Tank AY-101 wall thicknesses have been measured using ultrasonic (UT) images. I
nterest is in using the available data to estimate a worst-case minimum wall thi
ckness for regions of the tank that remain unexamined with the UT approach. Eac
h UT image can provide the wall thickness at a very large number of pixels, but
only the minimum value for each image is used in this work. Since the data aris
e in this manner, one of several candidate statistical extreme value distributio
ns should successfully fit these reported minimum values. Given a set of multipl
e UT images and the resulting minimum measured wall thicknesses, extreme value d
istributions are fit to the data with subsequent extrapolations made to estimate
the minimum measured wall thickness expected for the entire tank. Such a min
imum estimate incorporates both the variability in wall thicknesses and the uncertainty associated with the measurement method. Uncertainties of the estimate
d parameters are also derived and used in propagation of variance methods to obt
ain confidence bounds on the estimated minimum measured wall thickness as well.
Revised: November 11, 2002 |
Published: October 31, 2002