Chapter 4

POSSIBILISTIC SAFETY ANALYSIS UNDER RANDOM

AND FUZZY PARAMETERS

4.1 General

The reliability analysis of hybrid uncertain system in probabilistic format is presented in

the previous chapter. The probabilistic approach may not be a wise choice when the non-

probabilistic uncertain parameters (fuzzy) are more than the number of input random

parameters to characterize a structural system. In such cases, one will prefer possibilistic

approach of safety evaluation for hybrid uncertain system. The present chapter deals with

possibilistic safety analysis of hybrid uncertain system. For complete presentation of the

proposed reliability analysis procedure, the fundamentals of possibilistic safety analysis

are discussed first, followed by the present approach of safety analysis. Finally, the

procedure is elucidated with those four numerical examples considered in the previous

chapter.

4.2 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 as discussed in chapter 3 section 3.3. The mfs are usually

obtained from the numerical data along with the expert knowledge and experience. The

fundamental concept of fuzzy set theory based structural reliability analysis is

summarized here.

The uncertain input variables (xi) described by the fuzzy mf μ(xi) is transformed to

fuzzy output response through α-discretization. Normally, the vertex method is used. The

optimum output values zj are described by the mf μ(zj). The uncertain failure functions

π(zj) is considered to be the possibility distribution function and describe the possibility

of failure in respect of j-th limit-sate condition which may be directly defined or

determined from the membership functions. For example, if the failure criterion of a

~
steel member is considered to be the attainment of yield stress fy at a point in the cross

section, the failure function π(σ) is determined from the membership function μ(fy) i.e.

()  sup (f y )

f y 

(4.1)

Finally, the failure possibility for j-th limit state is obtained by evaluating the

membership function μ(zj) of the fuzzy results and the uncertain failure condition π(zj)

with the aid of,

 f  sup min(  ( z j ), ( z j ))

zj Z j (4.2)

Possibility was also defined as a special case of the plausibility measure, which is

used in the evidence theory. According to this definition, if the body of evidence about a

set of events is nested, then the plausibility of each event reduces to possibility. On the

other hand, if the body of the evidence consists of singletons, then plausibility reduces to
probability. The properties of the possibility, when it is defined as a special case of

plausibility and the properties of Zadeh’s possibility are almost same. Based on this

interpretation of possibility, Dubious and Prade (1997) has shown that a normalized


possibility distribution π(ω) can be effectively represented by a fuzzy number Q from its

membership function Q () as,

()   Q () (4.3)

And it is shown that the fuzzy number can be used to select a class of probability measure

P for which possibility and necessity represent the upper and the lower bounds of the

probability of each event.

4.3 Possibilistic Reliability Analysis

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.

A convenient way to represent the fuzzy variables is the α-discretization method

of its possibility distribution described by the mf as described in chapter 3 section 3.3.

The possibilistic approach of safety evaluation adopted in the present study is based on

the development proposed by Cremano and Gao (1997). The details of the forward and

inverse of the transformation and complete examination of the theory may be seen in

Cremano and Gao. The essential concept is briefly described here in order to outline the

procedure of possibilistic safety evaluation used here

For a safety margin M which introduces n uncertain variables (Xi)1in:

M=g(X1,………….Xn) (4.4)

Where g(X1,………….Xn) is the limit state related to the observed phenomenon. The

possibilistic approach for assessing component reliability will try to evaluate safety in

terms of possibility of failure

 f    M  0 . (4.5)

Let us consider a limit state M=g(R,S) composed of two variables R and S modeled as

non-interactive symmetrical Gaussian fuzzy numbers:

R   mR ,  R ,  R  L* R*
(4.6)
S   mS ,  S ,  S  L* R *

The membership function  M  . is easily built by means of the results from fuzzy

arithmetic.

  m  mS   x 
 M  x   L*  R  if mR  mS
 R  S 
(4.7)
 x   mR  mS 
*

R   if mR  mS
 R  S 

That directly gives the possibility of failure:

  m  mS  
 f  sup(  M  x  )  L*  R  if mR  mS
x 0  R  S  (4.8)
 1 if mR  mS

Since, L  x   L   x  , equation 4.7 can be written as follows:

* *

  m  mS  
 f  sup(  M  x  )  L*   R  if mR  mS
x 0  R  S  (4.9)
 1 if mR  mS
 ( mR  mS )
This calculus highlights a particular term,   ( R   S ) which is known as

possibilistic reliability index. Let us transform the two symmetrical Gaussian fuzzy

numbers into normalized Gaussian fuzzy numbers R0 and S0:

R0   0,1,1 L*R*
(4.10)
S0   0,1,1 L* R*

Then it becomes

R  mR   R R0
(4.11)
S  mS   S S0

The limit state g(R,S) = 0 keeps its linearity in the new space composed of R0 and S0:

gU  R0 , S0   mR  mS   R R0   S S 0  0. (4.12)

* *
By using the extension principle and the concept of crisp sets, two values R0 and S0

verifying

 M  0    S0 ( S0* )   R0 ( R0* ) (4.13)

exist. It then follows that

L* ( R0* )  R* ( S0* )  R0*   S 0*

(4.14)
and  f  L*  R0*  ,

Leading to possibilistic reliability index

R0*   
 mR  mS  .
(4.15)
 R S 
The possibilistic reliability index and the possibilistic design point are the

solutions of a minimization problem expressed as follows:

  min U  
according to
(4.16)
gU  U  0
Ui  U j , i, j.

The equality between absolute values comes from the fact that each U i has a degree

of possibility equal to the failure possibility. Since these fuzzy variables are non-

interactive and have identical distributions of possibility, then the corresponding values to

the degree of possibility f have identical absolute values.

4.4 The Possibilistic Safety Assessment under Hybrid Uncertainties

The reliability analysis of structures characterized by both the probabilistic and the

possibilistic variables by probabilistic approach is presented in the previous chapter. The

present chapter deals with safety analysis of such hybrid uncertain system in possibilistic

format.

As already discussed, the limit state function of the related reliability analysis

problem involves the probabilistic and also the possibilistic variables. And, to make the

analysis compatible with the reliability analysis in possibilistic format it is required to

express the performance function in terms of the fuzzy variables only. Thus, the random

variables need to be transformed to equivalent fuzzy variables for safety analysis under

mixed uncertain variables. The transformation condition should satisfy similar conditions

as discussed in chapter 3. As already discussed, various transformation approaches i.e.

the entropy based transformation, scaling of the mf, the least square curve fitting,

Housdrof distance equivalence etc. have been used for transforming random variable to

equivalent fuzzy variable or vice versa. The entropy based transformation can be directly

used for the transformation whereas other approaches the pdf is first scaled to
probabilistic fuzzy set of same shape as the original pdf of the random variables. The

entropy based transformation has been successfully used for probabilistic reliability

analysis. It is realized 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. The principle allows one to use

all the available information without unwittingly adding information that is not contained

in the evidence. 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.

Similar to above, to perform the transformation from random variables to fuzzy

variables, all non-normal variables are first transformed to equivalent normal variables.

Subsequently, the support of the fuzzy variables can be obtained by using Eqn. (3.20). As

earlier, the core value is assumed to be same as the mean of the equivalent normal

variable. The application is not restricted to symmetric triangular fuzzy distribution only.

Non symmetric triangular or trapezoidal fuzzy distribution property can also be readily

obtained by using the associated expression of the fuzzy entropy corresponds to the

desired mf of the fuzzy variables. Now, Eqn. (3.19) and (3.23) can be used to obtain the

support width of the transform fuzzy variable i.e. (b-a) in terms of standard deviation of

the normal random variable. Finally, the possibility of failure is obtained by using the

procedure described in section 4.3.

4.5 Numerical Study

The possibilistic safety evaluation of structures characterized by hybrid uncertain

parameters as presented in the previous section are elucidated through four numerical
examples which were considered in the previous chapter. 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.

4.5.1 Example One: A Reinforced Concrete Beam

The simply supported beam as detailed previously in section 3.5.1 has been now

considered to study the proposed safety evaluation procedure of hybrid uncertain system.

The numerical data of the problem remain same, i.e. the properties of random and fuzzy

variables are as depicted in table 3.1 (ref. chapter 3).

The possibility of failure is computed following section 4.3 by transferring the

random variables to equivalent fuzzy variables following the basic principle as described

in section 3.4 of chapter 3. The results are presented in Fig. 4.1 with increasing nominal

value of the concentrated load. The fuzzy variables are considered to be symmetric

triangular mf of different support width (as shown in Fig. 3.6 of chapter 3). The COV of

all the uncertain parameters are taken as 0.2 to develop this plot.
 1.2

1.0

P o ssib ility o f F ailu re

0.8

0.6
Support Widths

0.4 


0.2


0.0
0 50 100 150 200 250 300
Load (KN) Fig. 4.1:

The

variation of possibility of failure with increasing nominal value of the concentrated load

for symmetric triangular fuzzy variables for different support width

Keeping support width of symmetric triangular fuzzy variable to be same i.e.

w=3, the possibility of failure is computed for different COV (representing more

uncertainty). The possibility of failure with increasing value of the nominal load is shown

in Fig. 4.2 for different COV.

1.0
P o ssib ility o f F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
0 50 100 150 200 250
Load (KN)
 Fig. 4.2: The variation of possibility of failure with increasing nominal value of the
concentrated load for symmetric triangular fuzzy variables for different COV
The possibility of failure is also computed considering unsymmetrical triangular

mfs of the fuzzy variables (as shown in Fig. 3.11 of chapter 3). Note that for i-th such

fuzzy variable, xL  a  xi  w1 xi , c  xi and xR  b  xi  w2 xi . The left and right

end supports are now defined by two unequal parameters w1 and w2 . The variation of

possibilities of failures for different support widths are shown in Fig. 4.3. The COV of all

the uncertain parameters are assumed to be 0.2 to develop these results.

1.2

1.0
P o ssib ility o f F ailu re

0.8

0.6
Support Widths

0.4 


0.2


0.0
0 50 100 150 200 250 300
Load (KN)

Fig. 4.3: The variation of possibility of failure with increasing nominal value of the
concentrated load for unsymmetrical triangular fuzzy variables for different support
width
Now with fixed support width of the unsymmetrical fuzzy variables i.e. for

w1  1.0 and w2  2.0 , the possibilities of failures for different COV with increasing

nominal load are shown in Fig. 4.4

 1.0

P o ssib ility o f F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
0 50 100 150 200 250
Load (KN)

Fig. 4.4: The variation of possibility of failure with increasing nominal value of the
concentrated load for unsymmetrical triangular fuzzy variables for different COV.

4.5.2 Example Two: The Ten Bar Truss

The ten bar truss as shown Fig. 3.16 of chapter 3 (example 2 of section 3.5.2) is now

considered again as the second example to study the posiibilistic safety analysis. The

properties of the random and fuzzy variables remains same as summarized in table 3.2 of

chapter 3.

The possibility of failure results for symmetric triangular fuzzy mf (Ref. Fig. 3.6

of chapter 3) of different support width (w) are presented in Fig. 4.5 with increasing

nominal value of the vertical load. The COV of all the uncertain parameters are taken to

be 0.1 to develop this plot.

 1.2

1.0

P o ssib ility o f F ailu re

0.8

0.6
Support Widths
0.4 


0.2 

0.0
200 300 400 500 600 700 800
Load (KN)

Fig. 4.5: The variation of possibility of failure with increasing nominal value of the
vertical load for symmetric triangular fuzzy variables for different support width

The variation of possibility of failure with increasing nominal value of the vertical load

keeping same support width i.e. w=3.0 for different COV is shown in Fig. 4.6.

1.0
P ossib ility of F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
200 300 400 500 600 700 800
Load (KN)

Fig. 4.6: The variation of possibility of failure with increasing nominal value of the
vertical load for symmetric triangular fuzzy variables for different COV
 The possibility of failure is also computed considering unsymmetrical triangular

variation of the fuzzy variables (ref. Fig. 3.11). For an unsymmetrical triangular fuzzy

distribution with different support width (w1 and w2), the variation of possibility of failure

with increasing vertical load is shown in Fig. 4.7. The COV of all the uncertain

parameters are taken as 0.1 to develop this plot.

1.2

1.0
P o ssib ility o f F ailu re

0.8

0.6

Support Widths
0.4 


0.2

0.0
200 300 400 500 600 700 800
Load (KN)

Fig. 4.7: The variation of possibility of failure with increasing nominal value of the
vertical load for unsymmetrical triangular fuzzy variables for different support width
The possibility of failure for different COV with increasing nominal value of

vertical load is further obtained for unsymmetrical fuzzy variable with support width of

w1  1.0 and w2  2.0 and is shown in Fig. 4.8.

 1.0

Possibility of Failure
0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
200 300 400 500 600 700 800
Load (KN)

Fig. 4.8: The variation of possibility of failure with increasing nominal value of the
vertical load for unsymmetrical triangular fuzzy variables for different COV.

4.5.3 Example Three: The Reinforced concrete pile embedded in clay

The RC pile embedded in clay (example 3 of section 3.5.3) is further considered for

numerical study of possibilistic reliability analysis under hybrid uncertainty. The

properties of random and fuzzy variables are summarized in table 3.3 of chapter 3.

Possibilistic Reliability Analysis for Maximum Horizontal Displacement at Tip

The possibility of failure of the pile considering the limit state function for maximum

horizontal displacement at tip as described by Eqn. (3.42) of chapter 3 is computed

following section 4.3 by transferring the random variables to equivalent fuzzy variables

following the basic principle as described in section 3.4 of chapter 3. The possibilities of

failures with respect to limit state function correspond to horizontal displacement criteria

are computed and presented in Fig. 4.9 for increasing nominal value of the horizontal

load. The probabilities of failures by transferring the fuzzy variables to equivalent

random variables by entropy based transformation and the upper and lower bound

probabilities of failures (following section 3.4) are also shown in the same plot for

comparative study and consistency of the proposed transformation based results. The

COV of all the uncertain parameters are taken as 0.2. The support width of the triangular

fuzzy variables described as a function of standard deviation (σ) is taken as 3σ i.e. w= 3.

It can be readily observed from the plots that the consistency principle i.e. the possibility

of failure is greater than the probability of failure is satisfied for most conservative

estimate of probability of failure bases on the evidence theory (i.e. the upper bound

probability of failure results as obtained following section 3.4). As expected, the

probability of failure results as obtained by transformation of fuzzy variables to

equivalent random variables are within the probability of failure bounds obtained based

on the evidence theory.

1.0

0.8
Prob./Poss. of failure

0.6

0.4

Possibility of failure
0.2
Probability of failure
Upper bound
0.0 Lower bound

40 60 80 100 120 140 160 180 200

Load (kN)

Fig. 4.9: The possibility and probability of failure with respect to displacement limit state
with increasing nominal value of horizontal load for symmetric triangular fuzzy variables
for support width w=3.0
Again the possibility of failure results for symmetric triangular fuzzy mf (Ref. Fig. 3.6 of

chapter 3) of different support width (w) are computed and presented in Fig. 4.10 with

increasing nominal value of the horizontal load. In Fig. 4.10, the COV is assumed to be

0.1.

1.2

1.0
P o ssib ility o f F ailu re

0.8

0.6 Support Widths



0.4


0.2 

0.0
20 40 60 80 100 120 140 160
Load (KN)

Fig. 4.10: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different support width
The variation of possibility of failure with increasing nominal value of horizontal

load for different COV of all the uncertain variables is shown in Fig. 4.11 considering

same support width of fuzzy variable i.e. w=3.0.

 1.0

P ossib ility of F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
40 60 80 100 120 140 160
Load (KN)

Fig. 4.11: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different COV

The possibility of failure is also computed considering unsymmetrical triangular

variation of the fuzzy variables. For an unsymmetrical triangular fuzzy distribution with

different support width (w1 & w2), the variation of possibility of failure with increasing

nominal value of horizontal load is shown in Fig. 4.12. The COV of all the uncertain

parameters are taken as 0.1 to develop this plot.

 1.2

1.0

P o ssib ility o f F ailu re

0.8

0.6

0.4
Support Widths

0.2 


0.0
20 40 60 80 100 120 140 160
Load (KN)

Fig. 4.12: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different support width

With unsymmetrical fuzzy variable having support width of w1  1.0 and w2  2.0 , the

possibility of failure with increasing nominal value of horizontal load for different COV

of all the uncertain variables is shown in Fig. 4.13.

1.0
P ossib ility of F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
20 40 60 80 100 120 140 160
Load (KN)
 Fig. 4.13: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different COV.

Possibilistic Reliability Analysis for Maximum Bending Moment

The possibility of failure of the pile considering the limit state function for maximum

bending moment as described by Eqn. (3.43) of chapter 3 is computed following section

4.3 by transferring the random variables to equivalent fuzzy variables following the basic

principle as described in section 3.4 of chapter 3. The possibilities of failures with respect

to limit state function correspond to maximum bending moment criteria are computed

and presented in Fig. 4.14 for increasing nominal value of the horizontal load. The COV

of all the uncertain parameters are taken as 0.1. The support width of the triangular fuzzy

variables is taken as 3σ i.e. w= 3. The observations of results remain same as earlier i.e.

the possibility of failure results considering displacement limit state condition.

1.0

0.8
Poss. / Prob. of failure

0.6

0.4

Possibility of failure
0.2 Probability of failure
Upper bound
Lower bound
0.0
30 60 90 120 150 180 210
Load(kN)

Fig. 4.14: The possibility and probability of failure with respect to moment limit state
with increasing nominal value of horizontal load for symmetric triangular fuzzy variables
for support width w=3.0
 The possibilistic reliability results are further presented in Fig. 4.15 with

increasing nominal value of the concentrated load for different support width of

symmetric triangular fuzzy variable. The COV is assumed to be 0.1 to develop this plot.

1.2

1.0
P o s s ib ility o f F a ilu re

0.8

0.6
Support Widths

0.4


0.2 


0.0
20 40 60 80 100 120 140
Load (KN)

Fig. 4.15: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different support width

The variation of possibility of failure with increasing nominal value of horizontal load for

different COV of all the uncertain variables considering same support width of symmetric

triangular fuzzy variable i.e. w=3.0 is shown in Fig. 4.16. As expected, the possibility of

failure increases with increase of COV (more uncertainty).

 1.0

P o ssib ility o f F ailu re

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
40 60 80 100 120 140
Load (KN)

Fig. 4.16: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variable for different COV

The possibility of failure is again computed considering unsymmetrical triangular

variation of the fuzzy variables as shown in Fig. 3.11 of chapter 3. For an unsymmetrical

triangular fuzzy distribution with different support width (w1 & w2), the variation of

possibility of failure with increasing nominal horizontal load is shown in Fig. 4.17. The

COV of all the uncertain parameters are taken as 0.1 to develop this figure.

1.2

1.0
Possibility of Failure

0.8

0.6

0.4
Support Widths

0.2


0.0
20 40 60 80 100 120 140
Load (KN)
 Fig. 4.17: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different support width

Keeping support width of unsymmetrical fuzzy variable to be same i.e. for

w1  1.0 and w2  2.0 , the possibility of failure for increasing nominal value of

horizontal load for different COV of all the uncertain variables is shown in Fig. 4.18.

1.0
Possibility of Failure

0.8

0.6

0.4

10% COV
0.2 15% COV
20% COV
25% COV
0.0
40 60 80 100 120 140
Load (KN)

Fig. 4.18: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different COV.

4.5.4 Example Four: A Two Dimensional Portal Frame

A two dimensional two bay and two storied portal frame (example 4 of section 3.5.4) is

reconsidered as fourth example for numerical study of possibilistic reliability analysis

under hybrid uncertainty. The properties of random and fuzzy variables are summarized

in table 3.4 of chapter 3.

Possibilistic Reliability Analysis for Maximum Horizontal Displacement at Node 9

The possibility of failure of the frame considering the limit state function for maximum

horizontal displacement at node 9 as described by Eqn. (3.46) of chapter 3 is computed

following section 4.3 by transferring the random variables to equivalent fuzzy variables

following the basic principle as described in section 3.4 of chapter 3.

The possibility of failure results for symmetric triangular fuzzy mf (Ref. Fig. 3.6 of

chapter 3) of different support width (w) are computed and presented in Fig. 4.19 with

increasing nominal value of the concentrated load. In Fig. 4.19 the COV is assumed to be

0.1.

1.2

1.0
P o s s ib ility o f F a ilu re

0.8

0.6

0.4 Support Widths

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

Fig. 4.19: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different support width

The variation of possibility of failure with increasing nominal value of horizontal

load for different COV of all the uncertain variables for symmetric triangular fuzzy is

shown in Fig. 4.20 considering same support width of fuzzy variable i.e. w=3.0.
 1.2

1.0

P o s s ib ility o f F a ilu re
0.8

0.6

0.4 10% COV

15% COV
20% COV
0.2
25% COV

0.0
20 30 40 50 60
COV (%)

Fig. 4.20: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different COV

The possibility of failure is also computed considering unsymmetrical triangular

variation of the fuzzy variables. For an unsymmetrical triangular fuzzy distribution with

different support width (w1 & w2), the variation of possibility of failure with increasing

nominal value of horizontal load is shown in Fig. 4.21. The COV of all the uncertain

parameters are taken as 0.1 to develop this plot.

 1.2

1.0

Possibility of Failure
0.8

0.6

0.4 Support Widths


0.2 


0.0
20 30 40 50 60
Load (KN)

Fig. 4.21: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different support width

With unsymmetrical fuzzy variable having support width of w1  1.0 and w2  2.0 , the

possibility of failure with increasing nominal value of horizontal load for different COV

of all the uncertain variables is shown in Fig. 4.22.

1.2

1.0
P o s s ib ility o f F a ilu re

0.8

0.6

0.4 10% COV

15% COV
20% COV
0.2
25% COV

0.0
20 30 40 50 60
COV (%)
 Fig. 4.22: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different COV.

Possibilistic Reliability Analysis for Maximum Bending Moment at Node 3

The possibility of failure of the frame considering the limit state function for maximum

bending moment as described by Eqn. (3.47) of chapter 3 is computed following section

4.3 by transferring the random variables to equivalent fuzzy variables following the basic

principle as described in section 3.4 of chapter 3.

The possibilistic reliability results are presented in Fig. 4.23 with increasing nominal

value of the concentrated load for different support width. The COV is assumed to be 0.1

to develop this plot.

1.2

1.0
P o s s ib ility o f F a ilu re

0.8

0.6

Support Widths
0.4 

0.2 


0.0
20 30 40 50 60
Load (KN)

Fig. 4.23: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variables for different support width

The variation of possibility of failure with increasing nominal value of horizontal load for

different COV of all the uncertain variables considering same support width of symmetric
triangular fuzzy variable i.e. w=3.0 is shown in Fig. 4.24. As expected, the possibility of

failure increases with increase of load, support width as well as COV.

1.2

1.0
P o s s ib ility o f F a ilu re

0.8

0.6

0.4 10% COV

15% COV
20% COV
0.2 25% COV

0.0
20 30 40 50 60
COV (%)

Fig. 4.24: The variation of possibility of failure with increasing nominal value of the
horizontal load for symmetric triangular fuzzy variable for different COV

The possibility of failure is again computed considering unsymmetrical triangular

variation of the fuzzy variables as shown in Fig. 3.11 of chapter 3. For an unsymmetrical

triangular fuzzy distribution with different support width (w1 & w2), the variation of

possibility of failure with increasing nominal horizontal load is shown in Fig. 4.25. The

COV of all the uncertain parameters are taken as 0.1 to develop this figure.
 1.2

1.0

P o s s ib ility o f F a ilu re
0.8

0.6

0.4 Support Widths



0.2


0.0
20 30 40 50 60
Load (KN)

Fig. 4.25: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different support width

Keeping support width of unsymmetrical fuzzy variable to be same i.e. for

w1  1.0 and w2  2.0 , the possibility of failure with increasing nominal value of

horizontal load for different COV of all the uncertain variables is shown in Fig. 4.26.

1.2

1.0
Possibility of Failure

0.8

0.6

0.4 10% COV

15% COV
0.2
20% COV
25% COV

0.0
20 30 40 50 60
COV (%)
 Fig. 4.26: The variation of possibility of failure with increasing nominal value of the
horizontal load for unsymmetrical triangular fuzzy variables for different COV.

4.6 Summary and Observations

In the probabilistic safety analysis approach, all the possibilistic variables are transformed

to equivalent probabilistic variables to deal with hybrid uncertain system. However, the

main premises of the probabilistic repetitiveness of the collected data and large number

of data are violated. The approach tries to use information which does not have. On the

contrary, the possibilistic approach transforms the probabilistic information to equivalent

possibilistic information to perform a conservative safety assessment of hybrid uncertain

system. Obviously, such approach losses some of the essential part of the available

information. It is not very clear in open literature, which is more justified and it is felt to

need more study on this aspect. In the present study, the possibilistic safety analysis of

system characterized by hybrid uncertain parameters is studied. A general framework of

safety analysis of structures is presented in which the non-probabilistic parameters have

been treated as fuzzy variables. The probabilistic parameters are transformed to

equivalent fuzzy variables using the entropy based transformation concept. The

possibilities of failures obtained by the proposed transformation approach satisfy the

consistency principles. In fact, the possibility of failure is greater than the most

conservative upper bound estimate of probability of failure obtained based on the

evidence theory. The probability of failure obtained by the inverse transformation based

on the same entropy principle further shows that the probability of failures so obtained

are within the upper and lower bound solutions. It may be noted that the entropy

transformation directly gives equivalent normally distributed parameters without

additional requirement of transformation. The present study is restricted to transformation

of normal random variables to symmetrical triangular fuzzy variables. However,

transformation of random variables to unsymmetrical triangular or trapezoidal fuzzy

variables needs further study.