You are on page 1of 12

COMPUTATIONAL METHODS IN APPLIED MATHEMATICS, Vol.2(2002), No.2, pp.

113124
c Institute of Mathematics of the National Academy of Sciences of Belarus

NUMERICAL SOLUTIONS OF FUZZY DIFFERENTIAL EQUATIONS BY TAYLOR METHOD


SAEID ABBASBANDY
Department of Mathematics, Imam Khomeini International University Qazvin, Iran E-mail: abbasbandy@yahoo.com

TOFIGH ALLAH VIRANLOO


Department of Mathematics, Islamic Azad University Tehran, Iran E-mail: alahviranlo@yahoo.com Abstract In this paper, numerical algorithms for solving fuzzy ordinary dierential equations are considered. A scheme based on the Taylor method of order p is discussed in detail and this is followed by a complete error analysis. The algorithm is illustrated by solving some linear and nonlinear fuzzy Cauchy problems. 2000 Mathematics Subject Classication: 34A12; 65L05. Keywords: fuzzy dierential equation, p-order Taylor method, fuzzy Cauchy problem.

1. Introduction
Knowledge about dynamical systems modeled by dierential equations is often incomplete or vague. It concerns, for example, parameter values, functional relationships, or initial conditions. The well-known methods for solving analytically or numerically initial value problems can only be used for nding a selected system behavior, e.g., by xing the unknown parameters to some plausible values. However, in this case, it is not possible to describe the whole set of system behaviors compatible with our partial knowledge. We may set that the fuzzy input is somehow transformed into the fuzzy output dened by the corresponding crisp systems. This reasons us to refer such systems to as Fuzzy Input Fuzzy Output (FIFO) systems. Here, we are going to operationalize our approach, i.e., to propose a method for computing the approximate solution for a fuzzy dierential equation using numerical methods. Since nding this set of solutions analytically does only work with trivial examples, a numerical approach seems to be the only way to solve such problems. The topics of fuzzy dierential equations, which attracted a growing interest for some time, in particular, in relation to the fuzzy control, have been rapidly developed recent years. The concept of a fuzzy derivative was rst introduced by S. L. Chang, L. A. Zadeh in [4]. It was followed up by D. Dubois, H. Prade in [5], who dened and used the extension principle.

114 S. Abbasbandy and T. Allah Viranloo Other methods have been discussed by M. L. Puri, D. A. Ralescu in [10] and R. Goetschel, W. Voxman in [6]. Fuzzy dierential equations and initial value problems were regularly treated by O. Kaleva in [7] and [8], S. Seikkala in [11]. A numerical method for solving fuzzy dierential equations has been introduced by M. Ma, M. Friedman, A. Kandel in [9] via the standard Euler method. The structure of this paper is organized as follows. In section 2, some basic results on fuzzy numbers and denition of a fuzzy derivative, which have been discussed by S. Seikkala in [9], are given. In section 3, we dene the problem that is a fuzzy Cauchy one. Its numerical solution is of the main interest of this work. Solving numerically the fuzzy dierential equation by the Taylor method of order p is discussed in section 4. The proposed algorithm is illustrated by some examples in section 5 and the conclusion is in section 6.

2. Preliminaries
Consider the initial value problem x (t) = f (t, x(t)); x(a) = . a t b, (1)

Let Y (t) be an exact solution of (1) and Y (ti ) be approximated by yi = y(ti ), which in the p-order Taylor method is as follows: yi+1 = yi + hT (ti , yi ), i = 0, 1, . . . , N 1, and T (ti , yi ) =
i=0 p1

(2)

hi f (i) (ti , yi ), (i + 1)!

(3)

(b a) = ti+1 ti . (4) N A triangular fuzzy number v is dened by three numbers a1 < a2 < a3 , where the graph of v(x), the member function of the fuzzy number v, is a triangle with the base on the interval [a1 , a3 ] and the vertex at x = a2 . We specify v as (a1 /a2 /a3 ). We will write: (1) v > 0 if a1 > 0; (2) v 0 if a1 0; (3) v < 0 if a3 < 0; and (4) v 0 if a3 0. Let E be a set of all the upper semicontinuous normal convex fuzzy numbers with bounded r-level sets. It means that if v E, then the r-level set a = t0 t1 ... tN = b and h = [v]r = {s| v(s) r}, 0<r 1,

where

is a closed bounded interval which is denoted by [v]r = [v1 (r), v2 (r)]. Let I be a real interval. The mapping x : I E is called a fuzzy process and its r-level set is denoted by [x(t)]r = [x1 (t; r), x2 (t; r)], t I, r (0, 1]. The derivative x (t) of the fuzzy process x is dened by [x (t)]r = [x1 (t; r), x2 (t; r)], t I, r (0, 1],

Numerical solutions of fuzzy dierential equations by Taylor method

115

provided that this equation determines the fuzzy number, according to Seikkala [11]. Denote by the set of all nonempty compact subsets of R and by c the subset of consisting of nonempty convex compact sets. Recall that (x, A) = min x a
aA

is a distance of the point x R from A and that the Hausdor separation (A, B) of A, B is dened as (A, B) = max (a, B).
aA

Note that the notation is consistent, since (a, B) = ({a}, B). Now, is not a metric. In fact, (A, B) = 0 if and only if A B. The Hausdor metric dH on is dened by dH (A, B) = max{(A, B), (B, A)}. The metric dH on E is as follows: d (u, v) = sup{dH ([u]r , [v]r ) : 0 r 1}, u, v E.

3. Fuzzy Cauchy problem


Consider the fuzzy initial value problem x (t) = f (t, x(t)), t I = [0, T ], x(0) = x0 , where f is a continuous mapping from R+ R into R and x0 E with the r-level set [x0 ]r = [x1 (0; r), x2 (0; r)], r (0, 1]. The extension principle of Zadeh leads to the following denition of f (t, x) when x = x(t) is a fuzzy number: f (t, x)(s) = sup{x( )| s = f (t, )}, s R. From this it follows that [f (t, x)]r = [f1 (t, x; r), f2 (t, x; r)], r (0, 1], where f1 (t, x; r) = min{f (t, u)| u [x1 (t; r), x2 (t; r)]}, (5)

f2 (t, x; r) = max{f (t, u)| u [x1 (t; r), x2 (t; r)]}. by

(6)

The mapping f (t, x) is a fuzzy process and the derivative f (i) (t, x), i = 1, . . . , p is dened [f (i) (t, x)]r = [f1 (t, x; r), f2 (t, x; r)], t I, r (0, 1], provided that this equation determines the fuzzy number f (i) (t, x) E, where f1 (t, x; r) =
(i) (i) (i) (i)

min{f (i) (t, u)| u [x1 (t; r), x2 (t; r)]},

f2 (t, x; r) = max{f (i) (t, u)| u [x1 (t; r), x2 (t; r)]}.

(7)

116 S. Abbasbandy and T. Allah Viranloo Theorem 3.1. Let f satisfy |f (t, v) f (t, v)| g(t, |v v|); t 0; v, v R,

where g : R+ R+ R+ is a continuous mapping such that r g(t, r) is nondecreasing. Let the initial value problem u (t) = g(t, u(t)), u(0) = u0 , (8)

have a solution on R+ for u0 > 0 and u(t) = 0 be the only solution of (8) for u0 = 0. Then the fuzzy initial value problem (5) has a unique fuzzy solution. Proof. [11] Since x (t) = f
(i) (i1)

f (i2) f (i2) (t, x(t)) = (t, x(t)) + (t, x(t))f (i2) (t, x(t)), t x

(9)

from (7) it follows that f1


(i1)

(t, x; r) = min (t, x; r) = max

f (i2) f (i2) (t, u) + (t, u)f (i2) (t, u)| u [x1 (t; r), x2 (t; r)] , t u f (i2) f (i2) (t, u) + (t, u)f (i2) (t, u)| u [x1 (t; r), x2 (t; r)] . t u

(10)

f2

(i1)

4. Taylor method of order p


Let the exact solution [Y (t)]r = [Y1 (t; r), Y2 (t; r)] be approximated by [y(t)]r = [y1 (t; r), y2 (t; r)]. The Taylor method of order p is based on the expansion
p

x(t + h; r) =
i=0

hi (i) x (t; r), i!

(11)

where x(t; r) is Y1 or Y2 . We dene


p1

F [t, x; r] =
i=0 p1

hi (i) f (t, x; r), (i + 1)! 1 hi (i) f (t, x; r). (i + 1)! 2 n N are denoted by

(12)

G[t, x; r] =
i=0

The exact and approximate solutions at tn , 0

[Y (tn )]r = [Y1 (tn ; r), Y2 (tn ; r)] and [y(tn )]r = [y1 (tn ; r), y2 (tn ; r)],

Numerical solutions of fuzzy dierential equations by Taylor method

117

respectively. The solution is calculated at the grid points of (4). Using the Taylor method of order p and substituting Y1 , Y2 into (11) and considering (12), we have Y1 (tn+1 ; r) Y1 (tn ; r) + hF [tn , Y (tn ); r], Y2 (tn+1 ; r) Y2 (tn ; r) + hG[tn , Y (tn ); r]. Hence, we get y1 (tn+1 ; r) = y1 (tn ; r) + hF [tn , y(tn ); r], y2 (tn+1 ; r) = y2 (tn ; r) + hG[tn , y(tn ); r], where y1 (0; r) = x1 (0; r), y2 (0; r) = x2 (0; r). The following lemmas will be applied to show convergence of these approximations, i.e.,
h0

(13)

(14)

lim y1 (t; r) = Y1 (t; r), lim y2 (t; r) = Y2 (t; r).

h0

Lemma 4.1. Let the sequence of numbers {Wn }N satisfy n=0 |Wn+1 | A|Wn | + B, 0 n N 1,

for the given positive constants A and B. Then |Wn | Proof. See [9]. Lemma 4.2. Let the sequences of numbers {Wn }N , {Vn }N satisfy n=0 n=0 |Wn+1 | |Vn+1 | |Wn | + A max{|Wn |, |Vn |} + B, |Vn | + A max{|Wn |, |Vn |} + B, An |W0 | + B An 1 , A1 0 n N.

for the given positive constants A and B. Then, denoting Un = |Wn | + |Vn |, we have Un where A = 1 + 2A and B = 2B. Proof. See [9]. Let F (t, u, v) and G (t, u, v) be the functions F and G in (12), where u and v are constants and u v. In other words,
p1

0 0

n n

N, N,

n A 1, An U0 + B A1

F (t, u, v) =
i=0 p1

hi (i) min f1 (t, )| [u, v] , (i + 1)! hi (i) max f1 (t, )| [u, v] , (i + 1)!

G (t, u, v) =
i=0

118 S. Abbasbandy and T. Allah Viranloo or F (t, u, v) and G (t, u, v) are obtained by substituting [x(t)]r = [u, v] into (12). The domain of F and G is K = {(t, u, v)| 0 t T, < v < , < u v}.

Theorem 4.1. Let F (t, u, v) and G (t, u, v) belong to C p1 (K) and the partial derivatives of F and G be bounded over K. Then, for arbitrarily xed 0 r 1, the numerical solutions of (14) converge to the exact solutions Y1 (t; r) and Y2 (t; r) uniformly in t. Proof. It is sucient to show
h0

lim y1 (tN ; r) = Y1 (tN ; r),

h0

lim y2 (tN ; r) = Y2 (tN ; r),

where tN = T . For n = 0, 1, . . . , N 1, using the Taylor theorem, we get Y1 (tn+1 ; r) = Y1 (tn ; r) + hF [tn , Y1 (tn ; r), Y2 (tn ; r)] + hp+1 (p+1) Y (n,1 ), (p + 1)! 1 hp+1 (p+1) Y2 (n,2 ), Y2 (tn+1 ; r) = Y2 (tn ; r) + hG [tn , Y1 (tn ; r), Y2 (tn ; r)] + (p + 1)!

(15)

where n,1 , n,2 (tn , tn+1 ). Denoting Wn = Y1 (tn ; r) y1 (tn ; r), from (14) and (15) it follows that Wn+1 = Wn + h{F [tn , Y1 (tn ; r), Y2 (tn ; r)] F [tn , y1 (tn ; r), y2 (tn ; r)]} + Vn+1 = Vn + h{G [tn , Y1 (tn ; r), Y2 (tn ; r)] G [tn , y1 (tn ; r), y2 (tn ; r)]} + Hence, |Wn+1 | |Vn+1 | where M1 = max |Y1
(p+1)

Vn = Y2 (tn ; r) y2 (tn ; r),

hp+1 (p+1) Y1 (n,1 ), (p + 1)! hp+1 (p+1) Y2 (n,2 ). (p + 1)!

hp+1 M, (p + 1)! hp+1 M, |Vn | + 2Lh max{|Wn |, |Vn |} + (p + 1)! |Wn | + 2Lh max{|Wn |, |Vn |} + (t; r)|, M2 = max |Y2
(p+1)

(t; r)| for t [0, T ],

M = max{M1 , M2 }, and L > 0 is a bound for the partial derivatives of F and G . Therefore, from Lemma 4.2, we obtain |Wn | |Vn | 2hp+1 (1 + 4Lh)n 1 M , (p + 1)! 4Lh 2hp+1 (1 + 4Lh)n 1 (1 + 4Lh)n |U0 | + M , (p + 1)! 4Lh (1 + 4Lh)n |U0 | +

Numerical solutions of fuzzy dierential equations by Taylor method

119

where |U0 | = |W0 | + |V0 |. In particular, |WN | |VN | hp+1 (1 + 4Lh) h 1 (1 + 4Lh) |U0 | + M , (p + 1)! 2Lh
N
T

hp+1 (1 + 4Lh) h 1 (1 + 4Lh) |U0 | + M . (p + 1)! 2Lh


N

Since W0 = V0 = 0, we have |WN | M e4LT 1 p h, 2L(p + 1)! |VN | M e4LT 1 p h. 2L(p + 1)!

Thus, if h 0, we get WN 0 and VN 0, which completes the proof.

5. Examples
Example 5.1. Consider the fuzzy initial value problem [9] y (t) = y(t), t I = [0, 1], y(0) = (0.75 + 0.25r, 1.125 0.125r), Using the Taylor method of order p, we have
p

0<r

1.

y1 (tn+1 ; r) = y1 (tn ; r)
i=0

hi , i!

y2 (tn+1 ; r) = y2 (tn ; r)
i=0

hi . i!

The exact solution is given by Y1 (t; r) = y1 (0; r)et , Y2 (t; r) = y2 (0; r)et , which at t = 1 is Y (1; r) = [(0.75 + 0.25r)e, (1.125 0.125r)e], 0<r 1.

The exact and approximate solutions for p = 2 and p = 4 are compared and plotted at t = 1 in Figs. 1 and 2.
1

0.8

0.6

0.4

0.2

0 1.8

Euler 2nd-order Taylor Exact


2 2.2 2.4 2.6 2.8 3

3.2

Figure 1. h = 0.2

120 S. Abbasbandy and T. Allah Viranloo


1

0.8

0.6

0.4

0.2

0 1.8

Euler 4th-order Taylor Exact


2 2.2 2.4 2.6 2.8 3

3.2

Figure 2. h = 0.2

The Hausdor distances of the exact solution from the Euler solution, the 2nd- and 4th-order Taylor ones are d = 0.2587, d = 0.0175 and d = 3.4528e 005, respectively. Example 5.2. Consider the fuzzy initial value problem [9] y (t) = ty(t), [a, b] = [1, 1], y(1) = ( e .5(1 r), e + .5(1 r)), 0 < r We separate the problem into two steps. (a) t < 0. The parametric form in this case is y1 (t; r) = ty2 (t; r), y2 (t; r) = ty1 (t; r), y1 (t; r) = (1 + t2 )y2 (t; r), y2 (t; r) = (1 + t2 )y1 (t; r), y1 (t; r) = min{(1 + 2t + 2t2 + t4 )u| u [y1 (t; r), y2 (t; r)]}, y2 (t; r) = max{(1 + 2t + 2t2 + t4 )u| u [y1 (t; r), y2 (t; r)]}, y1 (t; r) = min{(2 + 4t2 + 4t(1 + t2 ) + 4t(1 + t2 )2 + (1 + t2 )4 )u| u [y1 (t; r), y2 (t; r)]}, y2 (t; r) = max{(2 + 4t2 + 4t(1 + t2 ) + 4t(1 + t2 )2 + (1 + t2 )4 )u| u [y1 (t; r), y2 (t; r)]}, with the initial conditions given. The unique exact solution is Y1 (t; r) = Y2 (t; r) = where A = e
(t2 a2 ) 2

1.

(3)

(3)

(4)

(4)

AB A+B y2 (0; r) + y1 (0; r), 2 2 A+B AB y2 (0; r) + y1 (0; r), 2 2

,B=

1 . A

Numerical solutions of fuzzy dierential equations by Taylor method

121

(b) t

0. The parametric equations are y1 (t; r) = ty1 (t; r), y2 (t; r) = ty2 (t; r), 2 y1 (t; r) = (1 + t )y1 (t; r), y2 (t; r) = (1 + t2 )y2 (t; r),

and [y (3) ]r , [y (4) ]r are calculated as in part (a) with the initial conditions given. The unique exact solution at t > 0 is Y1 (t; r) = y1 (0; r)e 2 ,
t2

Y2 (t; r) = y2 (0; r)e 2 .

t2

The exact and approximate solutions for p = 2 and p = 4 are compared and plotted in Figs. 3 and 4.
r 1

0.8

0.6

0.4

0.2

0 0

T Euler T 2nd-order Taylor T T Exact T T T T T T T T T T T T T T T


0.5 1 1.5 2 2.5 3

3.5

Figure 3. h = 0.05

0.8

0.6

0.4

0.2

0 0

Euler 4th-order Taylor Exact


0.5 1 1.5 2 2.5 3

3.5

Figure 4. h = 0.05

122 S. Abbasbandy and T. Allah Viranloo The Hausdor distances of the exact solution from the Euler solution, the 2nd- and 4th-order Taylor ones are d = 0.1284, d = 0.0677 and d = 2.8099e 008, respectively. Example 5.3. Consider the fuzzy initial value problem y (t) = k1 y 2 (t) + k2 , y(0) = 0,

where kj > 0 (j = 1, 2) are triangular fuzzy numbers [2]. The exact solution is given by Y1 (t; r) = l1 (r) tan(w1 (r)t), Y2 (t; r) = l2 (r) tan(w2 (r)t), with l1 (r) = w1 (r) = where [k1 ]r = [k1,1 (r), k1,2 (r)] and [k2 ]r = [k2,1 (r), k2,2 (r)], k1,1 (r) = .5 + .5r, k1,2 (r) = 1.5 .5r, k2,1 (r) = .75 + .25r k2,2 (r) = 1.25 .25r. The r-level sets of y (t) are Y1 (t; r) = k2,1 (r) sec2 (w1 (r)t), Y2 (t; r) = k2,2 (r) sec2 (w2 (r)t), which denes the fuzzy number. We have f1 (t, y; r) = min{k1 u2 + k2 | f2 (t, y; r) = max{k1 u2 + k2 | 2 f1 (t, y; r) = min{2k1 u3 + 2uk1 k2 | 2 f2 (t, y; r) = max{2k1 u3 + 2uk1 k2 | f1 (t, y; r) = min{
(i)

k2,1 (r)/k1,1 (r), k1,1 (r)k2,1 (r),

l2 (r) = w2 (r) =

k2,2 (r)/k1,2 (r), k1,2 (r)k2,2 (r),

u [y1 (t; r), y2 (t; r)], u [y1 (t; r), y2 (t; r)], u [y1 (t; r), y2 (t; r)], u [y1 (t; r), y2 (t; r)],

kj kj kj kj

[kj,1 (r), kj,2 (r)], [kj,1 (r), kj,2 (r)], [kj,1 (r), kj,2 (r)], [kj,1 (r), kj,2 (r)],

j j j j

= 1, 2}, = 1, 2}, = 1, 2}, = 1, 2},

f (i1) (t, u) (i1) f (t, u)| u [y1 (t; r), y2 (t; r)], u j = 1, 2}, f (i1) (t, u) (i1) f (t, u)| u [y1 (t; r), y2 (t; r)], u j = 1, 2},

kj [kj,1 (r), kj,2 (r)],

f2 (t, y; r) = max{

(i)

kj [kj,1 (r), kj,2 (r)],

for i = 1, 2. There are two nonlinear programming problems. They can be solved by the GAMS (the acronym stands for General Algebraic Modeling Systems) software. GAMS is designed to make the construction and solution of large and complex mathematical programming models more straightforward for programmers and more comprehensible to users of models from other disciplines, e.g., economists [1]. Thus, the Taylor method suggested in this paper can be used. The exact and approximate solutions are shown in Figs. 5 and 6 at t = 1.

Numerical solutions of fuzzy dierential equations by Taylor method r 1

123

0.8

0.6

0.4

0.2

0 0.5

Euler 2nd-order Taylor Exact b b bH HH Ha aa aP P


1 1.5 2 2.5 3 3.5 4

4.5

Figure 5. h = 0.1

0.8

0.6

0.4

0.2

0 0.5

Euler 4th-order Taylor Exact Q b b H HH Ha a a a P P


1 1.5 2 2.5 3 3.5 4

4.5

Figure 6. h = 0.1

The Hausdor distances of the exact solution from the Euler solution, the 2nd- and 4th-order Taylor ones are d = 1.8711, d = 0.7027 and d = 0.0073, respectively.

6. Conclusion
It is shown that the convergence order of the method proposed is O(hp ), while the Euler method from [9] converges with the rate O(h) only. In comparison with [9] the solutions of Examples 5.1, 5.2 have the higher accuracy in this paper.

124 S. Abbasbandy and T. Allah Viranloo

References
[1] A. Brooke, D. Kendrick, and A. Meeraus, GAMS: A Users Guide, 2nd edn., The Scientic Press, Redwood city, CA, 1992. [2] J. J. Buckley and T. Feuring, Fuzzy dierential equations, Fuzzy Sets and Systems, 110 (2000), pp. 4354. [3] R. L. Burden and J. D. Faires, Numerical Analysis, seventh edn., Brooks-Cole Publishing, 2001. [4] S. L. Chang and L. A. Zadeh, On fuzzy mapping and control , IEEE Trans. Systems Man Cybernet., 2 (1972), pp. 3034. [5] D. Dubois and H. Prade, Towards fuzzy dierential calculus: Part 3, Dierentiation, Fuzzy Sets and Systems, 8 (1982), pp. 225233. [6] R. Goetschel and W. Voxman, Elementary fuzzy calculus, Fuzzy Sets and Systems, 24 (1987), pp. 3143. [7] O. Kaleva, Fuzzy dierential equations, Fuzzy Sets and Systems, 24 (1987), pp. 301317. [8] O. Kaleva, The Cuachy problem for fuzzy dierential equations, Fuzzy Sets and Systems, 35 (1990), pp. 389396. [9] M. Ma, M. Friedman, and A. Kandel, Numerical solutions of fuzzy dierential equations, Fuzzy Sets and Systems, 105 (1999), pp. 133138. [10] M. L. Puri and D. A. Ralescu, Dierentials of fuzzy functions, J. Math. Anal. Appl., 91 (1983), pp. 321325. [11] S. Seikkala, On the fuzzy initial value problem, Fuzzy Sets and Systems, 24 (1987), pp. 319330.

Received 5 Dec. 2001 Revised 3 Aug. 2002