EDDY CURRENT INSPECTION FOR DEEP CRACK DETECTION AROUND FASTENER HOLES IN AIRPLANE MULTI-LAYERED STRUCTURES Teodor Dogaru Albany Instruments Inc., Charlotte, NC tdogaru@hotmail.com Stuart T. Smith Center for Precision Metrology, UNC Charlotte, NC Carl H. Smith and Robert W. Schneider NVE Corporation, Eden Prairie, MN Abstract. Solid-state magnetic sensors based on Giant Magnetoresistance (GMR) effect have been integrated in novel self-nulling eddy current probes that are capable of detecting cracks and flaws emanating from fastener holes in multi-layered structures of the airplanes. The high sensitivity of these magnetic sensors extends to the low frequencies required for deep penetration by the eddy currents. The methods proposed herein take advantage of the symmetry of the specimen to eliminate the edge and fastener signals. Shaped excitation coils properly positioned and scanned with respect to the hole are used to focus the eddy currents at the edge of the hole. Consequently, the perturbation of the eddy current flow due to the presence of a crack initiating at the edge is greatly enhanced. Corner cracks of 2.5 mm in length were detected in the second layer of a 13 mm thick two-layer structure by using GMRbased probes. The novel eddy current probes will provide an effective tool for detecting small cracks within thick wing splice joint structures. Keywords: eddy current testing, giant magnetoresistive sensors, deep crack detection, airplane structures. 1. Introduction Within aircraft industry there is an increasing need to inspect for cracks and flaws emanating from fastener holes located on the wing multi-layered structures of the airplanes. These structures are held together by rows of taper-lock fasteners. Cracks can occur around the fasteners holes, in each of the structure layers. It is important to detect these cracks at the initial stage of development. Presently lower-layer cracks are not found until the aircraft has been removed from service and some disassemble done. These cracks and flaws are causing troubleshooting and repair, which significantly add maintenance costs. Currently, the detection of deeply buried flaws is carried out by using either eddy current techniques or ultrasound methods. The drawback of ultrasound methods is that they are not effective in detecting lower-layer flaws. In contrast, by using eddy current techniques, the electromagnetic field is less perturbed by the presence of the interfaces between layers. Typically the fasteners are disposed in rows within airplane multi-layer structures. Depending on the direction of stresses during the flight, there are two types of cracks around fastener holes commonly encountered in practice: a. Longitudinal cracks that initiate and propagate along a fastener row. This type of cracks is the most critical, because they can propagate from a fastener hole to the adjacent hole ( zipping effect), potentially causing major structural failure. b. Transversal cracks that propagate perpendicular to the fastener row. For narrow structures, the cracks can propagate across the structure towards its edge. Recent advances in magnetic sensor technology made the electromagnetic nondestructive evaluation method very attractive to address this difficult problem. To detect deeply buried flaws, a low frequency electromagnetic field must be induced in the specimen under test. Traditional eddy current testing methods based on excitation-detection coils is fundamentally limited by the poor sensitivity of the detection coils at low frequencies. The use of magnetoresistive sensors has several advantages over inductive coils: High sensitivity from DC to megahertz domain and low noise provide the capability of detecting deeply buried flaws as well as surface cracks. Small dimensions (of order of tens micrometers) allow high-spatial resolution flaw detection. Being fabricated using planar technology, thin film magnetoresistive sensors can be manufactured in customized arrays. Suitably patterned arrays are very attractive for mapping the magnetic field without the need of scanning the area of interest. The low cost of the magnetoresistive sensors (compared to other high sensitivity magnetic sensors such as SQUID or fluxgate) makes them attractive for developing commercial eddy current probes. Both anisotropic magnetoresistive sensors (AMR) and giant magnetoresistive sensors (GMR) have been successfully used in prior work for detecting deep cracks under installed fasteners (Wincheski et all, 1999; Lebrun et 2 a 6 de Junho de 23 / June 2 to 6 23 Rio de Janeiro - RJ - Brasil
all, 1995; Avrin, 2). In most cases, a cylindrical excitation coil produces the eddy current in the vicinity of the hole. The perturbation of the eddy currents is detected by the magnetoresistive sensor placed usually on the axis of symmetry of the cylindrical coil. In many cases the signal created by the edge of the hole, by the fastener itself or by adjacent holes or edges, can obscure the signal produced by a small deeply buried crack. Therefore it is important to distinguish and/or separate the crack signal from all these parasitic signals. The new GMR-based eddy current probes and methods of detection presented in this paper use simple scanning techniques to reduce the inspection time. The probes were designed such that the influence of edges or adjacent holes is greatly reduced. Both linearly scanned and rotational probes have been successfully tested for detecting buried cracks around fastener holes. 2. GMR sensors overview Recent developments in thin-film magnetic technology have resulted in films exhibiting a large change in resistance with magnetic field. This phenomenon is known as giant magnetoresistance (GMR) to distinguish it from conventional anisotropic magnetoresistance (AMR). Whereas AMR resistors exhibit a change of resistance of less than 3 %, various GMR materials achieve a 1 to 2 % and greater change in resistance. GMR films have two or more magnetic layers separated by a non-magnetic layer. Due to spin-dependent scattering of the conduction electrons, the resistance is maximum when the magnetic moments of the layers are antiparallel and minimum when they are parallel. Commercial GMR sensors are based on a thin-film GMR magnetic multilayers deposited on silicon substrates. Wheatstone bridge sensors are fabricated from four photolithographically patterned GMR resistors, two of which are active elements. The sheet resistance of these thin films are between 1 and 15 ohms per square. Resistors of 5 kω can be formed as 2 µm serpentine traces covering less than a 8 µm square. To protect two of the four equal resistors in a Wheatstone bridge from the applied field, small magnetic shields of permalloy plated over them, allowing them to act as reference resistors. Since they are fabricated from the same material, they have the same temperature coefficient as the active resistors. The two remaining GMR resistors are both exposed to the external field. The bridge output is therefore twice the output from a bridge with only one active resistor. The bridge output for a 1 % change in the resistance of the GMR material is approximately 5 % of the voltage applied to the bridge. Additional permalloy structures plated onto the substrate act as flux concentrators to increase the sensitivity. The active resistors are placed in the gap between two flux concentrators as is shown in Fig. (1). These resistors experience a field which is larger than the applied field by approximately the ratio of the gap between the flux concentrators, D1, to the length of one of the flux concentrators, D2. In some sensors the flux concentrators are also used as shields by placing two resistors beneath them as is shown for R3 and R4. The sensitivity of a GMR bridge sensor can be adjusted in design by changing the lengths of the flux concentrators and the gap between them. In this way, a GMR material which saturates at approximately 3 Oe can be used to build different sensors which saturate at 15, 5, and 1 Oe. D2 D1 B R2 R4 R1 R3 A R1 R2 B A R3 R4 External field Figure 1. Configuration of GMR resistors in a Wheatstone bridge sensor. Flux concentrators are shown. D1 and D2 are the lengths of the gap between the flux concentrators and the length of one flux concentrator respectively. Recent research work at NVE in developing multilayer GMR material for high temperature applications has resulted in new materials with different magnetic properties. Some of these materials have significantly increased sensitivity as measured in percent change in resistance with field. They do, however, have lower sheet resistance and somewhat lower saturation field the field at which the resistance no longer changes. The sensitivity of the new material or change in percent resistance with field is approximately ten times that of the conventional material. This type of GMR sensors in bridge configuration has typically a sensitivity of about 1 mv/oe for 1 V applied to the bridge. 3. Rotational probes for detection of cracks around fastener holes An original technique developed at Albany Instruments to detect deeply buried cracks around fastener holes in the second-layer of an aluminum structures is based on a rotational GMR-based eddy current probe (Dogaru, 21). The
use of D-shaped excitation coils instead of circular coils significantly enhances the capability of these probes to detect cracks initiating from the hole edge. Axis of sensitivity D - coil Buried crack Rotation about the center of the hole SDT sensor Specimen surface Stainless steel fastener Figure 2. Schematic diagram of rotational, self-nulling probe comprising a D-shaped coil and an externally positioned sensor. Figure (2) shows a D-shaped coil placed with its straight edge intersecting the hole region. Using this probe configuration, most eddy current loops are constrained to follow the path along the hole edge where the crack is to be detected. Within this configuration, the sensor is external to the excitation coil and, being attached to it, both coil and sensor rotate simultaneously around the hole. The GMR sensor is positioned above the hole edge, with its sensitive axis tangential to the edge. In a first experiment, a corner crack of 2.5 mm length and 2.5 mm height at the edge of a hole was detected at 1 mm below the surface. The specimen consisted of a square aluminum plate of dimensions 151 mm x 151 mm x 1 mm, having a 19 mm diameter hole in the middle. A corner crack of 2.5 mm was machined at the bottom of the plate. The probe was circularly scanned over the edge of the hole. The output of the sensor resulted from two complete rotations of the probe around the hole is shown in Fig. (3). The crack is located at 27 degrees angular position on the plot, between the pair of peaks. The optimal frequency for the detection was 2 Hz. R-output voltage (mv) 16 14 12 1 8 6 4 2 9 18 27 36 45 54 63 72 Scanning position (deg) Figure 3. Plot resulted from two rotation of D-probe around the 19 mm diameter hole having a 2.5 mm crack at 1 mm depth; excitation frequency was 2 Hz. In the second experiment, an additional plate of 3 mm thickness was added at the top of the 1 mm plate. The plates have identical length and width and both contain a hole of 19 mm diameter in the middle. The crack of 2.5 mm was detected through the two-layer structure of 13 mm thickness. The optimum frequency for detection was 1 Hz. The result is shown in Fig. (4).
Y-output voltage (mv) 12 1 8 6 4 2-2 -4-6 -8-1 9 18 27 36 45 54 63 72 Scanning position (deg) Figure 4. Plot resulted from two rotation of D-probe around the 19 mm diameter hole having a 2.5 mm crack at 13 mm depth; excitation frequency was 1 Hz. Another experimental study was undertaken to map real cracks initiating at the edge of rivet holes. Specimens have been supplied by a leading aerospace manufacturer. First specimen was a 3 mm thick aluminum plate containing two cracks of length 2.65 mm and 2.75 mm respectively occurring either side of a 6.4 mm diameter hole. The second specimen contains two cracks of length.85 mm and.95 mm respectively occurring either side of a 4.9 mm diameter hole on a 1.6 mm thick plate. The cracks in the first specimen are difficult to observe visually while those in the second plate, being very short, are not visible at all. Magnified pictures of the cracks were taken using optical microscope Fig. (5). Figure 5. Microscope photos of a 2.65 mm long crack on the left side of the hole (left); a 2.75 mm long crack on the right side of the hole (right). Figure 6. Sensor s output magnitude when scanning around the hole containing two real cracks of 2.65 mm and 2.75 mm respectively on the opposite sides of the hole.
The eddy current test has been performed by scanning a GMR probe over a ring containing the edge of the hole. The GMR probe configuration was based on a circular excitation coil. For the first plate a 3-D map indicating the output voltage of the probe as a function of x-y position above the plate is presented in Fig. (6). A similar map for the second plate was also obtained. In both cases the maps indicate two pairs of peaks, which correspond to the presence of cracks. The cracks can be precisely located between these pairs of peaks. The excitation frequency used in these experiments was 3 khz. The results demonstrate the capability of the GMR probe to detect and locate cracks shorter than 1 mm initiating on the edge of holes. 4. Linearly scanned eddy current probes For cracks around holes disposed in rows, linear scanning methods are often preferable to the circular scanning. The methods presented in this paper use a single scanning of the fastener row, rather than a two-dimensional scanning, therefore reducing the inspection time. The principle of these new probes and methods is based on symmetry considerations. The scanning lines are always chosen such that the eddy current loops induced in the material are symmetric about the scanning line. In this way, in the absence of cracks, by using a proper orientation of the sensitive axis of the GMR sensor, the output of the sensor is theoretically zero. A crack will break the symmetry of the loops about the scanning line, creating a signal at the sensor. To obtain the symmetry, the scanning line coincides to the diameter of the hole to be inspected, and is perpendicular to the direction of the cracks. For transverse cracks, the scanning line will be the symmetry axis of the fastener row. Any transverse cracks will break the symmetry about this axis. For longitudinal cracks, the scanning line will be perpendicular to the fastener row. The eddy current probes are based on flat excitation coils manufactured from a ribbon cable. The current through the wires of the coil flows along the scanning direction of the probe. Two standard connectors can be attached to the ends of the ribbon cable, enabling various connections (through jumper wires) of different wires within the cable. Therefore different coil configurations can be designed on the same cable, by changing the jumper connectors. Due to the plastic insulation, the cable can slide on the surface of the specimen to be inspected without damaging the surface paint. Handheld probes can be easily manufactured and scanned in mechanical contact with the specimen surface. Thus the probe lift-off can be minimized and maintained at a constant value. An experimental study was conducted on laboratory specimen comprising rows of holes. The specimens were aluminum plates of the width 5 mm and the length of 28 mm. Ten holes of 6.3 mm diameter were drilled in each plate along the longitudinal symmetry axis of the plate. The distance between the centers of adjacent holes is 19 mm. In one specimen, two transverse notches were machined at hole #3 (on the left side) and hole #7 (on the right side). The top view of the specimen is shown in Fig. (7). The slots have the same height (1 mm) and lengths of 2.5 mm, and 2 mm respectively. Note that the height of cracks is less than one third of the thickness of the plate, emulating corner cracks. Figure 7. Plate containing a 2.5 mm left-side notch in hole #3 and a 2 mm right-side notch in hole #7. Holes are numbered 1 through 1 from left to right. Firstly, the probe was scanned along the specimen axis, with the cracks on the backside of the plate (3.2 mm under the surface). The excitation frequency was 2 khz. The out-of-phase component of the sensor s output is shown in Fig. (8). The left notch in hole #3 (2.5 mm in length) produced a positive peak, while the right notch in hole #7 (2 mm in length) a negative peak. Because the scanning was manual, the scanning speed could not be maintained constant. Consequently, the positions of peaks (or of the ripples corresponding to the defect-free holes) do not correspond precisely to the positions of holes indicated in Fig. (8). The ratio between the crack signal and the signal produced by defect-free holes was approximately 5.
Subsequently, a plate of thickness 1.6 mm was placed on the top of the specimen to test the capability of detection of defects buried at 4.8 mm below the surface in a two-layer structure. The result obtained using the same eddy current probe at a frequency of 1 khz is shown in Fig. (9). On this figure it can be observed that the cracks signal is about three times larger than the background signal caused by the defect-free holes. Sensor output (mv) 6 5 4 3 2 1-1 -2-3 -4-5 1 2 3 4 5 6 7 8 9 1 11 Holes position Figure 8. Output signal of the eddy current probe when scanned along a specimen containing two cracks of 2.5 mm (hole #3) and 2 mm (hole #7) respectively, at 3.2 mm depth. 4 Sensor output (mv) 3 2 1-1 -2 1 2 3 4 5 6 7 8 9 1 11 Holes position Figure 9. Output signal of the eddy current probe when scanned along a two-layer specimen containing two cracks of 2.5 mm (hole #3) and 2 mm (hole #7) respectively, at 4.8 mm depth. 5. Conclusions Novel methods based on rotational self-nulling probes and linearly scanned probes were developed. Tests were performed on both in-house manufactured specimens (containing slots machined around holes) and samples provided by leading airplane manufacturers. Using rotational probes, corner cracks of 2.5 mm around fastener holes were detected at 13 mm in the second-layer of a two-layer structure. Transverse notches of 2 mm in length, 1 mm in height were detected at 4.8 mm depth in a two-layer specimen containing a row of holes using linearly scanned probes. 6. References Wincheski, B. and Namkung, M., 1999, Development of very low frequency self-nulling probe for inspection of thick layered aluminum structures, Review of Progress in QNDE, Vol. 18, pp. 1177-1184. Lebrun, B., Jayet, Y. and Baboux, J.C., 1995, Pulsed eddy current application to the detection of deep cracks, Materials Evaluation, Vol. 53, No. 11, pp. 1296-13. Avrin, W.F., 2, Eddy-current measurements with magnetoresistive sensors: third-layer flaw detection in a wingsplice structure 25 mm thick, Proceedings of SPIE, Vol. 3994, pp. 29-36. Dogaru, T. and Smith, S.T., 21, Novel eddy current probes for detection of deep cracks around fastener holes, Fifth Joint NASA/DOD/FAA Aging Aircraft Conference, Orlando, USA.