C H A P T E R

5
INTERVAL ARITHMETIC

5.1. INTRODUCTION
each fuzzy set, and thus also
Fuzzy arithmetic is based on two properties fuzzy numbers; (1)
of
its a-cuts; and (2) a-cuts of each
each fuzzy number, can fully and uniquely be represented by
numbers for all a e [0, 1]. The interval analysis,
a
fuzy number are closed intervals of real facilitate our
wel-established area of classical mathematics, we
shall study in this chapter and to
presentation of fuzzy arithmetic in the next chapter.

INTERVAL NUMBERSs situation in which the

R. Consider a
or collection of all real
numbers is denoted by between two
Set x is located
is uncertain. It is only known that
within this interval
ofinterest (a real number) form an interval. All numbers
which
a, and az, where a, az,
ae
COnsidered as having equal rights in given
a
situation.
bounded interval
belongs to a closed
uncertain value x are called
Usua will a s s u m e that the the interval; a, and a,
a] meaning
x belongs to
Sx e
Sa, or x [a,
boundaries , of the interval. numbers x such that
or end points set of real
A is defined as the
number ¬ R,
I N I T I O N : An interval r E la, a2l, a, a
Sx Sa, i.e., .(1)
a
a,]= {x: aj Srs
a, x e R},
A =
[a, where x is a real number.
Sx S a, Geometrically,
ch reads: nthee set of all
set of all x such that a,
an
uncertain number x
e [a, a].
represents
closed interval Eq. (1) numbers.
S Drese
esented as a segment o n
the line of real
A
number A.
5.1: Interval
Fig.
 Applicatios
136 Fuzzy Set Theory, Fuzzy Logic and Their

5.3. CLASSIFICATION OF INTERVALS

be classified
into fwelve different types, listed beloy
below, wh
The intervals of real numbers can <
a and a, are real numbers, with a,

5.3.1. Empty Interval Ø.

(a, a) ={}
=

la, a) =(a, a) =
5.3.2. Point Interval or Singleton
to the real number
a =
[a, aj which is also

Ifa, =
a,
=
a, the
interval number reduces clet
a point interval or singleton.
0 [0, 0). 1 =
[1, 1]. 5
=
[5, 5] are point intervals.
For example,
Intervals
5.3.3. Proper and Bounded
sxsa} (open) R: aq
Ja,al (aj. a)= {r e
= e

Sx< a} (left-closed, right-open)

la. a la, a)= {r R:a,
R: aa < x s a} (left-open, right-closed)
Ja, al (a, a]= {r e
(closed).
la,a{rER:aq sraa}
5.3.4. Left-Bounded and Right-Unbounded Intervals
(a,o)= {x R:x>a,} lef-open)
e

la.o) ER:x2a,} (lefi-closed).

5.3.5. Lef-Unbounded and Right-Bounded Intervals
o , a) {r e R;x<a,} (right-open)
o , a] {x e R:xsa} (righ-closed).
5.3.6. Interval Unbounded at Both Ends
- , o)R (open ray)
5.3.7. Width of Interval
Let A = [a1, a ] , then

Width
of A= w{A) = wla, a]=a2-4
5.3.8. Magnitude of Interval
Let
A la, az), then =

Magnitude of A =

\4 |[4.
=

max (|.lal)
l l it lajzlal
l l if lalsll
5.3.9. Image of Interval
Let
A=[41, a,], then
Image of A=A [4, al =

5.3.10. Inverse of Interval

Let A = [ a , a ] , then
 Inverse ofA =
AT' =
[a,, a]

provided 0 [a1, a].

a
Ex. 1. JLet A ={2, 5] be an interval number, then find w{A), A}, A and A.
Sol Width of A = wMA) = w([2, 5) = 5 -2 3 .

Magnitude of A = 4| = |(2, 5]|

m a x (121, I5|) = 5.

Image of A =A = [2, S]

=-5,-2].
Inverse of A =
AT =
[2, 5]"
- [1/5, 1/2].

A A A
-5 -2 o312
Fig. 5.2: Interval numbers A, A and A.

5.3.11. Indusion of Intervals

Consider the interval numbers
A [a, az] and B [b;, b,) such that b, <
bz
=

a <a
Then, we write A c B, ie., la, azl c ib, b,] or B A, ie., lb, b] > la, al. meaning
that the interval fa, a] is contained or inclusion as set in the interval [b, ba]. This inclusion is
calledthe nestiag-property-of-intervals
The inclusion can be generalized for more than two intervals,
A[41, a,l, B= [b1, b,l,.,L = [h, 41, M=[m, m].
Assume that, m <, ..

<b a a <b, <. lh m

Then we have the inclusion
l , a) [b, b,] C... c , 4 ] c m, m].
Ex. 2. Consider the intervals
A = [1, 2] and B [ - 1 , 4].

Show that the interval A is nested into the interval B.

Sol Here, a=1, a, -2, b,=-1, b, =4.
Since, -1 <1<2<4, we have
[I, 2]-, 4].
Hence, the interval A is nested into the interval B.
.5.12. Equality of Interval Numbers
The interval
interval AA=[a, a] and B = [b, b,] are called equal, if a = b, and a, b
Theg, we write A B.

ARITHMETIC OPERATIONS
The
nat are operations of interval arithmetic are, for two intervals A =
[a1, a] and B
subsets of the real line ( , o).
=

[b1, b2J
138 Fuzy Set Theory, Fuzzy Logic and Their Applications

5.4.1. Addition of Intervals

A= la, a] = {x: a `rsa}
Let
and
B [b,, b,]- y:b, sys h},
then A+B=[a, a] + [b, b

la,+b,, a, + b,]
Sr+ysh, b,
+
when at a
x +y e [a, + az, b + bl.
or

5.4.2. Subtraction of Intervals

A- B =[a, a]- [b,, bJ
[a-b, a -b1].
The operation subtraction is in fact addition of the image of B to A,
A+B = [a, a ] + [b,, b J

la, a]+ -b, -b,]

5.4.3. Multiplication of Intervals
AB [a, a,] Tb,, b]
=
min(a,b,ab2, abj, ajbz)ymax(a,bi,4bz,abj,zb)
5.4.4. Division of Intervals

A:B- =A/B= l4, az]: [b, ba]

The operation division is multiplication of the inverse of B with A,

Division by an interval containing zero is not defined

under the basic interval arithmeul
For division by an interval including zero, first define

1/a, 0-]-o,1/a,]
and 1/0, a,] =[l/ag,[
For a0 < a2, We get
1/[a,a]=]-o, lla,]u[l/laz, *»
which as a single interval gives
1/[4, a]=]- o, co [: this loss useful
ínformation about (1ia, 1a2*
typically it is common to work with - o, 1/a,] and
So
[l/a,, [ as separate interval
Because several such divisions may oecur in an interval
useful to do the calculation with so-called multi-intervals ofarithmetic
the for
calculation, it is some

The corresponding
multi-interval arithmetic maintains a
disjoint set of intervals and
also

provides for overlapping intervals to unite. Since a number r e R can be

, you can combine interval and real numbers. interpreted as the in
 Tnterval Arithmetic
With the help of these
definitions, it is already possible
functions, such as
to calculate the range of simpi
Sa, b, x) = a x +b.
Ex. 3Leta =
[1, 2]. b= [5, 7] and [2, 3], calculate the range of function f (a, b, x).
x

Sol. S(a, b, x) =

(1.2] {2.3]) +[5, 7]

=
[1 2, 2 3] +[5, 71
-
[2, 6]+ [5, 71
[7, 13].
Interpreting this as a
functionf(a, b, x) of the variable x with interval parameters a and b,
then it is possible to find the roots of this function. It is then
s([1.2].[5, 7]. x) - ([1.2]-x) + [5, 7] - 0
[1, 2]x = -1,-5]

x=-7,-5]/[1, 2]
=[-7,-5][1/2, 1]
min(-3.5,-7,-2.5,-5), max(-3.5,-7,-2.5,-5)
-7,-2.5]
the possible zeros are in the interval [-7, -2.5
5.4.5. Multiplication of Positive lIntervals
As in the above example, the multiplication of intervals often only requires two multiplications.
It is in fact
AB [a.4,]lb1, ba
[a,b, a,b,).
The multiplication can be seen as a destination area of a rectangle with varying edges. The
Tesult interval covers all levels from the smallest to the largest. The same applies when one of the
Generally, multiplication can produce results
nervals is non-positive and the other non-negative. occurs, for example in a division,
as wide as [-o, o], for if 0. o is
example squared. This also
if the numerator and denominator both contain zero.

4 . Let A [1, 2] and B = [4, 6], then find A

= +B, A B, AB, AIB, A and B
Sol. A +B [1, 2] + [4, 6]
=
[1 +4, 2 +6]
[5, 8]
B
6 8
1
A+B
Fig. 5.3: A + B.

A-B [1, 2] [4, 6]

-

=
[1 -

6, 2 -4]
= [-5, -2].
A
6
-5 -2 0 12

A-B
Fig. 5.4:A -
B.
 Fuzzy Set Theory, Fuzy Logic and Their Applications

A:B [1, 2] 14, 6]

[4, 12. B
A 12
O 1 2 6
A B

Fig. 5.5: A.B.

A/B [1, 2/[4, 6]

-
[1, 21-[1/6, 1/4]
= [1/6, 1/2].

VB 4
011 1 2
62
Fig. 5.6: A/B.

A=[1,2]T
= [1/2, 1 ]

B [4,6]
[1/6, 1/4].

5.5. RULES FOR OPERATIONS

The addition and multiplication operations are commutative, associative and sub-distributive far
the intervals
A- [a, al. B [b, b,] and C = [ , c].
1. Commutative: A+B B+A
AB = BA

2. Associative: (4 +B)+C
C=At (B+C)
(AB)C A(BC)
3. Sub-distributive: A(B + C) AB + AC
terval
The point intervals 0 = [0, 0] and 1 = [1, 1] are the unique neutral elements for nc
addition and multiplication:
A =A +0 =0+A and A =A.1 =1.A
If A =
[a, al (point interval), then the distributive law holds:
A(B +C)= AB +AC. holds
Ex. 5.
Given A [1, 2], B [1, 2] and
=

C [-2,-11, then show that sub-distribunv

Sol. L.H.S A(B +C)
[1, 21([1,2]+[-2,-1))
[1, 2][-1, 1]
min (-1,1,-2,2), max(-1,1,-2, 2)]
-2, 2].
R.H.S= AB + AC
=
[1, 2][1, 2]+ [1, 2][-2,-1]|
 Interval Arithmetic 141
=
min (1, 2, 2, 4), max (1.2, 2, 4)
-
+min(-2,-l,-4,-2),max (-2,-1,-4,-2)]
[1, 4]+ 4, -1]
-[-3, 3].
.the following property, called
Howe

equality), holds
sub-distributive, involving the sign (inclusion or
A(B +C)c AB + AC.
OTE: When
NOT reduces to , the distributive law is
valid. The condition for this is
b]= and z e le1,
C2 C, ie., yz20 for all
yelb,
A(B +C) AB + AC if yz 2 0, Vy e B, z e C. =

E Let A =[1, 2], B =[-1, 1], C= [3, 5], then find the
additionand multiplication and sub-distributivity. commutativity, associativity of
Sol. Commutativity of addition:
A +B [1, 2] +[-1, 1]
= [0, 3],

B+A I-1, 1]+ [1, 2]

[0, 3].
Commutativity of multiplication:
AB [1, 2]-1, 1]
min(-1, 1,-2,2),max(-1,1,-2,2)]
-

= - 2 , 2],

BA -1, 1j[1, 2]
-

min(-1,-2,1,2), max(-1,-2,1,2)]
-2, 2.
Associativity of addition:
4+8)+ C (1.2]+[-1,1) +[8, 5]
10, 3]+ [3, 5
[3, 8],
A+(B + C) =[1, 2] + ([-1,1]+13,5])
-
[1, 2]+(2, 6]
[3, 8].
Ciativity of multiplication
AB)C- (1,2]1-11}) 13, 5]
-[-2,2][3, 5]
min (-6,-10,6, 10), max(-6,-10,6, 10)
= [-10, 10],

A(BC)-[1, 21(-1,1][3,3)
[1, 2] min(-3,-5,3,5), max (-3,-5, 3,5)
 Applications
and Their
FuzyLogic
PMLySetTheury,
[1.2-5.5 5, -10,10)
-10, 10), max(-5,
-

min(-5, ,
- [-10, 10).

Sub-distributivity
MB+C)-1,2]-1,11+[3,5])

[1,23/2, 61
6, 4, 12)
min(2,6,4, 12), max(2,
[2,12]. 51
AB+ AC 1, 2]1-1, 1]+1,2]13,
(-,1-2,2)
min(-1, 1, -2,2), max
5,6, 10), max (3, 5, 6, 10)J
+min(3,
= -2, 2]+ [3, 10)
- [1, 12].

[2. 12][1. 12)

Hence,
5.6. INCLusiON MONOTONIC
Interval arithmetic is inchusion monotoníc, if A cB and C C D, then
) A CcB+ D
(i) A - CcB - D
(ä) AC c BD
i AC c BID, if 0 e C, D.
EL1. Forthe intervals A 2, 31, B [1, 6], C= [4, 51, D = [3, 7] verify theincusi
monotonicity properties.
Sol. ( A+C [2, 3] + [4, 5
= [6, 8)

arid
B+D-[1,6]+ (3,7]
-[4,13)
6,3)C4, 13).
A-C [2, 3 -[4,5)
- (2-5,3-4]
-3,-11
B-D[1,6) [3,7] -

-{1 -7,6-3]
6,31
-3,-1c-6,31.
AC 2,34, 5)

min (8,10,12,15), max (8,

-(8, 15) 10,12,15)
BD [1,63, 71

min (3,7,18, 42), max

(3,7,18, 42)]
 Interval Arithmetic 143
[3, 42]
[8,15][3, 42].
(iv) AC [2, 3][1/5,1/4]1
=

min (2/5,1/2, 3/5,3/4), max

= [2/5, 3/4] (2/5,1/2,3/5, 3/4)
B/D [1, 6][1/7, 1/3]
=
min (1/7,1/3, 6/7,6/3), max
[1/7, 6/3]
(1/7,1/3,6/7,6/3)
(2/5, 3/4][1/7, 6/3].
5.7DISTANCE BETWEEN INTERVALSs
Consider the interval A =

[aj, a,] and B

between A and B is defined by
=

[b, b,] in the set of real

intervals I(R). Distance
d4, B) max(4-laz
If Aand B are point intervals, A= [a, a] and B - b|). . (1)
[b, b], then =

distance between real numbers is

d(la, al. [b, b)) max (la b|.la-b|) = -

=
la -

bl.
From Eq. (1), it follows that
d(A, B)= d(B, A),
dA, B) 0 if and only if A B. =
=

Ex.&Let A [1, 6] and B [3, 7], then find the

=

Sol. distance between intervals A and B.

dA, B) max
(1-3.6-7|)
=

max(2, 1)
2.
5,8. INTERVAL OPERATIONS IN Z
Let the set of all integers
An interval number
Z {..,-3, 2,-1, 0, 1, 2,
3,...
A e Zis defined as the set or collection of integers x,
A
or [a, a] {x:a, `isa, x e Z}, =

A
The arithmetic operations addition, {a, a + 1, a, t+2, a-2, a 1,
subtraction and
- a}. ,

nCin are
belong to Z. defined with Article 5.4.1, 5.4.2 and 5.4.3multiplication
Z for interval numbers A, B
correspondingly; the operation results
he operation
Dut to R, except indivision
the
is not defined in Z since
1/b, and 1/b, (Article 5.4.4) do not
Ex. 9. Given trivial case b,
b2 belong to
the intervals in Z
A =
-2, 1] {-2, -1,0, 1},
=

B [2, 6] {2, 3, 4, 5, 6},

=
then find A
+B, A -
B and AB.
Sol.
A+B -2, 1]+[2, 6)
[0, 71
{0, 1, 2, 3, 4, 5, 6, 7},
 ZY Set Theory, Fuzz Logic and Thetr Applicato
A- B
[2, 1]-[2,6}
-8,-1
-8, -7, 6, -5, 4, -3, -2,-1j
AB - 2 , 1][2,6)
min(4,-12, 2.6).max(4,-12, 2,6)
[-12,61 1, 2, 3,4, 5, 6:
-9,-8, -7, 6, -5, 4, -3,-2, -1, 0,
12,-11, -10,

5,9. INTERVAL OPERATIONS IN N the set of alll non-nega

for intervals in N, non-Tez
multiplication
Thearithmetic operations addition and
integers
N- 10, 1, 2, 3, . , Nc Z,
subtraction is not defined in N since 6.
the those for intervals in Z. The operation
same reason the image of an interel
are same as
to N. For the
resulting number may belong to Z but not
not defined (Article 5.3.9).
Ex. 10. Given the intervals in N
A =
[1, 4] =
{1,2,3, 4}, B =[3, 5] =
{3, 4, 5}, then find A+ B and AB.
Sol A+ B = [1, 4] + [3, 5]
= [4, 9]

14,5, 6, 7, 8, 9},
AB = [1, 4]13, 5]

-
min{3,5, 12,20},max(3,5, 12, 20}
= [3, 20]

{3, 4, 5, 6, 7,8,9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20).
5.10. NOTATION
To make the notation of intervals smaller in formulae, brackets can be used. So we can s
= , ] to represent an interval. For the set of all finite intervals, we can use
(R: 4,x]:x sx, and z,x, ¬RU]-o, l}
as an abbreviation. For a vector of
intervals ((1, E].)E [RI" we can also use abold font P
..,
In such a compact notation, we
should note that [x] should not be confused berweeu a s
called improper or single point interval
[x, x] and the lower and upper limit.
5.11. ELEMENTARY FUNCTIONS
Interval methods can also
apply to functions which do not ndwr
must also use other basic functions for just use simple arithmetc, i
properties. redefining intervals, using already known mo
For monotonic functions in one
If f:R
variable, the range of values is also easy.
R is
the interval y1, 2 ¬ monotonically risingg or falling in the interval x, x2], then all values
[¥j, *2) such that y, Sy,, one of the for ai
following inequalities apple
So)
The range corresponding to thess), fo)2fo).
or
interval y,
function to the endpoints y, and
y,
y,]sk,xlcan be calculated by pP Pplying
 mterval Arthmei

s )=[min{s).s(v}.max{f(n)»s(v))}
this
From this the following basic features for
interval functions can easily be defined:
Donential Function: a " = | a ,a" ], for a >I
LOgarithm: los. {l4*2l)=|log, , log, 21
forpositive intervals {x, *] and a >1,

0dd Powers: T=|*]} for odd n e N.

Lar even powers, the range of values being considered is important, and needs to be dealt
with before doing any multyplication. For example x" for x e [-1, 1] should produce the interval
r0, 1] when n = 2, 4, 6, But i f - 1 , 1l' is taken by applying interval multiplication of form

H.1 -1, 1] then the result will appear to be [-1, 1], wider than necessary.

5.12. LINEAR INTERVAL SYSTEMS

A inear interval system consists of a matrix interval extension [4] e [R" * " and an interval

vector [b]e [RY'. We want the smallest cuboid [x] e [R", for all vectors x e R" which there
is a pair (4, b) with A e [4] and b e [b] satisfying
Ax = b.

be such interval vector [x],

quadratic systems, in other words, for n m, there
=
can an
For
interval Gauss method. This replaces
which covers all possible solutions, found simply with the
known as Gaussian elimination becomes
the numerical operations, in that the linear algebra method
the interval entities [A] and [b] repeatedlyy
its interval version. However, since this method uses
Hence using the result of the
n the calculation,it can produce poor results for some problems. the entire
estimates, since although it contains
nterval-valued Gauss only provides first rough
solution set, it also has a large area outside it.
interval version of the Gauss-Seidel method.
A rough solution [x] be improved by an
can often
the linear equation
The motivation for this is that the i" row of
the interval extension of

therefore simultaneously

] and x
e-
e

Sowe can now replace bY

aji
Since the procedure is more diagonally
efficient for a
the vector [x] by each element.
SO =

1b], one can often try multiplying it by an

matrix, instead of the system A]
*
ninant
resulting matrix equation
Ppropriate rational matrix [M] with the
(M-[4)x= M-[b]
M=A" for the central matrix A e [A], then M-[4] is
e t to solve.
If one chooses, for example,
outer extension of the identity matrix.
 FuTzy Set Theory, Fuzzy Logic and Their Applicattons
EXERCISE 5
1. For the interval numbers
(a12, 71 (b)-6,- 1] (c) [4, 16] ()-3,91
ind the width, magnitude, image and inverse (if any).
and multiplication for.
associative operations for addition
* * V e y the commutative and

(a) A B =[0, 5]. C =[3, 5]

[-2, 1],
6) A [, 4], B [- 3, 1]. C - -1, 4.
3. Verify the distributive law for
[- 1, 2], C [2, 4]
=

(a) A =
[3, 3], B =

(6) A [ , 3], B [2, 4], C =[2, 3]1.

5], D =
13, 7] verify the inclusin
4. For the intervals A =
[2, 3], B [1, 6], C = [4, sion
monotonicity properties.
numbers:
5. Find the distance between the interval
B [5, 7
A -2, 3], B [2, 5]
i) A =
[1, 4), =

( A = [0, 8], B = [3, 6].

6, For the interval numbers in Z:

A = - 3 , - 1] = {-3, - 2, - 1, 0, 1, 2},
.
B [1,2]= {1, 2),
C [2,4]- {2, 3, 4),
carry out the following arithmetic operations:
Aa) A + B 6) B - A B-C d) A B
A ) A-C DAB + AC (g)AB + C) ( h ) A -A +B-B

ANSWERS
1. a) 5,7,-7, -2]. 6) 5,-1, [, 6),--
()12, 16, -16, -4], (4) 12, 9, [-9, 3], Not possible.
3. (a) Valid
(b) Valid.
5. (a) 4
(b) 4 (c) 3.
6. (a) -2, 1] {-2, -1, 0,
=

1} (6) [2, 5] =
{2, 3, 4, 53
(e)-3, 0] = {-3,-2, -1, 0} {-6, -5, -4, -3, -2,
(a) -6, -1]
=

(e) -12,-2] {-12,

, -2} D -18, -3] (-18, . , 3}1, 2,
=
..

(8) -18,-3] ={-18, -1, 0, 3

...
,
-3} ) B,3] 3 , -2, =