Phased Array Scanning of Artificial and Impact Damage in Carbon Fibre Reinforced Plastic (CFRP)
Channa Nageswaran, Colin R Bird and Reiko Takahashi
Paper presented at the BINDT conference in September 2005
Carbon Fibre Reinforced Plastic (CFRP) materials are increasingly being used in modern civilian aircraft for structural applications. There is an urgent need to detect and classify delamination defects and evaluate the threat to the integrity of the component in-service. This paper presents a study to evaluate the possible use of ultrasonic phased array (PA) technology.
Linear array ultrasonic probes are used to scan CFRP panels that contain both Teflon simulated delaminations and Barely Visible Impact Damage (BVID) caused by drop weight impacting. A comparison is made between the use of phased array scanning as opposed to conventional single-crystal transducer scanning. The advantages and disadvantages of using PA technology for thin composite material components are discussed.
This paper results from work carried out on the CRAFT Framework V project NANOSCAN (G4ST-CT-2002-5028). The aim of the project was to study the effectiveness of several current NDT techniques for the inspection of aerospace composites. The techniques evaluated include air-coupled ultrasonics, thermography, shearography and resonance testing. The materials of primary concern were CFRP laminates and honeycomb structures. A range of defect types can degrade the quality of both these composite materials and it was the aim of each technique to both detect and classify these defects.
1.2 Carbon fibre reinforced plastics and the critical defects
The manufacturing process (hand lay-up) can lead to several flaws:
- Foreign body inclusions (backing film, dust).
- Voids (entrapped air, moisture ingress).
- Incorrect lay-up order (fibre/ply misalignment).
- Incorrectly applied curing cycle.
- Bonding failure between plies during cure.
In service damage primarily leads to delaminations that could arise from voids and inclusions introduced during fabrication (but missed by QA procedures) which then act as initiator sites once the component becomes a load bearing structure. A major form of damage is induced by relatively low velocity impacts on the component (dropping tools on the wing skin, bird strike on aircraft during flight). These events lead to extensive subsurface delaminations without any obvious visual clues on the impacted surface; this is often referred to in industry parlance as Barely Visible Impact Damage (BVID).
1.3 Phased array (PA) equipment
From its initial NDT applications in the nuclear industry, phased array techniques are being considered for use in different fields and the NANOSCAN project evaluated the use of PA technology for composites in the aerospace sector.
The R/D Tech FOCUS 32/128 PA system was used in this project. It is able to build an active aperture using a maximum of 32 elements and address 128 channels. The Tomoview (Version 2.2 Release 9) data collection software was used for data representation and analysis.
All the specimens in this project are based on the aerospace grade prepregs ('pre-impregnated' continuous fibres in resin) AS4/8552 and HTS/6376 manufactured to aerospace specifications (HS-CP-5000)  . The AS4 (12K fibre count/tow) was infused in a part cured epoxy matrix to derive the 8552 system; it is tough in resisting impacts and is highly damage tolerant. It is amine cured to increase the toughness of the epoxy resin and to operate in temperatures of up to 121°C.
Impact damage was introduced to the Q-ID specimen using the Rosand Instrumented Falling Weight Impact Tester (IFWIT). Energy of impact was controlled by changing the height of the drop weight (potential to kinetic energy). The specimen was rigidly supported on a rectangular window of 100 mm x 110 mm; the impactor head was 20 mm in diameter and weighs 4.72 kg. After several trials 3 mm thick neoprene padding was placed over the impact point to prevent surface damage; the neoprene would have absorbed some of the impacting energy and hence the imparted energy was less than that calculated. Since the aim was to study the characteristics of the impact damage rather than the conditions and mechanisms by which it took place, this discrepancy is not expected to be an issue.
Artificial delamination damage was introduced using Teflon inserts before final cure. Porosity was introduced into several specimens during manufacture by not adhering to the recommended curing cycle. Table 1 summarises the specimen details.
Table 1. Specimen details and lay-up
( Key: QI - Quasi-isotropic; UD - Unidirectional; CP - Cross-ply; A - Artificial Delaminations; I - Impact Damage; P - Flawed Cure Cycle induced Porosity).
A number of specimens were destructively evaluated to measure and confirm the presence of impact induced and porosity defects. The length of delaminations and their through thickness positions, as measured using micrographs, were compared to ultrasonic evaluations (both PA and conventional).
A diamond saw was used to section specimens and subsequently polished to a 0.25 µm finish and mounted in CitoFix. A digital optical microscope is used for imaging (x1000 maximum magnification).
Three types of defects were investigated ultrasonically in this project. During hand lay-up of the specimen Teflon inserts were deliberately inserted between plies. These are known to artificially simulate inter-ply delamination and are also representative of manufacturing inclusions type defect. The acoustic impedance mismatch between the bulk CFRP and the Teflon should lead to their detection. The N-series panels are manufactured with entrapped air; this leads to the introduction of gross porosity where there is impedance mismatch between the material and air. Impact induced damage is known to lead to severe delamination below the impact surface. The panels are scanned after the introduction of impacts ranging from 30 to 70 Joules. We noted that on an 8 mm (64 plies) thick quasi-isotropic panel 45 Joules represented the threshold for initiation and perforation took place around 60 Joules ( Figure 1). However, as noted in the literature, the threshold values are dependent on many factors and researchers have found a very large variation [2,3] .
CIVA  and Quicksonic  modelling software are used to theoretically check the integrity of the incident PA sound field and remove unexpected grating lobes. An example of using CIVA is given in Figure 2 for the probe used in collecting the data presented in this paper (7 MHz, 16 element aperture, 0.6 mm pitch).
Fig. 1. Damage size related to energy of impact under present set-up scenario
Fig. 2. 16-element linear aperture with 0.6 mm pitch; calculated field for point focussing without significant grating lobes
All the data presented in this paper were collected in pulse echo mode (both conventional and PA) because more information can be derived about the through thickness features of the material (as opposed to through transmission). Furthermore, in-service inspections of aircraft components may allow one-sided access only and the scanning frames required for pulse echo testing are less complicated and more adaptable. Noting that delaminations occur between plies and are always parallel to the surface, 0° incidence testing offers the best inspection strategy (in terms of defect detection and coupling efficiency).
An important aspect of the present work is the need to immerse the specimens. When several elements of an array are phased time delays are applied to focus the resultant sound beam within a desired zone (in our case at 0° incidence on to the specimen). The number of elements (hence the active aperture) determines the distance required in the medium for effective beam forming to take place ( Figure 2). A practical value for the minimum focal range distance is given below:
is the distance in medium, A
is the active aperture length and λ is the wavelength in the medium. CFRP specimens are too thin to allow beam forming (1 to 8 mm); furthermore, the anisotropic properties (highly dependent on lay-up) re-distribute the sound energy unfavourably. Hence, to use a linear array in a phased configuration the specimen must be immersed in water so that the beam can be effectively formed for inspection.
The need for immersion is noted to be a disadvantage in both post-manufacture and in-service inspection as components could be susceptible to damage by moisture ingress.
3. Results and Discussion
3.1 Material characteristics
For the current study, an accurate value for the longitudinal velocity of the sound in the material was evaluated to be 2850 m/s with a standard deviation of 1.3%. Similarly the attenuation on the sound travel (combined absorption and scattering) was evaluated as a function of sample thickness and frequency; this then allows us to choose an appropriate frequency for the study.
3.2 Artificial delaminations
For aircraft in-service, the maximum allowable defect size (especially delamination) before it is considered to be beyond airworthiness depends on several factors:
- Design philosophy (Fail safe/Safe life).
- Material type.
- Impact risk zones (dropped tools, maintenance traffic, runway debris etc.).
Similarly the criteria relating to monolithic components is different to sandwich components due to differing damage morphology  . A typical critical delamination size in CFRP wing skin exposed to direct airflow is 10 mm x 10 mm for a modern fighter aircraft. The -6 dB drop method is used for sizing  ; here a locus line tracing a 6 dB drop in the amplitude from the maximum value over the defect represents the defect size.
Figure 3 shows a scan of the N1-D specimen. It shows the presence of the artificial delamination and the presence of through thickness gross porosity. The PA system is able to impose the -6 dB sizing criterion for both the delamination and the porosity regions automatically. The through thickness positions can also be evaluated accurately.
Fig. 3. 32-element linear array scan of N1-D showing artificial delamination and gross porosity (B-scan to the left, C-scan on the right)
3.3 Manufacture induced porosity
A study of the morphology of porosity in three different lay-ups shows that the weave of the material plays an important role in the degree of damage. Figure 4 shows C-scans of N2-D and N3-D that are of cross-ply and quasi-isotropic lay-up, respectively. All the panels have flawed cure cycle induced porosity. Three equal volumes of ultrasonic data are extracted for the three lay-up specimens of 1.4 x 180 x 100 mm 3 .
The maximum echo amplitude within this volume above 20% of the front wall echo was plotted on a two-dimensional map; hence the area of the total cross-section covered by porosity is evaluated by image thresholding ( Figure 5). The total volume of porous region is greatest in the cross-ply panel then the quasi-isotropic and least in the unidirectional laminate; furthermore the fibres tend to 'guide' the growth direction. Porosity is created due to entrapped air between the fibres and can be used to deduce the lay up of a laminate  .
Fig. 4. Entrapped porosity in three lay-up schemes
Fig. 5. A two-dimensional map of the porosity damage for unidirectional (top), cross-ply and quasi-isotropic laminates
The porosity levels in the unidirectional, cross-ply and quasi-isotropic laminates are 6.94%, 19.83% and 10.33%, respectively.
3.4 Impact induced delaminations
It was confirmed that impact damage exists at different levels through thickness; this is done by a time-of-flight (TOF) analysis ( Figure 6).
Fig. 6. TOF showing multiple level delaminations (55 J impact energy)
The delamination near the back wall is larger than the ones near the front wall (where the impact took place). When impacted there is rapid local bending at the impact point ( Figure 7).
Fig. 7. Local deformation scenario on impact
The in-plane shear stresses created between successive plies in different orientations leads to a break down in the bonding, creating delamination; the mismatch in bending stiffness has been directly related to the severity  . It is also possible to deduce the major orientation directions (0°, +/- 45°, 90°).
Figure 8 shows a PA amplitude scan. RF data is collected at a digitising frequency of 100 MHz. The probe is a 16 element 7 MHz linear probe. Gates can be used to select delaminations at different levels. The data file is modified to remove repeat echoes that result in a misleading picture of the damage inside the material and artificially extend delaminations; currently there is no algorithm in the software to perform this task and hence repeat echoes were manually identified and deleted.
Fig. 8. Corrected C-scan showing delamination at three levels
A study was carried out to confirm the -6 dB ultrasonically determined sizes of the delaminations in the sample by sectioning. A through thickness datum plane was identified accurately and corrected depth and size measurements were made of the delaminations. The sample was then sectioned and micrographs were taken at magnifications from x16 to x1000 (e.g. Figure 9).
Fig. 9. Delamination damage between two plies (45° (top) and 90°) showing typical measured wall gap in Q-ID-2 (Magnification x1000)
Table 2 shows that PA testing is able to locate and size multiple level delaminations within an acceptable margin of error (satisfying the minimum size criteria).
|Ultrasonic -6 dB Length|
It was not possible to transmit significant energy beyond the first incident delamination. For sound propagation in a medium (both planar and spherical) the particle vibration amplitude is related to the sound pressure using the following relation  .
is the sound pressure (Pa); ρ is density (kg/m 3
); ω is angular frequency (rad/s) and is particle displacement (m).
A further quantity of interest in understanding plane and spherical sound propagation is the intensity of the sound in W/m 2  .
Where Z (= ρc) is the acoustic impedance. The above is then rearranged for particle displacement, as follows:
This feature of sound propagation in solids is used to detect delaminations in CFRP. The gap between plies (where disbond leads to delaminations) is reported to be around 0.7 µm  , whereas we measured an average of 3.24 µm on sectioned specimens. Assuming bulk measured longitudinal velocity of 2850 m/s and a frequency of 5 MHz., the wavelength is 0.57 mm. Hence the particle displacement is around 0.00000114 mm (2x10 -6 of wavelength)  . This implies that at the face of the delamination the interface will see a free boundary and almost all the energy will be reflected back to the transducer as an echo.
Fig. 10. Delamination gap is sufficient for the propagating wave to interact with the incident wall as a free boundary
Due to this effect it is not possible to penetrate past the first incident delamination to those in its shadow; all the energy is then confined between the first incident delamination and the front wall to create repeat echoes. Multiple level delaminations are detected by virtue of the conical damage feature (those near the back wall being larger in area).
Components made out of CFRP are often thin (1 to 15 mm). Hence short broad band pulses are needed to resolve features along the axial direction. Conventional probes have a -20 dB pulse length between 200 and 800 ns whereas the PA probe used in this study had a measured value of 1300 ns. Since the delamination defects and porosity are parallel to the front wall and the components are thin, the beam steering feature of the PA is redundant. Similarly focussing within the thickness has been shown to do little to better illuminate defects near the back wall (for an 8 mm thick specimen).
3.5 Practical implementation
The specimens are immersed to allow sufficient coupling distance for effective beam forming. In-service inspection of aircraft components would require dismantling and provision for large immersion tanks that will increase the overall cost. When collecting RF, the data volume transfer is a limiting factor on the size of the array that can be used effectively.
A major known advantage of using array transducers (not necessarily working in phased array configuration) is the increase in scanning speeds that can be achieved. A scan of a 100 mm x 100 mm panel with an 8 MHz conventional and 7 MHz PA probes, motor speed of 5 mm/s and resolution of 0.6 mm x 1 mm was set up. The total scanning time for conventional was 33 minutes while it was 1 minute 46 seconds for PA. Noting that these results are highly probe (length of array) specific, it is a good illustration of the performance increases that can be achieved with an array.
- Both delamination and porosity damage can be effectively detected.
- Total time to complete scans can be reduced very significantly when an array is used (not necessarily in phased mode).
- It is possible to create live images of component sections to study damage growth characteristics and mechanisms (fatigue loading).
- Active aperture size of the probe can be altered electronically to improve sizing (beam spot size) depending on the precision required.
- Component needs to be immersed for effective beam forming; beam steering and focussing within the material volume are redundant.
- Broadband probes need to be used to increase axial resolution (current PA probes tend to have long pulse lengths to improve the lateral resolution).
- The EC for part funding the NANOSCAN project.
- All partners of the consortium for the successful completion of the project.
- EPSRC for funding the studentship at TWI/University of Birmingham.
- Colin Bird and John Rudlin for helpful suggestions concerning the subject matter.
References and footnotes
- Hexcel, Prepreg Datasheets, www.hexcel.com
- S Abrate, 'Impact on laminated composite materials', Appl. Mech. Rev., Vol 44, No 4, pp 155-190, 1991.
- P Robinson and G A O Davies, 'Impactor mass and specimen geometry effects in low velocity impact of laminated composites', Int. J. Impact Eng., Vol 12, No 2, pp 189-207, 1992.
- CEA, CIVA: Simulation software for Non Destructive Testing, www-civa.cea.fr, France.
- Imasonic, Quicksonic v.4.01, www.imasonic.com, France.
- MMS 13, 'Assessment & Criticality of Defects & Damage in Materials Systems - Task 1 Review', www.mms15.com, 2003.
- R A Smith, 'Ultrasonic defect sizing in carbon fibre composites - an initial study', Insight, Vol 36, No 8, 1994.
- R A Smith and B Clarke, 'Ultrasonic C-scan determination of ply stacking sequence in carbon-fibre composites', Insight, Vol 36, No 10, 1994.
- D Liu, 'Impact induced delamination - a view of bending stiffness mismatching', J. Compos. Mater., Vol 22, pp 674-692, 1988.
- P A Lagace, 'On delamination failures in composite laminates', Composite Structures: Testing, Analysis and Design, Narosa Publishing House, pp 111-132, New Delhi, 1992.
- J Krautkramer and H Krautkramer, 'Ultrasonic Testing of Materials', Springer-Verlag, 1990.