VALVE MEASUREMENT USING PHOTOGRAMMETRIC FEATURE
MODELING
Alan Voss
Tennessee Valley Authority
Chattanooga TN
USA
James Bethel
Purdue University
School of Civil Engineering
West Lafayette IN
USA
Key Words: Photogrammetry, Close-range, Features, Mirrors, Reflection, Valve, Nuclear Power
ABSTRACT
The assessment of critical components in nuclear power
plants is an important problem which presents some
unique challenges. For photogrammetric processing, the
difficulties relate to the lack of point features for use in a
conventional bundle adjustment. Viewing angles can be
unfavorable when looking inside a valve. Likewise illumi-
nation and reflections cause interpretability problems in
the photographs. We have chosen to use feature modeling
to address the lack of points. We have chosen mirror re-
flections and corresponding mathematical modeling to ad-
dress the view angle problem. Current research is address-
ing the illumination and interpretation problems. Results
to date relating to implementing the desired mathematical
models has been successful. Experimental results in actu-
ally producing valve measurements have so far exhibited
unacceptable discrepancies compared to manual measure-
ments. These discrepancies are being investigated.
1 INTRODUCTION
In 1989 the NRC, Nuclear Regulatory Commission (USA),
issued Generic Letter (GL) 89-10, Safety-Related Motor-
Operated Valve Testing and Surveillance. This required
that each nuclear utility develop a program to ensure that
the switch settings on all safety-related motor-operated
valves (MOV) are selected, set, and maintained in such
a way as to ensure that the MOV will operate under the
design basis conditions. GL 89-10 recommends that, when
practicable, all safety-related MOV’s be tested in-situ at
design basis conditions to demonstrate their capability to
function. In-situ testing at design basis conditions is diffi-
cult and not practicable in many cases.
The Electric Power Research Institute, EPRI, is cur-
rently conducting a comprehensive MOV research pro-
gram (MOV Performance Prediction Program) with the
objective of providing nuclear utilities with an analytical
method of predicting MOV performance under design ba-
sis conditions. This program includes a combination of
analysis and tests that address relevant aspects of valve
and operator performance. The methodology developed
by the EPRI MOV Performance Prediction Program will
require information on gate valve internal design features,
materials, and dimensions in order to predict valve perfor-
mance.
These dimensional measurements of the valve internals
are primarily made manually on the valve itself. This is
44
done using scales, calipers, micrometers, and verniers. The
problems associated with these measurement techniques
are chiefly associated with the contamination hazard from
close and prolonged contact with valve components. Per-
sonnel who have exceeded their limit for radiation expo-
sure are no longer permitted to perform these tasks. Be-
cause of these problems with the manual measurement
technique, it appeared that photogrammetry offered the
possibility for reduction in both time and proximity com-
pared to the manual method. Since only photographs
would have to be taken at the valve location, with the
actual measurement done later in a clean environment,
this should reduce the time of exposure. Likewise, the
excessive hand contact required by manual measurement
should be reduced with most of the activity involved with
the photography taking place at a distance of about one
meter. Another advantage to using photogrammetry is
that additional measurements can be made at a later time
without going to or disrupting the plant to disassemble
the valve. Also the photographs will provide a permanent,
archival record for comparison with later inspections to de-
termine if conditions are changing. There have also been
some problems associated with reliability of the manual
measurements. While photogrammetry does not neces-
sarily enhance this reliability, it does offer the possibility
for multiple observers to make duplicate or repeated sets
of observations in any cases where discrepancies or unex-
pected results are obtained. Cost has also been suggested
as a factor where photogrammetry may potentially pro-
vide an advantage. Certainly in other spheres, i.e. topo-
graphic mapping, antenna measurement, etc., photogram-
metric techniques have been accepted as preferable in cost
and performance to strictly manual techniques.
The desired accuracy of measurements for this pur-
pose has been stated as 0.006 inches (0.15 mm). From
larger distances, point accuracies have been reported as
a fraction of the object distance in the range of 1 part
in 20,000 to 1 part in 100,000. At a distance of one me-
ter, 1 part in 20,000 would be 0.05 mm which is about
one-third of the desired level. Thus it seems plausible
that this technique might be usable for the proposed task.
In many other close-range photogrammetric problems, the
position determination is focused on discrete, well-defined
points in space. For the present problem, however, the
dimensions required are such things as distance between
valve guide faces, radius of an edge on a beveled valve
seat, etc. These features do not lend themselves well to
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
traditional "p
moreover thei
locating the f
trast, poor ill
ing angles.
Thus ther
also some pot
racies attaina
surement task
2 FEAT
In this applic
range photogr
defined image
or control poi
hand, there a
in the images.
straight lines,
oriented in sy
for such linea
(1988, 1989) :
In our cas
of most inter
operated valv
linear feature:
cannot obtain
the feature. I
tifying charac
Fortunately w
it is only nece
feature, there
the case of str
equation wou
photograph.
In this equat:
the observed
ject space con
vector from t
the line and «
linear feature:
written for ea
where 1 is the
the center of |
plane, p is th
and r is the «
a valve seat v
As evident in
lack of well d:
able viewing
the construct
of the featur:
In order to se
to introduce
be positioned
