North-Holland
considered to be random (most of the load- three approaches can be used for this purpose.
related parameters, and some of the resistance- In the first approach, the problem can be
related parameters), some are considered to be considerably simplified if the fuzzy imprecision
deterministic (geometrical sizes of the struc- can be transformed to random uncertainty in a
tures), and some are neither deterministic nor suitable format that will facilitate the safety
random, but fuzzy (linguistic description of the evaluation process. The justification for this
structure reflecting the amount of degradation it approach is the most reliability formats are
has undergone). The class of problems where probability-based, the area is well-developed and
both random and fuzzy variables are present has it is easier to interpret probabilistic results. In
received very little attention from the profession this study, it is proposed that the transformation
so far. The objective of this paper is the of fuzzy imprecision to random uncertainty can
incorporation of fuzzy information to subjec- be made using the concept of entropy.
tively consider the present condition of the Conceptually, random uncertainty can be
structure and the evaluation of the correspond- transformed to fuzzy imprecision, and for
ing reliability of existing structures considering discussion purposes can be called the second
both the random and fuzzy variables. approach. However, as stated earlier, the area
needs further development before it can be
applied to practical problems. It will not be
2. Problem description discussed further in this paper.
In the third approach, a hybrid model can be
Evaluation of an existing structure is quite used to estimate reliability in the random-fuzzy
different than the process used when it was domain. Both approaches are discussed in the
initially analyzed, designed and built. The following sections.
mathematical model initally used to represent
the structure may not be applicable to the aged
structure. The exposure of the structure to the 3. Proposed method- entropy approach
real environment may have altered many of the
initial assumptions and structural parameters. The success of this approach depends on the
From a probabilistic point of view, it can be transformation of fuzzy imprecision to random
viewed as introducing another source of uncertainty and then formulating the problem in
uncertainty into the problem and must be a routine manner as is done in any conventional
accounted for accordingly. The information on reliability model by treating all the variables as
physical condition, e.g., the presence of an random [1,2, 11]. Thus, it is essential in this
excessive number of cracks or loss of material approach to transform fuzzy imprecision to
due to corrosion, may originate from visual random uncertainty. The concept of entropy is
inspection of the structure by experts. The used for this purpose.
experts' opinion could be linguistic in nature and Entropy is a measure of the uncertainty of a
must be incorporated in any risk evaluation random variable. Entropy is also defined for
procedure. Here, the parameter values obtained fuzzy variables as a measure of imprecision.
from a non-destructive evaluation procedure are Entropy can be thus looked upon as a measure
fuzzified with the subjective information on the of uncertitude. The premise of this transforma-
condition of the structure using expert opinions. tion is that this measure is invariant under
The parameters thus rendered fuzzy are used to transformation. It is therefore considered ap-
estimate the safety of the structure. The propriate to work with entropies to establish the
discussion of non-destructive evaluation schemes equivalence of fuzzy and random variables. This
is beyond the scope of this paper. concept has been used before [4, 8, 9] to update
It can be stated at this stage that in the the parameters of a random variable. Here an
evaluation of existing structures, some of the equivalent random variable is defined which has
uncertain parameters are random and others are an appropriate measure of uncertainty equiv-
fuzzy. Safety analysis in the fuzzy-random alent to the level of imprecision of a fuzzy
domain is expected to be complicated. At least variable. According to Shannon [12], the
A. Haldar, R.K. Reddy / Random-fuzzy analysis of structures 203
for the beam under consideration is then general cases where it is not possible to separate
fuzzy and random variables and the limit state
fBH2] (24) function is not available in close form.
pf=P(o>tof)=P(Q>t 6L /"
The method discussed here will now be
Here, of, B, H and L are all fuzzy variables. elaborated further with the help of an example.
Defining Z as
Numerical example
ofBH 2 A cantilever beam shown in Figure 2 is
Z =- - (25)
6L ' considered here. All the geometrical structural
parameters are shown in the figure. Initially,
(24) can be rewritten as
they are assumed to be deterministic. The stress
pf = P(Q >iZ). (26) at failure, of, is considered to be fuzzy and is
evaluated to be 50 ksi. Using the first approach,
Here Z is a fuzzy number and its membership
this value of of is considered to be the mean, and
function can be obtained from those of of, B, H
the equivalent standard deviation needs to be
and L using Zadeh's extension principle [10].
calculated using the fuzzy information, as
The interval of confidence Z((r) can be
discussed earlier. The fuzzy information must be
expressed as
collected on a case-by-case basis. However, for
Z(0() = [Zl(0t'), Zz(t~')] (27) illustrative purposes only, and to validate the
proposed algorithm, several arbitrary fuzzy
where
membership functions are assumed. Both discr-
of1( ol)B1( ol)H21(o,+) ete and continuous fuzzy variables are con-
Zl(~) - (28a)
sidered. Equivalent normal variable parameters
6L2(te)
are calculated for each case using the proposed
and algorithm. For comparison, Zadeh's [15] equiv-
alent random variable statistics are also deter-
Z~(~) Or2(ol)6L,(tr)
-
Bz( a) HZ~(a ) (28b)
mined. Zadeh's mean is the geometric mean of
the membership functions. The variance is the
A close-form expression for the membership second moment of the membership functions
function of Z involves the inversion of (28), about the mean.
which is extremely complicated. However, it can At first, of is considered to be a discrete fuzzy
be constructed numerically by varying a~ in [0, 1] variable with the following membership
for each one of the fuzzy variables; using (28), functions.
the corresponding tr-level values for Z can be Case 1:
obtained. This procedure is valid since the
2010.3, 25 10.5,3510.6, 40t0.8, 4510.9,
inversion of Zx(a0 and Z2(a') yields unique
results in this problem. It can be easily proved 5011.0, 55 10.9, 6010.8, 6510.6, 7510.5, 8010.3.
that this procedure would be valid in all cases if Case 2:
all the fuzzy variables are convex [10]. Once the
membership function of Z is available, the 1010.3, 20 10.4,2510.5,3010.7,3510.8,
probability of failure can be computed as a fuzzy 4010.9, 45 10.95, 5011.0, 5510.95, 6010.9,
number by the o~-level scheme described earlier
6510.8, 70 10.7, 7510.5,8010.4, 9010.3.
for Case 1.
The method suggested for multiple fuzzy
variables appears to be very simple. However, it
must be emphasized that the proposed algorithm
is applicable only when the limit state function is
I Q
Symmetric
1 0 50 100 50 19.95 50 2tl.41
2 10 50 90 50 15.96 50 16.33
3 20 50 80 50 11.97 50 12.25
4 30 50 70 50 7.98 50 8.17
5 40 50 61/ 511 3.99 50 4.08
Unsymmetric
1 0 50 80 50 15.96 43.3 16.50
2 2(I 50 65 50 8.98 45.0 9.35
3 30 50 55 50 5.00 45.0 5.40
4 35 50 90 50 10.97 58.3 11.61
5 45 50 811 50 6.98 58.3 7.73
208 A. Haldar, R.K. Reddy / Random-fuzzy analysis of structures
different values for nonsymmetrical membership approaches proposed here, i.e., the entropy-
functions according to the method suggested by based and hybrid, to estimate the reliability of
Zadeh. Since the mean values change according an existing structure. The cantilever structure
to Zadeh's method corresponding to the shown in Figure 2 is again considered here. All
geometric mean of the TFN, at least for the the geometrical structural properties are shown
nonsymmetrical cases, the corresponding stand- in the figure, and initially they are assumed to be
ard deviations are expected to be smaller. deterministic, except that the stress at failure, of,
However, that is not true in this case. This is considered to be a TFN. For the cantilever
observation points out the desirability of the beam considered here, the maximum stress
proposed algorithm, at least for continuous occurs at the fixed end and is given by
cases. 6QL
(7=--
From this discussion, it is established that the
BH 2
entropy approach is a viable alternative, i.e., the
fuzzy imprecision can be transformed to random The load is considered to be a random variable
uncertainty, and then the reliability of an represented by a normal distribution with a
existing structure can be estimated using the mean of 1.0 and a standard deviation of 0.2. The
conventional probability approach. probability of failure of the cantilever, i.e,
The stage is now set to compare the two a > ~f, is calculated using both methods for all 5
0.9
0.8
0.7
0.6
0.5
E
0.4
0.3
0.2
0.1
0
lO-IO 10-9 104 10-7 10~ 10-5 10-4 10-~ 10-2 lO-I 10o
probability of failure
Fig. 4. Membership function of pf.
A. Haldar, R.K. Reddy / Random-fuzzy analysis of structures 209
Table 4. Probabilities of failure for the hybrid approach, multiple fuzzy variable case
cases of Table 2 and the results are summarized variables are considered to be deterministic. By
in Table 3. changing the width of both the TFN's, the
For the hybrid approach, the fuzzy probability probabilities of failure are estimated from
of failure at various a'-ievels is calculated and is several cases using the proposed hybrid ap-
shown in Figure 4 for all five cases. The crisp proach. The results are summarized in Table 4
measure of probability for each case is also for some representative cases.
estimated using Figure 4 and is shown in column When the width of the TFN representing the
5 of Table 3. beam depth is zero, the problem reduces to a
Several important observations can be made case with one fuzzy variable. The probabilities of
from Table 3 and Figure 4. From Table 3, failure are found to be identical to the values
columns 2,4 and 5, it can be seen that as the given in Table 3. However, for a given width of
width of the TFN of the stress at failure the TFN of the yield stress, the probability of
increases, the corresponding probability of failure increases with an increase in the width of
failure for both methods also increases. This is the TFN representing the beam depth. The
expected since the increase in the width of the increase in the width of the TFN indicates larger
TFN indicates an increase in the uncertainty uncertainty in the estimation of the beam depth.
level of the fuzzy variable. With increased With larger uncertainty, the increase in the
uncertainty, a higher probability of failure is probability of failure is expected. This example
expected. demonstrates the applicability and robustness of
The probabilities of failure estimated by the the proposed algorithm.
two methods are shown in columns 4 and 5 of
Table 3. They are not expected to be identical.
The estimation of the crisp measure of Acknowledgment
probability for a fuzzy variable depends on the
definition used. However, the similarities of This paper is based upon work partly
these numbers are very encouraging and show supported by the National Science Foundation
that it is possible to reliably estimate risk in a under Grant No. MSM-8896267. Any opinions,
random-fuzzy environment using either of the findings and conclusions or recommendations
two methods suggested here. Further investi- expressed in this publication are those of the
gation is necessary in this area. authors and do not necessarily reflect the views
The hybrid approach is also applied to of the sponsor.
problems where multiple fuzzy variables are
present. For illustrative purposes only, the
cantilever beam discussed earlier can again be References
considered. However, in this case, in addition to
the yield stress, the depth of the beam is also [1] A.H.-S. Ang and W.H. Tang, Probability Concepts in
considered to be fuzzy. Both fuzzy variables are Engineering Planning and Design, Vol. I and II (John
represented by symmetric TFN's with modal Wiley & Sons, New York, 1975 and 1984).
[2] B.M. Ayyub and A. Haldar, Practical structural
values of 50 and 2, respectively. The load is reliability techniques, ASCE J. Structural Eng. U0 (8)
considered to be random as before and all other (1984) 1707-1724.
210 A. Haldar, R.K. Reddy / Random-fuzzy analysis of structures
[3] A. Bardossy and I. Bogardi, Fuzzy fatigue life of aging stochastic systems, ASCE J. Engng. Mechanics
production, Structural Safety 6 (1989) 25-38. 110 (5) (1984) 743-751.
[4[ C.B. Brown, Entropy constructed probabilities, ASCE [10] A. Kaufmann and M.M. Gupta, Introduction to Fuzzy
J. Engng. Mechanics 106 (4) (1980) 633-639. Arithmetic, Theory and Application (Van Nostrand
[5[ A. DeLuca and S. Termini, A definition of nonprob- Reinhold, New York, 1985).
abilistic entropy in the setting of fuzzy sets theory, [11] H.O. Madsen, S. Krenk and N.C. Lind, Methods of
Inform. and Control 20 (1972) 301-312. Structural Safety (Prentice-Hall, Englewood Cliffs, NJ,
[6] A. DeLuca and S. Termini, Entropy of L-fuzzy sets, 1986).
Inform. and Control 24 (1974) 55-73. [12] C.E. Shannon, The Mathematical Theory of Com-
[7[ L. Duckstein, I. Bogardi and A. Bardossy, A fuzzy munication (University of Illinois Press, Urbana, IL,
reliability model of water supply during droughts, Fall 1949).
National Meeting of the American Geophysical Union, [13] TriUes, E. and T. Rierra, Entropies in finite fuzzy sets,
Hydrology Session, 1988. Inform. Sci. 2 (1978) 159-168.
[8] T.-Y. Kam and C.B. Brown, Updating parameters with [14] J.T.P. Yao, Damage assessment of existing structures,
fuzzy entropies, ASCE J. Engng. Mechanics 198 (6) ASCE J. Engng. Mechanics 106 (4) (1980) 785-799.
(1983) 1334-1343. [15] L. Zadeh, Probability measures of fuzzy events, J.
[9] T.-Y. Kam and C.B. Brown, Subjective modifications Math. Anal. Appl. 23 (1968) 421-427.