Professional Documents
Culture Documents
Probability of Detection For Bolt Hole Eddy Current in Extracted From Service Aircraft Wing Structures
Probability of Detection For Bolt Hole Eddy Current in Extracted From Service Aircraft Wing Structures
Probability of Detection For Bolt Hole Eddy Current in Extracted From Service Aircraft Wing Structures
wing structures
P. R. Underhill, C. Uemura, and T. W. Krause
Abstract. Fatigue cracks are prone to develop around fasteners found in multi-layer aluminum structures on aging
aircraft. Bolt hole eddy current (BHEC) is used for detection of cracks from within bolt holes after fastener removal. In
support of qualification towards a target a90/95 (detect 90% of cracks of depth a, 95% of the time) of 0.76 mm (0.030”), a
preliminary probability of detection (POD) study was performed to identify those parameters whose variation may keep a
bolt hole inspection from attaining its goal. Parameters that were examined included variability in lift-off due to probe
type, out-of-round holes, holes with diameters too large to permit surface-contact of the probe and mechanical damage to
the holes, including burrs. The study examined the POD for BHEC of corner cracks in unfinished fastener holes extracted
from service material. 68 EDM notches were introduced into two specimens of a horizontal stabilizer from a CC-130
Hercules aircraft. The fastener holes were inspected in the unfinished state, simulating potential inspection conditions, by
7 certified inspectors using a manual BHEC setup with an impedance plane display and also with one inspection
conducted utilizing a BHEC automated C-Scan apparatus. While the standard detection limit of 1.27 mm (0.050”) was
achieved, given the a90/95 of 0.97 mm (0.039”), the target 0.76 mm (0.030”) was not achieved. The work highlighted a
number of areas where there was insufficient information to complete the qualification. Consequently, a number of
recommendations were made. These included; development of a specification for minimum probe requirements; criteria
for condition of the hole to be inspected, including out-of-roundness and presence of corrosion pits; statement of range of
hole sizes; inspection frequency and data display for analysis.
INTRODUCTION
Many aircraft operated by the Royal Canadian Air Force (RCAF) experience structural deterioration due to
fatigue cycling. During a typical flight, aircraft are subject to cyclic loading resulting in cyclic stresses that can lead
to physical damage to the aircraft structure. Continued cyclic loading is attributable to crack development and
damage accumulation that can lead to failure, at stresses well below the material’s ultimate strength. The damage
and failure caused by cyclic loading is known as fatigue.
To address the effects of fatigue cycling, the aerospace industry has adopted a damage tolerance approach in
aircraft maintenance and design. Damage tolerance analysis is used in the development of in-service inspection
programs that contribute to maintenance and life extension of the structure or component. Fatigue crack growth
prediction models have been developed to support these damage tolerance concepts. Fracture mechanics properties
such as fatigue crack growth data and fracture toughness are essential to conducting damage tolerance analysis of
primary aircraft structures [1]. Fracture mechanics techniques and fatigue crack growth prediction models are
essential tools in preventing the occurrence of failures, but the most critical component of the damage tolerance
approach is detecting the flaw or damage.
One of the most common types of fatigue crack growth is a crack growing radially from a bolt hole joining
multiple layers of aluminum. The standard way to inspect these sites is to use bolt hole eddy current in which a
probe mounted on a rotating scanner is passed through the hole and the eddy current signal is monitored for signs of
a crack or other defect. Probability of detection (POD) studies done on aluminum fastener holes show that cracks as
small as 0.43 mm (0.017”) can be reliably found using the right equipment and settings [2,3]. In the aircraft industry
the a90/95, the size of crack such that 90% of such cracks will be detected 95% of the time is usually taken to mean
“reliably detect”. The detection ability depends on a number of factors such as the probe type and the frequency
used to drive the probe. An important aspect of these POD studies is that they are almost all carried out on round
holes without any damage. There is some question as to how representative the holes used in this study are of real
holes in real structures. In a recent study, where fatigue cracks were grown in in-service material from which the
fasteners had been removed, the size of corner crack that could be reliably detected increased by a factor of 2 to
0.82 mm (0.032"), using 1600 kHz and automated C-Scan equipment [4]. The objective of this work was to conduct
160001-1
a second hit/miss probability of detection (POD) Bolt Hole Eddy Current (BHEC) study using multiple inspectors,
standard equipment and procedures to determine if similar results would be achieved again.
160001-2
FIGURE 2. Microscopic image - fastener hole with machined notch (circled in red).
FIGURE 3. Replica of a notch in a damaged hole indicating the two ways the crack length could be measured.
35
Number of notches
30
25
20
15
10
0
<5 5-10 10-15 15-20 20-25 25-30 30-35 >35
160001-3
Inspections were done from the opposite side of the hole to that which had the notches. This was done primarily
to hide the notches from the inspectors. The inspections reported here were conducted by 7 CGSB level 2 qualified
inspectors based at CFB Trenton, Ontario (i.e., people qualified to perform this inspection independently).
Inspectors were not asked to size the cracks, just indicate if a crack was present or not. Most of the inspections were
done in compliance with Gen-74E using Nortec 600 eddy current instruments, RA 2000 bolthole scanners and
contact probes. Gen-74E recommends a frequency of 400 kHz. Some duplicate inspections were done at 800 kHz.
Additionally scans were taken of all the holes using an automated system in the lab. This created a C-Scan record at
frequencies of 200, 400, 800 and 1600 kHz of each hole. The automated system replaced the Nortec 600 eddy
current instrument with an Olympus MS5800 instrument.
1.2
1
Fraction Detected
0.8
0.6
0.4
0.2
0
0 10 20 30 40 50
Size (Thou)
FIGURE 5. Hit Rate for the holes as a function of notch length.
160001-4
FIGURE 6. Hit/Miss POD of results for all inspectors combined.
0.30
0.25
0.20
False Call Rate
0.15
0.10
0.05
0.00
0 10 20 30 40 50
a90 (thou)
FIGURE 7. Comparison of False Call Rate and a90 for individual inspectors.
In order to understand these results better, the holes with the worst statistics were examined visually. The three
largest notches with the most misses and the three most commonly false-called holes are shown in Fig. 8. They are
characterized by significant out-of-roundness; Fig. 8 (a) M1, (b) F1 and (f) F3, large burrs (e) M3 and (d) F2, and
edge damage (c) M2.
160001-5
Worst Misses Worst False Calls
(a) M1: 037” long; 33% hit rate (b) F1: 75 % false calls
(c) M2: 034” long; 17% hit rate (d) F2: 58% false calls
(e) M3: 054” long; 33% hit rate (f) F3: 58% false calls
FIGURE 8: Worst misses on the left (a), (c) and (e) and false calls on the right (b), (d) and (f).
160001-6
Similar results were found for other holes, which exhibited either a large miss rate or a large false call rate. An
examination of the eddy current signals showed that many of the holes had very large horizontal (lift-off) signals.
It is not clear how the damage seen around these holes was incurred. It does not seem likely that the burrs were
introduced by the fastener removal process. The damage seen around holes like M2, shown in Fig. 8(c), seems old,
but that isn’t proof that it was created at assembly. Similarly the out-of-roundness was probably created when the
holes were originally drilled. It is clear that any inspection needs to, at the very least, flag the existence of damage so
that holes, which cannot be accurately inspected, are not declared defect free.
This study itself was not as rigorous as desirable, which motivates further work in this area. In particular, the
distribution of notches to smaller sizes than the final a90/95 of 0.97 mm (0.039”), the number of blanks being half that
recommended by the MIL handbook [6] and the high false call rate of 15% (target is 5%) suggest that the a90/95
reported here may be less than what might be anticipated were these additional factors taken into account in a follow
up study.
CONCLUSION
It is clear from this study that earlier studies on in-service material were not as representative of actual possible
conditions as desirable, in that they do not appear to have encountered holes with as much damage as was seen in
this work. The current work indicates that while the standard detection limit of 1.27 mm (0.050”) is met, anything
less than this may be optimistic when inspecting in-service material. In order for qualification of the BHEC method
to proceed, it is essential that a method of determining that a hole is sufficiently similar to a round defect free-hole,
so that it can be inspected with confidence, be established.
ACKNOWLEDGMENTS
The authors would like to thank Gaetan St-Amant at the Aeronautical Engineering Research Squadron (ATESS),
Trenton for technical assistance. This work was supported by the Aerospace Research Advisory Committee
(AERAC) through Directorate of Technical Airworthiness and Engineering Support (DTAES in the Department of
National Defence, Canada.
REFERENCES
160001-7