The use of eddy current (EC) arrays to detect damage in sandwich panels, such as disbonding of the carbon fiber reinforced polymer (CFRP) face-sheet to the core, is investigated. It is shown that the array is very sensitive to slight core crush and can readily find small dents and disbonds. At the same time, the eddy current array can look much deeper into the honeycomb to detect defects such as tears. The phase map of the EC signal can be used in some cases to distinguish between different types of damage. EC arrays offer the ability to rapidly scan large areas of CFRP panels.
Composite honeycomb panels have become widespread, especially in the aircraft industry, because of their excellent weight-to-stiffness ratio. There is concern, however, about the ability to detect barely visible impact damage (BVID). With aluminum face-sheets, impacts that have sufficient force to crumple the underlying honeycomb core typically also result in a permanent dent in the aluminum skin. Slight impacts normally do not result in the separation of the skin from the underlying honeycomb. With the carbon fiber reinforced polymer (CFRP) composite face-sheets, separation of the face-sheet from the underlying core has been observed to happen. In these cases, there is little or no sign of damage on the CFRP face-sheet, but the strength and stiffness of the panel has been compromised because the face-sheet is no longer attached to the core. There is further concern that what started out as small invisible disbonds may grow under fatigue loading into large disbonds, which lead to failure of the part.
A number of non-destructive testing (NDT) techniques have been used to investigate the integrity of composite face-sheet sandwich panels. One of the earliest of these was the tap test [1–5] in which a coin was dropped on the panel and the inspector listened to the sound to determine whether a disbond was present. This approach has been improved upon and built into several commercial instruments, including the woodpecker (Mitsui Heavy Industries), the rapid damage detection device (Boeing), and the computer aided tap tester (Iowa State University) . More sophisticated approaches, including the use of neural nets  and cluster analysis , have been used for analyzing the response of such instruments. In a review of the tap test for defect detection of sandwich panels, the Federal Aeronautics Administration (FAA) found that the results were highly variable .
Ultrasonic inspection of sandwich panels is common (e.g., Refs. [10–18]), and a very basic American Society for Testing and Materials (ASTM) standard practice guide (ASTM E2580) for inspection of flat panels has been developed. Guided wave inspection appears to be particularly suitable for detecting disbonds in BVID, partly because of its ability to inspect large areas at once. Major issues with ultrasonic inspection are that it requires skilled operators to set up and run the equipment and to interpret the data, and can be relatively slow when compared to methods like thermography .
Thermography is a rapidly evolving NDT technique that shows great promise in the detection of defects in sandwich panels with composite face-sheets. There are a wide range of thermographic techniques available that vary in how heat is generated and how the signal is analyzed. Yang and He  have published a comprehensive review of the various thermographic techniques as applied to composite panels. Duan et al.  compared pulsed thermography to ultrasonic C-scan of panels and found that flash thermography could actually detect a smaller defect size at 90% probability of detection (POD). A major drawback in thermography is the requirement for an expensive infrared camera. However, Strugala et al.  have shown that equivalent results can be obtained using thermochromic sheets and a conventional camera.
A variety of other techniques have also been proposed for determining damage in sandwich structures including shearography , vibro-acoustic modulation , and pulsed eddy current scanning . Interestingly, conventional eddy current has only been considered for aluminum face-sheets . In this case, for sandwich panels with aluminum honeycomb cores, the advent of eddy current arrays offers the possibility of rapid scanning of large areas to detect dents and disbonds using the lift-off signal. This paper examines the results obtained using conventional eddy current array probes on CFRP sandwich panels with aluminum honeycomb cores. The results are compared with flash thermography on a number of CFRP sandwich panels with BVID. The advantages and disadvantages of the two techniques are discussed and their suitability for rapidly scanning large panels to find and characterize BVID is examined.
2.1 Panel Construction.
Four specimens were constructed to examine the effectiveness of the eddy current arrays for inspection. All of these were based on a 12.5 mm (0.5 in.) thick 5052 aluminum honeycomb core with a cell size of 3.2 mm (1/8 in.) and a wall thickness of 0.02 mm (0.0007 in.). Construction of the honeycomb meant that in one direction, two walls were bonded together producing a wall of double thickness.
In the first of these specimens, six impacts were made on the bare honeycomb using 51 mm (2 in.) diameter spherical impactors. The results are shown in Fig. 2 and approximate dimensions of the indents are given in Table 1. A 1.4 mm thick sheet of cardboard was used as a face-sheet.
Furthermore, three 150 mm × 150 mm samples were constructed using CFRP face-sheets. The CFRP panels were laid up in 0/90 configurations of a unidirectional material. The number of plies and final thickness are described in Table 2.
Panels 1 and 2 were impacted at various energies with either a 25 mm (1 in.) or 51 mm (2 in.) indenter to make a number of small defects, as shown in Fig. 1. Panel 3 had indents made into the core of side 1 using a 63 mm (2.5 in.) indenter before the panel was assembled. It then had several further impacts made upon it. These three panels, therefore, contained a variety of dents, disbonds, and dents with disbonds in them.
Following construction and defect production on both sides, all three panels were scanned with a laser scanner on a Faro arm to produce a map of the physical dents in the surface. Figure 2 shows an example of one such map.
In addition, a thermographic image of each side was taken using a FLIR T620 thermal camera and a flash heating system . The thermographic image shows hot spots where the aluminum honeycomb core has become disbonded from the CFRP face-sheet. Figure 3 shows a thermographic image of the same specimen as in Fig. 2. It can be seen that the brightness of the spot in the bottom left, which corresponds to the largest dent in Fig. 2, is relatively weak. However, the two bright spots on the right of Fig. 3 do not even appear in Fig. 2. This is because the face-sheet in these locations has sprung back after becoming separated from the core.
In addition to the constructed samples, some scans were also taken on an ultrasonic standard, 76301-74D111295-100X, where X = 1 for a 76 mm (3 in.) thick sample, X = 3 for a 25.4 mm (1 in.) thick sample, and X = 5 for a 13 mm (1/2 in.) thick sample. The face-sheet on all three standards was five plies of 5 mil (0.2 mm) CFRP, and the adhesive to hold it together was FM300. All three standards had four nominal disbonds, one with a diameter of 38.1 mm (1.5 in.) and the other with a diameter of 25.4 mm (1.0 in.). The panels were divided into half with one side having a cell with a 4.8 mm (3/16 in.) cell size (3.1 lbs/sq ft.) and the other half having a cell with a 6.4 mm (1/4 in.) cell size (2.3 lbs/sq ft.). Each half had one of each size of disbond. It is not known whether the disbond areas were created by crushing of the core or by scalping the core before the panels were assembled. X-ray images of the panels suggest the latter .
2.2 Eddy Current Probes.
Eddy current array probes have the potential to quickly scan large areas of composite panels to find damage. The probes detect the aluminum honeycomb beneath the composite face-sheet and measure the distance (lift-off) from the probe to the honeycomb. If the honeycomb has been damaged, so that the core is crushed, then the probe-to-honeycomb distance is increased and this will show up in a C-scan image even if the composite surface has sprung back.
Two commercial eddy current probes from Olympus were used in conjunction with an Omniscan MX system to investigate the performance of the arrays for this application. The first was an Olympus SAB-067-005-032, which is a transmit–receive probe (a drive coil transmits a signal which a remote pick-up coil receives ) operating at 40 kHz or less. It has two banks of coils arranged in a double row as shown in Fig. 4. The array covers 67 mm (2.625 in.) of width as it scans.
A second array probe (FBB-051-150-032) was operated in an absolute bridge mode (measure changes in the impedance of a single drive coil ) over a frequency range of 100–1500 kHz. The coils in this probe were much smaller, having a diameter of approximately 9 mm. They were arranged in one bank of 32 coils in a close pack arrangement similar to that shown in Fig. 4. The array covers 51 mm (2 in.) as it scans.
As no reflection-type array (like transmit–receive but the pick-up is co-axial with the transmit coil )was available, the performance of a reflection-type array was simulated using a reflection probe (Olympus 9222199.01) mounted on the arm of a TecScan robotic system, which translated the probe and recorded both the horizontal and vertical components of the eddy current signal. The reflection probe had a diameter of 11.2 mm (the actual coil sizes are unknown). It was connected to an Olympus Nortec 600D eddy current system. The signal was set so that lift-off was vertical, and the gain was adjusted so that 1 mm of lift-off corresponded to 4 V (divisions). The probe was operated at three different frequencies, 10, 40, and 160 kHz. All gave very similar results but the signal-to-noise ratio improved slightly as the frequency increased. Since the probe had a nominal top frequency of 40 kHz (as given by the manufacturer) and the improvement in going to 160 kHz was minimal, that frequency was used in the results presented here. The TecScan scanner was adjusted to scan an area of 132 mm × 132 mm in the center of the 150 mm × 150 mm panels to keep the probe away from the edges of the panel. The raster scan took readings every 2 mm along a scan line and separated scan lines by 2 mm, producing a 66 × 66 grid of data. The height of the probe was approximately 0.25 mm above the face-sheet. Because the scanner and the face-sheet were not exactly parallel, the signal showed some side-to-side drift in some cases. This would not happen with an actual array probe as the coils would then be surface-riding.
Figure 5 shows the results from sample 3 side 2. The C-scan results are divided into four panels (d–g). The color for each panel has been adjusted so that the data fill the range. The top left panel (d) displays the horizontal signal. If the signal were pure lift-off, this panel would be blank, as the probe is set up to make lift-off strictly vertical. The bottom left panel (f) is the vertical signal and would be the signal displayed in a single C-scan on an instrument such as the Olympus Omniscan. The top right panel (e) shows the magnitude of the signal. Finally, the bottom right panel shows the phase of the signal (in an impedance plane display, phase is the angle relative to the horizontal voltage, conventionally defined as lift-off ). The three large indications down the left side of the C-scans are horizontal slits cut into the honeycomb at different depths from the surface. The phase coloring of these is very distinct from that of the round impact craters distributed throughout the images. The two long narrow indications in the bottom right of the images are from vertical slits that were made in the honeycomb before the panels were glued together. Again the phase indication is different from that of the impact circles.
The phase image shows that it is possible to distinguish between some kinds of defects using the eddy current data.
3 Results and Discussion
3.1 Transmit–Receive Array.
The transmit–receive array (Olympus SAB-067-005-032) was first tried on sample 1. An operating frequency of 40 kHz was used, and the lift-off signal was rotated to be strictly vertical. The results for scanning along the line of defects are shown in Fig. 6. Note that the X axis in the image is almost four times longer than the Y axis (255 mm versus 67 mm). Hence, the round defects appear oval in the image as presented. Defects 2–6 can be readily detected, given the selected color palette and range. However, when the probe was rotated by 90 deg and scanned along a line parallel to the double walls, only the largest defect could be detected.
Similar results were obtained when the probe was used to scan the ultrasonic standards. In this case, on the larger honeycomb, not only did the array show strong directionality but it also showed striping (see Fig. 7) indicating that it was being affected by the honeycomb structure. This highlighted a further problem with the arrays when the size of the coils approaches the size of the honeycomb cells.
This is discussed in more detail later (Sec. 3.3). The high directionality of the transmit–receive configuration makes it unsuitable for this application.
3.2 Absolute Bridge Array.
The absolute bridge array was scanned over several 150 mm × 150 mm sandwich panels. Typical results are shown in Fig. 8. A similar scan was obtained when the array was scanned at 90 deg, indicating that this probe did not suffer the directionality problem that the transmit–receive probe did. This was not surprising as the absolute coils have cylindrical symmetry and hence would not be expected to show a preferential direction in the plane of the honeycomb.
Because the coils were smaller than the cell size of the honeycomb, a strong texture due to the honeycomb was superimposed on the image. The strength of the signal of many of the smaller defects is barely larger than that of the honeycomb and they can be most readily seen because they have a solid color as opposed to the pattern produced by the honeycomb. The probe shows potential for this application but is not ideal for the task because of the small size of the coils. A model of the honeycomb and a single coil was constructed in comsol 5.4 to test the effect of the coil size (see Fig. 9). As can be seen, the larger coil creates a much more extended pattern of eddy currents, so the response does not vary a great deal as the coil moves over the undamaged honeycomb. On the other hand, the small coil only excites eddy currents in a few walls and the pattern changes substantially as the coil moves. This causes the smaller coil to create a much more textured C-scan. An even larger coil would give an even smoother response to translation but, of course, would show less response to damage in the honeycomb.
3.3 Reflection Probe.
Scans of sample 2 side 1, taken with the reflection probe on the TecScan arm, are shown in Fig. 10. Readings were taken every 2 mm along scan lines, which were, in turn, separated by 2 mm. Numerous interesting features are evident. There are eight round dents in the eddy current scans that correspond to impacts from a spherical indenter. According to Fig. 10(d), the deepest of these is in the lower left of the image. There are four other impacts that produce measureable indentations. The brightest points in the thermographic image are the two indents on the right and the topmost indent in the center. None of these correspond to significant indentations of the surface. Previous work  indicates that brightness in the thermographic image is associated with disbonding of the face-sheet from the core. When this happens, the face-sheet cannot loose heat by thermal conduction to the highly conductive aluminum core and hence appears brighter in the thermal image. In the eddy current phase image, all eight dents appear similar, going from orange to red to white to blue. This similarity occurs because the eddy current signal is primarily sensitive to lift-off and the core crush gives a small horizontal component that produces the phase shift. The eddy current signal does not distinguish whether the core is attached to the face-sheet or not and merely indicates that it is further away than the undamaged core and shares a similar type of damage (crushing).
The eddy current image also indicates four other features, three of which are on the left hand side of the image and one vertical narrow strip at the top right. In the phase image, these are all similar and show a transition from orange to yellow to green, and the opposite transition to the dents. These features all represent slits in the honeycomb created by a razor blade. The three slits in the side are at different depths from the face-sheet, while the vertical indication is a vertical slit, part way through the honeycomb from the opposite face. It is not surprising that the vertical slit is visible, since it would impede the flow of eddy currents, which are expected to be parallel to the face-sheet. The ability to detect the horizontal slits is attributed to the disruption of the main honeycomb structure, which is expected to support connected eddy currents to a greater depth of penetration on the inner honeycomb surfaces than a solid conductor would.
The same probe was used to scan the two “disbond” features in the standard 76301-74D111295-1005 to examine the effect of the increasing honeycomb cell size. In this case, because a smaller region was being scanned, a 1 mm step size was used and the scan lines were 1 mm apart. The results are shown in Fig. 11. As was the case with the absolute bridge probe, the structure of the honeycomb is much stronger relative to the lift-off features, because the probe covers much less of the honeycomb and, hence, does not average out the cell structure. The effect gets stronger as the ratio of probe size to cell size decreases as can be seen by observing the fluctuations in the line scans (strip chart display) to the right of the C-scan images. These line scans correspond to the vertical crosshair in the C-scan image.
In order to see what a larger diameter probe would do, an Olympus 9222160.01 reflection probe which has a 16 mm diameter was used under the same conditions as the Olympus 9222199.01 used above. The results are shown in Fig. 12. As expected, the larger diameter probe reduces the oscillations as the probe traverses the cells.
The larger probe was then used to scan the same sample as that shown in Fig. 10 (see Fig. 13). The features that were very distinct in that image became blurred and reduced in magnitude so that they were barely detectable. It would appear that it is better to use the smaller probe on all the samples than to use the large one. When the small probe is used on a larger honeycomb, the features of the honeycomb are superimposed on the defects, which can make measurements harder, but at least the defects are visible. When the larger probe is used, many of the defects fade from view.
The use of eddy current arrays to examine defects, especially disbonds, in sandwich panels with insulating CFRP face-sheets has been examined. It has been shown that transmit–receive arrays are not suitable for this application, because they have a very directional response to the honeycomb. Both absolute bridge and reflection array probes can work to detect dents and disbonds. Their cylindrical nature makes them immune to the directionality experienced by transmit–receive probes. However, the size of the coils relative to the size of the honeycomb is an important consideration. If the coil diameter is too large, it can blur out defects making them hard to detect. If the coil diameter is too small, the structure of the honeycomb is superimposed on the C-scan, which can make interpretation difficult. A coil diameter of 2–3 cell diameters appears to be optimal. No information has been determined about how thick a face-sheet can be. Coil diameter will limit the amount of lift-off that can be dealt with. Smaller coils can only work at smaller lift-offs. However, in most honeycomb sandwich panels, the face-sheets are thin, less than the cell size, so using an optimal size probe should make this consideration moot.
Eddy current C-scan, thermography, and laser scanners provide complementary information. Any two of them can give complete information about a sample. For example, thermography can detect disbonds but has trouble finding dents that are still bound. Eddy current array detects both disbonds and dents but cannot say which is which. Similarly, a laser scanner can find dents but cannot detect disbonds. Eddy current and thermography together can find the disbonds. Eddy current array has the advantage that is easy to set up and can scan large areas quickly.
The authors would like to thank the AERAC committee of DTAES (DND, Canada) for support of this research.