Page 439 - Pipeline Pigging Technology

Pipeline Pigging Technology

For instance, a good weld at one point in a pipeline can produce a very
different image from an equally-good weld at a different point on the same
pipeline. More sophisticated techniques have had to be used.
Possible faults are analysed using pattern-recognition and image-process-
ing techniques similar to those employed in medical scanning and satellite
imaging. Such techniques, originally developed for purposes like enhancing
blurred photographs, or teaching computers to recognize particular words,
are equally relevant to the interpretation of pipeline inspection data. Instead
of a blurred photograph, the on-line inspection device provides a record of
magnetic field variations in the pipeline; its sharpness is limited by the
response of sensors and electronics and the errors introduced during data
collection in the harsh conditions inside a pipeline.
British Gas has modified and developed existing techniques to cope with
the problems posed by pipeline inspection. The general approach has been
to measure various parameters to characterize a signal and then to use
statistical techniques to discriminate between significant and spurious data.
Much depends on choosing the appropriate image parameters to measure.
The experience of engineers who design and operate inspection vehicles has
proved invaluable for this purpose.
The data-reduction techniques employed are designed to operate in a
cascade fashion, so that only the simplest operations are applied to the bulk
of the inspection data, more complex steps being reserved for later stages in
the analysis sequence. Using various software tools, the operator may search
for particular types of feature, manipulate images on graphics terminals, and
test new signal-processing algorithms to identify any misclassification errors.
These techniques have been developed at the On-Line Inspection Centre and
by leading consultancy organizations working under contract.
The procedure may be modified when dealing with data from seamless
pipe in which the method of manufacture produces large variations in wall
thickness (often outside specified tolerance limits) over quite small areas of
pipe. In addition, the amount of metal-working associated with the forging
process also produces significant variations in the material's magnetic char-
acteristics. Such wall-thickness and magnetic variations are detected by
magnetic-flux leakage inspection vehicles, and can obscure or distort signals
from potential defects. A special de-blurring process has been developed by
British Gas which enables the "natural" variation in response to be recognized
and eliminated without distorting the signals from metal-loss defects. The end
product is corrected data which looks like that obtained from pipes manufac-
tured from controlled-rolled plate.

420