Quicker, more effective methods of corrosion prediction and classification
will help ensure a safe and operational transportation system for both civilian
and military sectors. This is especially critical now as transportation
providers attempt to meet the increased expense of repairing aging aircraft with
smaller budgets. These budget constraints make it imperative to find corrosion
and to correctly determine the appropriate time to replace corroded parts. If
the part is replaced too soon, the result is wasted resources. However, if the
part is not replaced soon enough, it could cause a catastrophic accident. The
discovery of models that limit the possibility of a costly accident while
optimizing resource utilization would allow transportation providers to
efficiently focus their maintenance efforts. While our concern in this study was
with aircraft, the results will also be useful to other transportation
providers. This paper describes the complexity challenges in the discovery and
comparison of empirical models to predict corrosion damage from non-destructive
test (NDT) data. The NDT data derive from eddy current (EC) scans of the United
States Air Force's (USAF) KC-135 aircraft. While we might suspect a link between
NDT results and corrosion, up until now this link has not been formally
established. Instead, the NDT data have been converted into false color images
that are analyzed visually by maintenance operators. The models we discovered
are quite complex and suggest that in data mining we can sometimes more
effectively handle noisy data through more complex models rather than simpler
ones. Our results also show that while a variety of modeling techniques can
predict corrosion with reasonable accuracy, regression trees are particularly
effective in modeling the complex relationships between the eddy current
measurements and the actual amount of corrosion. In particular, we show that the
nonlinearities in the relationship between NDT data and corrosion can be handled
with the segmentation approach used in regression trees. This approach seems to
work even better than smoother approximations to the nonlinear structure of the
data.
Copyright. All Rights Reserved.