Four piezoelectric AE sensors were mounted on spars of the aircraft wing representative structure prior to conducting a series of constant amplitude loading.
COVER SHEET
Title: Detecting and locating fatigue cracks in a complex wing-box structure using the acoustic emission technique: A verification study Authors: Daniel Gagar Marcias Martinez Marko Yanishevsky Bruno Rocha Jim McFeat Peter Foote Phil Irving
ABSTRACT Acoustic Emission (AE) is recognized for its potential as a Structural Health Monitoring (SHM) technique, offering several advantages, such as passive sensing and wide area monitoring, making it particularly attractive for aerospace applications. Realistic structures, like aircraft wings for example, consist of structural elements of complex geometry, manufactured with different materials. Individually these elements may influence the performance of the SHM technique in detecting and locating structural damage. In this research, the capabilities of the AE technique to identify and locate damage in complex structures were assessed. These assessments were performed without disclosure of the damage location to the AE equipment operator. Four piezoelectric AE sensors were mounted on spars of the aircraft wing representative structure prior to conducting a series of constant amplitude loading fatigue tests. One of the spars contained a fatigue crack seeded at a location unknown to the damage monitoring analyst. The sensors monitored AE generated from the spars and event location estimates were obtained using time of flight analysis with a wave velocity value determined experimentally. The AE monitoring successful detected and located the hidden fatigue crack; this was independently verified by three nondestructive testing (NDT) methods.
INTRODUCTION The AE technique is recognized as a Structural Health Monitoring (SHM) tool capable of fatigue crack detection in metallic structures via acoustic signals generated from in and around the crack region. The location of the crack can be determined using information of the difference time of arrival of these signals at different sensors in an array configuration [1]. _____________ Gagar, D., Irving, P., Foote, P., Cranfield Uni. IVHM Centre, Bedfordshire, UK MK43 0FQ. Martinez, M., Faculty of Aerospace Engineering, Delft University of Technology, P.O.Box 5058, 2600 GB Delft, The Netherlands. Yanishevsky, M., Rocha, B., National Research Council of Canada, Ottawa, K1A 0R6. McFeat, J., BAE Systems, Warton Aerodrome, Lancashire, UK W427D.
Source
Propagation
- Damage mode
- Geometric features - Material anisotropy - Dispersion
Detection
- Sensor & signal coupling
System & Algorithm - Signal detection method - AE event characterisation
Figure 1: Sources of variability in the output of the AE system The performance of the AE technique in detecting and locating damage is however influenced by complex transfer functions between the source and processed signal [2]. These influences can be grouped as: AE signal source [2; 3], propagation medium [4], signal-sensor coupling [5; 6] and data processing [2], as illustrated in Figure 1. AE generation from fatigue cracks can originate from different processes, such as fracture of inclusions, crack closure, or crack extension occurring during fatigue crack propagation [7]. Depending on material properties, structural geometry and loading configurations, the prevalence of AE from these sources may be affected, which can result in periods of reduced opportunity for crack detection [3; 8]. Also, the integrity of sensor coupling with the structure can compromise AE detection [6]. Material properties and component geometry can affect the dispersion properties of Acoustic Emissions, namely the wave velocity, which is used as an input in triangulation algorithms for determining damage locations. Wave velocity calibration methods have been developed for improving the performance of this technique in anisotropic test samples, as well as those containing complex geometric features [4]. The capability of the AE technique often has been demonstrated in relatively simple test components [4; 8]. However, realistic structures like aircraft wings, which consist of assemblies of different components, complex geometries, as well as different materials, will provide a more challenging environment for this technique. The National Research Council of Canada (NRC) has developed several test rigs capable of closely replicating realistic structures, such as an aircraft wing, in terms of materials used, geometries and fatigue loading conditions. This paper reports on experiments conducted to verify the performance of the AE technique in detecting and locating fatigue cracks in a realistic wing-box structure. The actual location of the seeded crack remained undisclosed to the AE analysts until the end of the tests, when Non Destructive Inspection (NDI) was independently carried out on the entire wingbox. FATIGUE TEST SETUP The layout of the fatigue test rig is illustrated in Figure 2. This rig consisted of a central pedestal to which two wing-box test structures were bolted on a horizontal plane. The rig was designed such that either wing-box could be easily removed and replaced. The wing-box utilised in this setup consisted of a network of metallic ribs and spars made from AA7075-T651 material, with Carbon Fibre Reinforced Polymer (CFRP) composite skins bolted to the top and bottom of the wing-box, in a similar configuration to that found on CF188 fighter aircraft. Two 25 kN (5500 lbf) loading actuators, equipped with 22 kN (5000 lbf) load cells, were attached to the tip of the wing-boxes. The wing-box structure was designed to accommodate an additional internal C-channel spar manufactured with a ‘hidden’ crack established at a fastener hole, not visible to the naked eye. Details of the crack seeding process can be found in
Yanishevsky et al. [9]. There were a total of 15 fastener holes in both the top and bottom flanges of the test spar with a pitch distance of 25.4 mm (1 inch). The test spar also contained a geometric cut-out feature in its web, akin to a fuel weep hole, as illustrated in Figure 3.
Figure 2: Test rig layout with wing-box test platforms and loading actuators
Figure 3: Side and plan views of the test spar sample with fastener holes and a geometric cut-out feature Simulations were performed using AFGROW [10] to estimate the number of cycles required to grow seeded fastener hole cracks at different stress levels with an initial length of 1.6 mm (0.063 inches). The results were used to select applied loads in order to create a gross stress level of 69 MPa (10 ksi) in the crack region. This stress level ensured that the 1.6 mm crack would grow to a selected target length within a reasonable time / applied number of cycles. Crack growth was monitored without dismantling the structure during fatigue tests, using a Borescope and an Eddy Current probe. It is important to note that while the inspection of the damage was being evaluated by an NRC NDT inspector, the AE operator was required to leave the premises, so that the damage location was kept undisclosed. Two tests were conducted with this setup. The first had constant amplitude sinusoidal tension-tension loads, alternating between 5.6 kN (1250 lbf) and 0, applied at a frequency of 0.667 Hz which generated a maximum tension gross stress of 69 MPa (10 ksi) in the region of interest on the top skin of the wing-box. The second had constant amplitude sinusoidal tension-compression loads, alternating between 5.6 kN (1250 lbf) and -1.4 kN (312 lbf), applied at frequency of 0.5 Hz, again with the maximum tension achieved on the top skin of the wing-box . ACOUSTIC EMISSION SYSTEM SETUP A 4-channel Physical Acoustics AE system was used with broadband piezoelectric sensors to record AE data generated throughout the tests with a sampling rate of 1MS/s. Pre-amplifier gain for each channel was set at 40 dB and the AE signal detection threshold was set at 45 dB.
Ch ord wise d irecti on
A two sensor AE event location setup (1D) was used to monitor the test sample with the sensors attached to the web of the spar using Dow Corning RTV 3140 silicone rubber. This pair of sensors on the test sample is also referred to as Location Group 1, consisting of Sensors 5 and 6. Another pair of sensors was installed on the adjacent spar, consisting of Sensors 3 and 4, which are referred to as Location Group 2. A schematic of the wing-box layout and the AE sensor arrays are illustrated in Figure 4, with fastener hole and sensor locations on the test spar and wing-box structure provided in Table I. By using AE event time of flight differences measured from both sensor location groups, a 2D localization of the damage was achieved, despite the influences of different materials and component geometries, which affect the propagation velocity of the AE travelling in paths between the two pairs of sensors on adjacent spars [11]. The sensors on the test spar were placed as far away from each other as possible to maximise the sensor coverage area in between sensors, which spanned from Holes 3 – 13. The sensors on the adjacent spar were located in the same spanwise positions. A high-pass amplitude filter was applied to the AE data in post-processing, setting a threshold to enable distinguish noise from rapid amplitude rises characteristic of AE bursts. This filter threshold was set at 75 dB in Test 1; 90 dB in Test 2.
Sensor 6 location group 1
Sensor 5 location group 1
Sensor 4 location group 2
Sensor 3 location group 2
Sp an wise dire ctio n
Figure 4: Schematic of wing-box structure with sensor positions Table I: Locations of the fastener holes and sensors on the test sample Hole/ Sensor Hole 3 Hole 4 Hole 5 Hole 6 Hole 7 Hole 8 Sensor 3 Sensor 4
Spanwise position on spar (mm) 70.8 96.2 121.6 147.0 172.4 197.8 60.0 347.0
Spanwise position on wingbox (mm) 417.8 443.2 468.6 494 519.4 544.8 407 694
Hole/ Sensor Hole 9 Hole 10 Hole 11 Hole 12 Hole 13 Sensor 5 Sensor 6
Spanwise position on spar (mm) 223.2 248.6 274.0 299.4 324.8 60.0 347.0
Spanwise position on wingbox (mm) 570.2 595.6 621 646.4 671.8 407 694
ACOUSTIC EMISSION SYSTEM CALIBRATION The AE system was equipped with a built-in Automatic Sensor Test (AST) function, which allows the piezo AE sensors to interrogate each other using acoustoultrasonic pulses, to verify the integrity of sensor coupling and obtain the time of flight between sensors. With the sensors mounted in the assembled wing-box structure, the AST function was used to send 50 pulse signals, with 5 µs length and 100 ms delays between pulses, between sensors in a Location Group. The average time of flight was
recorded. Given a distance of 290 mm (11.42 inches) between sensors in the same spar, the average wave velocity was calculated. The results for the different interrogation paths are shown in Table II. Table II: Average time delay and wave velocity with acousto-ultrasonic interrogation between the two pairs of sensors Sensor interrogation (Actuator-Sensor) 5–6 6–5 3–4 4–3
Average ∆t (µs)
Wave velocity (km/s)
105 103 118 102
2.76 2.81 2.45 2.84
RESULTS
(a) 1000 750 500 250 0
100 150 200 250 300 350
Obse rva tio ns
Obse rva tio ns
Test 1 consisted of a total of 66,033 fatigue load cycles. The results of AE event location estimates for Location Groups 1 and 2 are illustrated in Figures 5(a) and 5(b) respectively, assuming an average wave velocity of 2.72 km/s. In Figure 5(a), it is shown that the distribution of AE event location estimates was between 270 – 300 mm (10.63 – 11.81 inches) along the spanwise direction on the test spar. The greatest peak occurred at approximately 280 mm (11.02 inches), corresponding to a position between Holes 11 and 12, as shown in Table 1. AE events were also detected by Location Group 2, despite the adjacent spar, where these sensors were installed not containing a seeded crack, as shown in Figure 5(b). These were located to have originated between 280 – 310 mm (11.02 – 12.2 inches) in the spanwise direction, corresponding to the region between Holes 11 and 13. (b) 10 8 6 4 2 0
Di sta nce ( mm)
100 150 200 250 300 350
Di sta nce ( mm)
Di sta nce (mm)
Figure 5: Distribution of 1D AE event location estimates in Test 1, with detection threshold of 45 dB and high-pass filtering ≥ 75 dB using: (a) Location Group 1 (sensors 5 and 6); and (b) Location Group 2 (sensors 3 and 4) 400 200 0 0
200
400
600
800
1000
Di stan ce (m m )
Figure 6: Distribution of 2D AE event location estimates on wing-box structure in Test 1 using KDE
Using the AE arrival time delay measurements from both Location Groups 1 and 2, 2D damage location estimates were performed. These results are presented in Figure 6. Both the 2D location estimates, as well as their distribution determined using the Kernel Density Estimation (KDE) method are shown superimposed over a drawing of the wing-box structure plan view. It was observed that there was one main cluster of AE events with its peak displaced from the test spar by 50 mm (1.97 inches) in the chordwise direction of the wing-box. In the spanwise direction, this peak was located at approximately 600 mm (23.62 inches) from the tip of the wing-box, corresponding approximately to Hole 10. Test 2 was performed by applying additional 14,967 fatigue load cycles. The results for AE origin location estimates for Location Groups 1 and 2 are illustrated in Figures 7(a) and 7(b). Once again, an average wave velocity of 2.72 km/s was used. It was observed that the location distribution ranged from 240 – 300 mm (9.45 11.81 inches), along the spanwise direction on the test spar with its peak occurring at 250 mm (9.84 inches), corresponding approximately to Hole 10. Almost identical distributions were observed for Location Groups 1 and 2; however, fewer AE signals were detected by the sensors in Location Group 2. The 2D location estimates were also obtained, similarly to the previous test. These results are presented in Figure 8. It can be observed that there was significantly more scatter in damage location estimates as compared with Test 1. The peak of the location estimate distribution was displaced from the test spar by 100 mm (3.94 inches) in the chordwise direction. In the spanwise direction, the peak was located between 570 - 600 mm (22.44 – 23.62 inches) from the tip of the wing-box, corresponding to the region between Holes 9 and 11. (b)
2000 1000 0 60
150
250
350
Obs er va tio n s
Ob se rva ti o ns
(a) 3000
3000 2000 1000 0 60
D i sta n ce (m m )
150
250
350
Di sta n ce (m m )
D i stan ce (mm)
Figure 7: Distribution of 1D AE event location estimates in Test 2, with detection threshold of 45 dB and high-pass filtering ≥ 90 dB using: (a) Location Group 1 (sensors 5 and 6); and (b) Location Group 2 (sensors 3 and 4) 400 200 0 0
200
400
600
800
1000
Di stan ce (m m )
Figure 8: Distribution of 2D AE event location estimates on wing-box structure in Test 2 using KDE From the 1D location distribution in both tests, shown in Figures 5 and 7, it can be seen that more AE events were detected by sensors 5 and 6 between 260 mm
(10.24 inches) and 350 mm (13.78 inches) in the spanwise direction of the test spar, as compared to sensors 3 and 4. This suggests that the AE source in this region was closer to sensors 5 and 6. Based on the distributions of estimated AE event locations obtained in the various tests, Holes 9 - 13 were identified by the AE analyst as the possible locations of the seeded fatigue crack on the test spar. Non Destructive Inspection (NDI) was independently performed by the National Research Council of Canada (NRC) on the entire wing-box using Eddy Current, Ultrasonic and Fluorescent Penetrant techniques and the results for the test spar are given in Table III. These identified the location of the ‘hidden’ crack to be at Hole 11. Furthermore, a crack-like indication was also obtained for Hole 9, using the Eddy Current NDI technique. Table III: Non-destructive Inspection of test spar Hole # 9 11
Eddy Current inspection Yes Yes
Ultrasonic inspection No Yes
Dye-penetrant inspection No Yes
Notes Crack-like indication < 0.76 mm Confirmed 4.2 mm crack
DISCUSSION The location of the ‘hidden’ fatigue crack was within the distribution of AE event location estimates obtained using Location Group 1 in both Tests 1 and 2. Comparing the peak in each of the distributions with the position of the actual crack, the location accuracy was found to be within 6 mm (0.24 inches) in Test 1 and 24 mm (0.94 inches) in Test 2. The NDI results verified that the other holes had not develop fatigue cracks, apart from Hole 9 which showed a crack-like indication with the Eddy Current technique, which could not be confirmed using the ultrasonic and fluorescent penetrant methods. This potential crack, most probably forming at a later stage during testing, could have been responsible for biasing the damage location estimates towards Hole 9 on Test 2. No other cracks were found in other parts of the wing-box, although there were some signs of minor damage around some fastener holes. From the 1D location estimates obtained with Location Group 2, it was apparent that the sensors on the adjacent spar also registered the AE signals generated from the crack location on the test spar. It appeared that there was signal coupling at the interface between the spar and skin, which enabled AE signal transfer between adjacent spars by propagation through the bolted-on skins connecting them. This occurrence was confirmed by the 2D representation of the data obtained from both sets of sensors with the greatest density observed in similar locations in both Tests 1 and 2, which also correspond to the vicinity of the actual crack location on the span of the wing-box. The errors obtained can be attributed to combinations of factors illustrated in Figure 1, most significantly associated with the complex propagation paths between sensors on adjacent spars, which had not been characterised in terms of wave velocity at the outset of AE monitoring. Although the amplitude high-pass filtering threshold applied was greater in Test 2 than in Test 1, and the duration of Test 2 was about a quarter of that in Test 1, the number of AE events recorded in Test 2 were significantly more than those in Test 1. This suggests that the tension – compression loading in Test 2 had a significant influence on the quantity of AE generation, and caused significantly more scatter in
the location estimates in comparison to Test 1. This is to be expected as tension - compression loading, passing through zero load, is likely to generate local relative movement between structural components, generating associated Acoustic Emissions located remotely from the seeded crack. The value of wave velocity obtained from the calibration tests was a function of the thickness of the propagating medium [11], the layered assembly of the spar together with the skins in this case. Clearly, this had some influence on the dispersion properties of propagating AE signals; however, this was not completely understood from the results obtained in this study. Further work is required to model and understand the dispersion behaviour of AE signals under such conditions. CONCLUSIONS 1. The AE technique was successful in detecting and locating a fatigue crack in a complex aircraft wing-box representative structure, without prior knowledge of its location. The accuracy of the damage location estimates was found to be 6 mm (0.24 inches) in a tension-tension fatigue load condition (Test 1) and 24 mm (0.92 inches) in a tension – compression fatigue load condition (Test 2). 2. AE sensor coverage was accomplished in a region of the wing-box structure, despite a complex signal propagation path resulting from several components made of different material types and having complex geometries. ACKNOWLEDGEMENTS This work was sponsored by the Cranfield University Centre for IVHM and BAE Systems. REFERENCES [1] Holford, K. M. (2000), "Acoustic emission - basic principles and future directions", Strain, vol. 36, no. 2, pp. 51-54. [2] Holford, K.M. ( 2009), Acoustic emission in structural health monitoring. [3] Gagar, D. O. (2013), PhD Thesis: Validation and verification of the acoustic emission technique for structural health monitoring, Cranfield University. [4] Baxter, M. G., Pullin, R., Holford, K. M. and Evans, S. L. (2007), "Delta T source location for acoustic emission", Mechanical Systems and Signal Processing, vol. 21, no. 3, pp. 15121520. [5] Rocha, B. (2011), PhD Thesis: Structural Health Monitoring of Aircraft Structures, Universidade Tecnica de Lisboa, Instituto Superior Tecnico, Lisbon, Portugal. [6] Fasana, A. and Garibaldi, L. (2007), Measurement of acoustic emission signals: Influence of the couplant. [7] Miller, R. K. and McIntire, P. (eds.) (1987), Non-destructive testing handbook, 2nd Ed., American Society for Nondestructive Testing. [8] Daniel, I. M., Luo, J., Sifniotopoulos, C. G. and Chun, H. (1998), "Acoustic emission monitoring of fatigue damage in metals", Nondestructive Testing and Evaluation, vol. 14, no. 1-2, pp. 71-87. [9] Yanishevsky, M., Martinez, M., Mandache, C., Khan, M., Fahr, A. and Backman, D. (2010), "Artificial seeding of fatigue cracks in NDI reference coupons", Insight: Non-Destructive Testing and Condition Monitoring, vol. 52, no. 12, pp. 664-671. [10] Harter, J. A. (1999), AFGROW users guide and technical manual, 0704-0188, Air Force Research Laboratory. [11] Rose, J. L. (2003), "Dispersion curves in guided waves testing", Mater.Eval., vol. 61, no. 1, pp. 20.
