Sharp JX-9400 Technical Information Seite 182

  • Herunterladen
  • Zu meinen Handbüchern hinzufügen
  • Drucken
  • Seite
    / 211
  • Inhaltsverzeichnis
  • LESEZEICHEN
  • Bewertet. / 5. Basierend auf Kundenbewertungen
Seitenansicht 181
represented by a vector x and a matrix A. The question is: which is the
resulting error y on the vector y, which is the vector containing the final results?
If the matrix A and the vector x were known, we could write:
ðA þ AÞðy þ yÞ¼x þ x ð7:72Þ
and, taking Equation 7.71 into account, we could solve:
y ¼ðA þ AÞ
1
ðx AyÞð7:73Þ
This equ ation can be used many times in a Monte-Carlo process, varying each
time all the components of A and x at random, according to their probability
density function. This provides a series of vectors y from which an estimate
of the probability density functions of the components can be calculated.
However, this procedure is time consuming and, assuming a normal distribu-
tion of the measurement methods, simpler methods are available, which are
described next.
Complete error analysis
The requested final result is calculated using:
y ¼ A
1
x hence y
i
¼
X
j
ij
x
j
ð7:74Þ
where the coefficients
ij
are those of the inverse matrix A
1
. The error
calculated with the most simple (or the differential) method will then be:
y
i
¼
X
k
@y
i
@x
k
x
k
þ
X
k;l
@y
i
@a
kl
a
kl
¼
X
k
j
ik
x
k
X
k;l
@
ij
@a
kl
x
j
a
kl
ð7:75Þ
But since
@
ij
@a
kl
¼
ik
lj
ð7:76Þ
we get finally:
y
i
¼
X
k
j
ik
x
k
X
kl
X
j
ik
jl
x
j
kl
ð7:77Þ
If the variances and covariances s
2
xk;xl
, s
2
aki;amn
and s
2
aki;x
m
are known, the
covariance of the results s
2
yi; yj
is well estimated using a first order Taylor’s
expansion (Bevinton, 1969). We get:
s
2
y
i
; y
j
¼
X
klmn
@y
i
@a
kl
@y
j
@a
mn
s
2
a
kl
a
mn
þ
X
kl
@y
i
@x
k
@y
j
@x
j
s
2
x
k
x
l
þ
X
klmn
@y
i
@a
kl
@y
j
@x
m
þ
@y
j
@a
kl
@y
i
@x
m
s
2
a
kl
x
m
ð7:78Þ
Common Methods and Techniques 161
Seitenansicht 181
1 2 ... 177 178 179 180 181 182 183 184 185 186 187 ... 210 211

Kommentare zu diesen Handbüchern

Keine Kommentare