Phased Array inspection system applied to complex geometry Carbon Fibre Reinforced Polymer parts

Transcription

1 Phased Array inspection system applied to complex geometry Carbon Fibre Reinforced Polymer parts André Cereja Instituto Superior Técnico, Lisboa, Portugal May 2015 Abstract Following the growing use of composite materials in the aerospace industry, particularly carbon fibre reinforced polymer (CFRP), arises the need to develop procedures to guarantee the monitoring and evaluation of CFRP components. Moreover, the cost associated with destructive testing should be eliminated, using methods that do not harm the parts service life. Ultrasonic testing methods, namely the advanced phased array technique, can be a useful tool for dealing with the challenges engineers face, when having to perform evaluation procedures for components subject to fatigue and/or impacts. A simulation software, CIVA, is used before the physical inspection of both planar and complex geometry parts is performed. With CIVA, the phased array probes are selected, together with the inspection parameters, e.g. probe aperture and focal laws. A study is performed on the methods CIVA employs to calculate the acoustic field in the parts, as well as the homogenization algorithms behind the handling of the anisotropy of CFRP. The difficulties expected when performing the physical inspection, namely attenuation, are predicted. For the complex geometries, an evaluation is done on the potential to inspect these components. Afterwards, physical inspections are performed. Three components are analysed: a testing specimen with embedded defects, a component subjected to fatigue loading, and a reinforcing omega-stringer. The defects (delaminations) in the test specimen are identified and characterized, in size and depth. Lacks of resin and debondings are discovered in the fatigue specimen. The stringer, impacted with known energies, is analysed and the resulting flaws identified and measured. Keywords: carbon fibre reinforced polymer, non-destructive testing, phased array ultrasonic testing, selfadaptive algorithm. 1. Introduction Following today s transportation industry design requirements, namely in the aerospace sector, the demand for carbon fibre reinforced polymer (CFRP) is forecasted to double by 2015, following the aim for an improved fuel efficiency and the need to use high-strength lightweight materials. [1] Allied with the use of CFRP, comes the need to mitigate the risk of catastrophic component failure. It is crucial to evaluate the failure modes associated to the use of composites, either coming from the design, fabrication or service phase. Non-destructive testing (NDT) methods come as the most widely used option to identify and characterize composite defects. This work intends to respond to the need to analyse internal defects in CFRP parts, specifically in the complex geometries which are very common in the aeronautical industry. A CFRP composite is by definition anisotropic, i.e. it presents different mechanical properties in different directions. This anisotropy and the importance of some mechanical properties comes into play when applying NDT procedures, which see their process characteristics influenced. The generalized Hooke s law for an elastic anisotropic material in the Voigt-Kelvin notation is: We have that: {σ i } = [C ij ]{ɛ i } (1) {σ i } stress components [C ij ] material coefficients {ɛ i } strain components All the indexes refer to an orthogonal Cartesian coordinate system, (x 1, x 2, x 3 ). The compliance matrix is obtained from the stiffness matrix and is its inverse: [S ij ] = [C ij ] 1. An anisotropic material is characterized by 21 independent constants, since the stiffness matrix is symmetric (C ij = C ji ). If a material has one or more planes of symmetry for its properties, the number of constants is further reduced. [2] 1

2 There is a special derivation of the orthotropic constitutive model that is characterized by having similar properties in a plane, and different properties in the normal direction to this plane. This constitutive model, in which materials are called transversely isotropic, allows for the reduction of the independent material constants to five. [3] This model is applicable, for example, to a composite laminate, with several unidirectional layers in the same plane, oriented in multiple directions (0, 90, ±45 ). It can also be applied to a single fibre, having similar properties in the cross sectional plane and different ones in the longitudinal direction. If these transversely isotropic fibres are used in a unidirectional lamina, it can also be considered transversely isotropic (properties in directions x 2 and x 3 of Figure 1 are equal). Figure 1: Unidirectional lamina coordinate system The use of this model allows for a reduction of the number of tests necessary to find the material constants. These five constants are: E 1, E 3, ν 13, ν 21, G 13 This transversely isotropic model is subjected to a homogenisation routine in the commercial software CIVA NDE 11 (a simulation tool developed specifically for NDT applications), with the aim to predict the ultrasonic wave propagations and defect interactions in the composite components. The most widely used NDT procedure nowadays for the inspection of CFRP structures is ultrasonic testing (UT). This method is based on the transmission of high frequency sound waves; in the industry, these are usually between the 500 khz and the 20 MHz range. The use of this testing method targets the detection and characterization of defects in the composite, providing information about their location, depth, size and orientation. The sound wave velocity is a characteristic of the medium where it is travelling, and is thus dependent on parameters obtainable from the [C ij ] matrix. As sound travels in a medium, its intensity (power per unit of area) decreases accordingly to an inverse law. This means that the sound intensity at a given point is reduced by 50% as its distance from the source doubles. This intensity in ultrasonics is measured in decibel (db), and it is considered to have a level of 0 db for the limit of human hearing. Intensity is correlated to the amplitude of a signal, in the instruments used to carry out NDT. Another important aspect to take into account when performing an ultrasonic inspection, is the sound attenuation in the material. This attenuation is a decay in sound pressure that is caused by three main reasons: wave spreading, scattering and absorption effects. The wave spreading is a characteristic of the sound wave itself and it relates to the near and far field concepts. The scattering effects come from non-homogeneity of the materials, causing a reflection of the wave in the material boundaries, for example in the fibre-matrix interface. In the case of CFRP, the absorption comes mainly from the visco-elastic losses in the epoxy matrix. These losses increase with the increase in the sound wave frequency. The phased array advanced ultrasound technique (PAUT), concerns the use of multi-element probes, together with different control algorithms. The principle behind the phased array technique is to activate for each shot all or some of the transducer elements which, with the adapted delay laws, contribute collectively to the generation of the beam. [4] Contrary to the mechanical translation of a single element probe, this technique allows to perform electronic scanning steps, with a single transducer position. Moreover, both electronic and mechanical steps can be combined, increasing the overall inspection efficiency, reducing the overall inspection time, and thus, cost. To analyse the complex geometries of certain CFRP components, the algorithm Self-Adaptive Ultrasonic Technique (SAUL), implemented by the company M2M in their PAUT systems, is tested and validated. This algorithm measures the time an ultrasonic wave takes to hit a reflector and to return to the probe, known as time of flight (TOF), for the entire range of elements being used. Then, this reception law is applied as an emission law, so that the emitted sound wave matches the shape of the part to inspect. This process is done in real-time, iteratively, until the incident wave matches the front surface, and the iterations converge. The SAUL algorithm needs a maximum of four iterations to converge, and it can adapt to both concave and convex shapes. [5] 2. Methods One of the components evaluated in this work is a CFRP multiple ply test specimen, with 25 embedded Teflon R defects, which is represented in Figure 2. 2

3 On side A, the flat panel s frontal side, two types of monitoring sensors can be seen: the first type consists of 10 optic fibre sensors, indicated by 1. The other type of sensor consists of two sensors (2.1 and 2.2), embedded in the composite, and indicated by the green rectangles. On the back side of the panel it can be seen a series of electrical wires, held in place with tape, and used to collect the information from a set of eight strain gauges attached to the panel with adhesive. These strain gauges are positioned inside the blue rectangles; some are placed near the impact locations, marked by the red x symbols. The stringer in the bottom is the one analysed in this work, together with the impact damage it sustains, indicated by 4.1 and 4.2. After the impact tests are done, the panel is submitted to a series of fatigue tests. The panel is stressed in sequences of 5000 tensile-compression tests, with loads of ±10000 N. These tests are performed four times, with PAUT done between each test. Hence, the total fatigue cycles add up to Figure 2: CFRP calibration specimen specifications and defect locations. All dimensions are in millimetre. The analysed complex geometry component is an omega-shaped reinforcing stringer, with the generic A flat panel is also evaluated, subjected to fatigue shape and evaluated sections of Figure 4. and impact tests, which also has bonded omega stringers. This panel is displayed in Figure 3. Figure 4: Colour highlights for the several sections of the omega stringer. The CFRP components are modelled in CIVA following the material specifications available from the software s internal library, seen next, in Table 1. Epoxy Density Material constitutive model Cl Ct 1230 kg/m3 Isotropic 2488 m/s 1134 m/s Carbon Fibre 1670 kg/m3 Transversely Isotropic NA NA Table 1: Epoxy and carbon fibre material specifications from the CIVA library Figure 3: Flat CFRP panel with sensors and impact locations. Besides these parameters, CIVA also requires the definition of the composite fibre volume fraction, 3

4 V f = 61, 5% as well as a fibre filament diameter of 0,005 mm. To obtain the stiffness matrix of the composite laminates, essential for the calculation of the acoustic field, two steps are necessary: first, CIVA uses a homogenization algorithm that treats the information available from the material lamina properties, to calculate the [C ij ] entries for this single ply. Afterwards, and considering the information inserted about the stacking sequence, a second stiffness matrix is calculated via a second algorithm. The first algorithm, described in the paper submitted by Lonné et al. (2004), is based on a model that couples viscoelastic and scattering losses of energy of the sound beam travelling through a fibrous composite. This model treats the complex wave behaviour, due to a composite s anisotropy, and considers the attenuation occurring through viscoelastic and scattering losses. The material densities, fibre volume fraction and wave frequency are the main parameters that are computed to calculate a complex-valued wavenumber which is then used to calculate the entries of the stiffness matrix. Detailed information can be found in [6]. For laminates with several plies, CIVA uses another homogenization method, developed by Deydier et al. (2005), to simplify the properties of the laminate by reducing it to a homogeneous equivalent whose elastic constants are determined by a semi-analytical method. For further details, please refer to [7]. Regarding the components material specifications, and using both CIVA s default parameters and the information available from the fabrication of the component, the transversely isotropic stiffness matrix for the single CFRP ply is obtained: [C ij ] = 145, 31 5, 32 5, , 32 12, 62 5, , 32 5, 85 12, , , , 28 Using a symbolic calculation software, Maple TM 17, a calculation method is developed to obtain the engineering constants. This method takes the theory compliance matrix, inverts it to obtain the stiffness matrix and then equals the latter to the numerically filled matrix from CIVA. The Maple TM software symbolically manipulates the expressions to solve the non-linear equation system, obtaining the engineering constants. Considering the single ply, the following values are calculated: E 1 = 142, 250 E 3 = 9, 848 ν 13 = 0, 288 ν 21 = 0, 0199 G 13 = 5, 28 GPa GPa GPa These calculations are performed to understand the evolution of the material constants when modifying, for example, the V f. Considering the single ply composite, are obtained increasing linear variations of E 1, E 2, while ν 12 decreases. As for G 12 and G 23, these values increase exponentially. The obtained results are realistic, since it can be correctly inferred that a single laminate, having a higher fibre volume fraction, will possess higher mechanical properties, such as the Young s modulus. Hence, the calculations performed internally by CIVA are deemed correct. From the range of probes available in ISQ, a set of four PAUT are selected for this project: two linear array probes, with frequencies of 4 and 10 MHz, and two 2-D matrix probes, of 3,5 and 5 MHz. 3. Results and Discussion After modelling the shape and material of the CFRP calibration specimen, it is possible to obtain information from CIVA regarding the attenuation of the sound beam. CIVA provides information concerning the distribution of the energy of the sound beam of a section being evaluated, via a colour plot, in Figure 5. Figure 5: Focal spot without and with attenuation, 5 MHz probe. On the left, no attenuation is considered. A large focal spot can be seen, marked by the rectangle surrounding the area with the higher acoustic pressure (in light blue). The focal spot is calculated in relation to a -6 db amplitude loss. It is also visible that the sound beam in the cross-section on the left has a better penetration. Looking at the right image, where the attenuation 4

5 is considered, two changes can be observed: there is a reduction of 9,22 mm 2 in the focal spot size, and an upwards shift of the maximum acoustic pressure zone, causing a discrepancy with the initial objective of having a focal spot at a 5 mm depth on the part. Following this result, it is predicted the effect of the energy loss along the component s thickness, particularly the difference between the interface echo and the backwall echo amplitudes. Using half of the probe s aperture, two different setups are used: a simple electronic scanning, with null delay law, and multi-point focalization across the component s tickness. The simulation results are displayed next: Half aperture, no focalization Half aperture, with focalization 5 MHZ 10 MHZ -9,9 db -12,8 db -7,1 db -8,6 db The deeper a given defect is located in the part, the higher the amplitude loss for the higher frequency probe will be. The SAUL algorithm is not implemented in CIVA. Despite this, the generic omega stringer from Figure 4 is modelled in CIVA, in an attempt to better understand the inherent difficulties in inspecting complex geometries. The 2D-Matrix 5 MHz probe is used with half aperture, applying a multi-point focalization technique in nine points located at half thickness. An overview of the inspection setup is shown in Figure 7. Table 2: Amplitude differences between interface and backwall echoes The deepest focalization is done considering a point very near to the backwall of the P3 section of the CFRP calibration specimen, at 11,3 mm depth, to attest for the minimum possible differences between both echoes. A stronger backwall echo is obtained when focalization is used and, as such, the differences between interface and backwall echo are alleviated, thus obtaining better results and a higher signal-to-noise ratio. In Figure 6 are displayed the A-scan and B-scan obtained in CIVA for the defect number 19. These results are obtained with the 5 MHz 2-D matrix probe, using the simple electronic scanning technique, with half aperture for the emission, and full aperture for the reception. No focalization is used. Since no focalization is used, as expected the higher echo response comes from the interface water-cfrp, seen either in the large amplitude (in absolute value) of the echo, or by the light blue colour in the B-scan. Figure 7: Multi-point focalization inspection technique applied to the reinforcing omega stringer. Figure 6: A-scan and B-scan for defect number 19. When CIVA finishes calculating the averaged acoustic field for all the shots, the obtained results are those of Figure 8. It is very difficult to obtain good results for a complex geometry without the use of a solution that adapts the sound waves to the part s geometry. Even when focalizing the sound in specific points, the results obtained show that no backwall echo exists. The absence of the backwall echo harms the successful inspection of any component, rendering it impossible to know, for example, the depth of defects. 5

6 The minimum deviation that is possible to identify is as small as 0,1 mm. This value highlights the quality and resolution of the results that are possible to obtain via the use of PAUT. Dimension deviations W [mm] L [mm] Maximum 1,3 1,1 Minimum 0,0 0,1 Average 0,5 0,5 σ 0,4 0,3 Table 3: CFRP calibration specimen deviation values analysis Figure 8: Acoustic field results for the stringer. From the above results, it becomes evident the need to apply an adaptive inspection algorithm, such as SAUL. In the next page are shown the final acquisitions performed for the CFRP calibration specimen, with the 10 MHz linear probe. Figure 9 contains three views: two C-scans and a B-scan. The top C-scan reproduces the signal s amplitude. The B-scan uses the same scale, and derives from the A-A section cut indicated in the figure. The middle C-scan contains information on a time basis, with the recording of the time of flight of the sound beam. Analysing the results all defects can clearly be seen, both in plane view and along the thickness (represented in the B-scan). In the middle C-scan, as can be understood when observing the colour scale, blue defects are located near the top surface, while red defects are placed near the backwall. Around the defects, a white zone can be seen: this white colour represents a zone with no recorded echo. Due to the thickness of the defects, the fibres are deformed in this area, and the sound beam is reflected in such a way that it does not return to the probe. The same effect can be seen for the interface between the several sections of the part. Using the criteria for an echo loss of -6 db, the real dimensions of the defects are compared to the theoretical values. The obtained maximum, minimum, average and standard deviation results are shown in Table 3. Figure 10 shows the final acquisition for the fatigue panel, after the complete fatigue testing is performed. This acquisition is performed with the 5 MHz 2-D matrix probe. The signal s amplitude measurements are displayed on the left. On the right, the scale shows the thickness measurements, for a calculated wave longitudinal velocity C l = 3300 m/s. Several defects are identified and labelled: defects 1, 3 and 4 are delaminations caused by the impact tests, or by the presence of the embedded sensors in the component. All grow along the fatigue testing of the CFRP panel. The areas number 2 and 5 are characterized by having a large attenuation in the backwall echo. Both via visual inspection, and analysing the NDT results, it is concluded that this area is affected by manufacturing problems, causing the used CFRP prepreg to have insufficient resin. Defect number 6 is an adhesive debonding between an omega stringer and the flat panel, with the shape of a triangle. The final result of this work, is the study of the reinforcing omega stringer. In Figure 11 is presented the complete acquisition for the omega stringer impacted in two locations. Two defects can be clearly observed in the image, inside the yellow frames. These defects, in the form of near-surface delaminations, are the result of the impact tests. The larger defect (on the left, with 11,3 x 24,6 mm), is originated from a 13 J impact, while the defect on the right (with 10,5 x 22,2 mm) exists due to a 10 J impact. The black lines in the image, as the one inside the blue frame, are originated by the electric cables that power the strain gauges, and that do not allow the passage of sound into the composite. 6

7 Figure 9: PAUT of the CFRP calibration specimen - complete results. Figure 10: PAUT results for the flat panel. On the left, amplitude results; on the right, time of flight results. 7

8 Figure 11: PAUT results for the omega stringer. On the top, amplitude results; below, time of flight results. 4. Conclusions The use of CFRP composites in critical areas of an airplane poses an added responsibility to airline companies. Not only an inhomogeneous, anisotropic material is a challenge to produce and repair, but also the NDT methods and techniques that must be used for the evaluation and testing of these components are complex. This work intends to respond to the above challenge, via the development of ultrasonic inspection procedures that can be used to examine aerospace components in situ. Using CIVA, the author selected four phased array ultrasonic probes to use in CFRP components, both with planar or complex geometries. The selected probes are two 64 element, 2-D matrix probes, with frequencies of 3,5 and 5 MHz, and two linear 32 element probe, with 4 and 10 MHz. After satisfactory results are obtained from the CIVA software, the author proceeds to inspect three CFRP components. The high attenuation of the CFRP, both due to sound scattering in the fibre and viscoelastic losses in the epoxy matrix, poses an added challenge to the PAUT. The most critical scenario encountered occurs when using the 10 MHz probe to inspect a 11,4 mm section of the CFRP test specimen: for this setup, the attenuation of the backwall echo amounts to more than 40 db. Using a multi-point focalisation technique, the difference in amplitude between the interface and backwall echo is reduced. Several defects are characterized successfully: delaminations, debondings and lacks of resin. For the complex geometry of an omega-shaped reinforcing stringer, also made out of CFRP, only when using a self-adaptive algorithm (SAUL), the embedded defects are detected and characterized. Hence, it is proved the validity of the use of this algorithm, and its applicability in the NDT of the complex geometry parts, common in the aerospace industry. Concluding, this work can now be used by IST to continue the academic work related to carbon fibre composite component inspection. As a final point, it can also be said that ISQ now has the ability to supply the industrial aerospace market with the PAUT and SAUL methodologies validated in this work, with a commercial solution that works, and can satisfy its clients NDT needs. References [1] Reinforced Plastics magazine, May/June 2013, Elsevier Ltd, pp. 28 [2] Reddy, J. (2004), Mechanics of Laminated Composite Plates and Shells - Theory and Analysis. 2nd Edition. Florida: CRC Press [3] Jones, R. M. (1975), Mechanics of Composite Materials, 2nd Edition, Taylor & Francis [4] Imasonic SAS, PA_principle.php [5] Hopkins, D. L. et al. (2012), Surface-adaptive ultrasound (SAUL) for phased array inspection of composite specimens with curved edges and complex geometry, QNDE [6] Lonn, S. et al. (2004), Modeling of ultrasonic attenuation in unidirectional fiber reinforced composites combining multiple-scattering and viscoelastic losses, Review of Progress in Quantitative Nondestructive Evaluation Vol. 23 [7] Deydier, S. et al. (2005), Ultrasonic field computation into multi-layered composite materials using a homogenization method based on ray theory, Review of Progress in Quantitative Nondestructive Evaluation 8