainty in soil parameters accurately and preferred
by different design codes. The uncertainties in the geotechnical anal- 2.1. Geology of the site
ysis come from different sources such as sampling, laboratory test-
ing, and materials. Including the uncertainty of engineering The site is located 55 Kilometers towards North-West of Nasir-
geotechnical parameters in the design and analysis of foundations iya city. It is located in Mesopotamian Plain and close to Euphrates
is very seldom on the soil in this city. The reliability-based analysis River. The floodplain of Mesopotamia contains Quaternary fluvial
of foundation has been studied by several authors using the analyt- sediments of the interacting Tigris and Euphrates rivers, integra-
ical equation of BC based on an approximate solution [9–22]. This tion into a marshland north of Basra and the Shatt Al Arab delta
BC of soil can be determined based on soil parameters obtained plain between Basra and the Arabian/Persian Gulf. The floodplain
through laboratory testing or field tests such as SPT and CPT. Differ- of the Tigris and the Euphrates is a relatively flat area with a low
ent equations used to compute the BC such as Tand’s equation topographic relief and occupies a part of the foreland basin of the
depend on qcs obtained from the results of CPT are available in the Zagros fold and thrust belt [31,32]. Through this basin, the two riv-
literature [e.g., 23]. However, few studies used the direct solution ers which originate from Turkey, are draining from the northwest
to compute the reliability of the BC of the shallow foundations partic- to south-east. Yearly, and since 12000 BP, Million tons of sediment
ularly in the soil in this city. transported through revisers to the region of floodplain which con-
Characterizing the variability of soil parameters and their effect tains silt, clay, and sand. The silt forms 60% of the sediment and the
on soil BC was reviewed by Kayser and Gajan (2017) [24]. The mag- reminder contains clay and sand [33]. The silt and sand are depos-
nitude and coefficient of variation of qc play a paramount role in ited in marshes while clay passes down to Shatt-al-Arab.
determining the bearing capacity. Although the range of coefficient
of variation of qc is available in the literature [25] for different 2.2. CPT-based soil classification
types of soil (e.g., for clay, COV = 20% - 40%); it has to characterize
COV of qc for every case individually [26]. As the width of the foun- The CPT is used to determine the geotechnical engineering
dation increases, the influenced depth also increases making the properties of soils and delineating soil stratigraphy. It is increas-
shape of variation of the qc differs based on the type of soil. No lit- ingly used in Iraq in the important projects since it performs fast,
erature available as we know presents a study on the effect of the continuously measures the resistance and friction and directly
width of the foundation on the bearing capacity for the soil in this gives a detailed picture of the soil profile [3]. The test uses a cylin-
city. In addition to that, no range for the type of probability distri- drical penetrometer of 3.6 cm diameter with a cross-sectional area
bution of the qc at different depth is presented which make neces- of 10 cm2, combined with a cone with 60° cone apex [4]. It is
sary to find the suitable probability distribution for analysis before pushed into the soil at a rate of 1 cm/s. Since the geology of the soil
computing the bearing capacity. Many studies taking the parame- refer to the type of soil that is relatively soft to medium soil, and
ters of soil to compute the resistance taking the load as constant. the drilling is suggested up to 15 m, the type of projected area of
However, the load also uncertain and uncertainty of loads should 10 cm2 can perform the test successfully.
be taken into consideration. Monte Carlo simulation is used to The soil can be classified based on the components measured
compute the probability of failure and the reliability index to eval- through CPT which are qc and side friction (fs). The soil behavior
uate the bearing capacity. The reliability index equal to 3 was con- classification index (ISBT) is directly linked to the soil type [3,4]
sidered the most appropriate value for the ultimate BC of shallow which can be defined as.
foundations [11]. Reliability-based analysis and design for shallow qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
foundations have been investigated considerably by [27–29]. Some 2  2
ISBT ¼ ðð3:47Þ  logðqc =pa ÞÞ þ logRf þ 1:22 ð1Þ
of these studies considered the uncertainty of one parameter and
some of them considered the parameters as two variables. where pa = the atmospheric pressure, Rf % = (fs/qc)100, and qc = the
In this research, the reliability of BC of the shallow foundation is cone tip resistance. Several methods are available [34–37] to iden-
studied using a direct approach based on CPT. Square footings of tify the boundary of soil layers such as those based on probability
different width ranges from 1 m to 4 m are considered. The foot- concept [38], wavelet transform modulus maxima [39] clustering
ings intended to be constructed onto the soil in Nasiriyah are com- analysis and statistical analysis [40], and probabilistic soil classifica-
puted taking into account the probability analysis. An equation of tion system [41–44]. The method used in this paper is the statistical
BC based on qc for clay is used where the qc is assumed as a random approach which uses homogeneous soil units (HSUs) [40] since it is
variable. The model of BC equation is assumed perfect model (per- efficient and relatively simple. Selecting the suitable equation of
fect predictor). The probabilistic approaches use the Monte Carlo bearing capacity is based on the type of soil behavior.
simulation (MCS) as a more reasonable estimation. This task is per- Identification the HSUs is an important step in variability anal-
formed in two stages. The first stage contains choosing the best PD. ysis. The soil is considered homogeneous when it is fairly uniform
The qc below the base of the foundation is expressed into two in composition or behavior. The homogeneous soil units are pro-
depths the first 0.5 B below the base and the second represents vided by a statistical theory which is based on CPT [45–47]. Based
0.5–1.5 B below the base of the foundation. These two parameters on the CPT data, the mechanical response to the penetration can be
are presented using the best-fitted distribution. The Tand equation investigated and the homogeneous in soil behavior against pene-
[30] is used to compute the ultimate BC. The reliability index (b) is tration is recognized. Assuming of statistical homogeneity is essen-
also obtained for different applied vertical loads and foundation tial for applying statistical techniques. The data obtained from the
widths. Comparison between allowable bearing capacities based CPT test are screened based on two assumptions (a) the measure-
on a FOS is achieved with the BC based on b value. The study uses ment interval is 0.01 m and (b) the length of sounding is more than
sixteen case of foundation width and 10 cases of an applied load 2 m. The first assumption to ensure the accuracy and the second to
implemented by Matlab software developed for this study to use ensure the stationary. The accuracy of data used in this study

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 3

(0.01 m) is better than that used in literature which permit the 3. Methodology
accuracy of data to be less than 0.05 m. To identify the HSUs the
soundings are delineated by sections where COV of Ic is less than 3.1. BC for a shallow foundation based on CPT results
0.1 [48]. The degree of homogeneity of the soil unit depends on
the coefficient of variation of Ic which should not be greater than A direct relationship between qc measured by CPT and BC of
0.1 [40]. As the COV decreases the degree of homogeneity shallow foundations is available for more than two decades. The
increases. commonly used direct BC equation for shallow spread footing on
clay is that presented by Tand et al. (1986) [30]. It is used in this
study since the soil at the shallow depth is clay soil as discussed
2.3. Soil type at shallow depth
in Section 2.3. The following equation can be used to calculate
the bearing capacity of a shallow foundation
The CPTs were performed at four locations (Fig. 1) on the site of
 
the project which is located at 55 Kilometers towards North-West
qult ¼ rv o þ Rk : qc  rv o ð2Þ
of Nasiriyah, Iraq. The depth of CPT1, CPT2, CPT3, and CPT4 reaches
15 m depth. There are also several boreholes were drilled however
where Rk can be obtained using a graph obtained by Tand which
the focus in this article was on the CPTs. Two components can be
consists of two curves based on the embedment ratio Rd = De/B;
measured through the CPT: qc and sleeve resistance (fs) which rep-
De = depth of embedment and B = foundation width. The equivalent
resent the main elements in soil classification. The stratification of
embedment ratio is implemented to account for the effect of strong
soil can be obtained successfully through the soil profile obtained
or weak soil above the bearing elevation which can be calculated as
based on CPT taking into account both the development and sens-
follow [53–56].
ing distances that limit the thickness of identified layers [49,50].
Considering a 6 m depth below the base of foundations is suffi- X
n
qci Dzi
cient to study the BC of foundations with width ranges from 1 m to De ¼  ð3Þ
i¼1 qc
4 m. The depth of shallow soil, influenced by the vertical load
applied on the foundation, extends to a depth equal to about 1.5 
where qci is the reading of cone tip resistance at every depth; qc is
times the width of foundation from the base since the influence
the equivalent qc at 1.5 B below the base of foundation, and Dzi is
of the applied stress intensity reaches to this depth. Fig. 2 shows
the interval depth between every two successive reading.
the distribution of qc with the depth of 7.0 m from the ground sur-
The curves developed to indicate the value of Rk are limited to
face for soils at four locations CP1, CP2, CPT3, and CPT4. Figs. 3 and
several types of soil [30] such as fissured and intact clay while
4 show the fs distribution with the depth of soil and the relation
there are many types of soil is not covered such as silt soil, silty
between Rf% and the depth of soil for four CPTs, respectively.
clay soil, and clayey silt soil. Moreover, several cases are presented
For the qc at a depth of 0.5 m in the soil located at CPT1, the type
between intact clay and fissured clay where the curve of fissured
of soil is classified as sand mixtures (silty sand to sandy silt)
clay is the loa one. Therefore, the curve of fissured clay was used
according to the classification system developed by Robertson
as a conservative value and was curve-fitted by the following equa-
since Ic = 2.3 [51,52]. The soil that has Ic in the range of 2.05 and
tion based on the embedment ratio (Rd) presented as follow.
2.60 is classified as sand mixtures [51,52,23]. The soil layer at a
shallow depth between 0.5 m and 5 m, located at CPT1, is classified Rk ¼ 0:3121 þ 0:0476Rd  0:0039R2d ð4Þ
as clay soil (Table 1) since the Ic average is 3.1, rIc = 0.21 and the
COV = 0.07 [3]. The soil layer depth from 0.5 m to 7 m gives the The equivalent qc can be calculated based on the following
same type of soil. The Ic distribution with depth is presented in equation:
Fig. 5(a–d) which represent the type of the soil. The two lines of qffiffiffiffiffiffiffiffiffiffiffiffiffi
  
Ic = 2.60 and Ic = 3.6 shown in the Fig. 5(a–d) represent zone 3 qc ¼ qc1 qc2 ð5Þ
and zone 4 (clay, silty clay, clayey silt to silty clay) [8,40]. Identify-
ing the type of the soil is an essential step to select a suitable BC Which represents the qc mean at two intervals: the first interval is
equation and to make a robust analysis and design. under the base of foundation for depth of 0.5 B and the second

Fig. 1. CPT locations at the site of soil. (a) Sketch of four CPT at four locations (b) Google earth image.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
4 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

Fig. 2. The cone tip resistance (qc) versus depth of soil at shallow depth in four CPTs.

Fig. 3. Distribution of the fs (kPa) with depth of soil for four CPTs.

Fig. 4. Distribution of the Rf% with depth for four CPTs.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 5

Table 1
Distributions suggested for the qc1, and qc2.

Type Distribution for qc1 Distribution for qc2


    2     q l 2 
Normal
f qc1 qc1 ; lqc1 ; rqc1 ¼ r 1pffiffiffiffi ffi exp 0:5 qc1rlqc1 f qc2 qc2 ; lqc2 ; rqc2 ¼ r 1pffiffiffiffiffi
exp 0:5 c2rqc2qc2
qc1 2p qc1 qc2 2p
"
2 #    q l 2 
Lognormal   q l
f qc2 qc2 ; llnðqc2 Þ ; rlnðqc2 Þ ¼ r 1 pffiffiffiffiffi c2 lnðqc2 Þ
f qc1 qc1 ; llnðqc1 Þ ; rlnðqc1 Þ ¼ r 1 pffiffiffiffi ffi exp 0:5 c1r lnqc1 2p
exp 0:5 rlnðqc2Þ
2plnðqc1 Þ ln ðqc1Þ
lnðqc2 Þ

Logistic    q l     q l 2     q l     q l 2 


f qc1 ; lqc1 ; sqc1 ¼ exp  sq 1 = sqc11 1 þ exp  sq c1 f qc2 ; lqc2 ; sqc2 ¼ exp  sq c2 = sqc2 1 þ exp  sq c2
c1 QC c1 q c2 q c2 q

c1 c1 c2 c2

Weibull    kqc1 1 h  k i   kqc2


 kq 1 h   i
kqc1 qc1 qc2 c2 kqc2
f qc1 ; kqc1 ; kqc1 ¼ kqc1 kqc1 exp  qc1 =kqc1 qc1 f qc2 ; kqc2 ; kqc2 ¼ kqc2 kqc2 exp  qc2 =kqc2
a1 bðqc1 Þ
Gamma a a
ðqc2 Þa1 ebðqc2 Þ
f ðqc1 ; a; bÞ ¼ b ðqc1 ÞCðaeÞ f ðqc2 ; a; bÞ ¼ b CðaÞ
h i1=2 8 h i2 9
Inverse k½ðqc1 Þlqc1  2 h i1=2 <k ðqc2 Þlq =
f ðqc1 ; l; kÞ ¼ k
exp 2l2 ðq Þ
2pðqc1 Þ f ðqc2 ; l; kÞ ¼
3
k c2
qc1 c1 exp
2pðqc2 Þ3 : 2lqc2 ðqc2 Þ ;
2

 2  
Rayleigh q
f ðqc1 ; rÞ ¼ rq2c1 exp 2rc12 f ðqc2 ; rÞ ¼ rq2c2 exp
q2c2
qc1 1 qc2 2r22

Fig. 5. Distribution of the Ic with depth for four CPTs (a) CPT1, (b) CPT2, (c) CPT3, and (d) CPT4.

1 XL
nn
interval is between 0.5 B below the base of the foundation and 1.5 B.
qc2 ¼ q ð7Þ
These two quantities are presented below: n  nL i¼n cið0:5B1:5BÞ

1 X where nL is the number of qc data at the half the width of the founda-
nL

qc1 ¼ q ð6Þ tion, and n-nL is the number of qc data at the depth between half times
nL i¼1 cið00:5BÞ
the width of foundation and 1.5 times the width of the foundation.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
6 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

3.2. Common probability distributions 3.4. Monte Carlo simulation

Despite the frequent use of the lognormal distribution to repre- Monte Carlo simulation is a mathematical tool used to estimate
sent the geotechnical parameters [27] the other distributions such the probability of a specific event which is the consequence of a
as Gamma, Weibull, and normal can successfully represent succession of stochastic processes [13,61]. In this method, the
geotechnical parameters [57,58]. Using the experience to get an deterministic solution with random values of the parameters in
indication to choose the PD to represent the data is the worth step, the model has repeated millions of samples. Through this stage,
but the worst is to ignore the other distributions and do not give a the Pf is computed for the case of a different load less than the ulti-
chance to be selected. The visual inspection of the PDs with the mate BC [62,63]. The results are not always stable; they depend on
actual data and using GOF methods can provide the best distribu- the number of realizations for each trial solution. Since the vari-
tion. It is essential to select the best PD since it may affect the ables are just two random variables, using crude MCS represents
results obtained from applying the limit state function (LSF) for a suitable choice. The LSF of soil BC is defined as follows:
the engineering application and give different reliability index. In
this research, seven PDs are equally given a chance to represent G ¼ Rðqc Þ  S ð9Þ
the data. These PDs are expressed by normal (Nr), logistic (Lt), log-
where R is the resistance and S is an applied load of a structure,
normal (Lg), Gamma (Ga), Weibull (We), inverse Gaussian (IG), and
G = F(R, S) is the variable parameters. The Pf can be defined as
Rayleigh distribution (Ra).
The two variables qc1 and qc2 are included in the equation of BC Pf ¼ P ½G < 0 ð10Þ
of shallow foundation and considered as random variables. The
random variation of the first variable qc1 is distributed along the The qc obtained from the CPT may be represented by different
layer having a thickness equal to half the width of foundation distributions as the considered soil depth increases. Including
and the second variable is spread along thickness equal to the new data to the original data of qc as the depth increases may alter
width of the foundation. The seven distributions above- the PD. A set of samples of the qcs are included in the equation of
mentioned are used to simulate the two variables Table 2. Accord-
ing to the goodness of fit test, the best distribution of data can be
selected

3.3. Selecting the suitable probability distribution

Each vector of qc1 and qc2 in addition to the resulted BC are sim-
ulated using seven types of PDs. The BC depends on the average
values of the two vectors of qc1 and qc2 at layers having two thick-
nesses 0.5 B and 1 B below the base of foundation respectively.
According to the GOF test, the best distribution of data can be
selected utilizing a code written based on Matlab software. The
selected GOF method is used three times in any solved problem.
The first and second are to get the PDs of qc1, qc2 and the third is
to develop the PDs the bearing capacity.
Different types of GOFs are used in the literature [59,60] for
hypothesis testing and assessing the suitability of the PDs, the
coefficient of determination (R2) is used widely. It computes the
correlation between the observed and predicted cumulative prob-
abilities of the qc data. It is the ratio of the variance of predictable
data to that of observed data. R2 ranges from zero to 1. A higher
value of R2 indicates the best fit of the model to the observed
cumulated probabilities. The following equation defines the R2
for the qc1 data.
PN  0  2
i¼1 F ðqc1 ÞP  F qc1 P
R2 ¼ PN  0  2 PN   2 ð8Þ
i¼1 F ðqc1 ÞP  F qc1 P þ i¼1 F ðqc1 ÞE  F q0c1 P

where (qc1)P is the cumulative predictive probability (theoretical


CDF), (qc1)E is the empirical cumulative probability, and q‘c1 repre-
sents the average cumulative predictive probability. Fig. 6. The schematic picture for the stated problem of a shallow foundation.

Table 2
The statistical description of Ic for soil thickness 0.5–7 m.

CPT Thickness lIc rIc COVIc


1 0.5–5 3.10 0.21 0.070
2 0.5–5 3.22 0.20 0.060
3 0.5–5 3.13 0.13 0.040
4 0.5–5 3.09 0.19 0.060
1 0.5–7 3.08 0.234 0.076
2 0.5–7 3.30 0.257 0.078
3 0.5–7 3.15 0.170 0.054
4 0.5–7 3.12 0.264 0.085

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 7

BC are generated according to their probability distribution. The Pf G < 0 represents failure region. The LSF is presented as the differ-
can be determined based on the LSF which represents the mathe- ence between the ultimate soil BC (qult) and the applied stress
matical model G (.) based on the qc1 and qc2 as well as to the load (qapp)
(S). The probabilistic characteristics of the response on the square
foundation are obtained after the simulation utilizing the sampling G ¼ qult  qapp ð11Þ
set is performed. If the limit state function is less than zero i.e. the
applied load is greater than the resistance, the failure occurs. The G ¼ ½rv o þ Rk  ðqt  rv o Þ  qapp ð12Þ
trial is repeated many times until failure occurs. The function
h  i
G = 0, states the failure surface, G > 0, represents safe region, and G¼ rv o þ Rk  ðqc1 qc2 Þ1=2  rv o  qapp ð13Þ

Fig. 7. Flowchart of calculation bearing capacity based on cone tip resistance with the selection of the suitable probability distribution for qc1 and qc2 in addition to the
computing of the bearing capacity based on Monte Carlo simulation (m = 100,000).

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
8 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

The solution is repeated many times to ensure convergence of 1 XN


Nf
the statistical results. In every trial solution, the values can be ana- Pf ¼ IðGÞ ¼ ¼ l I ¼ l pf ð16Þ
N i¼1 N
lyzed. The Pf is then calculated by the following equation
Pf is the probability of failure; N is (total number of trial) the
Pf ¼ P½G 6 0 ¼ E½IðGÞ ð14Þ
conducted trial and Nf is number of trials for which qc(.) violence
where I (.) is an indicator function, which is defined by outside. The above equation is the main form of reliability. The
expectation of the indicator for the failure event is just the proba-
1 if G 6 0
IðGÞ ¼ ð15Þ bility that failure occurs hence, the variance is expressed in Eq. (17)
0 otherwise which is expressed in term of Pf as in Eq. (18).

Table 3
Ultimate bearing capacity computed for width equal to 1 to 4 m (CPT1).

No B (m) start(m) End(m) qt(MPa)Av. Deq Rd Rk qult


1 1 0.5 2 1752.578 0.981 0.98 0.36 620.847
2 1.2 0.5 2.3 1787.301 0.962 0.8 0.35 621.642
3 1.4 0.5 2.6 1755.315 0.98 0.7 0.34 594.236
4 1.6 0.5 2.9 1761.986 0.976 0.61 0.34 580.630
5 1.8 0.5 3.2 1743.535 0.986 0.55 0.34 574.030
6 2 0.5 3.5 1697.105 1.013 0.51 0.34 564.129
7 2.2 0.5 3.8 1688.431 1.018 0.46 0.33 570.728
8 2.4 0.5 4.1 1661.508 1.035 0.43 0.33 564.127
9 2.6 0.5 4.4 1655 1.039 0.4 0.33 560.825
10 2.8 0.5 4.7 1718.604 1.001 0.36 0.33 587.223
11 3 0.5 5 1754.362 0.98 0.33 0.33 600.422
12 3.2 0.5 5.3 1793.27 0.959 0.3 0.33 613.620
13 3.4 0.5 5.6 1755.407 0.98 0.29 0.33 576.412
14 3.6 0.5 5.9 1741.465 0.987 0.27 0.32 570.010
15 3.8 0.5 6.2 1738.134 0.989 0.26 0.32 570.008
16 4 0.5 6.5 1716.613 1.002 0.25 0.32 563.606

Table 4
Ultimate bearing capacity computed for width equal to 1 to 4 m (CPT2).

No B(m) start(m) End(m) qt(MPa)Av Deq Rd Rk qult


1 1 0.5 2 1630.463 1.103 1.1 0.36 594.06
2 1.2 0.5 2.3 1632.14 1.102 0.92 0.35 582.78
3 1.4 0.5 2.6 1582.599 1.136 0.81 0.35 558.48
4 1.6 0.5 2.9 1554.009 1.157 0.72 0.34 542.85
5 1.8 0.5 3.2 1503.412 1.196 0.66 0.34 521.73
6 2 0.5 3.5 1532.289 1.174 0.59 0.34 526.56
7 2.2 0.5 3.8 1576.643 1.141 0.52 0.34 536.94
8 2.4 0.5 4.1 1601.968 1.123 0.47 0.33 541.92
9 2.6 0.5 4.4 1580.344 1.138 0.44 0.33 532.62
10 2.8 0.5 4.7 1564.637 1.149 0.41 0.33 525.54
11 3 0.5 5 1579.104 1.139 0.38 0.33 528.18
12 3.2 0.5 5.3 1573.499 1.143 0.36 0.33 524.76
13 3.4 0.5 5.6 1555.916 1.156 0.34 0.33 517.77
14 3.6 0.5 5.9 1526.494 1.178 0.33 0.33 507.27
15 3.8 0.5 6.2 1501.785 1.198 0.32 0.33 498.36
16 4 0.5 6.5 1480.513 1.215 0.3 0.33 490.65

Table 5
Ultimate bearing capacity computed for width equal to 1 to 4 m (CPT3).

No B(m) start(m) End(m) qt(MPa)Av Deq Rd Rk qult


1 1 0.5 2 1896.201 0.919 0.92 0.35 675.66
2 1.2 0.5 2.3 1919.975 0.907 0.76 0.35 671.28
3 1.4 0.5 2.6 1884.081 0.925 0.66 0.34 651.33
4 1.6 0.5 2.9 1845.237 0.944 0.59 0.34 632.55
5 1.8 0.5 3.2 1786.333 0.975 0.54 0.34 608.88
6 2 0.5 3.5 1785.959 0.975 0.49 0.33 604.56
7 2.2 0.5 3.8 1779.775 0.979 0.44 0.33 599.19
8 2.4 0.5 4.1 1772.248 0.983 0.41 0.33 593.91
9 2.6 0.5 4.4 1764.018 0.988 0.38 0.33 588.87
10 2.8 0.5 4.7 1793.352 0.971 0.35 0.33 595.92
11 3 0.5 5 1815.936 0.959 0.32 0.33 601.11
12 3.2 0.5 5.3 1833.749 0.95 0.3 0.33 605.07
13 3.4 0.5 5.6 1816.128 0.959 0.28 0.33 598.11
14 3.6 0.5 5.9 1803.778 0.966 0.27 0.32 592.98
15 3.8 0.5 6.2 1798.695 0.969 0.25 0.32 590.22
16 4 0.5 6.5 1777.881 0.98 0.24 0.32 582.69

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 9

h i
Var ½IðGÞ ¼ E IðGÞ2  fE½IðGÞg2 ð17Þ a vast number of realization [59]. The simulation object is to com-
pute the mean, deviation, and probabilities associated with the
  response of BC of the soil for random inputs (qc1 and qc2) and study
Var ½IðGÞ ¼ E½IðGÞf1  E½IðGÞg ¼ Pf 1  Pf ð18Þ the effect of the sensitivity of input variables on the BC.

The coefficient of variation of the failure of bearing capacity can


be calculated based on the square root of variance and the expec- 3.5. The reliability index (b) and probability of failure (Pf)
tation of the I (G).
Most of the civil engineering applications interested with a Pf Traditionally a value of FOS equal to 3 based on experience is
are between 1/1000 and 1/100 000 (0.001–0.0 0001) which require used commonly in the design of shallow foundations. The

Table 6
Ultimate bearing capacity computed for width equal to 1 to 4 m (CPT4).

No B(m) tart(m) End(m) qt(MPa)Av Deq Rd Rk qult


1 1 0.5 2 2312.092 0.737 0.74 0.35 805.14
2 1.2 0.5 2.3 2341.821 0.727 0.61 0.34 802.5
3 1.4 0.5 2.6 2311.543 0.737 0.53 0.34 784.29
4 1.6 0.5 2.9 2219.538 0.767 0.48 0.33 748.86
5 1.8 0.5 3.2 2116.086 0.805 0.45 0.33 711.3
6 2 0.5 3.5 2134.927 0.798 0.4 0.33 713.04
7 2.2 0.5 3.8 2085.764 0.817 0.37 0.33 694.23
8 2.4 0.5 4.1 2064.445 0.825 0.34 0.33 684.69
9 2.6 0.5 4.4 2067.939 0.824 0.32 0.33 683.34
10 2.8 0.5 4.7 2106.078 0.809 0.29 0.33 693.15
11 3 0.5 5 2120.929 0.803 0.27 0.32 695.94
12 3.2 0.5 5.3 2139.382 0.796 0.25 0.32 700.11
13 3.4 0.5 5.6 2143.213 0.795 0.23 0.32 699.87
14 3.6 0.5 5.9 2150.739 0.792 0.22 0.32 700.95
15 3.8 0.5 6.2 2162.745 0.787 0.21 0.32 703.56
16 4 0.5 6.5 2142.842 0.795 0.2 0.32 696.33

Table 7a
The R2 of predicted (qc1) at CPT1.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.938043 0.957163 0.953602 0.947755 0.923059 0.95234 0.63624 0.957163 Lt
2 1.2 0.896769 0.931195 0.916562 0.909215 0.883658 0.914973 0.661582 0.931195 Lt
3 1.4 0.899453 0.935507 0.922668 0.914689 0.881199 0.921468 0.665068 0.935507 Lt
4 1.6 0.93754 0.956752 0.796821 0.882949 0.925055 0.687743 0.706797 0.956752 Lt
5 1.8 0.954878 0.965654 0.838655 0.916887 0.944883 0.724167 0.733855 0.965654 Lt
6 2 0.961758 0.975011 0.866475 0.91157 0.96092 0.793254 0.819527 0.975011 Lt
7 2.2 0.952329 0.964989 0.872452 0.902901 0.946397 0.824778 0.871498 0.964989 Lt
8 2.4 0.96746 0.974041 0.900802 0.928429 0.962544 0.854258 0.893383 0.974041 Lt
9 2.6 0.963038 0.971546 0.894982 0.923483 0.959661 0.850355 0.879325 0.971546 Lt
10 2.8 0.957881 0.970139 0.884388 0.914234 0.955906 0.839005 0.860591 0.970139 Lt
11 3 0.955929 0.971289 0.877805 0.90898 0.955667 0.831614 0.846943 0.971289 Lt
12 3.2 0.959567 0.976599 0.876555 0.909285 0.960841 0.827476 0.838632 0.976599 Lt
13 3.4 0.955968 0.976328 0.869256 0.902971 0.959123 0.81903 0.823866 0.976328 Lt
14 3.6 0.952955 0.974022 0.868194 0.901509 0.957234 0.819787 0.817438 0.974022 Lt
15 3.8 0.96201 0.98014 0.879889 0.913332 0.965967 0.831743 0.822019 0.98014 Lt
16 4 0.963311 0.981458 0.882262 0.915803 0.967443 0.834984 0.818571 0.981458 Lt

Table 7b
The R2 of predicted (qc2) at CPT1.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.947244 0.954478 0.886549 0.909337 0.941339 0.840851 0.880552 0.954478 Lt
2 1.2 0.948118 0.96101 0.874767 0.902065 0.944428 0.82388 0.857588 0.96101 Lt
3 1.4 0.970635 0.980268 0.893866 0.925571 0.969952 0.83825 0.852002 0.980268 Lt
4 1.6 0.960931 0.978115 0.894542 0.918965 0.968773 0.881509 0.805859 0.978115 Lt
5 1.8 0.966838 0.984141 0.902207 0.926259 0.978778 0.889415 0.782362 0.984141 Lt
6 2 0.988784 0.995036 0.942514 0.962483 0.995831 0.932402 0.756264 0.995831 We
7 2.2 0.9918 0.989611 0.982111 0.986383 0.993988 0.981333 0.683737 0.993988 We
8 2.4 0.985455 0.982395 0.984751 0.985514 0.98362 0.984679 0.684717 0.985514 Ga
9 2.6 0.975531 0.973456 0.976342 0.97652 0.974022 0.976304 0.689213 0.97652 Ga
10 2.8 0.909021 0.981442 0.9729 0.956566 0.886317 0.968867 0.79506 0.981442 Lt
11 3 0.892248 0.972183 0.962405 0.942208 0.883846 0.957485 0.823671 0.972183 Lt
12 3.2 0.891708 0.963155 0.958958 0.938928 0.891154 0.954176 0.848867 0.963155 Lt
13 3.4 0.919592 0.973343 0.971921 0.960081 0.92392 0.969702 0.896956 0.973343 Lt
14 3.6 0.923003 0.974272 0.975869 0.963536 0.927503 0.973713 0.901492 0.975869 LG
15 3.8 0.938516 0.97874 0.983414 0.974229 0.946353 0.981951 0.92682 0.983414 LG
16 4 0.938595 0.979504 0.985923 0.975923 0.948158 0.984468 0.932716 0.985923 LG

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
10 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

allowable soil BC is determined by dividing the ultimate BC by the sents the shortest distance between the middle points of the ran-
FOS. The FOS represents a general factor which is assumed to dom result and the LSF [66,67]. The reliability index i.e. safety
include all sources of variability and uncertainty inherent in the index of limit state can be computed based on Pf which is obtained
geotechnical analysis [64] but it does not mean precisely that the by applying MCS and using the following equation
uncertainties in the soil parameters are determined accurately
 
[65]. Therefore reliability-based analysis and design are used to b ¼ u1 1  Pf ð19Þ
study the effect of variability on the safety of geotechnical engi-
neering. The reliability index (b) is a measure of safety that takes where U1(.) is the inverse normal variable CDF. It is clear that the
into account the uncertainty in the input variables [61]. It repre- large b leads to small Pf and vice versa.

Table 7c
The R2 of predicted (qc1) at CPT2.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.881923 0.904956 0.905109 0.896315 0.878988 0.903185 0.764464 0.905109 LG
2 1.2 0.948147 0.959112 0.972892 0.966509 0.944931 0.973296 0.844462 0.973296 IG
3 1.4 0.97571 0.978101 0.942002 0.961524 0.974379 0.929805 0.934429 0.978101 Lt
4 1.6 0.968386 0.969441 0.92673 0.947397 0.962471 0.916261 0.957191 0.969441 Lt
5 1.8 0.96808 0.969602 0.945651 0.959599 0.968115 0.940331 0.971563 0.971563 Ra
6 2 0.975888 0.979992 0.966513 0.977064 0.981205 0.962579 0.984101 0.984101 Ra
7 2.2 0.976967 0.978983 0.952936 0.967802 0.977081 0.947062 0.969862 0.978983 Lt
8 2.4 0.978124 0.979108 0.94382 0.961255 0.975126 0.936222 0.95648 0.979108 Lt
9 2.6 0.976613 0.977662 0.93581 0.954681 0.972226 0.927077 0.941864 0.977662 Lt
10 2.8 0.969187 0.971825 0.927607 0.946481 0.964958 0.918907 0.92447 0.971825 Lt
11 3 0.962922 0.967545 0.916675 0.936827 0.958388 0.907113 0.904826 0.967545 Lt
12 3.2 0.965436 0.972237 0.917753 0.938727 0.960866 0.907838 0.895608 0.972237 Lt
13 3.4 0.961469 0.96942 0.908908 0.931267 0.95673 0.896922 0.8876 0.96942 Lt
14 3.6 0.961839 0.970924 0.904923 0.928314 0.957623 0.891684 0.876942 0.970924 Lt
15 3.8 0.96201 0.98014 0.879889 0.913332 0.965967 0.831743 0.822019 0.98014 Lt
16 4 0.963311 0.981458 0.882262 0.915803 0.967443 0.834984 0.818571 0.981458 Lt

Table 7d
The R2 of predicted (qc2) at CPT2.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.941105 0.945059 0.902035 0.916902 0.934488 0.894348 0.906447 0.945059 Lt
2 1.2 0.914797 0.92954 0.860854 0.878607 0.907995 0.848378 0.857588 0.92954 Lt
3 1.4 0.920764 0.936811 0.866389 0.885172 0.920418 0.852517 0.819943 0.936811 Lt
4 1.6 0.949061 0.957906 0.910893 0.924942 0.953651 0.902331 0.806398 0.957906 Lt
5 1.8 0.967639 0.966471 0.936994 0.953166 0.969216 0.912921 0.79959 0.969216 We
6 2 0.955484 0.963552 0.9086 0.928672 0.962243 0.883241 0.775747 0.963552 Lt
7 2.2 0.950655 0.961389 0.901125 0.92147 0.958535 0.87628 0.773174 0.961389 Lt
8 2.4 0.95092 0.961001 0.902339 0.922635 0.958534 0.876634 0.758836 0.961001 Lt
9 2.6 0.966227 0.966505 0.921471 0.945219 0.966977 0.872851 0.813579 0.966977 We
10 2.8 0.966839 0.965638 0.9326 0.95132 0.966269 0.894179 0.843786 0.966839 N
11 3 0.972439 0.971582 0.937994 0.956539 0.972521 0.901743 0.848071 0.972521 We
12 3.2 0.976081 0.975976 0.93559 0.956644 0.976435 0.898028 0.848757 0.976435 We
13 3.4 0.977484 0.975622 0.952636 0.967107 0.977693 0.924577 0.87224 0.977693 We
14 3.6 0.975076 0.973357 0.960931 0.970386 0.97572 0.943327 0.918958 0.97572 We
15 3.8 0.9786144 0.976955 0.969192 0.978958 0.981289 0.941930 0.948796 0.981289 We
16 4 0.9797746 0.978614 0.977512 0.984932 0.984217 0.955124 0.960655 0.984932 Ga

Table 7e
The R2 of predicted (qc1) at CPT3.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.917614 0.942025 0.940664 0.932964 0.904607 0.939534 0.708076 0.942025 Lt
2 1.2 0.96359 0.976257 0.980832 0.976625 0.954149 0.980761 0.770547 0.980832 LG
3 1.4 0.983432 0.986616 0.977678 0.982968 0.981689 0.976551 0.859895 0.986616 Lt
4 1.6 0.983231 0.984326 0.974847 0.981095 0.984185 0.973415 0.906079 0.984326 Lt
5 1.8 0.986503 0.990746 0.986867 0.990522 0.986743 0.985971 0.896081 0.990746 Lt
6 2 0.980943 0.988194 0.993941 0.992659 0.982832 0.994042 0.903691 0.994042 IG
7 2.2 0.975108 0.986415 0.994732 0.991035 0.976032 0.994949 0.893035 0.994949 IG
8 2.4 0.980409 0.991119 0.994749 0.993322 0.978339 0.994687 0.875922 0.994749 LG
9 2.6 0.989164 0.992497 0.993777 0.995147 0.987778 0.993484 0.86825 0.995147 Ga
10 2.8 0.991956 0.990824 0.990068 0.993191 0.99215 0.989462 0.862179 0.993191 Ga
11 3 0.991023 0.988023 0.983018 0.987907 0.992059 0.981974 0.84757 0.992059 we
12 3.2 0.986917 0.98351 0.97389 0.980248 0.988725 0.97243 0.830936 0.988725 Lt
13 3.4 0.982515 0.979144 0.966197 0.973452 0.985352 0.964355 0.81664 0.985352 we
14 3.6 0.978731 0.975837 0.960113 0.967967 0.982412 0.958013 0.804607 0.982412 we
15 3.8 0.97702 0.975611 0.954878 0.963794 0.981502 0.952421 0.791642 0.981502 we
16 4 0.980704 0.980401 0.957509 0.966883 0.984845 0.954907 0.783791 0.984845 we

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 11

For foundations, the reliability index which is used as a target 4. Results and discussion
by AASHTO is between 2 and 3.5 [68]. The Canadian Building Code
CBC uses a target reliability index of 3.5. For the case when b equal The study covers computing the BC of shallow foundation based
to 3, the performance level is above average and when b equal to 4, on two solutions: the first is the deterministic solution and the sec-
worthy performance is indicated, where the probability failure are ond is the probabilistic-based solution. The first solution implies
0.001 and 0.00003 [69,24]. The Pf is negligible when the safety fac- computing the BC based on deterministic qc1 and qc2. The
tor between 3 and 4. The negative Pf means the probability that probabilistic-based solution includes the uncertainty of qc1 and
FOS is less than one. qc2 in the BC equation. It considers the qcs as random variables

Table 7f
The R2 of predicted (qc2) at CPT3.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.954964 0.958017 0.976241 0.971158 0.950226 0.976468 0.842652 0.976468 IG
2 1.2 0.915161 0.915741 0.91083 0.913189 0.907618 0.908313 0.857588 0.915741 Lt
3 1.4 0.942155 0.941569 0.930024 0.935407 0.939807 0.926685 0.712043 0.942155 No
4 1.6 0.961137 0.959713 0.961711 0.961669 0.955761 0.961352 0.652329 0.961711 LG
5 1.8 0.97266 0.971339 0.967015 0.970016 0.969472 0.966188 0.721421 0.97266 No
6 2 0.977058 0.975523 0.968627 0.972627 0.975944 0.967514 0.692492 0.977058 No
7 2.2 0.984916 0.984724 0.973781 0.979 0.984288 0.972478 0.678341 0.984916 No
8 2.4 0.9842 0.982532 0.982167 0.984156 0.979011 0.981579 0.696084 0.9842 No
9 2.6 0.986556 0.986699 0.989205 0.989483 0.979676 0.988953 0.716838 0.989483 Ga
10 2.8 0.986098 0.988148 0.990644 0.990261 0.978727 0.990513 0.740221 0.990644 Ga
11 3 0.983376 0.988069 0.990219 0.989102 0.973925 0.990111 0.752459 0.990219 LG
12 3.2 0.979259 0.987615 0.98965 0.987453 0.966859 0.989524 0.764355 0.98965 LG
13 3.4 0.98124 0.990556 0.989454 0.98827 0.969472 0.989239 0.777691 0.990556 Lt
14 3.6 0.979519 0.991856 0.990386 0.988338 0.965126 0.990142 0.773716 0.991856 Lt
15 3.8 0.979272 0.992522 0.990825 0.988547 0.963749 0.990557 0.775696 0.992522 Lt
16 4 0.98082 0.993171 0.990963 0.989325 0.967213 0.990681 0.792498 0.993171 Lt

Table 7g
The R2 of predicted (qc1) at CPT4.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.920935 0.962688 0.955172 0.9454 0.903638 0.954725 0.735254 0.962688 Lt
2 1.2 0.967634 0.979032 0.968349 0.972112 0.961854 0.96787 0.837462 0.979032 Lt
3 1.4 0.973068 0.973479 0.949197 0.960606 0.972143 0.94595 0.92021 0.973479 Lt
4 1.6 0.966347 0.964756 0.950286 0.958024 0.967597 0.94884 0.951704 0.967597 We
5 1.8 0.974367 0.973793 0.957099 0.966024 0.975278 0.95546 0.940568 0.975278 We
6 2 0.977346 0.978627 0.955527 0.966592 0.976618 0.953304 0.918232 0.978627 Lt
7 2.2 0.969964 0.973086 0.945356 0.957303 0.967894 0.942725 0.893298 0.973086 Lt
8 2.4 0.969855 0.97344 0.938855 0.952431 0.967463 0.935428 0.877922 0.97344 Lt
9 2.6 0.97977 0.982122 0.946541 0.961108 0.97859 0.942196 0.883088 0.982122 Lt
10 2.8 0.984016 0.985711 0.953623 0.968374 0.983025 0.948402 0.889694 0.985711 Lt
11 3 0.98715 0.987889 0.953928 0.96933 0.986641 0.947752 0.886126 0.987889 Lt
12 3.2 0.985892 0.987082 0.946324 0.963563 0.986079 0.938976 0.869099 0.987082 Lt
13 3.4 0.981897 0.983989 0.938165 0.956531 0.982975 0.929867 0.853164 0.983989 Lt
14 3.6 0.977782 0.982279 0.928363 0.94809 0.979822 0.918489 0.845975 0.982279 Lt
15 3.8 0.973292 0.980414 0.920482 0.940925 0.976934 0.91001 0.830296 0.980414 Lt
16 4 0.973099 0.982352 0.918156 0.939146 0.978098 0.907435 0.819545 0.982352 Lt

Table 7h
The R2 of predicted (qc2) at CPT4.

No. B Nr Lt LG Ga We IG Rah Max R2 PD


1 1 0.968314 0.96866 0.953195 0.961638 0.96708 0.9496 0.938607 0.96866 Lt
2 1.2 0.935158 0.94641 0.878897 0.899582 0.934017 0.86635 0.857588 0.94641 Lt
3 1.4 0.955233 0.973265 0.894378 0.917649 0.970772 0.881816 0.740023 0.973265 Lt
4 1.6 0.97368 0.982719 0.928337 0.946676 0.983125 0.919491 0.765949 0.983125 We
5 1.8 0.936605 0.953458 0.866613 0.89067 0.931791 0.849486 0.85106 0.953458 Lt
6 2 0.955264 0.964016 0.894122 0.91659 0.950375 0.879237 0.878564 0.964016 Lt
7 2.2 0.953469 0.952904 0.933992 0.942624 0.949678 0.928641 0.933306 0.953469 No
8 2.4 0.958011 0.956956 0.956844 0.959704 0.959879 0.954729 0.951093 0.959879 We
9 2.6 0.969322 0.96785 0.972561 0.97447 0.973848 0.971092 0.95609 0.97447 Ga
10 2.8 0.978528 0.975793 0.97246 0.977653 0.980748 0.969838 0.939336 0.980748 We
11 3 0.990658 0.987713 0.977375 0.985216 0.991996 0.973849 0.922352 0.991996 We
12 3.2 0.988611 0.985662 0.970184 0.979404 0.98937 0.966336 0.916937 0.98937 We
13 3.4 0.992009 0.989133 0.973543 0.982861 0.992926 0.969662 0.902235 0.992926 We
14 3.6 0.990933 0.988507 0.967728 0.978377 0.99198 0.963154 0.881916 0.99198 We
15 3.8 0.988045 0.986944 0.9604 0.972121 0.989548 0.955433 0.859959 0.989548 We
16 4 0.991174 0.989357 0.968105 0.978257 0.992247 0.964268 0.873344 0.992247 We

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
12 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

and uses the MCS to compute the BC of the soil foundation many 4.1. Deterministic solution
times to determine the Pf. The data is obtained by performing
CPT on the soil in Nasiriyah. The following sections present the The study uses a direct approach taking the qc1 and qc2 mean to
solutions of the BC. determine the BC of a shallow square foundation. The schematic

Table 8
The parameters of distribution for qc1 and qc2 CPT 1.

qc1 (CPT1) qc2


B (m) N Mu sigma Distribution B (m) N mu sigma Distribution
1 50 2000.773 171.94 Lt 1 100 1557.735 311.376 Lt
1.2 60 1941.354 167.012 Lt 1.2 120 1670.926 313.353 Lt
1.4 70 1890.526 165.861 Lt 1.4 140 1638.044 289.882 Lt
1.6 80 1875.135 174.62 Lt 1.6 160 1689.898 261.767 Lt
1.8 90 7.543 0.179 Lg 1.8 180 1649.389 234.108 Lt
2 100 1821.313 262.037 Lt 2 200 1754.544 5.246 We
2.2 110 1727.987 332.176 Lt 2.2 220 1821.168 6.837 We
2.4 120 1678.563 334.045 Lt 2.4 240 35.524 47.411 Ga
2.6 130 1685.557 316.474 Lt 2.6 260 32.507 51.209 Ga
2.8 140 1708.757 302.667 Lt 2.8 280 1682.444 44853.73 IG
3 150 1741.3 286.18 Lt 3 300 1681.118 216.887 Lt
3.2 160 1775.94 284.765 Lt 3.2 320 1699.735 239.155 Lt
3.4 170 1795.046 276.229 Lt 3.4 340 1629.782 276.87 Lt
3.6 180 1795.836 267.736 Lt 3.6 360 1596.838 275.007 Lt
3.8 190 1778.342 262.589 Lt 3.8 380 1604.468 313.287 Lt
4 200 1770.318 254.67 Lt 4 400 7.332 0.354 Lg

Table 9
The parameters of distribution for qc1 and qc2 CPT number 2.

qc1 (CPT2) qc2


B (m) N m r Distribution B (m) N mu sigma Distribution
1 50 7.589 0.216 Lg 1 100 1345.289 305.328 Lt
1.2 60 1885.41 22206.986 IG 1.2 120 1475.589 300.308 Lt
1.4 70 1703.791 384.755 Lt 1.4 140 1532.021 273.791 Lt
1.6 80 1561.361 430.557 Lt 1.6 160 1594.033 250.179 Lt
1.8 90 1166.101 0 Ra 1.8 180 1688.989 4.821 We
2 100 1138.694 0 Ra 2 200 1779.351 5.202 We
2.2 110 1452.881 388.491 Lt 2.2 220 1750.979 240.674 Lt
2.4 120 1501.731 377.294 Lt 2.4 240 1765.904 238.631 Lt
2.6 130 1539.479 365.775 Lt 2.6 260 1804.788 4.433 We
2.8 140 1557.552 348.919 Lt 2.8 280 1589.714 481.344 No
3 150 1568.255 334.324 Lt 3 300 1782.57 3.927 We
3.2 160 1563.037 320.328 Lt 3.2 320 1772.691 3.917 We
3.4 170 1575.92 315.374 Lt 3.4 340 1727.499 3.624 We
3.6 180 1602.523 308.653 Lt 3.6 360 1653.739 3.168 We
3.8 190 1626.841 302.026 Lt 3.8 380 1586.812 2.865 We
4 200 1620.03 296.279 Lt 4 400 5.137 268.365 Ga

Table 10
The parameters of distribution for qc1 and qc2 CPT number 3.

qc1 (CPT3) qc2


B (m) N mu sigma Distribution B (m) N mu sigma Distribution
1 50 2135.97 238.659 Lt 1 100 1640.495 18146.909 IG
1.2 60 7.607 0.23 Lg 1.2 120 1817.371 289.562 Lt
1.4 70 1905.849 328.289 Lt 1.4 140 1889.97 232.976 Lt
1.6 80 1798.747 355.176 Lt 1.6 160 7.522 0.175 Lg
1.8 90 1760.841 333.771 Lt 1.8 180 1785.193 386.877 No
2 100 1740.99 14844.786 IG 2 200 1832.09 374.637 No
2.2 110 1716.667 15911.226 IG 2.2 220 1845.204 347.739 No
2.4 120 7.408 0.31 Lg 2.4 240 1818.133 342.036 No
2.6 130 11.218 157.031 Ga 2.6 260 25.448 69.418 Ga
2.8 140 11.384 158.085 Ga 2.8 280 7.467 0.208 Lg
3 150 2018.645 3.864 We 3 300 7.475 0.209 Lg
3.2 160 2039.347 4.017 We 3.2 320 7.479 0.212 Lg
3.4 170 2058.201 4.155 We 3.4 340 1728.586 211.462 Lt
3.6 180 2068.471 4.275 We 3.6 360 1687.823 202.092 Lt
3.8 190 2074.493 4.411 We 3.8 380 1669.176 198.48 Lt
4 200 2071.045 4.502 We 4 400 1631.201 205.134 Lt

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 13

view of the problem that represents a square foundation placed on ring to the best PD which opposite to the maximum R2 among
a soil having a distribution of qc with depth is shown in Fig. 6. The other distributions.
procedure of computing the BC using Tand’s equation is demon- Tables 8–11 show the selected PD for qc1 and qc2 for soils at four
strated as a part of the flowchart shown in Fig. 7. A Matlab code locations along with the parameters of PDs for qc1 and qc2 for four
was developed to calculate the BC based on the flowchart men- cases of CPTs. Since the influence of intensity of the applied stress
tioned above. The solution based on this equation was imple- qapp reaches to a depth equal 1.5 B below the base of foundation,
mented in the software designed by Robertson [4,52]. However, the range of the qc1 and qc2 through this thickness will be changed
it needs to provide the software with the value of the factor of and the number of data also altered. Therefore the qc1 and qc2 data
Rk. In addition to that, the software does not take into account may show different PD. The number of the qc1 data included in the
the reliability analysis. Two cases (fissured and intact clay) pre- analysis ranges from 50 to 200 readings while it reaches from 100
sented by Tand can be used manually to find the Rk. An equation to 400 readings for qc2. The type of distribution is selected among
developed by curve fitting for the fissured clay and also the equiv- seven distributions and presented in the fifth column against every
alent embedment ratio is functioned in the program code devel- width of the foundation. For CPT1, the best PD for qc1 is the Logistic
oped in this study. It was found that the ultimate BC average distribution (Lt) while the PD for qc2 show different types such as
computed for a square foundation for soils located at CPT1, CPT2, Lt, Lg, Ga, and Weibull distribution. Fig. 8(a) and (b) show the his-
CPT3, and CPT4 equal 583.281 kN/m2, 535.988 kN/m2, togram for qc1 and qc2 for the case of the width of the foundation of
612.021 kN/m2, and 719.831 kN/m2 respectively (Tables 3–6). 1 m along with seven types of PDs. It is evident that the Rayleigh
Notable, the width of the foundation has an effect on the values distribution does not represent the qc1 and qc2 data compared to
of ultimate soil BC solved by the deterministic solution. For others. All the distributions suggested in this paper used the best
instance, the qult for 1 m width footing on soil at CPT1 equal to fit distribution based on R2. Many data groups show that logistic
620 kN/m2, and at 4 m footing it is reduced to 563.606 kN/m2. and Weibull distribution are the best compared to others (Tables
The percentage of reduction in bearing capacity (kN/m2) due to 7a–7h).
increasing the width of the foundation was (620–563)/620 = 9%.
It is also found that the BC shows a noticeable change at different 4.2.2. MCS –based BC
CPTs. The percentage of qult reduction in other locations are 17.4%, The BC of the footing is calculated based on Eq. (2) where the qc1
13.8%, and 13.5% respectively. and qc2 data are considered as random variables. Thousands of real-
izations are repeated to determine the Pf based on LSF (Eq. (14))
4.2. Probabilistic solution where the failure is computed according to Eq. (15). For every case,
the program determines the summation of failure over the times of
4.2.1. Probability distribution type of qc1 and qc2 realization in addition to an empirical PD. The suitable PD of the
In this section, the reliability analysis considers the uncertainty data is determined based on the same flowchart (Fig. 7) along with
of qc1 and qc2 at a thick layer of 0.5 B (m) below the base of foun- their parameters. The computing of the Pf can be achieved relatively
dation and 1 B (m) below the previous layer, respectively. First, the easy based on the flowchart (Fig. 7). The easiness is because of con-
data of the qc1 and qc2 are included in the program written by the sidering a small number of randomly distributed variables (qc1 and
author (available upon request) using Matlab software to select the qc2) to compute the BC. The dimensions of the foundation are con-
suitable PD among seven distributions based on the flowchart pre- sidered deterministic and have not been included in the analysis
sented in Fig. 7. The flowchart states that the best distribution is since the uncertainty is small and the construction process can con-
selected according to GOF test based on R2 method since it is suf- trol that through construction. Including the uncertainty of geo-
ficient and can recognize the types of PDs very well. Then the metric variables in the analysis makes the solution expensive and
parameters of the distribution are computed through best fitted. needs extensive analysis. In addition to that, the transformation
Notably, the qc1 at CPT1 can be simulated successfully by different error is not considered since the BC equation is directly related to
distributions, among them the logistic distribution shows the most the parameter, i.e., qc measured through CPT. The program devel-
representing one (Table 7a). Tables 7b–7h show the accuracy of oped in this study efficiently computes the BC for the soil taking
representing the PDs for the both qc1 and qc2 data for sixteen case the variability of the qc1 and qc2 data into account.
of square foundation with width from 1 m to 4 m at four CPT Table 12 shows the result of reliability analysis for the square
soundings by presenting the value of R2 for every case with refer- foundation of 1 m to 4 m width. PDs for qc1 and qc2 data with

Table 11
The parameters of distribution for qc1 and qc2 CPT number 4.

qc1 (CPT4) qc2


B (m) N mu sigma Distribution B (m) N mu sigma Distribution
1 50 2462.283 288.486 Lt 1 100 2122.588 476.353 Lt
1.2 60 2314.284 363.16 Lt 1.2 120 2427.63 445.602 Lt
1.4 70 2141.716 447.793 Lt 1.4 140 2573.162 316.142 Lt
1.6 80 2267.914 2.707 We 1.6 160 2673.107 5.428 We
1.8 90 2252.433 2.822 We 1.8 180 2328.172 453.021 Lt
2 100 1991.634 416.899 Lt 2 200 2350.262 477.76 Lt
2.2 110 2012.975 392.181 Lt 2.2 220 2157.421 844.479 No
2.4 120 2062.86 381.235 Lt 2.4 240 2339.428 2.783 We
2.6 130 2135.06 395.732 Lt 2.6 260 5.797 347.699 Ga
2.8 140 2213.703 424.766 Lt 2.8 280 2258.188 3.002 We
3 150 2275.288 429.209 Lt 3 300 2229.364 3.18 We
3.2 160 2325.676 425.171 Lt 3.2 320 2227.774 3.231 We
3.4 170 2373.379 421.647 Lt 3.4 340 2194.147 3.344 We
3.6 180 2397.046 418.777 Lt 3.6 360 2189.337 3.495 We
3.8 190 2423.089 407.441 Lt 3.8 380 2175.074 3.792 We
4 200 2432.793 394.203 Lt 4 400 2130.73 3.701 We

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
14 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

data. After determining the suitable PD for qc1 and qc2 data, they
will be simulated according to the PD as shown in Fig. 8(a) for
qc1, and Fig. 8(b) for qc2. The analysis requires simulating qc1 and
qc2 according to the best-fitted logistic distribution for a founda-
tion with width equals 1 m. Fig. 8(c) shows the histogram of the
soil BC with the probability distributions demonstrating that most
of the PDs can represent the data except Rayleigh distribution. The
number of realization of 100,000 can give approximately accepted
results.
Fig. 9 shows the relation between cumulative probability and
soil BCs at two levels of confidence, 90% and 95% using cumulated
density function (CDF) where the BC equal 436.93 kN/m2 and
496.589 kN/m2 respectively. The two values of BC at the two levels
of confidence are less than the BC predicted by the deterministic
approach which is 574.155 kN/m2. In addition to that, the BC at
50% confidence is approximately the same value of bearing capac-
ity obtained by the deterministic solution. The study concludes
that the BC obtained using a probabilistic approach is less than that
obtained by a deterministic method based on the mean values of
both qc1 and qc2. This finding confirms that obtained by Griffiths
and Fenton (2001) [70]. Dividing the ultimate BC obtained by
deterministic approach to that obtained by probabilistic approach
(equivalent FOS) at two levels 90% and 95% confidence, gives
(FOS90 574.155/436.93 = 1.314) , and (FOS95% = 574.155/496.589 =
1.156). Comparing the traditional FOS = 3 with that obtained at
two levels of confidence reveals that the FOS = 3 is conservative.
It is attributed to that the traditional FOS that assumed to be three
depends on experience while the FOS computed based on the prob-
abilistic approach can give an accurate result giving the powerfully
and robustness of their results. This conclusion confirms the find-
ing obtained by Shahin and Cheung (2011) [71] where the BCs
computed by using software @ Risk (Palisade, 2000) based on
Terzaghi and also by using the developed excel sheet based on
the reliability of 90% and 95% are more than the allowable BC based
on a FOS = 3.

4.2.3. Effect of varying the width of the foundation


The study considered sixteen cases of different width of founda-
tions from 1 m to 4 m. The effect of the width of the foundation on
the values, the PD, and the COVs of qc1, and qc2, are investigated.
Notably, the increase in the width of the foundation may affect
the number of qc1 and qc2 data. Ten values of qc1 are added to
the origin number as the width of foundation increases by 0.2 m
while qc2 data increases by 20 data as the width of foundation
increases by 0.2 m (Table 8–11). The mean value and standard
deviation of qc1 and qc2 decrease or increase based on their varia-
tion as the width of the foundation increases. For the case of qc1 at
CPT1, the qc1 data is not changed dramatically and the PD is not
changed as width increases; the PD still shows the logistic distribu-
tion for all cases. For qc2 at CPT1, the distribution of qc2 is altered,
and different distributions are produced. This conclusion is out-
lined for other data (Tables 8a, b, c, d). Here the width of the foun-
dation is a significant factor affect the result of BC. As the width of
the foundation increases the depth influenced by the applied stress
Fig. 8. (a) The histogram for qc1 at 0.5 B m below the shallow foundation of width also increases where the influence depth equal to 1.5 B. Increasing
1 m and depth 0.5 m with seven types of the best-fitted probability distributions for the affected depth means the considered qc1 and qc2 data increases
CPT performed at location CPT1. (b) Random points of qc2 at 0.5 B m below the too, and their variation is changed. The changing depends on the
shallow foundation of width 1 m and depth 0.5 m based on logistic distribution for
type of soil and the soil stratification. In the cases studied since
CPT performed at location CPT1.
the soil is clay soil, the variation change in the qc is slightly signif-
icant. With respect, the COVs, adding any data to the original data
various values of depth give different PDs for BCs. For example, the may change the values of qc1 and qc2 and adjust the COVqc where
foundation with a width of 3.8 m, qc1 represented by a logistic the alteration is fundamental in the sensitivity analysis of BC.
probability distribution, and qc2 having lognormal distribution give The effect of varying the width of the foundation on the reliabil-
a BC with a gamma distribution (Table 12). Fig. 8(a) and (b) show ity index is also investigated. Fig. 10(a, b, c, d) show that the soil BC
the histogram of qc1 and qc2 data and the best fitted PDs for the is changing as the width is varied even when the soil qcs have the
case, B = 1 m, to visually demonstrate the best PD that can fit the same type of probability distribution. The COV of the soil BC also

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 15

Table 12
Detailed values of the reliability index, qc1, qc2, qult, with statistical results (100 kN/m2).

B(m) b Pf mqc1 rqc1 PDqc1 mqc2 rqc2 PDqc2 mqult rqult PDqult COVqc1 COVqc2 COVqult
1 3.291 0.0005 2000.773 171.940 Lt 1557.735 311.376 Lt 626.722 69.526 Lt 0.086 0.200 0.111
1.2 3.432 0.0003 1941.354 167.012 Lt 1670.926 313.353 Lt 627.937 66.708 Lt 0.086 0.188 0.106
1.4 3.353 0.0004 1890.526 165.861 Lt 1638.044 289.882 Lt 605.573 59.975 Lt 0.088 0.177 0.099
1.6 3.540 0.0002 1871.731 188.756 Lt 1689.898 261.767 Lt 604.584 58.156 Lt 0.101 0.155 0.096
1.8 3.540 0.0002 1880.575 210.312 Lt 1649.389 234.108 Lt 594.668 55.835 Lt 0.112 0.142 0.094
2 3.540 0.0002 1813.411 272.989 Lt 1754.544 5.246 We 574.558 58.129 Lt 0.151 0.003 0.101
2.2 3.432 0.0003 1727.987 332.176 Lt 1821.168 6.837 We 572.722 62.777 Lt 0.192 0.004 0.110
2.4 3.090 0.001 1678.563 334.045 Lt 35.524 47.411 Ga 556.436 62.833 Lt 0.199 1.335 0.113
2.6 3.090 0.001 1685.557 316.474 Lt 32.507 51.209 Ga 553.918 59.815 Lt 0.188 1.575 0.108
2.8 3.121 0.0009 1708.757 302.667 Lt 1694.133 248.036 Lt 555.763 66.875 Lt 0.177 0.146 0.120
3 3.090 0.001 1734.689 293.207 Lt 1718.234 281.627 Lt 561.540 69.664 Lt 0.169 0.164 0.124
3.2 3.062 0.0011 1769.732 291.481 Lt 1746.991 313.467 Lt 568.977 72.192 Lt 0.165 0.179 0.127
3.4 2.834 0.0023 1789.323 282.719 Lt 1670.511 343.158 Lt 559.521 74.111 Lt 0.158 0.205 0.132
3.6 3.353 0.0004 1790.584 273.956 Lt 7.396 0.363 Lg 18.490 30.745 Ga 0.153 0.049 1.663
3.8 3.291 0.0005 1773.574 268.367 Lt 7.390 0.389 Lg 16.514 34.080 Ga 0.151 0.053 2.064
4 3.540 0.0002 1766.007 260.160 Lt 7.365 0.392 Lg 16.873 32.872 Ga 0.147 0.053 1.948

respectively. The logistic distribution represents percentage 56%


of the PDs that represent the BC. It is found that the PDs that rep-
resent the soil BC may be logistic, Weibull or normal distribution
based on R2. It is important to conclude that using one type of dis-
tribution or suggested to use one type or specified type to repre-
sent data is not precise. Some codes suggest assuming a
lognormal distribution, but the best is to select the best one among
several PDs as the study presented in this paper verifies it.

4.2.4. Performance function characterization (Safety margin)


For the statistical analysis of the performance function, ultimate
BC and the FOS imply identifying the location and variability in
addition to skewness and kurtosis. The skewness can identify the
symmetry around the mean value, and the kurtosis states the
heavy and light tail of the distribution. Fig. 11(a–f) show the distri-
bution of the performance function with the probability density
which is not always compatible with the normal distribution as
Fig. 9. The cumulative probability distribution for bearing capacity incorporating
assumed by [72]. The distribution of performance function (qult-
uncertainties of qcs and B = 2 m.
qall) follows the logistic distribution with skewness equal to
0.0359 which means that the distribution is positively skewed
varies as the width of foundation changes (Table 11). For a square since it extends to the more substantial values by 0.0359 more sig-
foundation founded on soil at the location of CPT 1, subjected to a nificant than, the smaller value (Fig. 11c). The distribution is not
pressure equal 100 kN/m2, all the foundations with different width normally symmetric. Since the skewness (SK) value is between
can give b higher than 3 except the foundation having a width 0.5 and 0.5, then the distribution can be classified as approxi-
equal 3.4 m, where the b is less than 3. It is found that the soil mately symmetric. If the value is between 0.5 or 1 or 0.5 and
resistance equals the applied stress of 550 kN/m2, i.e., b equal 0 1 the distribution is moderately skewed. When the skewness is
when the width of foundation from 2.4 m to 2.8 m and 3.4 to Less than 1 or greater than 1, the distribution is highly skewed.
3.8 m when the soil is at the location of CPT1 (Fig. 10(a). For the The distribution has a peak, not flatness. It is a shape parameter
soil at location CPT2, the square foundations with width values that characterizes the degree of symmetry. Since the SK is greater
from 1 m to 4 m gives b equal to zero. The width of the foundation than zero when the tail towards the high values indicating an
has slightly significant effect on the BC since as the width increases excess of low values (positively skewed). The distribution show
in our case, the qc decreases or being approximately constant. It light skewness.
can be concluded that the dimensions of foundations constructed The kurtosis equal to 3.8037 which is greater than 3. It means
on a clay soil may slightly affect the soil BC. This type of foundation that the peakness of the distribution is higher than normal where
and the BC can resist the loads applied to the low and the medium normal distribution gives kurtosis more than 3. The height of the
building. distribution and the sharpness of the central peak relative to the
The PDs of the input random variables qc1 and qc2 may give dif- bell shape can be estimated based on the kurtosis value where
ferent PDs for the output (BC). Since the width of the square foun- the kurtosis value equal to 3 is the standard sharpness of the nor-
dations is changed, some of the values may make the data of qc1 mal distribution
and qc2 shows different PDs such as lognormal distribution which The COV for BC between 0.15 < COV < 0.35 is classified as mod-
may give different PDs for soil BC such as Gamma distribution erate, COV < 0.15 is least variable and COV > 0.35 as is most vari-
(Table 12). The percentage of logistic distribution that represent able [48]. Since the COV of the BC (COVBC) obtained through a
qc1 for sixteen cases with different width for every location test probabilistic approach equal to about 0.11 for the case of CPT at
(e.g. CPT1) is 100% while it represents 56% for qc2. The PD for the location 1, is less than 0.15, it can be classified as the least variable.
soil BC for the square foundation founded on soil at CPT1 shows The COVBC average is ranked as the COV of the BC at CPT 2 equal to
81% logistic distribution. The PDs that represent the qc1 and qc2 0.214858. For CPT 3 as an average value of COV for all width the
are a logistic distribution with a percentage of 75%, and 44% COV equal to 0.25, it is also classified as a moderate variation of

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
16 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

Fig. 10. The relation between reliability index and width of foundation for different cases of applied stress for a square foundation (1 m  1 m) at CPT 1 with depth 0.5 m. (b)
for CPT2 (c) CPT3 (d) CPT4.

BC. The last CPT 4 shows COV for BC equal to 0.20778926 which is It is expected that as the values of qc1 and/or qc2 are changed by
also classified as fair. It can be concluded that the COVBV computed adding new values for the influence zone, the statistical character-
for the shallow foundation at the soil in Nasiriyah based on qc1 and istics and the type of PDs of the BC of the shallow foundation are
qc2 can be classified as fair. also altered. The b obtained by using MCS of the LSF of BC is com-
pared with the target b of 3 based on (AASHTO). To investigate the
sensitivity of qc1, it needs keeping the average and standard devi-
4.2.5. Sensitivity analysis
ation of qc1, the width of footing, applied stress and qc2 as constant.
Geotechnical reliability depends on applied loads and soil prop-
As the variation of the main parameter qc1 decreases, the BC confi-
erty which are represented by statistical parameters such as mean,
dence increases. At a width of foundation equal to 3 m, applied
standard deviation, and COV. The data obtained from CPT can give
stress equal to 100 kN/m2, qc1 = 1734.689 kN/m2 and qc2 = 1718 -
approximately sufficient data for statistical analysis. Controlling
kN/m2; the ultimate BC was 561.54 kN/m2 and reliability index is
the Pf can be achieved by studying the effect of these parameters
3.011 which is greater than 3. For CPT 4, making B = 3 m,
on the Pf. The sensitivity study objects to identify which variable
qc1 = 2275.288 kN/m2, qc2 = 2229.3 kN/m2, the qult = 677.9 kN/m2
influences the BC. The following section studies the effect of qc
and b= 3.29 which is greater than target value of b = 3. It is obvious
and the applied vertical loads on the bearing capacity.
that the qcs have a significant effect on the BC.

4.2.5.1. Effect of the qc. The sensitivity of the BC of shallow founda-


tions to qc1 and qc1 data is determined by considering both the 4.2.5.2. Effect of the applied load. In this subsection, the study pre-
width of the footing and the applied stress as constant parameters. sents the investigating of the impact of the vertical load applied

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 17

Fig. 11. probability distribution for performance function for samples of cases (a) COV = 0.1, FOS = 3; (b) COV = 0.2, FOS = 3;(c) COV = 0.3, FOS = 3; (d) COV = 0.5, FOS = 3; (e)
COV = 0.7, FOS = 3; (f) COV = 0.9, FOS = 3.

to the shallow foundation. The investigation contains ten cases of parameters: scale (r) and location parameter (l) (Eq. (20)) is used
the applied from 100 kN/m2 to 550 kN/m2. The b and FOS, both to simulate the applied load. For CPT1, the b obtained for the width
are used to measure the safety of the geotechnical structure. Since of foundation equal to 3 m and vertical stress 200 kN/m2 meet the
there is no load test results neither project design documentation, required code and the target reliability index. For sensitivity anal-
a range of vertical applied load are assumed with COV equal to 0.1 ysis comparing between different cases with constant qc1 and qc2,
to 0.9. The PD of extreme value type 1 (Gumbel) having two the applied vertical stress affect the ultimate BC.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
18 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

Fig. 12. The probability distributions for qult (Logistic distribution) and the qapp(extreme value type 1) , COV = 0.6, B = 1 m, Df = 0.5 m, with statistics for both qult and qapp.
The black line refer to the intersection between the two probabilities. (a) COV = 0.1; (b) COV = 0.2; (c) COV = 0.5; (d) COV = 0.6.

!!
  1 qapp  lqapp qapp  lqapp 0.2. The recommendation is that the risk increased as the COVapp
f qapp ; lqapp ; rqapp ¼ exp   exp 
rqapp rqapp rqapp increases beyond 0.2.
Referring to the graph presented in Fig. 15, the number of fail-
ð20Þ
ure points are increased for the case of a factor of safety decreased
Fig. 12(a, b, c, d) show both the PDs for BC based on uncertain from 3 to 1.5. The applied loads in these two figures are assumed
values of qc1 and qc2 and the PDs for the qapp for different values constant. The recommendation is to use the COVapp in the analysis
of COV for applied stress (COVapp). It is clear that the Pf decreases since they affect the bearing capacity of a shallow foundation.
as the COVapp increases from 0.1 to 0.2, beyond that, the Pf
increases. As the COVapp increases more than 0.2, some of the val- 5. Conclusions
ues of qapp become negative where they are not included in com-
puting the Pf in the same time the dispersion to the right is also The BC was computed for the shallow foundation of square
increased making the Pf increased. shape at 0.5 m depth constructed on a clay soil based on qc
Fig. 13 shows the limit state curve which divides the region of obtained through performing CPT by using two approaches deter-
the graph between qc1 and qc2 into two parts: the success region ministic and probabilistic methods. Different size of the square
(safe) and failed region (unsafe). The applied stress assumed to footing is considered at four locations in the project in Nasiriyah,
be varied according to COV which make the limit state vibrated. Iraq. Different types of PDs are used to simulate the variation of
The number of MCS is 100,000. It is obvious that the number of qc, and the MCS solution was used for analysis utilizing Tand’s
failed points increases as the COV increases. It increases from equation as a direct approach. The following points are outlined
333 to 433 failed points when COV changed from 0.3 to 0.5. At through performing this research.
0.7 the number of failed points was 682. The accuracy is also pre-
sented on curves which are computed according to the equation of 1. Through analyzing the results of CPT soundings, it was found
confidence [59] that the type of soil is clayey soil since the Ic is located in the
region of clay soil in the chart presented by Robertson.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Moreover, since the COV of Ic with depth was less than
½L; U 1a ¼ Pf  za=2 Pf ð1  P f Þ=n ð21Þ
10% (similar unit soil), the type of the BC equation founda-
tion constructed on clay soil was Tand’s equation
where a = 0.05, [L, U]1–0.05 is the confidence 95% two sided confi- 2. Deterministic solution using Tand’s equation gave an ulti-
dence interval, and z0.05 = 1.96. mate BC equal to about 600 kN/m2 for the four locations
Fig. 14 shows the relation between Pf with the COVapp. The for sixteen case of the width of square foundation ranges
ranges of COVapp are taken from 0.01 to 0.9. Four FOS is chosen from 1 m to 4 m where the cone tip resistance averaged is
2, 2.2, 2.5 and 3. It was found that the Pf increases with the increas- computed at a depth equivalent to 1.5 times the width of
ing of COVapp beyond 0.2. The Pf is sensitive for the COVapp beyond the foundation.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 19

Fig. 14. The relation between probability failure and COV of the applied stress for
the square foundation of width, B = 1 m, depth of foundation, Df = 0.5 m, for soil at
location CPT1).

Fig. 13. Monte Carlo simulations of qc1 and qc2 (100000 simulations). Points with
symbol (+) appears in the unsafe region are failed points. (a) COV = 0.1, FOS = 3 (b)
COV = 0.3, FOS = 3.0 (c) COV = 0.5 (d) COV = 0.7.

3. The various types of PD for the data of cone tip resistance


can be best fitted however the logistic distribution is the
most distribution can simulate the variability of cone tip Fig. 15. Monte Carlo sampling (a) FOS = 1.5 COV = 0.3, (b) COV = 0.1, FOS = 1.5.
resistance successfully.
4. The BC of shallow foundation obtained through a probabilis- 6. The different PD for qc can give different distribution for BC.
tic approach is less than the ultimate BC obtained by the It was found that about 90% of cases of qc classified as logis-
deterministic solution. tic distribution and 40% classified as logistic distribution for
5. The width of the foundation has a slight significant effect on BC.
the BC despite the influence of stress intensity reaches to 7. The reliability index is higher than the target when the
more depth and make the cone tip resistance depth con- applied stress is about 100–150 kN/m2 while the ultimate
tribute in the computation since the soil is clay and the cone limit state is fulfilled at load equal 550 kN/m2 where the reli-
tip resistance has not been increased notably with depth. ability index equal to zero.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
20 R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx

8. The variation of BC computed for a shallow square founda- [24] M. Kayser, S. Gajan, Application of probabilistic methods to characterize soil
variability and their effects on bearing capacity and settlement of shallow
tion based on a direct approach based on CPT is classified
foundations: state of the art, International Journal of Geotechnical Engineering
as moderately variation. 37 (1) (2014) 264–269.
9. The applied load has a significant effect on the BC. It is sim- [25] K.-K. Phoon, F.H. Kulhawy, Characterization of geotechnical variability, Can.
ulated using Gumble distribution with different COV. Using Geotech. J. 36 (4) (1999) 612–624.
[26] G.A. Fenton, D.V. Griffiths, Reply to the discussion by R. Popescu on "Bearing
COV equal to 0.3 may affect the result. capacity prediction of spatially random c - / soils, Canadian Geotechnical
10. The best fit distribution for a FOS is found to be logistic Journal 369 (2004) 368–369, https://doi.org/10.1139/t03-080. NRC Research
distribution. Press..
[27] G.A. Fenton, D.V. Griffiths, Bearing-capacity prediction of spatially random c &
#150; / soils, Can. Geotech. J. 40 (1) (2003) 54–65.
[28] G.L. Sivakumar Babu, A. Srivastava, D.S. Murthy, Reliability analysis of the
bearing capacity of a shallow foundation resting on cohesive soil, Can.
Acknowledgment Geotech. J. 43 (2) (2006) 217–223.
[29] A.B. Salahudeen, ‘‘Cone penetration based probabilistic assessment of shallow
foundation settlement,” vol. 1, 2018, pp. 85–100.
The author genuinely thankful for the civil engineering depart-
[30] K.E. Tand, E.G. Funegard, J.-L. Briaud, ‘‘Bearing Capacity of Footings on Clays:
ment in the college of engineering, University of Thi-Qar through CPT Method,” Use of In- Situ Tests in Geotechnical Engineering (GSP No. 6),
the support of the university to conduct the research and give the American Society of Civil Engineers, Reston, Va., 1986, pp. 1017–1033.
general facilities for research. The author also thankful to the Engi- [31] E. Garzanti, A.I. Al-Juboury, Y. Zoleikhaei, P. Vermeesch, J. Jotheri, D.B. Akkoca,
A.K. Obaid, M.B. Allen, S. And_o, M. Limonta, M. Padoan, The Euphrates-Tigris-
neering Consultant Bureau of College of Engineering (ECB) in Thi- Karun river system: provenance, recycling and dispersal of quartz-poor
Qar University to make the data available for research purposes. foreland-basin sediments in arid climate, Earth-Sci. Rev. 162 (2016) 107e128.
[32] J. Jotheri, M. Altaweel, A. Tuji, R. Anma, B. Pennington, S. Rost, C.E. Watanabe,
Holocene fluvial and anthropogenic processes in the region of Uruk in
References Southern Mesopotamia, Quat. Int. (2017).
[33] A.A.M. Aqrawi, G. Evans, Sedimentation in the lakes and marshes (Awhar) of
the tigris-euphrates delta, southern mesopotamia, Sedimentology 41 (1994)
[1] A. Eslami, M. Gholami, Analytical model for the ultimate bearing capacity of
755–776.
foundations from cone resistance, Sci. Iranica 13 (3) (2006) 223–233.
[34] M. Abu-Farsakh, Z. Zhang, M. Tumay, M. Morvant, Computerized cone
[2] N. Ravichandran, V. Mahmoudabadi, S. Shrestha, Analysis of the bearing
penetration test for soil classification, Transp. Res. Rec. 2053 (2008) 47–64.
capacity of shallow foundation in unsaturated soil using Monte Carlo
[35] B.-C. Jung, P. Gardoni, G. Biscontin, Probabilistic soil identification based on
simulation, Int. J. Geosci. 8 (10) (2017) 1231–1250.
cone penetration tests, Géotechnique 58 (7) (2008) 591–603.
[3] T. Lunne, P.K. Robertson, J.J.M. Powell, Cone penetration testing in geotechnical
[36] P.K. Robertson, Interpretation of cone penetration tests — a unified approach,
practice, Blackie Acad. Prof. 10 (1997) 352.
Can. Geotech. J. 46 (11) (2009) 1337–1355.
[4] P.K. Robertson, K.L. Cabal, in: Guide to Cone Penetration Testing, Gregg Drill.
[37] J. Li, M.J. Cassidy, J. Huang, L. Zhang, R. Kelly, Probabilistic identification of soil
Testing, Inc., 2012, pp. 1–134.
stratification, Géotechnique 66 (1) (2016) 16–26, https://doi.org/10.1680/
[5] P.W. Mayne, F. Illingworth, Direct CPT method for footings on sand using a
jgeot.14.P.242.
database approach. Proc. 2nd Intl. Symposium on Cone Penetration Testing,
[38] Y. Wang, K. Huang, Z. Cao, Probabilistic identification of underground soil
Vol. 3 (CPT’10), Omnipress: 315-322. 2010, www.cpt10.com.
stratification using cone penetration tests, Can. Geotech. J. 50 (7) (2013) 766–
[6] R.B. Kelly, J.A. Pineda, L. Bates, L.P. Suwal, A. Fitzallen, Site characterisation for
776.
the Ballina field testing facility, Géotechnique 67 (4) (2017) 279–300.
[39] J. Ching, J.S. Wang, C.H. Juang, C.S. Ku, Cone penetration test (CPT)-based
[7] Z. Ouyang, P.W. Mayne, Effective friction angle of clays and silts from
stratigraphic profiling using the wavelet transform modulus maxima method,
piezocone penetration tests, Can. Geotech. J. (2017).
Can. Geotech. J. 52 (12) (2015) 1993–2007, https://doi.org/10.1139/cgj-2015-
[8] P.K. Robertson, K.L. Cabal, in: Guide to Cone Penetration Testing for
0027.
Geotechnical Engineering, Gregg Drill. Testing, Inc., 2010, p. 138.
[40] M. Uzielli, G. Vannucchi, K.K. Phoon, Random field characterisation of stress-
[9] C. Cherubini, Reliability evaluation of shallow foundation bearing capacity on
normalised cone penetration testing parameters, Géotechnique 55 (1) (2005)
c’ f’ soils, Can. Geotech. J. 37 (1) (2000) 264–269.
3–20, https://doi.org/10.1680/geot.55.1.3.58591.
[10] X. Fu, Z. Liu, Reliability analysis for vertical bearing capacity of shallow
[41] J.Y. Ching, S.S. Wu, K.K. Phoon, Statistical characterization of random field
foundations, Yantu Lixue/Rock Soil Mech. 21 (4) (2000) 345–359.
parameters using frequentist and Bayesian approaches, Can. Geotech. J. 53 (2)
[11] K.C. Foye, R. Salgado, B. Scott, Resistance factors for use in shallow foundation
(2016) 285–298.
LRFD, J. Geotech. Geoenviron. Eng. 132 (9) (2006) 1208–1218.
[42] J.Y. Ching, K.K. Phoon, Characterizing uncertain site-specific trend function by
[12] C. Cherubini, Shallow Foundation Reliability Design, First International
sparse Bayesian learning, J. Eng. Mech., ASCE 143 (7) (2017) 04017028.
Symposium on Geotechnical Safety & Risk, Tongji University, Shanghai,
[43] Y. Wang, T. Zhao, Statistical interpretation of soil property profiles from sparse
2007, pp. 71–90.
data using Bayesian Compressive Sampling, Geotechnique 67 (6) (2017) 523–
[13] G.A. Fenton, X. Zhang, D.V. Griffiths, Reliability of shallow foundations
536.
designed against bearing failure using LRFD, Georisk 1 (4) (2007) 202–215.
[44] R. Jamshidi Chenari, H. Kamyab Farahbakhsh, S. Heidarie Golafzani, A. Eslami,
[14] K.K. Phoon, Reliability-based design in geotechnical engineering -
2018. Non-stationary realisation of CPT data: considering lithological and
Computation and applications. 2008, SN - 9781482265811, PB - CRC Press,
inherent heterogeneity. Georisk: Assessment and Management of Risk for
2008.
Engineered Systems and Geohazards, pp.1–14.
[15] Y. Honjo, Challenges in geotechnical reliability based design, Proc. of the 3rd
[45] S. Lacasse, F. Nadim, Uncertainties in characterising soil properties, in: C.D.
Int. Symp. on Geotechnical Safety and Risk, 2011, pp. 11–27.
Shackleford (Ed.), Uncertainty in the Geologic Environment: From Theory to
[16] C. Onisiphorou, ‘‘Reliability analysis of bearing capacity for shallow
Practice, Geotechnical Special Publication No. 58, ASCE, New York, 1996, pp.
foundations based on Eurocode 7”, 3rd International Symposium on
49–75.
Geotechnical Safety and Risk, Munich, Germany, 2011, pp. 463–469.
[46] K.K. Phoon, Towards reliability-based design for geotechnical engineering.
[17] Xue, Jianfeng; Nag, D. Reliability analysis of shallow foundations subjected to
Special Lecture for Korean Geotechnical Society, Seoul 9 (July) (2004) 1–23.
varied inclined loads. In: Vogt, Norbert; Schuppener, Bernd; Straub, Daniel;
[47] M. Uzielli, P. Simonini, S. Cola, Statistical identifica-tion of homogeneous soil
Bräu, Gerhardt (Hrsg.): Geotechnical Safety and Risk. ISGSR 2011. Karlsruhe:
units for Venice lagoon soils, Proceedings of the 3rd International Conference
Bundesanstalt für Wasserbau. 2014 S. 377–384.
on Site Characterization, Taiwan, April 1-4, 2008, Taylor & Francis, The Nether-
[18] A. Ahmed. (2012). Simplified and Advanced Approaches for the Probabilistic
lands, 2008.
Analysis of Shallow Foundations. THESE DE DOCTORAT.
[48] M.E. Harr, in: Reliability-based Design in Civil Engineering, McGraw-Hill, New
[19] Z. Luo, S. Atamturktur, Y. Cai, C.H. Juang, Simplified approach for reliability-
York, NY, 1987, p. 32.
based design against basal-heave failure in braced excavations considering
[49] E. Ganju, M. Prezzi, R. Salgado, Algorithm for generation of stratigraphic
spatial effect, J. Geotech. Geoenvironmental Eng. 138 (4) (2012) 441–450.
profiles using cone penetration test data, Comput. Geotech. 90 (2017) 73–84.
[20] M. Chew, Y. Ng, F. Kok, Shien, S. Ng. The effect of soil variability on the ultimate
[50] F.S. Tehrani, M.I. Arshad, M. Prezzi, R. Salgado, Physical Modeling of Cone
bearing capacity of shallow foundation. Journal of Engineering, Science and
Penetration in Layered Sand, J. Geotech. Geoenviron. Eng. 144 (1) (2018)
Technology 10 (2015) 1–13.
4017101.
[21] B.K. Low, K.K. Phoon, Reliability-based design and its complementary role to
[51] P.K. Robertson, Soil classification using the cone penetration test, Can.
Eurocode 7 design approach, Comput. Geotech. 65 (2015) 30–44, https://doi.
Geotech. J. 27 (1990) 151–158, https://doi.org/10.1139/t90-014.
org/10.1016/j.compgeo.2014.11.011.
[52] P.K. Robertson, Cone penetration test (CPT)-based soil behaviour type (SBT)
[22] Zaskórski, Łukasz & Puła, Wojciech. Safety assessment of a shallow foundation
classification system — an update, Can. Geotech. J. 53 (12) (2016) 1910–1927,
using the random finite element method, Fifth International Symposium on
https://doi.org/10.1139/cgj-2016-0044.
Geotechnical Safety and Risk (ISGSR) 2015 Rotterdam, the Netherlands 13-16
[53] U.S. Army Corps of Engineers. (1992). Engineering and design:Bearing capacity
October 2015, DOI: 10.3233/978-1-61499-580-7-677.
of soils. Em 1110-1-1-1905.
[23] P.K. Robertson, A.V. Da Fonseca, B. Ulrich, J. Coffin, characterization of
[54] FHWA-SA-91-043, 1992 report.
unsaturated mine waste: a case history, Can. Geotech. J. 54 (12) (2017).

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011
R.R. Shakir / Engineering Science and Technology, an International Journal xxx (2018) xxx–xxx 21

[55] CFEM, in: Canadian Foundation Engineering Manual, third ed., Canadian [64] D.V. Griffiths, G.A. Fenton, D.E. Tveten, Probabilistic geotechnical analysis:
Geotechnical Society, BiTech. Publishers, Vancouver, 1992, p. 512. How difficult does it need to be?, in: R. Poettler (Ed.), Proceedings of an
[56] P.W. Mayne, in: NCHRP Synthesis 368: Cone Penetration Test, Transportation international conference on, probabilistics in geotechnics: technical and
Research Board, National Academies Press, Washington DC, 2007, p. 118. economic risk estimation, Pub. VGE, Essen, Germany, 2002, pp. 3–20.
[57] M.S. Seyedein, R.J. Chinari, A. Eslami, Investigation on probability density [65] J.M. Duncan, Factors of safety and reliability in geotechnical engineering, J.
function for cone penetration test data. In: The 2012 world congress on Geotech. Geoenviron. Eng. 126 (4) (2000) 307–316.
advances in civil, environmental, and materials research (ACEM’12), (pp. 26- [66] A.M. Hasofer, Reliability index and failure probability, J. Struct. Mech. 3 (1)
30). Seoul, Korea (2012). (1974) 25–27.
[58] I. Laufer, Statistical analysis of CPT tip resistances, Periodica Polytechnica Civil [67] B.K. Low, W.H. Tang, Reliability analysis using object-oriented constrained
Engineering 57 (1) (2013) 45–61, https://doi.org/10.3311/PPci.2141. optimization, Struct. Saf. 26 (1) (2004) 69–89, https://doi.org/10.1016/S0167-
[59] G.A. Fenton, D.V. Griffiths, Risk Assessment in Geotechnical Engineering, John 4730(03)00023-7.
Wiley and sons, Hoboken, New Jersey, 2008. [68] AASHTO, 2012. AASHTO LRFD bridge design specifications, customary U.S.
[60] T.B.M.J. Ouarda, C. Charron, F. Chebana, Review of criteria for the selection of units. AASHTO load and resistance factor design bridge design specifications.
probability distributions for wind speed data and introduction of the moment [69] D.G. Kamien, 1997. Engineering and Design: introduction to probability and
and L-moment ratio diagram methods, with a case study, Energy Convers. reliability methods for use in geotechnical engineering. Engineering
Manage. 124 (2016) 247–265. Technology. Letter no. 1110-2-547. Department of the Army. Washington: U.
[61] D.S. Youssef Abdel Massih, A.-H. Soubra, B.K. Low, Reliability-based analysis S. Army Corps of Engineers.
and design of strip footings against bearing capacity failure, J. Geotech. [70] D.V. Griffiths, G.A. Fenton, Bearing capacity of spatially random soil: the
Geoenviron. Eng. 134 (7) (2008) 917–928, https://doi.org/10.1061/(ASCE) undrained clay Prandtl problem revisited, Geotechnique 51 (8) (2001) 731.
1090-0241(2008) 134:7(917). [71] M.A. Shahin, E.M. Cheung, ‘‘Probabilistic analysis of bearing capacity of strip
[62] A.B. Salahudeen, J.M. Kaura, Leonardo Electronic Journal of Practices and footings,” Department of Civil Engineering, Curtin University, Perth, Australia
Technologies (LEJPT) (30) (2017) 127–148. ISGSR 2011-Vogt, Schuppener, Straub & Bräu (eds) Ó 2011 Bundesanstalt für
[63] J. Spross et al., Reliability aspects of rock tunnel design with the observational Wasserbau ISBN 978-3-939230-01-4.
method Reliability aspects of rock tunnel design with the observational [72] A.S. Alawneh, O.K. Nusier, A.A.E. Al-Mufty, Reliability based assessment of
method, Int. J. Rock Mech. Min. Sci. 98 (July) (2017) 102–110. shallow foundations using mathcad, Geotech. Geol. Eng. 24 (3) (2006) 637–660.

Please cite this article in press as: R.R. Shakir, Probabilistic-based analysis of a shallow square footing using Monte Carlo simulation, Eng. Sci. Tech., Int. J.
(2018), https://doi.org/10.1016/j.jestch.2018.08.011

Você também pode gostar