Professional Documents
Culture Documents
1 s2.0 S0307904X11003751 Main
1 s2.0 S0307904X11003751 Main
a r t i c l e i n f o a b s t r a c t
Article history: In this paper, we investigate linear first-order fuzzy matrix differential dynamical systems
Received 1 December 2010 where the coefficients matrix is described by a fuzzy matrix. We show some properties of
Received in revised form 2 April 2011 the matrix differential dynamical systems, and their phase portraits are described by
Accepted 23 May 2011
means of examples.
Available online 30 May 2011
2011 Elsevier Inc. All rights reserved.
Keywords:
Fuzzy differential equations
Fuzzy dynamical systems
Fuzzy matrices
1. Introduction
The theoretical framework of fuzzy differential equations (FDE) has been on active research filed over the last few years
[1–15]. The concept of fuzzy derivative was first introduced by Chang and Zadeh [4]. It was followed up by Dubois and Prade
[16], who defined and used the extension principle. A comprehensive approach to FDEs has been the work of Seikkala [12],
and Kaleva [5,6], especially in its generalized form given by Buckley and Feuring [2]. Their work is important as it overcomes
the existence of multiple definitions of the derivative of fuzzy functions, see [16,17,7,9,12]. Also, Ref. [2] compares the var-
ious solutions to the fuzzy initial value problem that one may obtain using the different derivatives.
For further investigation, the fuzzy dynamical systems based on FDEs are also widely applied to fuzzy control systems
[18], bifurcations of fuzzy nonlinear dynamical systems [19], artificial system [20] and many other fields.
In this paper we use a complex number representation of the a-level sets of the fuzzy system following Pearson [21], who
investigates a property of linear differential equations where the initial state is described by a vector of fuzzy numbers. The
property is related directly to the matrix defining the original non-fuzzy system by passing to a complex number represen-
tation of the a-level sets of the fuzzy system which may avoid shortcomings mentioned above. Moreover, such approach
could be useful in concrete computations and in the future theoretical development of the topic. Then we present the solu-
tions of linear first-order fuzzy matrix differential dynamical systems as follows:
(
_
MðtÞ ¼ AMðtÞ;
ð1Þ
Mð0Þ ¼ M0 ;
where A is a square fuzzy matrix, M(t) is an unknown matrix function and M0 2 Rnm is a crisp known matrix.
We focus on discussing properties of the two-dimensional dynamical systems and describe their phase portraits with
examples.
The paper has been arranged as follows: in Section 2 we provides some preliminaries and notions, while, in Section 3, the
solutions of the fuzzy linear matrix differential dynamical systems are gained and proofed. The properties of the nonnegative
0307-904X/$ - see front matter 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2011.05.054
B. Ghazanfari et al. / Applied Mathematical Modelling 36 (2012) 348–356 349
matrix and nonhomogeneous linear matrix dynamical systems are also discussed in this section. The properties of two-
dimensional fuzzy matrix dynamical systems and their phase portraits are Shown in Section 4 by some examples.
2. Preliminaries
The basic definition of fuzzy numbers are given in [17,22], in this section, we review some of them.
1. u is upper semi-continuous,
2. u(x) = 0 outside some interval [c, d],
3. there are real numbers a, b : c 6 a 6 b 6 d for which
3.1 u(x) is monotonic increasing on [c, a],
3.2 u(x) is monotonic decreasing on [b, d],
3.3 u(x) = 1, a 6 x 6 b.
1. u(a) is a bounded left continuous monotonic increasing function over [0, 1],
ðaÞ is a bounded left continuous monotonic decreasing function over [0, 1],
2. u
3. uðaÞ 6 u ðaÞ; 0 6 a 6 1.
ðkuÞðxÞ ¼ uðx=kÞ; k – 0:
The collection of all the fuzzy numbers with addition and multiplication is denoted by E1 and is a convex cone. The fuzzy
numbers of Definition 2.2 as shown in [22], can be embedded into the Banach space B ¼ C½0; 1 C½0; 1 where the norm
is defined as
kðu; v Þk ¼ sup max juðaÞj; max jv ðaÞj : ð2Þ
06a61 06a61
is the distance between u and v. The function D(u, v) is a metric on E1. This metric is equivalent to the one used by Puri and
Ralescu [9] and Kaleva [5]. It is shown [10] that (E1, D) is a complete metric space.
Definition 2.5. f : R ! E1 is called a fuzzy function. If for arbitrary fixed t0 2 R and e > 0, a d > 0 such that
jt t0 j < d ) Dðf ðtÞ; f ðt0 ÞÞ < e
The concept of fuzzy differentiation was introduced by Dubois and Prade [16], Puri and Ralescu [9] proposed two ap-
proaches for finding the fuzzy derivative. The first, based on the H-difference notation regards E1 as a universe. The second
approach suggested also by Goetschel and Voxman [17], considers E1 as a subset of a larger Banach space with the metric D
and yields the following definition.
Definition 2.6. Let f : R ! E1 be a fuzzy function and let t0 2 R: The derivative f0 (t0) of f at the point t0 is defined by
f ðt 0 þ hÞ f ðt 0 Þ
f 0 ðt0 Þ ¼ lim ð4Þ
h!0 h
provided that this limit taken with respect to the metric D, exists. The elements f(t0 + h), f(t0) at the right-hand side of Eq. (4)
are observed as elements in the Banach space B ¼ C½0; 1 C½0; 1 [22]. Thus if f ðt0 þ hÞ ¼ ða; a the differ-
Þ and f ðt0 Þ ¼ ðb; bÞ;
ence is simply
f ðt 0 þ hÞ f ðt 0 Þ ¼ ða b; a
bÞ: ð5Þ
Clearly [f(t0 + h) f(t0)]/h may not be a fuzzy number for all h. However, if it approaches f0 (t0) (in B) and f0 (t0) is also a fuzzy
number (i.e. in E1), this number is the fuzzy derivative of f(t) at t0. In this case if f ¼ ðf ; f Þ then
where A is a fuzzy square matrix, M(t) is an unknown matrix function. Let A be an n n fuzzy matrix and M be a n m
matrix,
_ ¼ dM ¼ dM 1 dM2 dM m ;
M
dt dt dt dt
T
_ j ¼ dM j ¼ dm1j dm2j dmnj ;
M
dt dt dt dt
where M j 2 Rm and M i 2 Rn are the jth column and ith row of M respectively, and ðMÞij ¼ Mij ¼ mij , and the initial condition
M0 is a crisp matrix.
We choose to solve (7) in parametric form then the system (7) can be written as
(
_ a ðtÞ ¼ Aa M a ðtÞ;
M
ð8Þ
Ma ð0Þ ¼ M0 ; 0 6 a 6 1:
B. Ghazanfari et al. / Applied Mathematical Modelling 36 (2012) 348–356 351
Lemma 3.1. Assume that each element of the matrix M in (7) at the time instant t is a fuzzy number where
ija ðtÞ i ¼ 1; . . . ; n;
mija ðtÞ ¼ ½mija ðtÞ; m j ¼ 1; . . . ; m; ð9Þ
then the systems (7) can be described by 2mn differential equations.
8 k n o
> _ k i ija ðtÞ ;
> M ja ðtÞ ¼ min ðAuÞ : u 2 ½mija ðtÞ; m
>
>
>
> n o
< _k
M ðtÞ ¼ max ðAuÞk : ui 2 ½m ðtÞ; m
ja
ðtÞ ; ija ija ð10Þ
>
>
>
> M ja ð0Þ ¼ M ja0 ; k ¼ 1; 2; . . . ; n
>
>
:
M ja ð0Þ ¼ M ja0 ; j ¼ 1; 2; . . . ; m
k Pn l
where ðAuÞ ¼ l¼1 akl u is the kth row of Au. From (7) and (10) we have
X
n
_ kja ðtÞ ¼
m akl ul ; for j ¼ 1; 2; . . . ; m; ð11Þ
l¼1
where
(
uj ¼ mkja ðtÞ; akj P 0;
uj ¼ mkja ðtÞ; akj < 0;
X
n
_ kja ðtÞ ¼
m akl v l ; for j ¼ 1; 2; . . . ; m ð12Þ
l¼1
and
(
vj ¼ m kja ðtÞ; akj P 0;
v j ¼ mkja ðtÞ; akj < 0:
Define new complex variables as follows [14]:
kja ðtÞ k ¼ 1; . . . ; n;
mkja :¼ mkja ðtÞ þ im j ¼ 1; . . . ; m; ð13Þ
pffiffiffiffiffiffiffi
where i :¼ 1.
Hence, the system (7) can be rewritten with MðtÞ ¼ MðtÞ þ iMðtÞ as follows:
(
_ ðtÞ ¼ A ðM ðtÞ þ iM ðtÞÞ;
_ ðtÞ þ iM
M a a a a a
ð14Þ
Ma ð0Þ ¼ M 0 ; 0 6 a 6 1:
þ þ þ
Let a~ij ¼ ðaij Þ þ
a ; ðaij Þa ; Aa ¼ Aa ; Aa , where Aa ¼ ðaij Þa nn ; Aa ¼ ðaij Þa nn .
a ðtÞ ¼ exp t
Ma ðtÞ þ iM Bðl; aÞ M a0 þ iMa0 ; ð16Þ
l¼0
where the elements of the matrix B are determined from those of A as follows:
(
eaij ðl; aÞ; aij ðl; aÞ P 0;
bij ¼
gaij ðl; aÞ; aij ðl; aÞ < 0;
in which e is just the identity operation and g corresponds to a flip about the diagonal in the complex plan, i.e., "a + bi 2 C,
e : a þ bi ! a þ bi; g : a þ bi ! b þ ai: ð17Þ
Proof. 8
According to Theorem 1 in [15] for every column j of M we have that
>
<M_ ðtÞ þ iM _ ðtÞ ¼ S 1
Bðl; aÞðMja ðtÞ þ iM ja ðtÞÞ;
ja ja
l¼0
>
:
M ja ð0Þ ¼ M j0 ; 0 6 a 6 1;
352 B. Ghazanfari et al. / Applied Mathematical Modelling 36 (2012) 348–356
Theorem 3.3. M a ðtÞ þ iM a ðtÞ, is the solution of problem (14) if and only if M a ðtÞ þ iM a ðtÞ is also the solution of problem (15).
Proof. According to Theorem 2 in [15] for every column j of M we have that M ja ðtÞ þ iM ja ðtÞ, is a solution of column j of (14) if
and only if M ja ðtÞ þ iM ja ðtÞ is also the solution of column j of (15).
Consequently ðM 1a ðtÞ M ma ðtÞÞ þ iðM 1a ðtÞ M ma ðtÞÞ is a solution of (14) if and only if ðM 1a ðtÞ M ma ðtÞÞþ
iðM 1a ðtÞ M ma ðtÞÞ is also a solution of (15). h
þ þ ~ ~
where Aa ¼ A þ
a ; Aa , in which Aa ¼ ðaij Þa nn ; Aa ¼ ðaij Þa nn . And f ðtÞ is a continuous matrix function, the elements of f ðtÞ
are fuzzy numbers.
Similarly, problem (24) can be rewritten as
8
>
<M_ ðtÞ þ iM _ ðtÞ ¼ S 1
Bðl; aÞðM a ðtÞ þ iMa ðtÞÞ þ ðf a ðtÞ þ if a ðtÞÞ;
a a
l¼0 ð25Þ
>
:
M a ð0Þ þ iM a ð0Þ ¼ Mð0Þ þ iMð0Þ;
where the elements of the matrix B are determined from those of A as follows:
(
eaij ðl; aÞ; akj ðl; aÞ P 0;
bij ¼
gaij ðl; aÞ; akj ðl; aÞ < 0:
Theorem 3.5. If /2(t) is any fundamental matrix solution of (15) , then the solution of the nonhomogeneous linear systems (25) is
given by
Z t
Ma ðtÞ þ iM a ðtÞ ¼ /2 ðtÞðM a0 þ iMa0 Þ þ /2 ðt sÞðf a ðtÞ þ if a ðtÞÞds; ð26Þ
0
4. Applications
Example 4.1. Let C ¼
l1 0 and M ¼ x1 y1 z1 w1 and defining the coefficient value l to be a fuzzy number
1
0 l2 x2 y2 z2 w2
about 3 and the coefficient value l2 to be a fuzzy number about 3 which be done by setting for example
8
< 0;
> x < 2;
l1 ð0Þ ¼ ðx 3Þ2 þ 1; 2 6 x 6 4;
>
:
0; x > 4;
354 B. Ghazanfari et al. / Applied Mathematical Modelling 36 (2012) 348–356
and
8
< 0;
> x < 4;
l2 ð0Þ ¼ ðx þ 3Þ2 þ 1; 4 6 x 6 2;
>
:
0; x > 2;
h pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffii
½l1 ð0Þa ¼ 3 1 a; 3 þ 1 a
and
h pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffii
½l2 ð0Þa ¼ 3 1 a; 3 þ 1 a ;
so
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
a1 ¼ ð1 lÞ 3 1 a þ l 3 þ 1 a
and
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
a2 ¼ ð1 lÞ 3 1 a þ l 3 þ 1 a
0 1 2 30 1
x_ a1 y_ a1 z_ a1 w_ a1 a1 0 0 0 xa1 ya1 za1 _ a1
w
B_ C
B xa1 _ a1
y z_ a1 _ a1 C 6
w 0 a1 0 07 B x a1
y za1 wa1 C
B C¼6
6
7B a1
7B
C
C;
B x_ y_ a2 z_ a2 _
wa2 A C 4 0 0 0 a2 5@ xa2 ya2 za2 twa2 A
@ a2
x_ a2 _ a2
y z_ a2 _ a2
w 0 0 a2 0 xa2 a2
y za2 w a2
0 1 2 ea1 t 0 0 0
30 1
xa 1 ya1 za1 _ a1
w x10 y10 z10 _ 10
w
B x 6 7
B a1 a1
y za1 a1 C
w C 6 0 e a1 t
0 0 7B x 10
y z10 10 C
w
B C¼6 7B 10
ea2 t ea2 t 7B
C
C:
6 ea2 t þea2 t
@ xa 2 ya2 za2 wa2 A 4 0 0 2 2 @
5 x20 y20 z20 w20 A
xa2 a2
y za2
wa2 ea2 t ea2 t ea2 t þea2 t x20 20
y z20 20
w
0 0 2 2
17 12 11 15
Then the phase portrait for this example is given in Fig. 1 with a = 0.5 and initial value M 0 ¼ :
15 13 14 15
20
upper
lower
15 exact
10
0
x2
−5
−10
−15
−20
−250 −200 −150 −100 −50 0 50 100 150 200 250
x1
20
upper
lower
15 exact
10
5
x2
−5
−10
−15
−20
−20 −15 −10 −5 0 5 10 15 20
x1
Example 4.2. Let C ¼
l 1 and M ¼ x1 y1 z1 w1 and defining the coefficient value l to be a fuzzy number about
0 l x2 y2 z2 w2
3 which be done by setting for example
8
>
< 0; x < 4;
lð0Þ ¼ ðx þ 3Þ2 þ 1; 4 6 x 6 2;
>
:
0; x > 2;
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
½lð0Þa ¼ ½3 1 a; 3 þ 1 a;
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
let a ¼ ð1 lÞð3 1 aÞ þ lð3 þ 1 aÞ where l 2 [0, 1]. First we calculate
! !
x_ 1 þ ix_ 1 _ 1
y_ 1 þ iy z_ 1 þ iz_ 1 _ 1
_ 1 þ iw
w gðaÞ 1 x1 þ ix1 1
y1 þ iy z1 þ iz1 1
w1 þ iw
¼ ;
x_ 2 þ ix_ 2 _ 2
y_ 2 þ iy z_ 2 þ iz_ 2 _ 2
_ 2 þ iw
w 0 gðaÞ x2 þ ix2 2
y2 þ iy z2 þ iz2 2
w2 þ iw
0 1 2 30 1
x_ a1 y_ a1 z_ a1 w_ a1 0 a 1 0 xa 1 ya1 za1 wa1
B_ C
B xa1 _ a1
y z_ a1 _ a1 C 6 a 0 0 1 7
w 6 B x
7B a1 a1
y za1 a1 C
w C
B C¼6 7B C;
B x_ y_ a2 z_ a2 w_ a2 C 4 0 0 0 a 5@ xa2 ya2 za2 wa2 A
@ a2 A
x_ a2 _ a2
y z_ a2 _
a2
w 0 0 a 0 xa2 a2
y za2 a2
w
0 1 2 30 1
xa 1 ya1 za1 wa1 P Q Pt Qt x10 y10 z10 w10
B x a1
y za1 a1 C
w 6Q P Qt Pt 7B x 10
y z10 10 C
w
B a1 C 6 7B 10 C
B C¼6 7B C;
@ xa 2 ya2 za2 wa2 A 4 0 0 P Q 5@ x20 y20 z20 w20 A
xa2 a2
y za2 a2
w 0 0 Q P x20 20
y z20 20
w
eat þeat eat eat
in which P, Q denote 2
; 2
, respectively. Then the phase portrait for this example is given in Fig. 2 with a = 0.5 and
17 12 14 16
initial value M0 ¼ :
15 13 14 16
5. Conclusion
In this paper we extend the notion of first order linear fuzzy differential dynamical systems to the matrix dynamical sys-
tems with fuzzy matrices. Also, we discussed some properties of matrix dynamical systems with fuzzy matrices and their
phase portraits.
356 B. Ghazanfari et al. / Applied Mathematical Modelling 36 (2012) 348–356
References
[1] S. Abbasbandy, T. Allah Viranloo, Numerical solution of fuzzy differential equation by Runge–Kutta method, Nonlinear Studies 11 (1) (2004) 117–129.
[2] J.J. Buckley, T. Feuring, Fuzzy differential equations, Fuzzy Sets Syst. 110 (2000) 43–54.
[3] P.C. Chakrivat, M.S. Kamew, Stiffly stable second multi-step methods with higher order and improved stability regions, BIT 23 (1983) 75–83.
[4] S.L. Chang, L.A. Zadeh, On fuzzy mapping and control, IEEE. Trans. Syst. Man Cybernet. 2 (1972) 30–34.
[5] O. Kaleva, Fuzzy differential equations, Fuzzy Sets Syst. 24 (1987) 301–317.
[6] O. Kaleva, The Cauchy problem for fuzzy differential equations, Fuzzy Sets Syst. 35 (1990) 389–396.
[7] M. Ma, M. Friedman, A. Kandel, Numerical solution of fuzzy differential equations, Fuzzy Sets Syst. 105 (1999) 133–138.
[8] S. Ch. Palligkinis, G. Papageorgiou, I. Th. Famelis, Runge–Kutta methods for fuzzy differential equations, Appl. Math. Comput. 209 (2009) 97–105.
[9] M.L. Puri, D. Ralescu, Differential for fuzzy function, J. Math. Anal. Appl. 91 (1983) 552–558.
[10] M.L. Puri, D. Ralescu, Fuzzy random variables, J. Math. Anal. Appl. 114 (1986) 409–422.
[11] H.H. Rosenbrock, Some general implicit processes for the numerical solution of ordinary differential equations, Comput. J. 23 (1963) 329–330.
[12] S. Seikkala, On the fuzzy initial value problem, Fuzzy Sets Syst. 24 (1987) 319–330.
[13] X. Wu, J. Xia, Extended Runge–Kutta-like formulae, Appl. Numer. Math. 56 (2006) 1584–1605.
[14] J. Xu, Z. Liao, Z. Hu, A class of linear differential dynamical systems with fuzzy initial condition, Fuzzy Sets Syst. 158 (2007) 2339–2358.
[15] J. Xu, Z. Liao, J.J. Nieto, A class of linear differential dynamical systems with fuzzy matrices, J. Math. Anal. Appl. 368 (2010) 54–68.
[16] D. Dubois, H. Prade, Toward fuzzy differential calculus: Part 3, differentiation, Fuzzy Sets Syst. 8 (1982) 225–233.
[17] R. Goetschel, W. Voxman, Elementary calculus, Fuzzy Sets Syst. 18 (1986) 31–43.
[18] B. Chen, X. Liu, Reliable control design of fuzzy dynamical systems with time-varying delay, Fuzzy Sets Syst. 146 (2004) 349–374.
[19] L. Hong, J. Sun, Bifurcations of fuzzy nonlinear dynamical systems, Commun. Nonlinear Sci. Numer. Simul. 11 (2006) 1–12.
[20] I. Rashkovsky, M. Margaliot, Nicholsons blowflies revisited: a fuzzy modeling approach, Fuzzy Sets Syst. 158 (10) (2007) 1083–1096.
[21] D.W. Pearson, A property of linear fuzzy differential equations, Appl. Math. Lett. 10 (3) (1997) 99–103.
[22] W. Cong-Xin, M. Ming, On embedding problem of fuzzy number spaces. Part: I, Fuzzy Sets Syst. 44 (1991) 33–38.