Information Technology Reference
In-Depth Information
As memberships add one, without loss of generality, the polynomials p i may be
assumed to be homogeneous in
μ
, i.e. , composed only of monomials whose degree
in the variables
μ
is the same. Indeed, any monomial in p i can be multiplied by
( i μ i )
q , for any q , in order to incorporate as many powers of
μ
as required to
make p i homogeneous (Sala and Ariño 2007 ).
The sector-nonlinearity methodology in Tanaka andWang ( 2001 ) can be extended
to a polynomial case. A first idea is bounding by constants the non-polynomial non-
linearities (Tanaka et al. 2009b ). But there exists the possibility to bound a nonlin-
earity within two polynomials (instead of two constants ): the result would still be a
polynomial fuzzy system.
Lemma 5.1 (Chesi 2009 ; Sala and Ariño 2009 ) Consider a sufficiently smooth func-
tion of one real variable, f
, so that its Taylor expansion of degree “n” exists, i.e.,
there exists an intermediate point
(
x
)
ψ(
x
) ∈[
0
,
x
]
, so that:
n
1
f [ i ] (
f [ n ] (ψ(
0
)
x
))
x i
x n
f
(
x
) =
+
(5.4)
i
!
n
!
i
=
0
where f [ i ] (
denotes the i th derivative of f and f [ 0 ] (
x
)
x
)
is defined, plainly, as f
(
x
)
.
Assume also that f [ n ] (
x
)
is continuous in a compact region of interest
. Then, an
equivalent fuzzy representation exists in the form:
f
(
x
) = μ 1 (
x
) ·
p 1 (
x
) + μ 2 (
x
) ·
p 2 (
x
)
x
(5.5)
μ 1 (
) + μ 2 (
) =
1 and p 1 (
)
,p 2 (
)
where
x
x
x
x
are polynomials of degree n.
Proof Denote the Taylor approximations as:
n
1
f [ i ] (
0
)
x i
f n (
x
) :=
(5.6)
(
i
) !
i
=
0
For later developments, denoting f 0 (
x
) =
0, denote
f [ n ] (ψ(
f
(
x
)
f n (
x
)
x
))
T n (
x
) :=
=
(5.7)
x n
n
!
x n . In the region of interest
so the Taylor remainder is f
(
x
)
f n (
x
) =
T n (
x
)
,
is bounded because f [ n ] (
T n (
)
)
x
x
is continuous in
, as assumed in the lemma.
Denoting:
ψ 1 :=
sup
x
T n (
x
),
ψ 2 :=
inf
x
T n (
x
),
(5.8)
we may write T n (
x
) = (μ(
x
1 + (
1
μ(
x
))ψ 2 )
with:
Search WWH ::




Custom Search