Various Types of Defects Detection in Flat and Curved Laminated Composite Plates Using Nonintrusive Lamb Wave System

Composite materials are widely used in aerospace and aeronautical industries, and the interest in using more advanced composites is growing high [1] because of their advantages for lightweight, high strength, and engineering design flexibility. The large use of composites necessitates the development of appropriate nondestructive evaluation (NDE) methods to detect and quantify defects and damage in composites [2–4]. The National Aeronautics and Space Administration (NASA) Aeronautics Research Mission Directorate initiated research to reduce the certification time of composite components under the Advanced Composites Project [4,5] with one of the three technical challenges being developing NDE methods to quantitatively characterize defects in as-manufactured parts and damage incurred during or after manufacturing [2]. Various defects such as fiber waviness or undesirable material may appear in composite structures during the manufacturing process [6]. Wrinkles or waviness are common when adding new layers [7] or during curing with temperature gradients [2]. Wrinkles/waviness can be in-plane or out-of-plane [8] and will cause significant degradation of the composite strength and can weaken the composite structure performance. Debonding occurs when an adhesive stop adhering to an adherend or if physical or mechanical forces that hold the bond together are broken while delamination is a failure in a laminate which leads to separation of the layers or plies. All these undesirable defects affect the mechanical properties and structural overall performance [6], and may further jeopardize the safety of space structures during operation. Thus, nondestructive evaluation methods that are suitable for a wide range of defects are highly needed [1,9] for ensuring safety and reliability of aeronautic vehicles [5,10].

Lamb waves can propagate relatively long distances in plate-like structures with lower energy loss compared to bulk waves, which enables their ability for large area inspection [11,12], and they are sensitive to various defects [13,14]. With these advantages, Lamb wave-based NDE methods have been widely applied for defect detection in plate-like composite structures over the last decade [15–19]. Park et al. adopted time-reversal algorithms to reconstruct the Lamb waves excited by a piezoelectric patch in a quasi-isotropic composite plate toward a reference-free diagnosis technique [15]. Purekar and Pines used piezoelectric sensor (PZT) arrays to excite and receive Lamb waves for damage detection purposes in a composite plate with simulated delamination [16]. Rogge and Leckey demonstrated Lamb wavefield analysis and wavenumber imaging can be used to identify the depth and size of near-surface impact delamination in composite plates [17]. Hall and Michaels studied the multipath-guided wave imaging method using sparser arrays on a composite plate with simulated damage, resulting in improved image quality compared to their previous work [18]. Sohn et al. applied a filtering imaging method and highlighted the internal delamination using standing waves trapped within a composite plate [19]. Tian et al. quantified impact delamination dimensions in composite plates through wavenumber-based imaging methods [20].

Recently, the scanning laser Doppler vibrometer (SLDV) has attracted great interest in Lamb wave-based defect detection, visualization, and quantification [14,21–26]. Based on the Doppler effect, the SLDV can acquire velocities/displacements of Lamb waves on the plate surface where the laser inspection spot aims at. By scanning over a high-density set of points over the wave propagation area, the SLDV can acquire a high-resolution wavefield, which contains a wealth of information about the Lamb wave propagation and interactions with a defect in plates. Such high-resolution wavefields have been used for visualization and quantification of various types of defects, such as cracks, delamination, and debonding [27–31]. However, most of the work reported in the literature for Lamb wave-based defect detection in composites is mainly about delamination on flat plates or slightly curved ones [14,17,20].

This paper investigates the detection and evaluation of wrinkles and delamination defects in either flat or curved laminated composite plates using PZT-SLDV Lamb waves and multidimensional wavefield analysis. Laboratory experiments are first set up to provide minimally invasion to the plates being inspected, and the Lamb wave wavefields are obtained. For defect visualization and evaluation, two methods, adaptive wavefield imaging and wavenumber imaging, are developed and applied for the flat and curved plate scenarios, respectively. The inspection results show that the imaging methods can not only visually show the location of the defects but also roughly quantify their sizes. The remainder of this paper is organized as follows: Sec. 2 presents the nonintrusive PZT-SLDV inspection experimental setup. Section 3 presents the visualization and localization of delamination in a curved plate, while Sec. 4 presents that of wrinkle defects in a flat composite plate. Section 5 then concludes the paper with findings, discussions, and future work.

This section presents the nonintrusive experimental setup of the PZT-SLDV Lamb wave inspection system. The overall experimental setup of the PZT-SLDV system is given in Fig. 1. A Steminc SM412² PZT transducer (0.5 mm thick and 7 mm diameter) is used for Lamb wave actuation, while the Polytech PSV-400-M2 SLDV system is employed for wavefield sensing. Traditionally, the PZT transducer is bonded on the surface with permanent adhesive as couplant, while reflective tapes are attached on the specimen surface for SLDV signal enhancement [23]. Removing the PZT transducer or the reflective tape requires great care and may more or less still cause adverse modification of the material surface conditions. In order to implement truly nondestructive evaluation on the composite structures, nonintrusive setup is explored and implemented in this study, with selected honey (Fig. 2(a)) as couplant for PZT actuation and wipe-off reflective spray (Fig. 2(b)) for SLDV measurement enhancement.

To attach the PZT transducer, a syringe is used to apply a tiny drop of honey at the target surface and the PZT is then placed on top of the honey drop. Gentle thumb pressure is applied on the PZT transducer for about two minutes so that the honey will distribute evenly between PZT and the specimen. The attached PZT is then left for a couple of hours before the inspection so that the honey will harden and cure. Figure 2(c) illustrates the PZT transducer attached to the surface with honey. Removable transparent tape is used to hold the wire in place on the plate (as shown in Fig. 2(c)) to help the PZT stay on without sliding down if the specimen is placed vertically. On the other hand, the wipe-off reflective spray is applied with the spray nozzle being held 12 in. away and normal to the specimen surface as shown in Fig. 2(d). About 20 layers of spray are applied to the surface in order to achieve full signal level from the SLDV.

For Lamb wave inspection, three-count Hanning window smoothed sinusoidally is used as excitation. It is generated by a function generator (Tektronix AFG3022C) at a selected frequency, magnified by an amplifier (NF HSA 4014) (40 times gain in this study), and then sent to the PZT actuator. Through in-plane piezoelectric coupling [11], Lamb waves are excited and propagate in the plates. On the sensing side, SLDV works on the Doppler frequency-shift effect on light waves and measures particle velocity along its laser beam. For Lamb wave measurements, the SLDV head is placed normal to the specimen to obtain the out-of-plane Lamb wave velocity at a sensing point. In addition, the SLDV head is set up at the optimal standoff distance (506 mm, 711 mm, and 915 mm) to the specimen (711 mm in this study), With its scanning capability, SLDV will obtain the motions over a user-defined scanning line or area grids and output the multidimensional time-space Lamb wavefield v(t,x), where t is the time and x is the space vector (x,y) of the scanning point. More details of SLDV wavefield sensing can be found in Ref. [20].

This section presents the inspection of the curved specimen with delamination. To evaluate if the curved specimen can be inspected the same way as a flat one for damage detection purpose, the curved plate is treated as flat during the inspection using the nonintrusive PZT-SLDV inspection system. Studies have shown that when there is delamination in laminated composites, strong trapped waves with new wavenumbers associated with the ply location of the delamination are present and reflected back and forth inside the delamination area [20]. Wavenumber imaging method has been proved effective for delamination detection and quantification in flat or slightly curved composites [14,17,20]. Using the current PZT-SLDV system, the SLDV measurement will be the projection of the actual wavefield due to the curvature of the plate and thus will affect the wavenumber component as well as its strength. In this study, we explored the possibility of using the wavenumber imaging method using the current setup for delamination evaluation on the curved specimen without considering the curvature effect on wavenumbers. The wavenumber imaging algorithms and the inspection results are presented in this section.

Using multidimensional Fourier transform (FT) [32], the SLDV acquired time-space wavefield v(t, x) can be converted to frequency–wavenumber representation to reveal intrinsic wave characteristics related to the propagation and wave–defect interactions [22,32]. Mode conversion and/or new wavenumbers might be introduced due to the defect and further modify the distribution of the wavenumbers, which presents how the Lamb waves are modified at defect due to the wave-defect interactions. Hence, a wavenumber distribution map can be used to directly indicate the presence of the defect as well as its location in the structure.

The curved composite plate is manufactured with embedded delamination between different layers. The top surface and side view of the curved plate are shown in Figs. 3(a) and 3(b), respectively. The plate measures about 101 mm high and 7 mm thick, and 160-mm top surface arc length. Two area inspections are performed on the plate with one for wing W1 and another for W2 as shown in Fig. 3(c). A three-count toneburst amplified to 40 Vpp is used as excitation for each inspection. The excitations are located at the center of each wing O₁ (40 mm away from the left edge and 50 mm away from the bottom edge) for W1 and O₂ (40 mm away from the right edge and 50 mm away from the bottom edge)for W2, so that the wave propagation can cover the entire wing. Considering that the measurement on different surfaces may vary due to the nonsymmetric defect location across the thickness/layer with respect to each surface for composite plates [20], the measurement has been conducted on both the top and bottom surfaces. The inspection frequency is set as 150 kHz as it provides strong wave–defect interaction. The SLDV head is placed normal to the surface where the excitation is located as shown in Fig. 3(d). Due to the curvature of the plate, the SLDV will not maintain normal to the plate surface of the whole scan area, as illustrated in Fig. 3(d). Such scanning angle effect and resulted loss in wave strength are not considered in this study since our study goal is to explore if the present measurement and analysis methods are applicable for delamination detection on plates with such amount of curvature. For each inspection, the scanning area covers the accessible surface of the inspected wing: around 90 mm × 90 mm within x–y plane. The SLDV spatial resolution is 1 mm, and the sampling rate is 10.24 MHz. Note the inspection area on the top and bottom surface differs subtlety due to the plate curvature.

Figures 4(a) and 4(b) plot the wavefield snapshots at 45 µs for W1 and W2 inspection on the top surface. The wave–delamination interactions can be immediately noticed at two locations along y = 50 mm (marked as D2) and 82 mm (marked as D1) from the wavefield plots. Similar wave–defect interactions are observed: (a) defect-induced waves at D1 as shown in Fig. 5(d) with significantly smaller wavelength compared to the 10-mm nominal wavelength of the incident waves (Fig. 4(c)); (b) defect-induced waves at D2 (Fig. 4(e)) with obvious higher intensity while no obvious wavelength change compared to the incident waves (Fig. 4(e)), indicating the possibility of the presence of a defect.

Further evaluation of the delamination in the curved plate is conducted with the weighted wavenumber imaging method given in Sec. 4.1. The resulted wavenumber images are plotted in Figs. 5(a) and 5(b). From the image for W1 (Fig. 5(c)), a strip-shaped defect (D1) is highlighted with large wavenumbers in the range 40 mm to 90 mm along y = 82 mm, which is consistent with the shorter wavelength observed in the wavefield snapshot (Fig. 4(d)) given the inverse relationship between wavenumber and wavelength [20]. In addition, another strip-like defect (D2) can also be observed in the range x = 40–90 mm at y = 50 mm through the indication via slightly wavenumber variations, which is consistent with the wavefield observation that no obvious wavelength changes (Fig. 4(e)). In the wavenumber image for W2 (Fig. 5(b)), D1 (y = 80 mm) and D2 (y = 50 mm) are indicated in the range 60–110 mm along x-axis. By comparing the imaging results of W1 and W2, we can conclude that D1 and D2 are localized in the range 40–110 mm along x-axis with D1 at 82 mm and D2 at 50 mm along y-axis.

The inspection wavenumber imaging results obtained on the bottom surface are presented in Figs. 5(c) and 5(d). Similar subtle wavenumber variations are observed at the D2 location detected earlier on the top surface, which further confirms the D2 range and shape. In addition, another strip-shaped defect (D3) is observed, located at y = 18 mm with a similar range along x-axis, 40–110 mm by considering the results of both Figs. 5(c) and 5(d). Through comparing the inspection results on the top and bottom surfaces, we can conclude that (1) three strip shape defect exists in the plate in the range of 40–110 mm along the x-axis with D1 at 82 mm, D2 at 50 mm, and D3 at 18 mm along y-axis, and (2) along the thickness direction, D1 is close to top surface since it is only clearly highlighted in the results obtained on the top surface, while D3 is close to the bottom surface since it is only visualized in the results measured on the bottom surface. D2 shall be located within the middle plane since the interaction is more difficult to capture on both surfaces resulting in lower imaging resolution.

In this section, out-of-plane wrinkle defects are evaluated in a flat laminated composite plate. A simulation study has shown that wavenumber-based methods are not sensitive to wrinkle defects since no new wavenumbers are introduced, while the first arrival (S0) shows sensitivity to wrinkles [2,3]. Wavefield imaging method based on the first arrival has been used by Michaels [35] for bond quality detection in a bonded aluminum plate since S0 is almost non-dispersive in the low-frequency-thickness range [36,37], and it is not complicated by the slower modes and edge reflections [35]. The imaging results in Michaels’ study are significantly improved compared to the wavefield imaging method using the whole wave signal. Within the acquired multidimensional time-space wavefield in this study, strong interactions between the wrinkle and the first arrival are observed. Hence, an adaptive wavefield imaging method has been developed using the time of flight information of the first arrival in the acquired waves and applied to evaluate the wrinkle defect. The adaptive wavefield imaging algorithms and the inspection results are presented in this section.

Very often, directly acquired wavefield does immediately show the wave–defect interactions, indicating the existence of structural discontinuity [19,20,31,33]. However, such detection requires experiences of wavefield analysis, and immediate visualization of discontinuity in the structure is highly desired. Studies have shown that evaluating the wave energy along the wave propagating distance provides a rapid yet effective way to generate an inspection image of the plate [38,39]. The wave energy can be represented by quantities such as the peak amplitude or the root-mean-square value of the waveform at the position [35]. However, it has shown that the resolution of such images are limited and the revealed defect features are not clear enough for quantification purpose [35].

The flat composite plate is manufactured with a total of 20 plies and 3.83 mm thickness. The overall dimension of the plate is measured at 306 × 260 (unit: mm). Significant wrinkles can be observed on both top and bottom surfaces, as illustrated in Fig. 6(a) (indicated by the arrows), and also can be seen from the side along the thickness direction as in Fig. 6(b). No other information about this wrinkle plate is available such as the material properties as well as the detailed layup. To assist the inspection, Cartesian coordinates are employed with the origin set at the left bottom corner of the plate. We have chosen to study the area with more visible wrinkles and referred to it as “wrinkle region,” in contrast to the area with less visible wrinkles as “reference region,” as shown in Fig. 6(c). A PZT actuator is placed at location (160, 110) on the specimen (Fig. 6(c)) to excite the waves. The actuator is about 30 mm away from the wrinkle edge as shown in Fig. 6(b), which is far away enough from the wrinkle edge so that the “blind zone” near the source is avoided and it allows the two modes to separate when they arrive at the wrinkles. Area scans (95 mm by 60 mm in the range x = 60–160 mm and y = 80–140 mm) are performed using the area scan setup given in Fig. 6(c). Similar to Sec. 3, the measurement has been conducted on both the top and bottom surfaces in the wrinkle region. A line scan within the reference area has also been performed for inspection parameter setup. The SLDV scanning spatial resolution is 1 mm, and the sampling rate is 10.24 MHz.

For Lamb waves NDE, as frequencies go higher, more wave modes are likely to be excited and cause complexity in subsequent damage evaluation. Therefore, low frequency is often employed [41]. A few quick wave acquisitions at different excitation frequencies through line scans (Fig. 6(c)) are first performed in the reference region to identify a proper excitation frequency. The scans are along x-axis from 10 to 100 mm away from the PZT actuator. The representative acquired time-space wavefield at 120 kHz, 210 kHz, and 300 kHz as well as a waveform at a certain propagating distance is presented in Fig. 7. From Fig. 7(a), we can see that there are two wave packets (S0 and A0) in the 120-kHz Lamb waves but the first arrival (S0) at about 30 µs has very weak strength. And in the wavefield, we can barely observe any wave–wrinkle interactions. On the other side, the first arrival (S0) of the waves at 210 kHz in Fig. 7(b) exhibits much stronger strength. And more obvious wave–wrinkle interactions (S0 to A0 mode conversion) can be observed in the wavefield, indicating that S0 is more sensitive to the wrinkle defect than A0. When the excitation goes up to 300 kHz, the two modes are more difficult to separate and more edge reflections can be observed as shown in Fig. 7(c) due to its faster group velocity. By comparing the results in Fig. 7, 210 kHz will be ideal for the wrinkle inspection since sufficient first arrival will be excited and less complexity will be introduced. Thus, 210 kHz will be used as the excitation frequency for the inspection of the wrinkle plates.

To implement the imaging algorithm, the group velocity of the first arrival in the acquired waves at 210 kHz needs to be determined by finding out its time of flight (TOF) at a given propagation distance. To estimate the group velocity more accurately, eleven signals from 70 to 90 mm at a 2-mm increment along the scanning line are used. Three example waveforms at propagating distance 70, 80, and 90 mm are plotted in Fig. 8(a). Hilbert transform is applied to extract the envelopes of these waveforms, and the TOFs of the first arrivals are extracted indicated by the peak arrival time (e.g., TOF is 47.5 µs at 90 mm in Fig. 8(a)). All of the acquired TOF and corresponding propagating distance are plotted in Fig. 8(b) (stars). The plot shows a linear relationship that can be curved fitted through linear regression (red line, derived with linear curve fitting as y = 5.249x − 158.23). The group velocity of the first arrival is therefore obtained, which is about 5.249 mm/µs.

Inspection using the subject PZT-SLDV system is performed on the wrinkle region on both top and bottom surfaces of the test plate using the setup given in Fig. 6(c). The resulted time-space wavefield snapshots at 25 µs are presented in Fig. 9. Circular wavefront can be observed in the incident waves leaving the PZT actuator for both results. If there is no defect, the circular waves are expected to propagate and attenuate with propagating distance with no abrupt change [20]. However, in the wavefields, the circular wavefronts are guided and modified due to the wrinkle geometry (with the shape straight and vertical along y-direction) to straight crest ones when they arrived at and passed through the wrinkles [42,43]. These straight crest waves are most likely standing waves formed at the wrinkle area due to the reflections back and forth between the wrinkles. Comparing the waves measured on the top (Fig. 9(a)) and bottom (Fig. 9(a)), the wave interaction with defects are occurring in a similar range (90–130 mm along x-direction). It matches with the visually observed wrinkle range in the plate. The detection of wrinkles from both top and bottom indicates the wrinkles might be manufactured through the thickness and can be inspected from either side.

The wavefield data have been further processed with the adaptive wavefield imaging method outlined in Sec. 3.1, and the results are given in Figs. 10(a) and 10(c) for top and bottom surfaces, respectively. Typically, for the wavefield image generated for a pristine plate, the energy concentrates at the source as shown in Fig. 10 and then gradually decreases with the distance [44]. However, it can be observed in the resulted images that the wave energy has two distinguished distributions in the range 80–130 mm and 130–160 mm. In the range 130–160 mm, the strong energy from the incident waves is observed with a circular shape. While in the range 80–130 mm, strip-shaped wave energy distribution is observed. The zoom-in images of the wrinkle range are presented in Figs. 10(b) and 10(d) for top and bottom surfaces, respectively. Clear wrinkle patterns are visualized now with peaks and valleys represented by the bright and dark colors in both images. The wrinkle patterns are discernibly visualized in both images within the range 80–130 mm along x direction, and the identified wrinkle ranges agree well with that observed through visual examination given in Fig. 6(b).

In this paper, two types of composite defects, delamination in a curve plate and wrinkles in a flat plate, are evaluated using a nonintrusive PZT-SLDV system without detailed manufacturing information about the structures. For the evaluation of delamination in the curved plate, promising results are obtained with the current setup. Wavelength change is observed compared to the incident waves in the time-space wavefield, indicating wavenumber changes in the wavenumber domain. The wavenumber imaging method is therefore employed to evaluate the delamination in the curved plate. The resulted images indicate three delamination defects, including their location, overall sizes, and shapes. In addition, the defect location along the thickness direction is indicated with D1 near the top surface, D2 in the middle, and D3 close to the bottom surface. While for the wrinkle plate, the first arrival of the waves shows clear wave interactions with the defect and thus is used to evaluate the wrinkles through an adaptive wavefield imaging method. The resulted images clearly present straight wave patterns caused by the wrinkle defect as well as their location and ranges.

Although both the delamination and wrinkles in the two plates are localized in the resulted images, still some improvements can be achieved for both plates: (1) delamination D2 is not visualized as clearly as the other two delaminations, and one possible reason could be that the selected excitation location is too close to the defect and (2) the left edge of the wrinkle region on the wrinkle plate is not identified since most of the wave energy scattered after interacting with wrinkles as well as attenuated after propagating long distance. Future work can be focused on trying additional actuator locations such as left and right edges of the plate to cover more inspection views and/or allow wave propagation before the defect or use multiple actuator arrays to excite stronger waves. In addition, the window size effect as well as image fusion can be studied to further improve the wavenumber image resolution. Moreover, comparison with other imaging methods such as ultrasonic C-scan and further implementation of the PZT-SLDV system on other composite defect evaluation such as porosity is also of the authors’ interest.

The authors would like to thank the financial supports from NASA Advanced Composite Project under Award Nos. NNL09AA00A and 80LARC17C0004. The material is based upon work supported by NASA under Award Nos. NNL09AA00A and 80LARC17C0004. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Aeronautics and Space Administration.