The following example presents the results of an evaluation of an eddy current inspection for corrosion damage in C/KC-135 lap joints taken from Hoppe, et al. [2000].  For the example, the corrosion damage metric was taken to be thickness loss as thickness loss is an important criteria in judging severity of corrosion damage and eddy current is sensitive to thickness loss in the top layer of the lap joint.

Both real and engineered specimens were used for the capability demonstration.  Several pieces from C/KC-135 and Boeing 707 fuselages were acquired.  The specimens represented areas of interest on the aircraft and were expected to contain representative amounts of crevice corrosion.  The specimens included fuselage lap joint and doubler sections that were anticipated to contain corrosion, as determined by disassembly of adjacent pieces of the skin.  The specimens also included areas of little or no corrosion.  The specimens that were selected incorporated the type, material, size and spacing of fasteners, thickness and lay-up of skins, presence of substructure, and specimen curvature variability that were expected to be experienced in typical aircraft inspections.

An engineered specimen was designed and manufactured for measuring the spatial resolution of the eddy current system.  Spatial resolution of the system was necessary in to order to ascertain inspection regions of complete independence of the eddy current response.  This specimen was constructed with several sets of lines of different widths machined in to the back surface of the front layer of an assembly of aluminum layers.

Specimens of a skin configuration were inspected using the eddy current system.  NDI responses were recorded at independent sites within each specimen producing an inspection output profile of the specimen.  Because thickness loss due to corrosion is variable within a specimen, the responses at the independent sites represent different samples of response at different thickness losses.  The process is illustrated in Figure 3.1.8.  The eddy current output at a point, P, in a response image is a function of the corrosion in a small region (or cell), C, on the specimen.  The set of non-overlapping cells represents the collection of independent inspection opportunities from which probability of detection as a function of thickness loss can be calculated.

Figure 3.1.8.  Schematic Diagram of Specimen and Inspection Output Images

After completion of the inspection of a specimen, the actual corrosion profile of the specimen was determined.  The specimens were carefully disassembled by drilling out the fasteners and prying apart the layers.  Corrosion products were chemically removed using a high concentration nitric acid exposure protocol.  Measurement of remaining thickness was accomplished using calibrated topographic radiography.  The inspection system output images and actual thickness loss profiles were registered to specimen features, such as fasteners and lap joint edges, in order to relate measured to actual thickness loss across each specimen.

Data pairs of real and EC measured thickness loss were generated for the independent inspection cells.  The data pairs are plotted analogously to the â versus a plot of crack detection POD estimation.  Figure 3.1.9 is an example of thickness loss versus EC response for one of the structural configurations.  The scatter of the EC responses about the mean trend determines the POD as a function of thickness loss.  Figure 3.1.10 shows the POD function for a threshold chosen to yield 90 percent detection for a 10 percent thickness loss.  Also shown in Figure 3.1.10 is the 95 percent confidence bound on the POD function.

Figure 3.1.9.  Example Eddy Current Response for Cells of Different Thickness Loss

Figure 3.1.10.  POD versus Percent Thickness Loss with Response Detection Threshold Set to Yield POD of 90 percent at 10 percent Loss