Sharp JX-9400 Informations techniques Page 174

  • Télécharger
  • Ajouter à mon manuel
  • Imprimer
  • Page
    / 211
  • Table des matières
  • MARQUE LIVRES
  • Noté. / 5. Basé sur avis des utilisateurs
Vue de la page 173
the a posteriori information. T his new distributio n:
ðzÞ¼C exp
1
2
fðzÞ
T
C
1
T
ðzÞþðz z
p
Þ
T
C
1
z
ðz z
p
Þg
hi
ð7:41Þ
From this distribution, the z vector presenting the maximum likelihood can be
found. It is the vector, z, that minimizes the exponent:
ðzÞ
T
C
1
T
ðzÞþðz z
p
Þ
T
C
1
z
ðz z
p
Þð7:42Þ
This most probable vector contains the identified model parameters and the
most probable values of the measured quantities. Practically, this vector is
found using numerical methods looking for the minimum of the exponent
given above. More references on such methods are Mitchell and Kaplan
(1969) and Nelder and Mead (1965).
Error analysis
This method allows us to obtain the a posteriori estimate, C
i
z
, of the co-
variance matrix of the distribution, (z). For that purpose, the model, (z), is
linearized around the most proba ble vector, z
s
. The a posteriori covariance
matrix is then:
C
i
z
¼ðF
T
s
C
1
T
F
s
þ C
1
z
Þ
1
¼ C
z
C
z
F
T
s
ðF
s
C
z
F
T
s
þ C
T
Þ
1
F
s
C
z
ð7:43Þ
where F
s
is a matrix having the dimensi on N M, with M ¼ Nn þ n þ N, N
being the number of measurements and n the num ber of parameters to be
identified. F
s
contains the derivatives of the model, (z), evaluated at the
point, z
s
:
F
s
¼
@
1
@z
1

@
1
@z
M
  
@
n
@z
1

@
N
@z
M
0
B
B
B
@
1
C
C
C
A
z
s
ð7:44Þ
Error analysis
Purpose of the error analysis
The accuracy of any measurement depends on the conditions in which the
measurement is done, on the quality of the measuring instrument and on the
skill of the people making the measurement. Measurements canno t be perfect,
accuracy cannot be infinite, and any measurement result includes some uncer-
tainty. That means that the result is not absolute, but it is always possible to
state that the actual value is contained, with a given probability, within some
confidence limits, or vice versa, that the probability that the actual value is
outside some limits is lower than a certain value. Since this confidence interval
may be large, there is no sense in giving the result of a measurement without
any information on its reliability.
Common Methods and Techniques 153
Vue de la page 173
1 2 ... 169 170 171 172 173 174 175 176 177 178 179 ... 210 211

Commentaires sur ces manuels

Pas de commentaire