Further Developments in Ultrasonic Phased Array Inspection of Aging Aircraft

Further Developments in Ultrasonic Phased Array Inspection of Aging Aircraft Irene G Pettigrew 1, David I A Lines 2, Jesse A Skramstad 3, Robert A Smith 4 and Katherine J Kirk 1 1 Microscale Sensors, Institute of Physical Research, School of Engineering & Science, University of Paisley, Paisley, PA1 2BE, UK 2 Diagnostic Sonar Ltd, Kirkton Campus, Livingston, EH54 7BX, UK 3 NDT Solutions Inc., 1682 County Road K, New Richmond, WI 54017, USA 4 NDE Group, QinetiQ Ltd, Cody Technology Park, Farnborough, GU14 0LX, UK Abstract The high performance imaging capability of ultrasonic phased arrays has encouraged their recent widespread adoption for inspecting aircraft components. A critical inspection on aging aircraft is the detection of radial cracks around fasteners. In a current application, the array is mechanically scanned over the fasteners and volumetric RF waveforms data, comprising ultrasonic frames of 45º shear beams, are recorded for each position of the array. Multiple passes with the array skewed at-20º, 0º and 20º are required to find cracks at the specified orientations. Experimentally, this paper confirms that multi-angle inspection is necessary because the magnitude of reflection from a crack decreases with inspection angle. This paper also reports a beam skew technique that can be applied during processing enabling single pass data acquisition for enhanced second layer crack detection. Introduction In military aircraft defect detection there are two basic structure types. Firstly, fuselage which is a multi-layer structure consisting of thin sheets, typically containing small, closely spaced fasteners and most likely to suffer from lower stresses which cause defects such as long critical through-thickness cracks. The second is a machined wing skin/spar structure usually made from two thick plates with larger wider-spaced fasteners. The double layer structure is susceptible to high stresses causing small but critical corner cracks and may have a third layer (strap) used as a repair or support mechanism. There are several issues that have to be addressed upon every new inspection. Primarily, if similar aircraft structures exist they have to be individually identified and detectability issues regarding the location of cracks, crack orientation with respect to the spar and determining the critical crack length are important. 1 of 21

Commercially available systems have approached the crack detection challenge using transient eddy currents or ultrasonic phased arrays. The benefits of Full-Waveform Capture (FWC), where the RF waveform is recorded for every point on the component surface, has been reported [1]. The speed of their electronic scan means that arrays are particularly suited to this technique and a number of such systems are now in use for rapid 100% area coverage. In previous work [2, 3], an automated analysis program was developed to decide whether or not features found during an ultrasonic inspection of a wing skin/spar/strap structure using phased arrays with FWC were cracks or false calls. The sample was mainly designed for second layer crack detection with known cracks (0.5 mm to 2.5 mm in length) at known depths and angles. Here, the array was mechanically scanned over the fasteners and volumetric FWC data, comprising ultrasonic frames of 45º shear beams, were recorded for each position of the array. Multiple passes with the array skewed at - 20º, 0º and 20º were needed to find cracks at the specified orientations. The previous research revealed that for the third layer cracks, the coupling was not sufficient enough to detect defects. Inspection perpendicular to the spar showed the crack at 50 orientation in the second layer was not detected with a ±20 skew. A further extension of the FWC technique, termed Full Raw Data (FRD) collection and processing, has also been described [4]. FRD collection involves acquiring a coherent RF data set of multiple transmit-receive combinations from different locations along one or more arrays. The increased quantity of data requires special techniques to maintain inspection speed and these have been implemented in DSL s modular FlawInspecta system. This system has successfully found 0.5 mm second layer cracks [5]. The FRD approach allows many operations, such as beam steering and adjustable aperture sizing, shading and focusing, to be performed on captured data when normally their values are fixed at acquisition time. The system is also able to implement novel techniques that are only possible with FRD acquisition, including enhanced resolution by dynamic transmit focusing and non-linear beamforming. Introducing this phased array system capable of FRD capture, to reduce down-time and increase inspection time due to its fast electronic scanning compared to a conventional single element transducer, allows the elimination of multi-angle inspection through the introduction of beam skewing as a post-processing operation for second layer crack detection. Finite Element Analysis is used to illustrate phased array beam profiles to ensure the adoption of the optimal experimental set-up. This paper reports on using single pass data acquisition and advanced processing for second layer crack detection. 2 of 21

Theory elements elements elements (c) Figure 1 Array configurations with stepped scanned, electronically focused and (c) steered beams 3 of 21

Fig. 1 shows a schematic diagram of a linear array connected to a multiplexer (muxer) and a pulser-receiver unit where by exciting a small group of elements at the same time results in the generation of a plane wavefront. The RF waveform is collected when echo wavefronts arrive back at the same group of elements and processed back through the multiplexer. This allows the scan to be processed by region and therefore beams can be moved out to cover a larger inspection area. Electronic focusing (illustrated in Fig. 1) is an improvement on the step scanning because the elements are no longer connected to the multiplexer. In its place a differential delay profile is used for excitation to each element such that the focal position can be determined by the delay profile. As the transmit excitation is conducted, the echoes move out and return from a progressively further range and so the same reciprocal operation is performed on receive but in this case the delay profile can be changed with time in order to track the focal location from a variable depth. This is termed dynamic receive focusing. If the focusing characteristics and a differential delay that produces an effective beam skew are combined (Fig. 1(c)), the beam can be steered at angles away from the 0 incidence and thus steer and focus through the desired field of interest. With dynamic focusing, by progressively expanding the aperture and changing the delay profile with time the angle subtended throughout the field must be kept within the acceptance angle of the individual elements. Electronic beam skewing can be achieved since acquired backscattered pulse-echo waveform data can provide time-of-flight information, so it is possible to locate defects in the direction of beam propagation by interpreting the waveforms in the time-domain. With FRD capture, continuous full volumetric data is collected. Conventional detection of defects at a specific orientation usually requires the array to be positioned for scanning that direction. However, FRD capture allows a data set to be collected in one orientation and beam skewing to be applied in the post processing stage. All array systems exhibit grating lobe and sidelobe artifacts but the FRD reconstructed images will show no additional anomalies as long as the FRD data does not saturate. By considering array parameters such as element pitch and by increasing the number of array elements, these effects may be minimised. In this study only a small skew angle is required to produce the required effective horizontal skew angle because fastener cracks are most likely to be found, at maximum, at angles of ±50 from the scanning orientation. A key advantage of FRD capture is that the facility for dynamic focus on transmit and receive does not cause a significant effect on the SNR of the steered beam. Fig. 2 illustrates a LabVIEW finite element simulation of an intensity plot showing the amplitude of the ultrasonic field with and without beam skew and then repeated to show the effect of beam skew with the element pitch reduced by half for a 45 beam angle. Varying the element pitch of the 5 MHz array demonstrates that although a reduced element pitch of 0.53 mm increases the main lobe width, a tighter focus is achieved with reduced grating lobes. 4 of 21

(c) (d) Figure 7.13 LabVIEW FE simulation for a 45 beam angle element pitch 1.1 mm with no beam skew, element pitch 1.1 mm with 6.5 beam skew (c) element pitch 0.5 mm with no beam skew and (d) element pitch 0.5 mm with 6.5 beam skew. In all cases, element width 0.5 mm A H A θ R i (skew = 0) R i S θ R i (skew = 0) H O S O H O H θ A H R i S O 90-i 5 of 21

Figure 3 Schematic of skew angles where A H is the array height, A θ is the array angle, S θ is the skew angle, H θ is the horizontal skew angle, Ri (skew = 0) and Ri are the distances to the angle of incidence, S O is the skew angle offset and H O is the horizontal offset With variable input parameters of array height (A H ), array angle (A θ ), skew angle (S θ ), stand-off velocity (v s ) and target velocity (v t ), the derived angles corresponding to skew angles of the array (Fig. 3, 4 and Table 1) can be obtained using the following equations: A R i (skew =0) = H.(1) cos A θ R i = A H cos A cos S θ θ..(2) A S O = H tan( Sθ ) cos A...(3) θ H O = A H tan ( A θ )......(4) Using (3) and (4) H θ = tan 1 S H O O. (5) The angle of incidence of the skewed beam is 1 AH i = 90 - cos... (6) Ri Using Snell s Law and (6), the angle of refraction is v R = sin 1 v T s sin i.. (7) From the above equations, the horizontal distance x, from the array to target can be derived as Td x = Ri cos(90 I) + where I would take into account the skew angle (8) tan(90 R) or 6 of 21

AH Td x = +..(9) tan( 90 I) tan(90 R) Array normal R i A H i interface r T d Target Figure 4 Schematic diagram of the array and target position at depth T d with unknown horizontal distance x, from the array to the target The standard array consists of 64 element pairs with a 0.53 mm pitch producing a 33 mm effective array length and the full capability of the array will produce an array length of 67 mm. To allow the full capability of the array, an adapter is used to electronically halve the element pairs to enable the optimal element pitch. Using equations (1) - (9), Table 1 illustrates that the defects in the specimen in the third layer of the specimen should be detectable at a depth of 20 mm. Table 1 Skew angle calculations for maximum target depth of 20 mm KNOWN VARIABLES Array angle ( ) Array height (mm) Skew angle ( ) 19 25 6.5 v (stand-off) (mm/μs) v (target) (mm/μs) Target depth (mm) 1.48 3.2 20 CALCULATED PARAMETERS x Distances to angle of incidence Skew Angle Offset R i (skew =0) (mm) R i (mm) S o (mm) 26 27 3 Horizontal Offset Horizontal offset angle, H o (mm) H θ ( ) 9 19 Incidence (i) ( ) Refraction (R) ( ) Scan length, x (mm) 20 42 27 7 of 21

Specimens 361 362 363 364 365 367 366 (c) 368 369 370 Formatted: Centered Back of Skin (8 mm deep) 1.5 mm 1 mm 2 mm 361 363 365 367 369 360 362 1.5 mm 2.5 mm 1 mm 2 mm 3 mm Back of Spar (12/14 mm deep) 2 mm 2 mm 3 mm 30 3 mm 40 3 mm 50 30 40 364 366 368 1 mm 2 mm 1.5 mm Back of Strap (18 mm deep) 5 mm 10 mm 2.5 mm 3 mm Back of Spar (12 mm deep) Front of Spar (16/20 mm deep) Left of this line, spar is 4 mm thick, right of the line it is 6 mm thick (d) Figure 5 Wing skin/spar/strap specimen with top-down view, side view, (c) close-up view of fasteners and (d) defect locations and profile of fasteners. Proprietary of QinetiQ Ltd. 8 of 21

0 & 180 330 & 150 A 1 st 9 2 nd 5, 23 B 3/16 th inch ~ 4.7625 mm 1 st 2, 6, 8, 10 2 nd 14, 18, 22 1 st 7 2 nd 15 C 1 st 3, 10, 11 2 nd 21 SCANNING ORIENTATION 330 270 210 0 30 90 180 150 Formatted: Centered D 1 st 5 2 nd 19 1 st 4, 2 nd 16, 20 F E 30 & 210 Number Inch mm Fastener 1 2 3 4 5 0.250 0.100 0.050 0.125 0.075 6.35 2.54 1.27 3.175 1.905 2, 3, 22, 23 6, 7, 18, 19 10, 14, 15 4, 5, 20, 21 8, 9, 16 Figure 6 P-3 wing plank splice joint specimen containing 24 fasteners and fastener defect map indicating which defects should be seen at which orientations 9 of 21

Experimental Set-up DSL s modular array hardware system (Flawinspecta) was used with a 5 MHz linear array probe canted at 19 in a water-filled acoustic stand-off to produce 45 shear waves The scanning position of the array was recorded with a freehand string pot encoder at a step size of 0.1 mm (Fig. 7). Figs 7 and (c) indicate the scanning orientation relative to the wing skin/spar/strap and P-3 wing skin/spar specimens respectively. The effect of reducing the element pitch from 1 mm to 0.53 mm is shown in the B-scan of a carbon fibre composite specimen with side-drilled holes (Fig. 8). 90 0 90 0 String pot (freehand C-scan) sample guidance to ensure linearity array stand-off (c) scan direction Figure 7 Experimental set-up phased array canted at an angle of 19 housed in a water-filled stand-off attached to a c-scan trolley, probe and scanning orientation indicated relative to the wing skin/spar/strap sample and (c) ultrasonic scanning set-up for the P-3 wing skin/spar specimen. 10 of 21

Front wall Indication of side-drilled hole Back wall Figure 8 B-scan images of a carbon fibre composite specimen with side-drilled holes using a 5 MHz array with 1 mm element pitch and with an adapter to electronically reduce the element pitch to 0.53 mm. Element width is 0.5 mm Experimental Results Preliminary study of single skin specimen 90 80 70 60 50 40 30 20 crack fastener 10 0 Figure 9 Photograph of single skin Boeing sample with scanning orientation with respect to the crack indicated 11 of 21

The aim of scanning the single skin Boeing fastener sample from 0 to 90 in 10 increments (Fig. 9) was to investigate if a loss in amplitude (signal drop-off) occurs with inspection angle. With a specular reflector the typical behaviour is an exponential decrease in signal intensity with inspection angle, but a crack can have the potential to behave like a corner reflector. Fig. 10 is assuming that angles which have no noted amplitude from crack detection contain such a strong corner reflector signal within the B- scan image that it is difficult to distinguish between the echoes. Fig. 10 is displayed on an inverse axis to illustrate the decline in signal intensity from the crack with inspection angle varied from 90 to 0 (crack orientation is 0 ). This study has shown that there exists a signal drop off with inspection angle. The concern with using the 5 MHz probe to image the 0.5 mm crack is that the resolution and sensitivity of the transducer results in a strong presence of grating lobes on the C- scan images. This may become a problem if the amplitude of the lobes becomes visually greater than that of the reflection from the fastener and could potentially result in false calls. By using half the element pitch of 0.53 mm, previously unwanted side lobes are reduced and there is an increase in resolution and sensitivity. These results confirm that of the 5 MHz array with full pitch, demonstrating a reduction in signal amplitude with angle of inspection. From both sets of results, it can be deduced that a crack at a particular orientation with respect to scanning direction may be difficult to detect and that beam skewing for enhanced resolution may be a solution. Possible concerns are that cracks may contain debris giving false interpretation of size, water couplant filling fasteners and distorting signals, gating position, acquisition quality and coupling. 90 80 70 Angle (degree) 60 50 40 30 20 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Magnitude of reflection (normalised) 12 of 21

90 80 70 60 Angle (degree) 50 40 30 20 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Magnitude of reflection (normalised) Figure 10 Amplitude variation with angle of inspection for 5 MHz array with effective pitch 1 mm and 0.53 mm for inspection of Boeing single skin fastener specimen Previous research of second layer crack detection Fig. 11 illustrates fully optimised and processed images acquired using Andscan data acquisition software with the Flawinspecta technology, these scans were previously obtained and used in the automated analysis program. To explain the features within these scans the previous authors produced schematic diagrams (Fig. 12) to demonstrate reflections expected for different passes/ propagation paths in the specimen [2]. Figure 11 C-scan images displaying conventional pulse-echo with fixed focus and as but with a larger aperture of 16 mm that performs multi-zone transmit focusing. The green circle overlay is an indication of fastener positions. Proprietary of QinetiQ Ltd. 13 of 21

(c) (d) (e) (f) Figure 12 Ultrasonic second layer crack detection schematic showing array positions half skip to back of first layer, full counter sink, (c) half-skip to back of second layer, (d) full skip to front of second layer, (e) crack at back of second layer and (f) crack at front of second layer. Proprietary of QinetiQ Ltd. Preliminary Scans of wing skin/spar/strap specimen Preliminary scans of the wing skin/spar/strap specimen were obtained to compare the resolution using a 5 MHz array with 1 mm and 0.53 mm element pitch and a 10 MHz ultrasonic array (Fig. 13). The main objective of acquiring these C-scan images was to investigate if it was possible to replicate the images acquired by the previous authors [2] by using the 5 MHz array with 0.53 mm element pitch. Figs. 13 and show a 10 MHz inspection with the beam focused in the first layer with apertures of 16 mm and 31 mm respectively. Following these images are Figs. 13 (c) and (d) obtained with the 5 MHz array with 0.53 mm element pitch and lastly (e) shows the same array but with an element pitch of 1 mm. The reduction in element pitch of the 5 MHz array allows the resolution of the C-scan images to be similar to that of the 10 MHz array and therefore allowing any indications on the C-scan images to be analysed using the approach in Fig. 12. 14 of 21

365 366 367 368 369 369 365 366 367 368 369 361 362 363 364 365 366 367 368 369 (c) 361 362 363 364 365 366 367 368 369 (d) 366 367 368 369 (e) Figure 13 FWC C-scan images of wing specimen (Fig. 7) scanned at a 90 orientation displaying timeof-flight (TOF) and amplitude (AMP) data combined and 10 MHz array (4 mm focus) with aperture of 16 mm and aperture of 31 mm, (c) and (d) 5 MHz array with 0.53 mm pitch with (c) aperture of 16 mm (4 mm focus) and (d) aperture of 31 mm (12 mm focus) and (e) 5 MHz array (4 mm focus) with 1 mm pitch with an aperture of 31 mm FRD scans of wing skin/spar/strap specimen Due to the size of the data files required for FRD capture, it was not possible to scan the full fastener area of the wing skin/spar/strap specimen in one sweep using single channel data acquisition. Therefore, to reduce scan time and the scan file size, the specimen was scanned in sections. To optimize defect detectability, indications (excluding geometric features from the fastener reflection) visible on the B-scan images were time-gated for further analysis to identify potential defects within the layers. The fastener resolution decreased due to focusing within the layers (as in Fig. 16). Therefore, in the first instance 15 of 21

the focus was set to the minimum value to see fastener position and to have an idea of the fastener size and resolution with the adapter used on the 5MHz array. In Fig. 14, beam skewing at ±6.5 (20 effectively in the sample), enables vertical cracks at the top and bottom of the fasteners and the angle cracks at 40 and 50 numbered 368 to 366 in the strap (third layer crack detection) to be matched to the specification (Fig. 6). With second layer crack detection, by skewing -6, the vertical crack in 364 is dominant in the C-scan in Fig. 15. The other features suggest the cracks at 30 and 40 are also visible. By gating the features around fastener 361, the vertical cracks can be observed when a 6 beam skew is applied (Fig.16). 40 367 366 368 SKIN STRAP 50 40 STRAP (c) 0 SKIN 0 (d) STRAP 0 0 STRAP STRAP (e) (f) Figure 14 C-scan images of fasteners 366-368 with no beam skew (4 mm focus) no beam skew (18 mm focus) TOF&AMP (c) no beam skew (18 mm focus) AMP (d) -6 beam skew (4 mm focus) (e) 6 beam skew (18 mm focus) TOF&AMP (f) 6 beam skew (18 mm focus )AMP 16 of 21

40 365 0 30 40 364 366 0 SPAR SPAR 0 Figure 15 C-scan images of fasteners 364-366 with no beam skew (12 mm focus) and -6 beam skew (12 mm focus) TOF&AMP. 361 0 0 362 180 SKIN SKIN Figure 16 C-scans of fasteners 361-362 with no beam skew (4 mm focus) and 6 beam skew (8 mm focus) TOF&AMP. FRD scans of P-3 wing skin/spar specimen Initially investigated was inspection angle with depth variance for optimized imaging without using FRD capture. A focal depth variation study at 4 mm, 5 mm, 6 mm, 8 mm and 10 mm showed that the focal depth was not significant in detection of defects but the 5 mm scans provided better results than those scanned with a 10 mm focal depth. The gain applied in all cases was 44 db. Scans also taken at 10 MHz did provide higher resolution results but were not necessary as defects were also seen using 5 MHz imaging with a 0.5 mm element pitch. To eliminate scanning at multiple angles and indeed focal depths, the FRD approach was implemented as above in the wing skin/spar/strap structure (Figs. 17 to 21). Thresholding could have been applied to eliminate background noise and make the images appear clearer but this could have caused problems if the threshold level was similar to that of weak defect responses. These experimental results has provided evidence that by applying beam skew in post-processing is an effective technique for detecting all defects in the P-3 wing skin/spar specimen. Figure 17 Ultrasonic inspection of fasteners 1-5 using FRD capture (aperture 16 mm and focus 4 mm) with highlighted regions indicating time-gate positions 17 of 21

Figure 18 Ultrasonic inspection of fasteners 6-10 using FRD capture (aperture 16 mm and focus 4 mm) with highlighted regions indicating time-gate positions (c) (d) Figure 19 Ultrasonic inspection of fasteners 11-15 using FRD capture (aperture 16 mm) with highlighted regions indicating time-gate positions 4 mm focus with no beam steer, 4 mm focus and 6 beam steer, (c) 8 mm focus and 6 beam steer and (d) 4 mm focus and 6 beam steer scanned in opposite direction 18 of 21

(c) (d) (e) (f) (g) Figure 20 Ultrasonic inspection of fasteners 16-20 using FRD capture (aperture 16 mm) with highlighted regions indicating time-gate positions 8 mm focus with 6 beam steer, no steer, (c) 8 mm focus with -6 beam steer, (d) as but with a Hanning filter, (e) and (f) as but with 3.5 beam steer and (g) as but no beam steer 19 of 21

(c) (d) Figure 21 Ultrasonic inspection of fasteners 21-24 using FRD capture (aperture 16 mm) with highlighted regions indicating time-gate positions 8 mm focus with no beam steer, as with 6 beam steer, (c) 5 mm focus with no beam steer and (d) as with 3.5 beam steer Conclusions From a theoretical perspective, the advantages and disadvantages of implementing beam skew rather than manual orientation movement have been described and illustrated by geometrical analysis leading to the development of LabVIEW software that calculates the effective beam skew in the test specimen. This paper has additionally presented the concept of extended beam skew for crack detection on two wing structures. Experimentally, through inspection of a single skin fastener sample with a phased array acquisition system, confirmation can be given to verify that there exists a signal magnitude decrease with inspection angle. A phased array acquisition system using Full Raw Data (FRD) capture has been implemented to show that by using one single pass at a single orientation, advanced post-processing allows for the possibility to detect fastener cracks at a wider range of angles. This effectively would increase inspection speed, area coverage and in the long term, equipment durability. Future Work One of the main improvements for scanning efficiency would be to introduce a multichannel data acquisition card for optimal scanning speed. Awareness of the technique should allow for the application of the beam skew technique to blind in-service samples. 20 of 21

References 1. D I A Lines, J Skramstad, R A Smith, 2004, Rapid, low-cost full-waveform mapping and analysis using ultrasonic arrays, Proceedings of WCNDT 2004, Montreal, 30 August 3 September, 2004. 2. R A Smith, D Edgar, L Jones and D Percivall, An Ultrasonic Solution for Second Layer Crack Detection, Proceedings of NDT 2004, Torquay, 14-16 th September, 2004. 3. R A Smith, D Edgar and L D Jones, "Rapid second-layer crack detection using phased arrays and automated analysis," Proceedings of 8th Joint DoD/FAA/NASA Aging Aircraft, Palm Springs, Jan 2005. 4. D I A Lines, I G Pettigrew, J A Skramstad, K J Kirk, S Cochran, 2006, Rapid Distributed Data Collection and Processing with Arrays the next step beyond Full Waveform Capture, Proceedings of Aging Aircraft 2006, Atlanta, 6 9 March, 2006. 5. D I A Lines, J M Bending, R A Smith, 2006, "Novel Applications of Ultrasonic Arrays for Aerospace NDT", Proceedings of Aerospace Testing Expo - Europe, Hamburg, 4-6 April 2006. 21 of 21