Sharp JX-9400 Technical Information Seite 172

  • Herunterladen
  • Zu meinen Handbüchern hinzufügen
  • Drucken
  • Seite
    / 211
  • Inhaltsverzeichnis
  • LESEZEICHEN
  • Bewertet. / 5. Basierend auf Kundenbewertungen
Seitenansicht 171
where "
y
and "
x
are the experimental errors on y and x respectively and
deduce a corr esponding value of a using Equation 7.16. This recipe does not
show clearly which quantity is minimized by the fit.
Another more physical way is so-called ‘orthogonal’ regression. It mini-
mizes the real (orthogonal) distance between the measured points and the
regression line drawn with the scales on the axes inversely proportional to
the experimental errors, using variables weighted by the experimental errors:
Y
i
¼
y
i
"
y
and X
i
¼
x
i
"
x
ð7:29Þ
Writing the regression line with these coordinates:
Y ¼ A þ X ð7:30Þ
the minimized residual is:
SSR
?
¼
X
N
i ¼1
ðY
i
X
i
AÞ
2
2
þ 1
ð7:31Þ
The slope in original (x, y) coordinates is given by the following relations. Let
us define:
n ¼
s
2
y
s
2
x
2s
xy
ð7:32Þ
If
x
and
y
are respectively the estimated standard deviations of the abscissas,
x
i
, and ordinates, y
i
, the slope is given by:
n
?
¼
x
y
n
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
2
þ 1
p

ð7:33Þ
Between the two possibilities, the sign of n
?
must be chosen equal to the sign of
n given by Equation 7.16. T his sign is always positive in pressurization tests.
The coefficient, a, is obtained with the relation of Equation 7.16.
Bayesian identification
In section ‘Linear least square fit’, above, it was said that the usual regression
techniques for the identification of the coefficients can strictly be used only
when one of the variables is well controlled. In this case only, the relations
given in ‘Confidence in the coefficients’, can be used to get a good estimate of
the confidence intervals of the obtained coefficients.
When there are uncertainties on both axes, for example in pressurization
tests, these methods are not strictly valid, since they do not give any infor-
mation on the relation between the two coefficients and their uncertainties. If
several measurements of the same leak are perfor med, several pairs of
coefficients, C and n, for the relatio n:
Q ¼ C p
n
ð7:34Þ
Common Methods and Techniques 151
Seitenansicht 171
1 2 ... 167 168 169 170 171 172 173 174 175 176 177 ... 210 211

Kommentare zu diesen Handbüchern

Keine Kommentare