Synopsis

1.1 General

In engineering applications it is important to model and adequately treat all the available
information during the analysis and design phase. Typically, the information are originated
from different sources: field measurements, experts' judgments, objective and subjective
considerations. Over these features, the influences originated from the human errors,
imperfections in the construction techniques, influence of the boundary and environmental
conditions are added. All these aspects can be brought back to one common denominator:
presence of uncertainty. The uncertainty can be viewed as a part or class of imperfection in
the information that attempts to model a system behaviour in the real world. It is the gradual
assessment of the truth content of postulation e.g. in relation to the occurrence of a defined
event. French (1986) discussed different types of uncertainty and imprecision (including
physical randomness of data, choice of a model, numerical accuracy of calculations or lack of
clarity in the objectives) and their consequences in the process of modelling and analysis.
Normally, the uncertainty is viewed in two categories, namely aleatory and epistemic. The
aleatory uncertainty is classified as objective and irreducible uncertainty with sufficient
information on the input uncertain data. These are inherently connected to the problem at
hand and cannot be influenced by the designer. The epistemic uncertainty is a subjective and
reducible uncertainty that stems from the lack of knowledge about the input uncertain data. It
arises from the cognitive sources involving the definition of certain parameters, human
errors, inaccuracies, manufacturing and measurement tolerances etc. In brief, the objective
uncertainty is concerned with the tendency of an event to occur and the subjective
uncertainty is concerned with the ability to occur. In real application it is classified in such a
way that a mathematically founded and realistic description is ensured in the structural
analysis and safety evaluation (Möller et al. 1999).
The profession has accepted the fact that the existence of uncertainty cannot be
avoided in the analysis and design of engineering system. It is now well recognized that
when the existence of uncertainty is taken into account it leads to a more cost effective
design rather than when it is planned to eliminate or greatly reduce them for final design that
will be safe, reliable and robust against uncertainty. Thus, the consideration of uncertainty in
engineering analysis and design process is gaining increasing importance in the profession.
The uncertainty quantification in a typical engineering decision making process requires: (i)
Characterization of the uncertainty involve in various system parameters and external
environment; (ii) Propagation of this uncertainty through engineering models and
computational tools. The first step of characterization of the uncertainty involves the
development of methodologies to model the uncertainty of both the epistemic and the
aleatoric type. Regardless of the type being considered, the characterization process depends
on the Experimental Research and Expert Judgement. The outcome of the process is in the
form of probability distribution function (pdf), membership function (mf), interval bounds
etc. depending on the quantity and quality of data feasible to acquire. The propagation of
uncertainty mainly involves two aspect i.e. the response analysis of system considering the
uncertain input parameters and the associated safety analysis compatible with the decision
making process. The response of structural model under excitation can be mathematically
expressed by a set of equation with suitably prescribed boundary and initial conditions as:
  Y ( x, t )   F  x, t  (1)

Where Y (x, t ) is the output of the system due to input F (x, t ) and  is the differential
operator, representing the mechanical system. The model as defined by Eqn. (1.1) involves
two independent variables i.e. the position vector x and the time variable t. Depending on the
nature of the variables involve in a typical problem, various types of analysis problems may
arise. Those can be broadly classified as depicted in table 1.1. Nowadays, the response
analysis under uncertainty is generally performed through numerical model and various
methods like the stochastic finite difference method (SFDM), the stochastic finite element
method (SFEM), the stochastic boundary element method (SBEM), the fuzzy finite element
method (FFEM) etc. are emerged for this. The response surface method (RSM) based
various metamodel strategies are also very common to approximate expensive computer
simulations (Box and Draper 1987, Jin et al. 2001). The safety assessment is performed
through reliability analysis either in the probabilistic or in the possibilistic approach
depending on the quality and quantity of information available about the input uncertain
parameters.
Table 1.1: Classifications of structural analysis problem
Level System Input Output Remark
Operator(Ω) (F) vector (Y)
1 Deterministic Deterministic Deterministic Conventional deterministic
structural analysis
2 Deterministic Deterministic Uncertain Random vibration
3 Uncertain Uncertain Deterministic Stochastic system (SFEM,
SFDM, SBEM, FFEM etc.)
4 Uncertain Uncertain Uncertain Most general models

1.2 Safety Analyses of Structures under Uncertainty

In last twenty years or so, many new methods have been developed to deal with the
imperfection in input data. The large number of models reflects that there exist many aspects
of imperfection and the probability theory is not the unique normative model that can be
coped with all of them. In fact, numerous methods are used to deal with the uncertainty in
natural sciences and engineering. These include from the probability theory and its variants
(Bayesian theory, reliability theory), to multi-valued logic, fuzzy and related possibility
theory, interval analysis etc. Rouvray (1997) presented an accessible history of those
methods.

1.2.2 The Probabilistic Approach

For a long time, the probability theory was the only approach used to quantify the uncertainty
of input parameter necessary to model the engineering system. Even now, the probabilistic
methods are almost exclusively used in the industry. The applications vary from the
structural design to biomechanics, to water quality control, to estimate the output of drainage
systems, designing integrated circuits, traffic control and traffic flow estimation and so on. In
probabilistic approach, the assurance of performance is referred to as the reliability. The
computation of probability of failure (pf) by above equation is the fundamental of safety
evaluation in probabilistic approach. The full distribution approach requiring the joint pdf of
random variables is almost impossible; moreover, evaluating the multiple integral is a
formidable task. Thus, the Monte Carlo Simulation (MCS) technique is used as the robust
alternative to compute the pf for both the explicit and implicit limit state function with known
pdf of the random variables. But it requires a number of deterministic analyses ranging
between few hundreds to tens of thousands depending on the magnitude of pf. Normally, the
second moment based approximation methods i.e. the first order reliability methods (FORM)
and the second order reliability methods (SORM) are proposed in the literature to obtain a
reasonable estimate of pf with significantly lower computational effort. The second moment
methods require fewer deterministic analyses provided that the closed form expressions of
the sensitivity derivatives of the performance function with respect to the random variables
are available. Otherwise, one may need to opt for the RSM to approximate the failure surface.
The second moment methods may yield erroneous estimate of pf for which there are multiple
most probable point as in the case of structural dynamic problems. In this regard, the most
significant criticism on the widespread use of classical reliability methods is that the
information input in the analysis has to be in a precise probabilistic format and the limit state
function through which this information is propagated is a precise model. The statistical
distributions of the parameters, good information on correlations etc. are seldom known for
all random variables in real life design problems. Moreover, the probabilistic approach is
based on the accuracy of the probabilistic model of the random parameters. It has been
shown (Ben-Haim and Elishakoff, 1990) that even small errors in the statistical parameters
may have large effects on the computed pf, especially when these probabilities are very small.
The development in the field is quite extensive. In fact there are numerous well known texts
(Ditlevsen and Madsen 1996, Melchers 1999, Haldar and Mahadevan 2000).

1.2.2 The Possibilistic Approach

The reliability evaluation of structural system is evaluated based on the classical probability
theory in which some of the variables are considered to be random and rests are assumed as
deterministic. The limit of applicability of such approach is attained when insufficient
reliable data are only available to describe the real life systems with the aid of pdf. The
effectiveness of probabilistic approach is thus lost with non-probabilistic uncertainty in the
system input parameters. Moreover, the real world problems are more complex than their
corresponding mathematical model and to compensate this gap, some linguistic explanation
occasionally adds to the results obtained through the models. In order to quantify such
information, it is possible to apply the uncertainty measure on the basis of existing available
data with the additional information from expert knowledge and experience. This has lead to
the development of various possibilistic methods. The related developments are summarized
in the following.

1.2.2a The Possibility/Fuzzy Set Theory Based Approaches

The possibility is an alternative approach to the probability, initially introduced to model the
uncertainties when the available information is linguistic. Various methods have been
developed to deal with such subjective uncertainty. It is based on the possibility distribution
defined by the mf obtained from the numerical data along with the expert knowledge and
experience. The possibilistic safety evaluation algorithms mainly tried to explore the entire
range of the uncertain variables to estimate the pf. In a typical interval analysis or  -cut in
fuzzy set approach, repeated standard probabilistic reliability analysis is performed to obtain
the worst failure distribution pattern. Brown (1979) first apply the fuzzy measure concept
with classical structural reliability theory to obtain more reliable failure modes.
Subsequently, the outline of fuzzy theory and its contribution to structural reliability
assessment is enormous (Shiraishi and Furuta (1983), Yao and Furuta (1986) Furuta (1995,
Möller and Beer 2005) which is further detailed in the next chapter.

1.2.2b Interval Analysis and Convex Modeling

In many cases, for example in preliminary design phases, even though some experimental
data are available, it is not enough to construct the pdf, reliably. The available data can be
used, particularly in combination with the engineering experience, to set some tolerances or
bounds only. In such cases, the uncertain design variables can be well modelled using the
non-probabilistic convex models of uncertainty. Examples of convex models include
intervals, ellipse or any convex sets. This kind of mathematical model is called uncertain but
bounded (UBB) model. A quantitative non-probabilistic reliability measure based on the
convex model was first introduced by Ben-Haim (1995) and the subsequent developments on
the structural reliability analysis in which the bounds on the magnitude of uncertainty is only
required are notable (Penmetsa and Grandhi 2002, Qiu et al. 2008, Karanki et al. 2009). The
interval method of analysis seems to be a logical alternative when the parameters required to
create the probabilistic models cannot be precisely determined due to lack of data. Interval
analysis considers rectangular model which encloses all the possible combinations of the
uncertain variables i.e. consists of all the possible probabilities that are consistent with the
available information. It is basically a worst scenario method since all the UBB variables
vary independently and thus may reach their extreme values simultaneously lead to overly
conservative design. The ellipsoid model considers all the variables to be correlated with
each other, which excludes the extreme combination of the uncertain parameters and thus
avoids over-conservative designs. However, in reality, only part of these UBB variables is
actually correlated while some others vary independently. Therefore, a more realistic option
is to divide all the UBB quantities into groups and treat them with a multi-ellipsoid convex
model (Luo et al. 2008).

1.3 Probabilistic or Possibilistic Safety Analysis

The advantages of probability-based safety analysis are now well known and have been
accepted broadly by the profession (Sexsmith 1999). In fact, it has become an essential
aspect of structural safety analysis. Various second moment based approximations
(FORM/SORM) approaches, well matured and established in the profession can provide good
estimate of pf with significantly lower computational effect. However, the second moment
methods may yield erroneous estimate of the failure probability for problem having multiple
MPP (e.g. as in case of structural dynamic problems). The errors using FORM/SORM in
view of the large uncertainty in selecting the appropriate stochastic model and its parameters,
various RSM based MCS methods are becoming more attracting in the light of enormous
development in the field of computational science. In this regard, the most significant
criticism on the widespread use of classical reliability methods is that the information input
in the analysis has to be in a precise probabilistic format and the limit state function through
which this information is propagated is a precise model. The concept of predictive reliability
index that provides a measure of uncertainty in estimated reliability index due to parameter
uncertainties are worth mentioning (Kiureghian 2008). It has been shown (Ben-Haim and
Elishakoff, 1990) that even small errors in the statistical parameters may have large effects
on the computed pf, especially when these probabilities are very small. The statistical
distributions of the parameters, good information on correlations etc. are seldom known for
all random variables in real life design problems. Scarcity of data available necessitates
strong assumptions which may be sensitive enough in terms of safety of the system. Thus
possibilistic methods requiring less information yet can provide a measure of reliability is
becoming attractive particularly, at early stage of design. The subjective uncertainty
representing the design imprecision and inexactness in choosing among design alternatives
usually dominates the preliminary design configuration. Moreover, the probabilistic approach
is difficult to make more conservative to protect the design to inaccuracies made due to lack
of information. But, it is easy to increase the degree of conservatism in possibilistic approach
[Sophie 2000]. However, the issue of switching from probabilistic to possibilistic approach
by justifying whether the information is little is not very clear. It is also important to note that
the possibility theories are of little use in design of system with a large number of failure
modes that are known to be independent (Chen et al 1998). As the number of modes increase
the size of failure region increases lead to larger probability of failure, but the possibility of
failure remains the same as it is the possibility of a single mode (the maximum one).
Possibility of failure effectively imposes a factor of safety (>1) on the probability of failure
in general. This conservatism would certainly ensure performance but could adversely affect
the optimum cost. With the progress of iterative design process, these types of uncertainty
reduced gradually, but the objective uncertainty remains throughout the design process.
Though, the valuable comparisons between these two methods are available in the literature
(Sophie 2000, Nikolaidis et al. 2004), no work seems to be available to find the relationship
between the two approaches. Moreover, as such there is no consensus about what method
should be used in many real life problems. The general understanding is that if the
uncertainties are modelled accurately, the probabilistic methods are better than its counterpart
for efficient design (Nikolaidis et al. 2004). But, it has been repeatedly invoked in literature
that possibility is a better choice in cases of scarcity of data as it is not only safe but also
simpler to apply.

1.4 The Safety Assessment under Hybrid Uncertainty

The probabilistic and the possibilistic methods of reliability analysis of structures as
mentioned in the previous section have been developed independently i.e. the former is to
consider the random uncertainty and the later is for non stochastic uncertainty. Though, the
valuable comparisons between these two methods are available in the literature (Sophie
2000, Nikolaidis et al. 2004), no work seems to be available to find the relationship between
the two approaches. Moreover, as such there is no consensus about what method to be used
in many real life problems. The general understanding is that if the uncertainties are
modelled accurately, the probabilistic methods are better than its counterpart for efficient
design. But, the scarcity of data available may need to make strong assumptions which can
be sensitive enough in terms of safety of the system. It has been repeatedly invoked in the
literature that the possibility is a better choice in such case as the approach is not only safe
but also simpler to apply. Interestingly, the development in the field of reliability analysis are
based on the fundamental assumption that the input system parameters information are either
all possibilistic or all probabilistic. Various transformation methods have been emerged in
the literature to transform the possibilistic variables to equivalent probabilistic variables or
vice versa. The concept of establishing the equivalence between the fuzzy and random
variables through entropy principle (Haldar and Reddy 1992, Rahman and Zahaby 1997,
Zhenyu and Chen 2002, Marano et al. 2008), scale the fuzzy mf of the system response to a
pdf Smith et al. (2002), Balaji (2006), equivalent class of pdf compatible with the
corresponding fuzzy descriptions Ferrari and Savoia (1998), Savoia (2002).

1.5 Objective of The present study

The focus of the present study is on the reliability analysis of structures characterized by both
the probabilistic and the possibilistic variables. It is generally realized that either the
probabilistic or the possibilistic approach is not compatible to tackle hybrid uncertainties.
But, the concept of various transformation methods can be readily used to unify the mixed
variables so that the existing probabilistic or the possibilistic reliability analysis format will
be compatible. The fuzzy variables need to be transformed to equivalent random variables or
vice versa for safety analysis under mixed uncertain variables. The concept of various
transformation methods seems to be potential to unify the mixed variables so that the existing
probabilistic or possibilistic reliability analysis format will be compatible. Relying on the
fundamental concept of such transformations, the feasibility of reliability analysis of hybrid
uncertain system either in the probabilistic or in the possibilistic format is the subject of the
present study. This will provide the necessary flexibility to the designers to model the
structural parameter uncertainty either probabilistically or possibilistically depending on the
nature and quality of the input data. The specific objectives of the present research work are
briefly describe in the following

1.5.1 Probabilistic Safety Analysis under Random and Fuzzy Uncertain Parameters
The focus of the present section of the study is on the probabilistic approach of reliability
analysis of structures characterized by both the probabilistic and the possibilistic variables.
The probabilistic variables are described by the associated pdf and the possibilistic variables
are described by the associated fuzzy mf. The limit state function of the related reliability
analysis problem involves the probabilistic and also the possibilistic variables. To make the
analysis compatible with the reliability analysis in the probabilistic format, one needs to
express the performance function in terms of the random variables. A structural system in
which some variables are random and some are fuzzy, the properties of the equivalent
random variables are obtained following the fundamental concept of transformations. If the
transformed variables are not normal, those are further transformed to equivalent normal
variables by applying suitable transformation like the Rackwitz-Fiessler algorithm. Now, the
standard second moment format is readily applied to obtain the reliability index and the
corresponding probability of failure. For a comparative study and consistency of
transformation approach of reliability analysis of hybrid uncertain systems, a conservative
upper bound and lower bound of pf is also computed following the equivalent class of
probability distribution boundaries following the interpretation of the possibility distribution
based on the evidence theory. The lower CDF and the pdf for the fuzzy variables are
obtained from the mf and those are transformed to equivalent normal variables. To obtain the
lower bound pf, the same procedure as used to obtain the upper bound value is followed.
Finally, the reliability computation under hybrid uncertainty is elucidated through numerical
examples to illustrate the capability, consistency and suitability of the transformation
approach to evaluate the safety of hybrid uncertain system. The first two examples (a simply
supported reinforced concrete beam and a ten bar truss) have explicit limit state functions.
The third (a reinforced concrete cast in situ bored pile embedded in clay) and the last one
(two storied portal frame subjected to lateral and vertical load), the limit state functions are
not explicit function of the uncertain parameters. The Response Surface Method (RSM) is
used to approximate the limit state functions.

1.5.2 Possibilistic Safety Analysis under Random and Fuzzy Uncertain Parameters
The present section focused on the possibilistic reliability analysis of structures characterized
by both the probabilistic and the possibilistic variables. Similar to the previous section, the
fundamental concept of transformations are used for safety analysis. However, to perform the
possibilistic reliability analysis, the random variables defined by the associated pdf can are
transformed to equivalent fuzzy variables following the transformation concept. As already
discussed, various transformation approaches have been used for transforming random
variable to equivalent fuzzy variable or vice versa. The entropy based transformation is
hinged on sound mathematical basis and the theory of expressing the uncertainty information
is well established and applied in various fields of engineering. In the present study, the
entropy based transformation approach is applied to transfer the random variables to
equivalent fuzzy variables so that the limit state function can be expressed in terms of the
fuzzy variables only. Subsequently, the possibilistic safety analysis is performed following
the procedure as outlined in the previous section. The bounds on the probability of failure
based on the evidence theory are also computed to study the consistency of the possibility of
failure results obtained by the transformation based algorithm. The possibilistic safety
evaluation of structures characterized by hybrid uncertain parameters is elucidated through
four numerical examples which were considered in the previous section. However, the safety
analysis is now performed in possibilistic format. The purpose of the example problems are
to demonstrate the capability of the transformation approaches to tackle the presence of
mixed uncertain parameters to perform the possibilistic safety analysis of structure.

1.5.3 Safety Analysis of System under Random and Bounded Uncertain Parameters
The present section deals with the second moment safety evaluation format for structural
system characterized by hybrid uncertainties i.e. both probabilistic and interval type uncertain
parameters. The available literatures on the application of interval analysis approach in the
framework of set theoretical description are limited to deal with response evaluation and
optimization. But it is appeared to be potential to apply in solving structural reliability
analysis problem. The approach is extended here to reliability analysis problem with an
objective of evaluation of reliability of structures involving both random and bounded
variables. The bounds of performance function (limit state function) of a typical structural
reliability analysis problem with UBB parameters are presented first by making use of
interval mathematics. Subsequently, the application of the approach to fuzzy variables has
been elaborated next. The safety evaluation of systems under random and bounded uncertain
parameters is investigated through two numerical examples (a simply supported reinforced
concrete beam and a reinforced concrete cast in situ bored pile embedded in clay) which were
considered in the previous section. However in present case, the information available about
the uncertain variables are not sufficient to describe those probabilistically and can be
modelled as UBB type.

1.6 Representative Results of the Numerical Studies

The reliability computations are elucidated through various numerical examples as
mentioned in the previous sections to illustrate the capability, consistency and suitability of
the transformation approach to evaluate the safety of hybrid uncertain system. To provide a
synoptic view of the detailed numerical studies, representative results from the following
three examples are presented in this section: safety evaluation of a simply supported
reinforced concrete beam subjected to a concentrated load at centre for probabilistic approach
under hybrid uncertain system; safety evaluation of a two storied two bay portal frame,
subjected to horizontal and vertical load for possibilistic approach under hybrid uncertain
system; and a reinforced concrete pile embedded in clay, subjected to horizontal and vertical
load as well as moment at tip for safety evaluation considering UBB and random parameters.

1.6.1 A Reinforced Concrete Beam: Probabilistic

The beam is subjected to a concentrated load at centre and reinforced with steel bars of area
(As) =1400 mm2, having length of 4.0 m. The characteristic strength of concrete (fck) and the
yield stress of steel (Fy) are considered as random variables and assumed to be normally
distributed. The load (P), width (b), depth (D) and the cover of reinforcements (t) of the
beam are modelled as fuzzy variables. The following mean values of the concrete beam are
considered in the numerical study: Fy = 320 N/mm2, fck =20N/mm2, B=300mm, D=50mm2. In
the present study, the reliability is defined with respect to the flexural mode of failure.
Considering the ultimate flexural strength of the beam, the equation of failure surface is,
 0.77 Fy . As 
g  Fy . As .  D  t  1    Me , (2)
 f ck .b.  D  t  

Where, M e  PL / 4 is the moment due to external load. The probability of failure of the
beam considering the above limit state function is computed in the FORM framework by
transferring the fuzzy variables to equivalent random variables. The reliability results are
presented in Fig. 1 with increasing nominal value of the concentrated load. The support width
of the symmetrical triangular fuzzy variables described as a function of the standard
deviation is taken as two, unless mentioned otherwise. The coefficients of variation (COV) of
all the uncertain parameters are taken as 0.2. Following the interpretation of the possibility
distribution based on the evidence theory (Dubious and Prade 1991), an equivalent class of
probability distribution where the lower bound and the upper bound of the probability are
shown to be the possibility and the necessity are also obtained. Based on this, the upper and
lower bound pf are also computed and are shown in the same figure to study the consistency
of the proposed transformation based results. As expected the width of the bounded solution
increases as the level of uncertainty increases. However, the pf are within the bounded
solution as obtained based on the conservative evidence theory. The variation of pf with
increasing load is shown in Fig. 2 by considering unsymmetrical triangular variation of the
fuzzy variables. The left and right end supports are now defined by two unequal parameters
w1 and w2 . In Fig.2, w1 1.0 and w2  3.0 and the COV of all the uncertain parameters are

taken as 0.2.

0.5 Upper Bound 0.5

Lower Bound Upper Bound
Scaling Lower Bound
Probability of Failure (pf)

0.4 Entropy Scaling

Probability of Failure (pf)

0.4
Entropy

0.3 0.3

0.2 0.2

0.1
0.1

0.0
0.0
50 100 150 200
50 100 150 200
Load (KN)
Load (KN)

Fig. 1: The variation of pf with increasing Fig. 2: The variation of pf with increasing
load for symmetric triangular fuzzy variables load for unsymmetrical fuzzy distribution
for support width w=2.0 with support width w1=1.0 and w2=3.0

1.6.2 A Two Storied Portal Frame: Possibilistic

The horizontal load (X1), storey height (X2) and applied superimposed load (X5) are
considered to be fuzzy variables. The modulus of elasticity of material (X3) and depth of
beam in each storey (X4) are considered to be random variables. The nominal values are
considered as: X1=20kN X2=3m X3=25000Mpa X4=450mm X5=5 kN/m2. The COV for all
variables are considered to be 0.1. The reliability is computed with respect to the maximum
horizontal displacement and the maximum bending moment. Since the limit state equations
are not explicit function of the uncertain parameters, the RSM is used to approximate those.
The response analysis is performed by using STAAD Pro Software. The maximum
horizontal displacement and bending moment values are extracted from the analysis results
corresponding to each set of input variables as per Design of Experiment (DOE) to construct
the response surfaces. The fuzzy variables (X1, X2 and X5) are described by the symmetric
triangular fuzzy distribution. The remaining variables i.e. X3 and X4 are assumed to be normal
random variables.
The possibility of failure of the frame considering the limit state function for
maximum horizontal displacement is computed by transferring the random variables to
equivalent fuzzy variables. The possibility of failure results for symmetric triangular fuzzy
mf of different support width (w) are computed and presented in Fig. 3 with increasing
nominal value of the concentrated load. For an unsymmetrical triangular fuzzy distribution
with different support width (w 1 & w2), the variation of possibility of failure with increasing
nominal horizontal load is shown in Fig. 4. The COV of all the uncertain parameters are
taken as 0.1 to develop this figure.

1.2 1.2

1.0 1.0
Possibility of Failure

Possibility of Failure

0.8 0.8

0.6 0.6

0.4 Support Widths 0.4 Support Widths

 
 
0.2 0.2
 

0.0 0.0
20 30 40 50 60 20 30 40 50 60
Load (KN) Load (KN)

Fig. 3: The variation of possibility of failure Fig. 4: The variation of possibility of failure
with increasing nominal value of the with increasing nominal value of the
horizontal load for symmetric triangular horizontal load for unsymmetrical triangular
fuzzy variables for different support width fuzzy variables for different support width

1.6.3 A Reinforced Concrete Pile: Bounded

A reinforced concrete cast in situ bored pile embedded in clay, subjected to horizontal and
vertical load as well as moment at tip of the pile is taken up to elucidate the safety evaluation
considering random and UBB type parameter. The statistical properties of the random and
fuzzy variables are described in table 1. The pile diameter (X1), modulus of elasticity(X2),
applied horizontal load (X11), vertical load (X13) and moment (X12) at tip of the pile are
considered to be random variables. The length of pile in various soil layers (X3 to X6) i.e. in
soft clay, moderately stiff clay, stiff clay, very stiff clay and the corresponding values of
modulus of sub-grade reaction (X7 to X10) are considered to be fuzzy variables.
The reliability is computed with respect to the maximum horizontal displacement and
bending moment developed in the pile. Since the limit state functions are not explicit
function of the uncertain parameters, the RSM is used to approximate the limit state
functions. To obtain the response of the pile necessary for constructing resposne surfaces for
each limit state function considered in the present safety assessment study, the pile is
considered as a frame structure in finite element model and discretized by 24 numbers of two
noded frame element. The response analysis is performed by using STAAD Pro Software.
The maximum horizontal displacement and bending moment values are extracted from the
analysis results corresponding to each set of input variables as per DOE to construct the
response surfaces.
Table 1 Properties of the random and fuzzy variables of R.C. Pile embedded in clay

Notation Units (,) Notation Units (,)

X1 (mm) (500.0, 50.0) X7 (KN/m3) (15000 , 1500)
X2 (N/mm2) (29580.4,2958.0) X8 (KN/m3) (27000,2700)
X3 (mm) (1500.0 ,150.0) X9 (KN/m3) (35000,3500)
X4 (mm) (2000.0,200.0) X10 (KN/m3) (70000,7000)
X5 (mm) (1500.0 ,150.0) X11 (KN) (25,2.5)
X6 (mm) (3500.0 , 350.0) X12 (KN-m) (80 ,8)
X13 (KN) (100 ,10)

The radiuses of bounds are expressed as the percentage of the nominal property values and
COV of random variables are taken as 0.1. The nominal value of load is considered to be 30
KN to develop this plot. The reliability is also computed assuming that the interval variables
are uniformly distributed. The reliability results are presented in Fig. 5 with increasing value
of interval length expressed as percentage of mean value.
 Upper Bound 6.0
5.85
Lower Bound Upper Bound
Uniform 5.5
Lower Bound
5.80 Uniform
5.0
Reliability Index

5.75
4.5

Reliability Index
5.70 4.0

3.5
5.65
3.0
5.60
2.5
5% 10% 15% 20% 25%
2.0
x(%)
20 30 40 50 60 70 80 90 100 110
Load (KN)

Fig. 5 The variation of reliability index with

increasing value of interval length. Fig. 6: The variation of reliability index with
increasing load for Δx =10% and COV=10%.

Keeping value of interval length (Δx) and COV of random and UBB variables to be 0.1, the
bounds of reliability index as well as reliability index based on uniform assumption are
shown in Fig. 6 with increasing value of nominal load.
Bounded and Fuzzy
The effectiveness of proposed safety evaluation approach under hybrid uncertainty is

investigated by considering the same example problem taken with same properties of random

and fuzzy variables as depicted in table 1. Each fuzzy variable is converted to a bounded

variable at ith α-cut level. Once fuzzy variables are converted to a bounded variable at ith α-cut

level, the bounds on the reliability index and the associated probability of failure is readily

obtained following standard second moment reliability analysis format. The variation of

deffuzzified reliability index with increasing uncertainty level and different load is shown in

Fig. 7. The support width as defined by the weight factor w is taken as 3.0. The value of

interval length expressed as percentage of mean value is assumed to be 10% to develop this

figure. The distribution of reliability index for different support widths having same COV is
depicted in Fig. 8. The COV of random and fuzzy variables are assumed to be 10% to

develop this plot.

6.0 P=30KN Support width = 

1.0
P=20KN Support width =
5.5 Support width=
P=50KN
0.8
5.0

4.5
0.6
Defuzzified Beta

Membership
4.0
0.4
3.5

3.0
0.2
2.5
0.0
2.0

10% 15% 20% 25% 30% 5.3 5.4 5.5 5.6 5.7 5.8 5.9
COV (%) Reliability Index

Fig. 7 The variation of defuzzified reliability Fig. 8 The distribution of reliability index for
index with increasing COV for different loads different support width for COV =10%.
for Δx =10%

1.7 Summary and Conclusion

The concept of establishing the equivalence between the fuzzy and the random variables are
found to be potential to perform the safety evaluation either in the probabilistic or the
possibilistic framework under hybrid uncertainties. Relying on such emerging transformation
concepts, the safety analysis of structural system characterized by hybrid uncertain
parameters are demonstrated in present chapter. The transformation approaches satisfy the
consistency condition that the possibility of an event should be greater than or equal to its
probability. The entropy based transformation directly provides the equivalent normally
distributed parameters, whereas in scaling approach there is an additional requirement that
the probabilistic parameters need to be transformed to equivalent normal variables. The
probability of failure obtained by the proposed transformation approaches are in conformity
with the failure probability bounds derived based on the evidence theory. From the numerical
study it is seen that the trend of failure probability obtained by both the scaling and the
entropy based approach are the same. However, there is a definite difference in the reliability
index obtained based on the two transformation approaches. The bounds on failure
probability based on the evidence theory are too far apart and too conservative which is
obvious as the evidence theory gives conservative estimate of the upper bound and the lower
bound of the CDF. These bounds are computed to serve as a check whether the
transformation based algorithm produces reliability results are within the bounds obtained
from a different approach following the consistency principle. The authors are of the opinion
that the entropy based transformation is hinged on sound mathematical basis and the theory
of expressing the uncertainty information is well established and applied in various fields of
engineering. It is safe to apply entropy based transformation approach when no specific
judgment is available regarding the distribution pattern of the uncertain variables. The
conventional safety analysis algorithms normally assume all the variables to be of single type
i.e. either all probabilistic or all possibilistic. As such this approach introduces gross
assumptions at the very beginning in modelling of the system, whereas in the proposed
approach there is always some chance to lose some of the information during transformation
stage. But, the present approach offers more flexibility to the designer in realistic modelling
of the system. However, it is not very clear in open literature, which is more justified and it
is felt to need more study on this aspect.
The lower bounds of reliability obtained by the interval analysis approach are smaller than
those predicted by assuming that the UBB variables are uniformly distributed probabilistic
variables. Similarly, the upper bounds yielded by the present interval method are larger than
those obtained by the probabilistic assumption. Thereby, the reliability of structures with the
UBB parameters obtained by the interval analysis method contains those obtained based on
the most conservative uniform pdf assumption. This kind of results is in consistent with the
membership distribution of fuzzy reliability. In other words, it ensures that the possibility
distribution is always greater than the probability distribution. The fuzzy reliability
distribution and the defuzzified reliability index for different load levels or uncertainty levels
as obtained by the proposed approach are also found to be consistent. However, the present
analysis is based on the assumption of monotonic responses i.e. response lies at the end
points of the UBB parameters. If the uncertain parameters are not so, alternative method
should be applied. It may be noted that the proposed approach is based on the first order
Taylor series expansion. Consequently, as the uncertainty level of the parameters becomes
larger, the predicted fuzzy membership distribution of reliability will be increasingly
inaccurate. A second order perturbation analysis may be explored to improve the accuracy.
However, it needs further study.

1.8 Organization of the Thesis

The entire work is presented in six chapters. The details of the fundamental developments in
the field of safety analysis of system characterized by mixed uncertain parameters and the
specific objectives of the present work are introduced in chapter one, followed by a state-of-
the-art review of related literature in chapter two. The issues related to the reliability analysis
under probabilistic uncertainty, possibilistic uncertainty and hybrid uncertainties are
addressed here. The third chapter deals with the theoretical developments of the probabilistic
safety analysis under random and fuzzy uncertain parameters and its efficiency is
investigated with the help of four informative examples. Subsequently, the fourth chapter
deals with the theoretical developments of the possibilistic safety analysis under random and
fuzzy uncertain parameters and its efficiency is investigated with the help of same four
example problems of previous chapter. Chapter five deals with the safety analysis of system
under random and uncertain bounded parameters. Finally, chapter six draws the significant
conclusions based on present study and identifies important recommendations towards future
direction of studies.

1.9 References
1. Anoop M. B. , Balaji R. K., Gopalakrishnan S., Conversion of probabilistic information
into fuzzy sets for engineering decision analysis, Computers and Structures 84 (2006) 141–
155.

2. Ben Haim, Y., and Elishakoff, I., Convex models of uncertainty in applied mechanics,
Elsevier Amsterdam, 1990.

3. Ben-Haim Y. (1995) A non-probabilistic measure of reliability of linear systems based

on expansion of convex models, Struct Safety, 1995, 17, 91-109.

4. ,Box G. E. P., Draper, N. R. Empirical Model Building and Response Surface, Wiley,
New York, 1987.
 5. Brown C. B. (1979) A fuzzy safety measure, entropy constructed probabilities. J of
Engng Mech ASCE, 105(5), 855-871.

6. Chen, Q., Nikolaidis, E., Cudney, H., Rosca, R. T., Comparison of Probabilistic and
Fuzzy Set-Based Methods for Designing under Uncertainty, 40th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference
and Exhibit, St. Louis, MO, 1998, 2860-2874.

7. Ditlevsen O., Madsen H.O., Structural Reliability Methods, John Wiley & Sons Ltd.
1996

8. Ferrari P. and Savoia M., Fuzzy number theory to obtain conservative results with
respect to probability, Computer Methods Appl. Mech. Engg., 160, 1998, 205-222.

9.French, S. (1986) Decision theory: An introduction to the Mathematics of Rationality,

Ellis Horwood Ltd., Chichester.

10. Furuta H., Reliability and optimization of structural systems, Proc. of the VI IFIP WG
7.5 working Congress, ed. Rackwitz, Augusti, Borri, Chapman and Hall, 1995.

11. Haldar A. and Mahadevan S. (2000a) Reliability Assessment Using Stochastic Finite
Element Analysis. John Wiley and Sons: USA.

12. Haldar A. and Mahadevan S.(2000b) Reliability Assessment Using Stochastic Finite
Element Analysis. John Wiley and Sons, USA.

13. Haldar A. and Reddy R. K., A random fuzzy analysis of existing structures, Fuzzy sets
and systems, 48, 1992, 201-210.

14. Karanki D. R., Kushwaha H. S., Verma A. K., Ajit S., Uncertainty analysis based on
probability bounds (p-box) approach in probabilistic safety assessment, Risk Anal 2009,
29(5), 662–75.

15. Kiureghian A. D. (2008) Analysis of structural reliability under parameter

uncertainties, Probab Eng Mech 23,351–358.

16. Luo Y., Kang Z., Luo Z., Li A., Continuum topology optimization with
nonprobabilistic reliability constraints based on multi-ellipsoid convex model, Struct
Multidiscip Optim, 2008, 37(2), 107-119.
 17. Marano G. C., Quaranta G., Mezzina M., Fuzzy time-dependent reliability analysis of
RC beams subject to pitting corrosion, J. of Materials in Civil Engng. ASCE, 2008, 20(9),
578-587.

18. Melchers, R. E., Structural reliability analysis and prediction, John Wiley and Sons
Limited, Chichester, West Sussex, England, second edition, 1999.

19. Möller B., Beer M.(2005) Fuzzy Randomness. Uncertainty in Civil Engineering and
Computational Mechanics, Springer-Verlag, Berlin.

20. Moller B., Beer M., Graph W. and Hoffman A., Possibility theory based safety
assessment, computer-aided civil and infrastructure engineering, 14, special issue on fuzzy
modeling, 1999, 81-91.

21. Nikolaidis E., Chen S., Cudney H., Raphael T. H., Rosca R. T., Comparison of
probability and possibility for design against catastrophic failure under uncertainty, J .of
Mechanical Design ASME, 2004, 126, 386-394.

22. Penmesta R. C. and Grandhi R. C., Efficient estimation of structural reliability for
problems with uncertain intervals, Comp.Struct., 80, 2002, 1103-1112.

23. Qiu et al. 2008

24. Rahman. M. S., Khalid M. and Zahaby E., Probabilistic liquefaction risk analysis using
fuzzy variables, Soil dynamics and Earthquake Engg., 16, 1997, 63-79.

25. Rouvray (1997) The treatment of uncertainty in the Phusical sciences, In Fuzzy Logic
in Chemistry edt. by Rouvray (1997), pp1-30,Academic Press, London, 1997 - 364 pages

26. Savoia, M., Structural reliability analysis through fuzzy number approach, with
application to stability, computers and structures, 80, 2002, 1087-1102.

27. Sexsmith R. G.(1999). Probability-based safety analysis - value and drawbacks, Struct
Safety 21, 303-310.

28. Shiraishi N., Furuta H. (1983) Reliability analysis based on fuzzy probability, J. of
Engng. Mech. ASCE , 1983,109(6), 1445-1459.

29. Smith S. A., Krishnamurthy T. and Mason B. H., Optimized vertex method and hybrid
reliability, AIAA-2002-1465.
 30. Sophie Q. C., Comparing Probabilistic and fuzzy set approaches for design in presence
of uncertainty, phd thesis, Virginia Polytechnique Inst. and state University., 2000.

31. Yao J., Furuta H.(1986) Probabilistic treatment of fuzzy events in civil engineering,
Prob Eng Mech, 1(1), 58-61.

32. Zhenyu L., Chen Q. (2002) A new approach to fuzzy finite element analysis. Comp
Methods in Applied Mech.,191, 5113-5118.

1.10 Publication Based on the Present Work

Publ List
SRESA
Springer
IJSHUEm
IJNME
IJRQSE
Conf. BARC