Journal of Fuzzy Set Valued Analysis 2013 (2013) 1-8

Available online at www.ispacs.com/jfsva

Volume 2013, Year 2013 Article ID jfsva-00153, 8 Pages
doi:10.5899/2013/jfsva-00153
Research Article

Some applications of random fuzzy alternating renewal

processes based on fuzzy simulation

Behrouz Fathi Vajargah1∗ , Sara Ghasemalipour1

(1) Department of Statistics, Faculty of Mathematical Sciences, University of Guilan, Iran

Copyright 2013 ⃝
c Behrouz Fathi Vajargah and Sara Ghasemalipour. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract
This paper studies an alternating renewal process when on-off times are random fuzzy variables. We provide a the-
orem concerning the long-term average expected service times of customers. Here, we do not use the α0 -pessimistic
value and the α0 -optimistic value of fuzzy variables, we simulate the expected value of random fuzzy variables based
on fuzzy simulation to calculate the long-term average expected service times of customers per unit time. Some ex-
amples is provided to illustrate the results.

Keywords: Fuzzy alternating renewal processes, Random fuzzy variables, Fuzzy simulation, The long-term average expected ser-
vice times.

1 Introduction

The alternating renewal process is one of the renewal processes in renewal theory. It is widely used in inventory
systems and queueing systems. In classical alternating renewal processes, one of the assumptions is that the on times
and off times are independent random variables. The notion of a fuzzy set was introduced by Zadeh [1] who defined
it as a generalized characteristic function. Recently, fuzzy set theory has been widely developed and used to renewal
processes. Popova and Wu [2] studied renewal reward processes with random fuzzy interarrival times and rewards.
The concept of random fuzzy variables provided by Kwakernaak [3],[4] and Puri and Ralescu [5] is a particular fuzzy
set.
Using random fuzzy theory introduced by Liu and Liu [6], Shen [7] established an alternating renewal process with
two states: on times and off times. They provided a theorem on the limit value of the average chance of a given
random fuzzy event in terms of system being on at time t by letting on times and off times be random fuzzy variables.
As another extension of classical alternating renewal processes, an alternating renewal processes only with fuzziness
is established in this paper by characterizing the on times and off times as unrelated convex fuzzy variables.
In section 2, we introduce some basic definition about possibility measure and random fuzzy variables. In section 3,
we discussed about fuzzy alternating renewal process and some theorem is proved. In section 4 we illustrate the fuzzy
simulation and in section 5, some examples is described.

∗ Corresponding author. Email address: [email protected]

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 2 of 8

2 Definitions and preliminaries

Definition 2.1. [8] Let Θ be a nonempty set, and P(Θ) shows the power set of P(Θ). Let Pos be a set function defined
on the family of subsets of the P(Θ). The set function Pos is said to be a possibility measure if it satisfies the following
conditions:
1. Pos{Θ} = 1.
/ = 0.
2. Pos{0}
3. Pos{∪i Ai } = supi Pos{Ai } for any collection Ai in P(Θ).
4. Let Θi be nonempty set on which Posi {.} satisfies the first three axioms, i = 1, 2, ..., n, respectively, and Θ =
Θ1 × Θ2 × ... × Θn . Then

Pos{A} = sup Pos1 {θ1 } ∧ Pos2 {θ2 } ∧ ... ∧ Posn {θn }, (2.1)
(θ1 ,...,θn )∈A

for each A ∈ P(Θ).

The triplet (Θ, P(Θ), Pos) is called a possibility space.
Definition 2.2. Let (Θ, P(Θ), Pos) be a possibility space, and A be a set in P(Θ). Then the necessity measure of A is
defined by:
Nec{A} = 1 − Pos{Ac }. (2.2)
The details about the possibility measure and the necessity measure can be found in [9].
Definition 2.3. Let (Θ, P(Θ), Pos) be a possibility space, and A be a set in P(Θ). Then the credibility measure of A is
defined by:
1
Cr{A} = (Pos{A} + Nec{A}). (2.3)
2
Definition 2.4. A fuzzy variable ξ is defined as a function from Θ to the set of real numbers, and the possibility
distribution µξ of ξ is defined by:
µξ (r) = Pos{θ ∈ Θ|ξ (θ ) = r}. (2.4)
Definition 2.5. Let ξ be a fuzzy variable defined on the possibility space (Θ, P(Θ), Pos). Then the set

ξα (r) = {ξ (θ )|θ ∈ Θ, Posθ ≥ α }, (2.5)

is called α -level set of ξ .

Definition 2.6. Let ξ be a fuzzy variable defined on the possibility space (Θ, P(Θ), Pos), and α ∈ (0, 1]. Then

ξα′ = inf{r|Pos{ξ ≤ r} ≥ α }, ξα′′ = sup{r|Pos{ξ ≥ r} ≥ α }, (2.6)

are called α -pessimistic value and α -optimistic value of ξ , respectively.

Definition 2.7. (Liu and Liu [8]). Let ξ be a fuzzy variable on the possibility space (Θ, P(Θ), Pos). Then the expected
value E[ξ ] is defined by:
∫ ∞ ∫ 0
E[ξ ] = Cr{ξ ≥ r}dr − Cr{ξ ≤ r}dr, (2.7)
0 −∞
provided that at least one of the two integrals is finite.
In particular, if the fuzzy variable ξ is positive (i.e. Pos{ξ ≤ 0} = 0), then
∫ ∞
E[ξ ] = Cr{ξ ≥ r}dr. (2.8)
0

Proposition 2.1. Let ξ be a fuzzy variable defined on the possibility space (Θ, P(Θ), Pos). Then we have
∫ 1
1
E[ξ ] = [ξα′ + ξα′′ ]d α . (2.9)
2 0

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 3 of 8

Proof. If ξ is normalized, i.e., there exists a real number r0 such that µξ (r0 ) = 1 and if r0 > 0, then the equation (2.7)
can be rewritten as ∫ +∞ ∫ r0
1
E[ξ ] = [r0 + Pos(ξ ≥ r)dr + r0 − Pos(ξ ≤ r)dr] (2.10)
2 r0 −∞
∫ 1
1
= (ξα′ + ξα′′ )d α , (2.11)
2 0
The same result can be obtained when r0 > 0.

Definition 2.8. A random fuzzy variable is a function from the possibility space (Θ, P(Θ), Pos) to the set of random
variables.

Definition 2.9. Let ξ be a random fuzzy variable on the possibility space (Θ, P(Θ), Pos). Then the average chance of
random fuzzy event ξ ≤ 0 is defined as
∫ 1
Ch{ξ ≤ 0} = Cr{θ ∈ Θ|Pr{ξ (θ ) ≤ 0} ≥ p}d p. (2.12)
0

3 Random fuzzy alternating renewal processes

Consider a system that can be in one of two operating states: on or off. Initially, it is on and it remains on for a
time ξ1 ; it then goes off and remains off for a time η1 ; it then goes on for a time ξ2 ; then off for a time η2 , and so
forth. Without losing generality, suppose that ξi , i = 1, 2, ... are random fuzzy variables defined on possibility space
(Θi , P(Θ), Posi ) and ηi are random fuzzy variables defined on possibility space (Γi , P(Γ), Pos′i ). If the random fuzzy
vectors (ξi , ηi ), i = 1, 2, ... are iid, then the process depicting by the sequence {(ξi , ηi , i ≥ 1)} is called a random fuzzy
alternating renewal process, defined on the possibility space (Θi , P(Θ), Pos), where (Θi , P(Θ), Pos) is the infinite
product possibility space characterized by
Θ = Π∞ i=1 (Θi , Γi ) (3.13)
and
Pos{(θ1 , γ1 ), (θ2 , γ2 ), ...} = Pos1 {θ1 } ∧ Pos′1 {γ1 } ∧ Pos2 {θ2 } ∧ Pos′2 {γ2 } ∧ .... (3.14)
For any θ = ((θ1 , γ1 ), (θ2 , γ2 ), ...) ∈ Θ. For any fixed θ ∈ Θ, ξi (θ ) and ηi (θ ) are independent random variables, also
E[ξi (θ )] and E[ηi (θ )] are just the expected values of ξi (θ ) and ηi (θ ). However, when θ is varied all over in Θ,
E[ξi (θ )] and E[ηi (θ )], as functions of θ , are fuzzy variables and their α -pessimistic values and α -optimistic values
can be written as
E[ξi (θ )]′α = inf{r|PosE[ξi (θ )] ≤ r ≥ α },
E[ξi (θ )]′′α = sup{r|PosE[ξi (θ )] ≥ r ≥ α },
E[ηi (θ )]′α = inf{r|PosE[ηi (θ )] ≤ r ≥ α },
E[ηi (θ )]′′α = sup{r|PosE[ηi (θ )] ≥ r ≥ α },

Definition 3.1. A positive random variable ξ is said to be lattice if and only if there exists d ≥ 0 such that
∞
∑ Pr{ξ = nd} = 1. (3.15)
n=0

Theorem 3.1. Let {(ξi , ηi , i ≥ 1)} be a sequence of i.i.d positive random fuzzy vectors. Assume that the distribution
functions of ξi (θ ) and ηi (θ ), for any given θ ∈ Θ, are nonlattice, the α -pessimistic values and the α -optimistic values
of the fuzzy variables E[ξi (θ )] and E[ηi (θ )], i = 1, 2, ... are continuous at the point α0 , α0 ∈ [0, 1], and E[ξ1 + η1 ] < ∞,
then
E[ξ1 (θ )]′α0
lim Pα′ 0 (t) = , (3.16)
t→∞ E[ξ1 (θ )]′α0 + E[η1 (θ )]α′′ 0

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 4 of 8

and
E[ξ1 (θ )]′′α0
lim Pα′′0 (t) = , (3.17)
t→∞ E[ξ1 (θ )]′′α0 + E[η1 (θ )]α′ 0
where Pα′ 0 (t) and Pα′′0 (t) are the α0 -pessimistic values and the α0 -optimistic values

P(t) = Pr{system being on at time t}. (3.18)

Proof. Let Ai = {θi ∈ Θi |µ (θi ) ≥ α } and Bi = {ϑi ∈ Γi |µ (ϑi ) ≥ α }. Then for all θi ∈ Ai and ϑi ∈ Bi we have

E[ξi (θi1 )] ≤ E[ξi (θi )] ≤ E[ξi (θi2 )]

E[ηi (ϑi1 )] ≤ E[ηi (ϑi )] ≤ E[ηi (ϑi2 )]

Conditioning on the time of that last renewal prior to or at time t yields

P1 (t) = Pr{process A being on at time t|SA = 0}Pr{SA = 0}

∫ ∞
+ Pr{process A being on at time t|SA = y}dFSA (y)
0
and
P3 (t) = Pr{process C being on at time t|SC = 0}Pr{SC = 0}
∫ ∞
+ Pr{process C being on at time t|SC = y}dFSC (y),
0
where SA represents the time of the last renewal prior to or at time t in process A, SC the time of the last renewal prior
to or at time t in process C, FSA (y) the distribution function of SA and FSC (y) the distribution function of SC .
It is clear that
P1 (t) ≤ P3 (t).
Furthermore, using the result of stochastic alternating renewal processes the theorem is proved (see [7]).
Remark 3.1. If {(ξi , ηi , i ≥ 1)} degenerates to a sequence of i.i.d random vectors, then the results of (3.16) and (3.17)
in theorem 3.1 degenerate to the form
E[ξ1 ]
lim Pr{system being on at time t} = , (3.19)
t→∞ E[ξ1 ] + E[η1 ]
which is just the conventional result in stochastic case.
Remark 3.2. If {(ξi , ηi , i ≥ 1)} degenerates to a sequence of fuzzy vectors with the same membership function, then
for each α0 ∈ [0, 1], the α0 -pessimistic and α0 -optimistic values of ξ1 and η1 degenerate to four real numbers. The
results of (3.16) and (3.17) in Theorem 3.1 respectively degenerate to the form
ξα′ 0
lim Pα′ 0 = , (3.20)
t→∞ ξα′ 0 + ηα′′0

and
ξα′′0
lim Pα′′0 = , (3.21)
t→∞ ξα′′0 + ηα′ 0
We are concerned with Pon , the long-run proportion of time that the system is on. If we let

Xn = ξn + ηn , n ≥ 1 (3.22)

then at time X1 the process starts over again. That is, the process starts over again after a complete cycle consisting
of an on and an off interval. In other words, a renewal occurs whenever a cycle is completed. Therefore, write
Sn = ∑ni=1 Xi .

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 5 of 8

Theorem 3.2. Let {(ξi , ηi , i ≥ 1)} be a random fuzzy alternating renewal process. Assume that S(t) be the total on
time prior to or at time t. Then
E[S(t)] ξ1
lim = E[ ]. (3.23)
t→∞ t ξ1 + η1
Proof. It follows from Proposition 2.1 that
∫ 1
1
Ch{system being on at time t} = (Pα′ (t) + Pα′′ (t))d α , (3.24)
2 0

By Theorem 3.1,

E[ξ1 (θ )]′α
lim Pα′ (t) = , (3.25)
t→∞ E[ξ1 (θ )]′α + E[η1 (θ )]α′′
and
E[ξ1 (θ )]′′α
lim Pα′′ (t) = , (3.26)
t→∞ E[ξ1 (θ )]α′′ + E[η1 (θ )]′α
Furthermore, it follows from the definition of the limit that there exist two real numbers t1 and t2 with t1 > 0 and t2 > 0
such that for all t ≥ t1

E[ξ1 (θ )]′α
0 ≤ Pα′ (t) ≤ , (3.27)
E[ξ1 (θ )]′α + E[η1 (θ )]′′α
and for all t ≥ t2
E[ξ1 (θ )]′′α
0 ≤ Pα′′ (t) ≤ . (3.28)
E[ξ1 (θ )]′′α + E[η1 (θ )]′α
Therefore, for any t ≥ max(t1 ,t2 ),

E[ξ1 (θ )]′α E[ξ1 (θ )]′′α

0 ≤ Pα′ (t) + Pα′′ (t) ≤ 2 + + . (3.29)
E[ξ1 (θ )]′α + E[η1 (θ )]′′α E[ξ1 (θ )]′′α + E[η1 (θ )]′α

Then ∫ 1
ξ1 1 E[ξ1 (θ )]′α E[ξ1 (θ )]′′α
E[ ]= ( + )d α . (3.30)
ξ1 + η1 2 0 E[ξ1 (θ )]α + E[η1 (θ )]α E[ξ1 (θ )]′′α + E[η1 (θ )]′α
′ ′′

From Fatou’s lemma and (3.28), proof is completed.

4 Fuzzy simulation

In order to evaluate the expected value of a fuzzy variable, Liu and Liu [8] designed a fuzzy simulation for both
discrete and continuous cases.
(a) Discrete fuzzy vector: assume that f is a function, and ξ = (ξ1 , ..., ξm ) is discrete fuzzy vector whose joint possi-
bility distribution function is defined by 

 µ1 , u = u1

µ2 , u = u2
µξ (u) = (4.31)

 ...

µn , u = un
where µu = min1≤i≤m µ (i) (ui ) and u = (u1 , ..., um ) ∈ ℜm and µ (i) are the possibility distribution function of ξi for
i = 1, 2, ..., m.
Let ai = f (ui ). Without loss of generality, we assume that a1 ≤ a2 ≤ ... ≤ an ,then the expected value is given by
n
E[ f (ξ )] = ∑ ai pi , (4.32)
i=1

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 6 of 8

where
1 1 i
pi = (∨nj=i µ j − ∨n+1
j=i+1 µ j ) + (∨ j=1 µ j − ∨ j=0 µ j )
i−1
(4.33)
2 2
where (µ0 = µn+1 = 0) for i = 1, 2, ..., n.
(b) Continuous fuzzy vector: assume that ξ is a continuous fuzzy vector with a possibility distribution function µ . In
this case, we can estimate the expected value by the formula (4.32).

5 Application Examples
Now, for illustrating the method, we consider some examples. We present an application practical of this model
how using fuzzy simulation method to estimate the expected value.
Example 5.1. Let us consider a random fuzzy alternating renewal process that ξi and ηi are trapezoidal fuzzy vari-
ables, ξi = (1/2/3/4) and ηi = (3/4/5/6). From the Theorem 3.2 and the fuzzy simulation method that discussed in
section 4, we simulate the long-term average expected on times E[ S(t)
t ] with the various number of iterations and the
results are shown in Table 1 and Figure 1.

Table 1: Results obtained using fuzzy simulation method

Number o f iteratins 100 500 1000 5000 10000 50000

Expected on times 0.3557 0.3565 0.3568 0.3568 0.3569 0.3569

Figure 1: The solutions with different iterations

Table 1 and Figure 1 show that the mean of expected on times after 50000 times is equal 0.3569, in fact it remains at
0.3569, level.
Example 5.2. Consider a machine that can be in one of two operating states: on or off. Let the on times can be
modeled by i.i.d. positive random fuzzy variables with exponential distribution ξi ∼ exp(λi ) that λi = (2/4/6), i ≥ 1
in minutes. The off times can be modeled by i.i.d. positive random fuzzy variables with exponential distribution
ηi ∼ exp(µi ) that µi = (1/3/5), i ≥ 1 in minutes.
For calculating the long-run expected value of on times, it follows from Theorem 3.2 that limt→∞ E[S(t)]
t = E[ λ λ+1µ ].
1 1
We used fuzzy simulation method from section 4 and obtain the following results in Table 2 and Figure 2.
It is obvious, that the long-run average expected on times with fuzzy simulation is equal 0.2837 after 50000 times
iteration. In this method we obtain the real solution without using α -cuts and interarrival arithmetic. This is an
advantage of this method. Then the long-run average expected off times is equal 1-0.2837=0.7163.

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 7 of 8

Table 2: The Long-run average expected on times with fuzzy simulation

Number o f iteratins 500 1000 5000 10000 20000 30000 50000

Expected on times 0.2662 0.2750 0.2818 0.2824 0.2830 0.2837 0.2837

Figure 2: The long-term average expected on times with different iterations

6 Conclusion

In this paper studied an alternating renewal process when on-off times are random fuzzy variables. A theorem
about the long-run average expected on times is proved. We simulated the expected value of fuzzy random variable
to calculate the long-term average expected on times per unit time. We can improve the solution with increasing the
number of iterations or improveing the random numbers in simulation algorithm

References

[1] L. A. Zadeh, Fuzzy sets, Information and Control, 8 (1965) 338-353.

http://dx.doi.org/10.1016/S0019-9958(65)90241-X
[2] E. Popova, H. C. Wu, Renewal reward processes with fuzzy rewards and their applications to T-age replacement
policies, European Journal of Operational Research, 117 (1999) 606-617.
http://dx.doi.org/10.1016/S0377-2217(98)00247-1
[3] H. Kwakernaak, Fuzzy random variablesI. Definitions and theorems, Information Sciences, 15 (1978) 1-29.
http://dx.doi.org/10.1016/0020-0255(78)90019-1
[4] H. Kwakernaak, Fuzzy random variablesII. Algorithms and examples for the discrete case, Information Sciences,
17 (1979) 253-278.
http://dx.doi.org/10.1016/0020-0255(79)90020-3

[5] M. L. Puri, D. A. Ralescu, Fuzzy random variables, Journal of Mathematical Analysis and Applications, 114
(1986) 409-422.
http://dx.doi.org/10.1016/0022-247X(86)90093-4
[6] Y. K. Liu, B. D. Liu, Expected value operator of random fuzzy variable and random fuzzy expected value models,
International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 11 (2003) 195-215.
http://dx.doi.org/10.1142/S0218488503002016

International Scientific Publications and Consulting Services

Journal of Fuzzy Set Valued Analysis
http://www.ispacs.com/journals/jfsva/2013/jfsva-00153/ Page 8 of 8

[7] Q. Shen, R. Q. Zhao, W. S. Tang, Random fuzzy alternating renewal processes, Soft Computing, 13 (2009)
139-147.
http://dx.doi.org/10.1007/s00500-008-0307-y

[8] B. Liu, Y. Liu, Expected value of fuzzy variable and fuzzy expected value model, IEEE Transactions on Fuzzy
Systems, 10 (2002) 445-450.
http://dx.doi.org/10.1109/TFUZZ.2002.800692
[9] B. Liu, A survey of credibility theory, Fuzzy Optimization and Decision Making, 5 (4) (2006) 387-408.
http://dx.doi.org/10.1007/s10700-006-0016-x

International Scientific Publications and Consulting Services