Characterization of barely visible impact damage (BVID) in polymer matrix composites (PMCs) is necessary to use slow crack growth damage tolerance models and evaluate remaining life of PMC components. Azimuthally scanned angled-beam pulse-echo ultrasound is investigated as a complimentary technique to normal incidence ultrasound inspection of BVID in PMCs to characterize delamination fields. It is found that there is a correlation between signals present in the azimuthally scanned angled-beam pulse-echo ultrasound C-scans and transverse cracks seen in X-ray computed tomography inspection. These transverse cracks are not readily identifiable as transverse cracks in normal incidence C-scan inspection.
For the past several years, Air Force Research Laboratory has explored methods to characterize barely visible impact damage (BVID) in polymer matrix composite (PMC) structures using single-sided inspection and field level nondestructive evaluation tools. The motivation for this research is to develop characterization capabilities that assist damage-tolerant lifting approaches, which are already used by the USAF for metallic structures [1–3]. One approach that has been investigated is angled-beam (also known as oblique angle) pulse-echo ultrasound [3–6]. While the characterization of the delamination substructure is a primary input for progressive damage models, detecting and characterizing other defects for inclusion in the models, such as matrix cracks, would also improve the fidelity of the model.
This paper leverages work in the ultrasonic polar scan in composites [7,8] and backscattered ultrasound in composites . Martin and Andrews were the first to recognize the angular dependence of defects in composites using backscattered ultrasonic testing (UT) in an azimuthal scan configuration. They imaged a single point as a function of azimuthal angle (Az) and plotted the data as a B-scan . To date, all the previous work has looked at responses from a single point [7–10] or a single azimuthal angle C-scan [3–6]. Specifically, this paper considers the possibility that multiple angled-beam C-scans conducted at different specimen azimuthal angles can provide quantitative information on the location and orientation of defects in impacted composites with BVID that could complement current normal incidence UT NDE. Figure 1 shows this schematically for delaminations and connecting transverse cracks. Results of characterizing a BVID delamination field with multiple C-scans as a function of the azimuthal angle at a constant incidence angle (Inc) to identify and locate preferentially oriented defects is presented.
2 Materials and Methods
A flat laminate PMC panel was used in this work (Fig. 2). The specimen was fabricated from unidirectional prepreg of IM7 carbon fiber and 977-3 epoxy using a stacking sequence of [−453/903/453/03]S. The layup was chosen because there would be fewer impact delaminations, and the delaminations would be spaced farther apart in the thickness direction than a more typical quasi-isotropic layup, like [−45/90/45/0]3S. The desire for fewer and more widely spaced delaminations was to make it easier to discriminate the individual delamination and crack signals from each other. The panel dimensions were 100 × 150 × 3.2 mm. BVID was induced in the panel with a 10 J drop impact. A small triangle of copper tape in the upper right corner of the specimen and small drops of epoxy in a square shape around the delamination field were used as fiducial marks. These epoxy drops were placed on the top surface of the specimen to ensure the UT and X-ray computed tomography (XCT) data sets could be registered.
2.1 Ultrasonic Modeling.
A commercially available ultrasonic modeling software package (civa ut with fidel 2d plugin) was used to simulate the B-scan response of the transverse matrix crack expected in the experimental setup. The composite specimen simulated had a [−453/903/453/03]S layup and used nominal material properties of IM7 carbon fiber and 977-3 epoxy matrix with a fiber volume fraction of 0.6 . The simulation used the same transducer and inspection parameters as the ultrasonic experiment, a 2.25 MHz, 6.35 mm diameter, and 50 mm nominal focus transducer with an incident angle of 24 deg (counterclockwise) in pulse-echo configuration. Two delaminations three lamina apart were connected by a transverse crack with a 45-deg (clockwise) orientation. A schematic of the specimen configuration for the transverse crack case is shown in Fig. 3. The no crack case omitted the transverse crack and maintained the positions of the two delaminations. In the simulations, the transducer was scanned from left to right across the specimen configuration shown in Fig. 3.
2.2 X-Ray Computed Tomography.
XCT was performed using a commercial system (Zeiss Xradia Versa 520). The detector has a resolution of 2000 × 2000 pixels, and 1601 projections per tomograph were collected with a tungsten filament source at 130 kV and 8 W. The stitching features of the system were used enabling four tomographs to be collected and stitched together (two tomographs in the horizontal direction and two tomographs in the vertical direction) into one final tomograph. The final tomograph was 3768 × 3808 × 3322 pixels and 91 GB in file size. On a workstation with 40 threads and 64 GB of RAM, the entire tomograph could not be opened without cropping it down first in a preprocessing step. The cropped data set of just the volume of interest was 3768 × 3322 × 365 pixels, with a voxel edge length of 15.328 µm and a final volume extent of 57.76 × 50.92 × 5.59 mm. The exported image stack in a 16-bit TIFF file format is 8.5 to 11.6 GB depending on whether the x-y, x-z, or y-z image stack is saved. Commercial image software (dragonfly pro) was used for the preprocessing tomograph cropping and TIFF image stack export.
2.3 Ultrasonic C-Scans.
UT was performed using a custom five-axis (x, y, and z, transducer incident angle, specimen rotation angle) ultrasonic scanning system. The specimen was held in a custom fixture to maintain an air back wall condition, and this fixture was described in Ref. . C-scans were performed with a 2.25 MHz, 6.35 mm diameter, 50 mm nominal focus transducer. The inspections were performed in pulse-echo configuration with a 24-deg incident angle. C-scans were taken at azimuthal angles from 0 deg to 360 deg in 45-deg increments resulting in nine C-scans as a function of azimuthal angle. A C-scan using normal incidence pulse-echo UT was also performed with the same transducer. The transducer position was adjusted to focus on the front wall. Each C-scan contains 16-bit amplitude values for 5120 time point A-scans in a 500 × 500 pixel spatial grid, with a spatial resolution of 0.2 mm and a time resolution of 0.004 µs. This produced a 100 mm × 100 mm × 20.48 µs UT data set for each C-scan collected that was 2.5 GB in size when stored as a binary file.
3 Results and Discussion
3.1 Ultrasound Modeling.
Simulation of inspection B-scans showed readily discernible differences between the response of the specimen configuration without a connecting transverse matrix crack, Fig. 4(a), and with a connecting transverse matrix crack, Fig. 4(b). The response from the crack occurred at position = 20.5 mm, time = 62.9 µs, with amplitude = 0.944 arbitrary units. In comparison, the no crack response occurred at position = 19.5 mm, time = 63.8 µs, with amplitude = 0.514 arbitrary units represents 54% of the amplitude of the transverse crack case. If both responses are experimentally detectable, the transverse crack response should be almost twice that of the no crack response. The amplitude of the crack signal is also compared to the trailing edge response of the delaminations. All of the comparisons are given in Table 1. The fact that the crack response is expected to be more than four times the amplitude of the trailing edge response is encouraging for detection and identification.
3.2 Ultrasound Experiments.
Azimuthally scanned angled-beam pulse-echo C-scan images from the 10 J drop impacted specimen with BVID are shown in Fig. 5, with the normal incidence C-scan as a reference. The normal incidence C-scan, Fig. 5(e), shows the typical lemniscate-spiral shape of impact delamination field in a PMC. Figures 5(a)–5(d) and Figs. 5(f)–5(i) show angled-beam pulse-echo UT amplitude C-scans as a function of azimuthal angle. The images in Figs. 5(a)–5(d) and Figs. 5(f)–5(i) have been rotated so that the damage fields are in the same orientation for easy comparison. This orientation is the same as the orientation of the specimen in Fig. 2. Each azimuthal angled-beam pulse-echo ultrasound scan in Fig. 5 shows that each azimuthal angle is sensitive to different regions of the BVID field. This is a positive result, indicating that azimuthally scanned angled-beam pulse-echo ultrasound can be used to identify preferentially oriented defects.
Different azimuthal angles should be sensitive to different crack orientations, and in particular, azimuthal angles that are 180 deg apart from complementary pairs that should enable one to capture the transverse cracks on either side of the delamination field shape, Fig. 1. An example of such a complimentary azimuthal angled-beam pulse-echo ultrasound scan pair (with the normal incidence C-scan for reference) is shown in Fig. 6. The yellow arrows in Figs. 6(b) and 6(c) indicate the location of a strong signal in the 24 deg Inc/90 deg Az C-scan, and the green arrows in Figs. 6(b) and 6(c) indicate the location of a strong signal in the 24 deg Inc/180 deg Az C-scan.
In order to demonstrate that the signal orientation dependence is from crack orientation, the optical, normal incidence and azimuthal angled-beam pulse-echo ultrasound scans need to be spatially aligned to each other. For this correlation, a single azimuthal angle is used as an example. The impact specimen optical image, Fig. 7(a), is correlated with the normal incidence, Fig. 7(b), and azimuthal angled-beam pulse-echo ultrasound, Fig. 7(c), by identifying the fiducial marks present in each. The vertical lines of fiducial marks are identified in Figs. 7(a)–7(c) by the green and blue vertical dashes at the ends of the vertical lines of fiducial marks. A common scan line (red dashed line) that passes through two fiducial marks, for ease of identification, is chosen for further analysis. In the ultrasound data, this common scan line corresponds to a particular scan line or index line of ultrasound data, which is plotted as a B-scan.
After the alignment process identifies the equivalent scan lines in the ultrasound data, the normal incidence and azimuthal angled-beam pulse-echo ultrasound C-scans and B-scans can be compared as shown in Fig. 8. Figures 8(a) and 8(b) show the normal incidence and 24 deg Inc/90 deg Az C-scans, respectively, and they are provided for easy reference. Again, green vertical dashes indicate the left line of fiducial marks, blue vertical dashes indicate the right line of fiducial marks, and a red horizontal dashed line indicates the location of the B-scan in each of the C-scans. Figures 8(c) and 8(d) show the normal incidence and 24 deg Inc/90 deg Az B-scans, respectively, taken at the red dashed line position in the corresponding C-scan. Figure 8(c) shows a band of signal at ∼2 µs, which corresponds with the top surface of the specimen. The band of signal occurring at ∼4 µs is consistent with the expected back wall response from the back surface of the specimen. Signals occurring after ∼4 µs are multiple reflections of earlier occurring signals.
In Fig. 8(c), the fiducial marks are seen as a slight disturbance in the front wall return signal (∼2 µs) at index axis positions 76 mm and 22.4 mm with a lack of return signal later in time. The lack of signal later in time may appear strange for a fiducial mark in intimate contact with the specimen and made from a material of similar acoustic properties. As such, the epoxy dot fiducial marks should have minimal acoustic impedance mismatch and good sound transmission. However, this observed response is consistent with the expected physics. The epoxy dot fiducial marks are dome-shaped droplets of epoxy on the surface of the specimen, and they cannot be considered simply as a layer of material for wave propagation purposes. The geometry of the fiducial mark must be considered as well. The dots are ∼2–3 mm in diameter, which is close to the expected focal spot size. The portion of the beam that is traveling perpendicular to the fiducial mark surface and the specimen surface is transmitted through. However, this is a small amount of the total beam. The much larger portion of the beam is interacting at some angle to the surface of the fiducial mark. Essentially, it is an oblique angle interaction. The ultrasound interacting at an oblique angle with the surface of the fiducial mark is diffracted at various angles as a result. The diffracted signals either have no return path to the transducer or a return path that results in a signal that occurs much later than the rest of the normal incidence signals. Thus, the epoxy dot fiducial marks shadow all defect signals that may occur after them in the specimen. Epoxy dots work well for fiducial marks for ultrasound as shown here, but caution is needed as they should be placed so that they will not interfere with features of interest.
Considering the right side of the normal incidence B-scan, Fig. 8(c), and moving left, the expected behavior of scanning over the stair-step delaminations of a “cone” delamination field is seen. Delaminations detected in the normal incidence B-scan are located at ∼3.3 µs and ∼10.2–35.4 mm, ∼3.1 µs and ∼35.4–44.0 mm, and ∼2.8 µs and ∼44.0–54.6 mm. Another delamination is located at ∼4.0-µs and ∼54.6–75.0 mm; however, this delamination is seen as a reduction in back wall signal amplitude because the delamination is so close to the back wall that the delamination and back wall return signals are interacting with each other. Figure 8(d) is the B-scan taken at the red dashed line in Fig. 8(b), and it is over the same region of the specimen as the normal incidence B-scan shown in Fig. 8(c). The signals from the fiducial marks are centered at ∼1.8 µs and ∼21.4 mm and ∼1.8 µs and ∼76.0 mm, which are both to the right of the indicated fiducial mark lines (green and blue vertical dashes). This observed behavior is due to the angled-beam pulse-echo scan preferentially capturing the side of the domed epoxy dot fiducial facing the transducer as the transducer is scanned across the specimen. The signals occurring between ∼2.4 and 9.7 µs and ∼29.0–68.2 mm correlate with delamination and crack signals. Several of these signals will be correlated to specific features in the XCT data in Sec. 3.4.
3.3 X-Ray Computed Tomography.
The XCT x-z image slice corresponding to the ultrasound B-scan is shown in Fig. 9(a). To interpret XCT results, materials that attenuate X-rays more (e.g., PMC and fiducial marks) appear lighter in XCT image slices and those that attenuate less (e.g., air, defects) appear darker. The black region of the images is empty space included to make the overall image field rectangular. The light gray to the nearly white rectangular region in the center is the specimen, and the dark gray triangular regions above and below the specimen are the region of air surrounding the specimen. The damage features visible in Fig. 9 (top) consist of horizontal delaminations (dark horizontal line-like features found in the specimen region) and transverse matrix cracks (dark vertical or angled features found in the specimen region). Several locations are marked in Fig. 9 (top) as well: “1” is the right-hand side epoxy dot fiducial mark, “2” is a delamination edge with transverse crack, “3” is a delamination edge with two transverse cracks, and “4” is a delamination edge with no transverse crack. Investigation of the damage features observable in the XCT image slice yielded Fig. 9 (bottom), which shows an enlarged image of a region of Fig. 9 (top), a region marked in Fig. 9 (top) with a red box. The damage features of interest are at location “3”: a delamination edge with a “<” shaped configuration of two transverse cracks, “Crack A—Vertical”: a transverse crack not at a delamination edge with a nearly vertical orientation, “Crack B—Preferential”: a transverse crack not at a delamination edge with a favorable orientation for detection by the 24 deg Inc/90 deg Az angled-beam pulse-echo ultrasound scan, and location “2”: a transverse crack at a delamination edge with an orientation favorable to detection by the 24 deg Inc/90 deg Az angled-beam pulse-echo ultrasound scan.
3.4 Correlation of Ultrasound and X-Ray Computed Tomography.
Damage features identified in the normal incidence UT scan, XCT, and azimuthal-scanned angled-beam pulse-echo UT are positionally correlated using the fiducial marks placed on the specimen, and the results are presented in Table 2. The positions of the delamination edges are easily identified in the XCT data, as shown in Fig. 9, as well as in the normal incidence B-scan image, Fig. 8(c), and discussed in Sec. 3.2. Table 2 presents that the positions for the delamination edges agree well with the XCT and normal incidence data as expected. The delamination edges are much more difficult to identify in the azimuthally scanned angled-beam pulse-echo UT data. Figure 10 replots a portion of the data shown in Fig. 8(d) to make it easier to identify the signals. The delamination edge positions calculated from the azimuthally scanned angled-beam pulse-echo UT data consistently overestimate the distance of the delamination edge in comparison with XCT and normal incidence UT. Consistent overestimation indicates a form of systematic error. The source of this error is likely due to the differences in the oblique response with varying beam characteristics in both depth and beam width.
The amplitudes of the experimental azimuthally scanned angled-beam pulse-echo return signals can be compared to the SA-FDTD model results (Sec. 3.1). Results are presented in Table 3. One difference to note is the experimental data include a trailing edge, Fig. 9 (top) location 4. With five identified signals and using one signal as a reference, four amplitude comparisons to the SA-FDTD model results can be made. Two of the indication amplitude ratios correlated well with the model results. These were (see Table 3) “Crack B—Preferential” and “2—Preferential,” which correspond to the physical defects shown in Fig. 9 (bottom) with the respective labels. The amplitude ratios of “Crack B—Preferential” and “2—Preferential” were 4.0 and 5.7, respectively, and they agree well with the model result of 4.4. These two damage features most closely resembled the modeled damage case of a nearly 45 deg transverse matrix crack oriented for maximum reflection of the ultrasound energy back to the transducer. The features “3—Combination” and “Crack 1—Vertical” in Table 3 correspond to the physical defects shown in Fig. 9 (bottom) with the respective labels. These features do not correspond well to the modeled defect. Feature “3—Combination” has two transverse cracks oriented in an arrowhead pointing left configuration. This configuration is expected to interfere with the return signal of the azimuthal angled-beam pulse-echo ultrasound. The trailing edge ratio for this defect is 1.6 when the model suggests 4.4 for an ideal case. “Crack 1—Vertical” is a transverse crack that is nearly perpendicular to specimen top surface, and it results in a low amplitude ratio of 0.9 in comparison to the idealized model result of 4.4 for a transverse crack at 45 deg. This result is unexpected as this configuration is similar to a “corner trap” configuration, which should give a good signal return for angled-beam UT. The reasons for this are not clear. It is hypothesized that this particular crack is oriented in a plane that is not perpendicular to the plane of the XCT image. This skewed type of crack orientation relative to the XCT image would cause some amount of the angled-beam UT signal to be directed away from the source. Additional analysis is planned to verify this. Overall, these amplitude comparison results support the hypothesis that azimuthal pulse-echo UT can be used to identify preferentially oriented transverse matrix cracks in PMCs.
Azimuthally scanned angled-beam pulse-echo ultrasound has been demonstrated for locating and characterizing transverse cracks in PMCs. By combining the information of several azimuthally scanned angled-beam pulse-echo UT C-scans with a normal incidence pulse-echo UT C-scan, both delaminations and transverse cracks can be located and characterized. This combined method offers a more complete nondestructive, single-sided inspection than previously possible. While multiple single-element transducer C-scans are time consuming, it should be possible to collect normal incidence and multiple oblique angle C-scans simultaneously using a 2D phased array probe as well. The normal incidence and multiple oblique angle C-scan collection time using a 2D phased array should approach the time it takes for a single-element transducer normal incidence C-scan of the same area. However, this remains to be proven.
Damage-tolerant life management of PMCs requires accurate BVID characterization as an input into slow crack growth damage tolerance models. This work offers a readily implementable solution for improved characterization of BVID in PMCs, and it offers the possibility for an optimized inspection in the future. Combining normal incidence and oblique angle ultrasound inspections to characterize BVID delamination fields will help enable damage-tolerant life management and thereby eliminate unnecessary repair and replacement of PMC structures. This combined method will achieve this goal by improving the accuracy of the BVID characterization that is an input to the slow crack growth damage tolerance model.
The authors wish to acknowledge the impact testing performed by Ms. Sarah Wallentine (AFRL) and Mr. Norm Schehl (UDRI). The authors also wish to acknowledge the many fruitful discussions with Dr. Michael Uchic and the other members of the RXCA research team.
Portions of this work are funded by the U.S. Air force (Contract Nos. FA8650-14-D-5224 TO 0001, FA8650-14-D-5224 TO 0002, and FA8650-15-D-5231 TO 001; Funder ID: 10.13039/100006831).