Damage Assessment of Low-Velocity Impacted Sandwich Composite Structures Using X-Ray Micro-Computed Tomography

Sandwich composite structures ofer signifcant versatility in structural system design but are susceptible to low-velocity impact damage, impacting their structural robustness. Tis study focused on nondestructive testing, particularly using X-ray micro-computed tomography, to assess damage on these structures, comprised of thin glass fbre reinforced polymer face sheets and a polyvinyl chloride foam core, under low-velocity impacts. Impacts were induced by a constant mass of 5.61 kg, dropped from various heights, generating impact energies between 2 and 22J. Tis resulted in varied damage levels, from indentations to full perforations. Te X-ray micro-computed tomography technique was chosen for its ability to detect internal damage. However, the system’s efcacy in accurately assessing damage depends on numerous factors like focus-to-detector distance, focus-to-object distance, and spatial resolution of the detector, among others. Te system yielded an approximated resolution range of 10–25 μ m for a focal spot size of 4 μ m and the resolution range of 11–26 μ m for a spot size of 7 μ m. To this end, the system was able to reveal damage inficted across the specimen through captured and reconstructed images. Te quality of the reconstructed images was validated using ImageJ2 software by comparing with the processed images. Te median flter was found to deliver images that closely resembled the original ones, albeit with a slight reduction in quality. Damage types varied based on impact energies. Low-level impacts caused matrix cracking and delamination at the foam interface. Medium-level impacts led to intralaminar and interlaminar damage, fbre fractures, and signifcant damage to the foam core through shearing and crushing. High-level impacts resulted in near or full perforations, with more pronounced delamination at the bottom interface, and fbre fractures in the impact zone, displaying a distinctive diamond-like damage pattern. Tese fndings can be instrumental in developing a predictive impact damage model.


Introduction
Within the framework of this study, damage assessment is understood as the evaluation of low-velocity impact damage (LVID) to gather both qualitative and quantitative data on sandwich composite structures.Diferent damage assessment methods and their advantages are discussed.Many existing studies have primarily focused on damage assessment in laminates, while research on sandwich composite structures remains scarce despite their growing utilisation in various felds.Unlike laminates, which form a simpler monolithic structure, sandwich composite structures are more prone to LVID, which compromise their load-bearing capacity.
Furthermore, sandwich foam cores are made of diferent materials.Core materials substantially afect damage initiation characteristics because they have lower mechanical properties than the skins or face sheets due to their low density [1].Tis highlights the fact that independent investigations need to be conducted for diferent core materials to determine their damage initiation thresholds.Te sandwich composite structures used in this work consisted of thin glass fbre reinforced polymer (GFRP) face sheets and polyvinyl chloride (PVC) foam core of low density.

Low-Velocity Impact (LVI).
Te notion of low-velocity impact has been clarifed in existing literature to distinctively separate it from other velocity categories.A common consensus among researchers establishes that low-velocity impacts occur at speeds below 10 m/s [2,3].Tis study will adhere to this definition, ensuring that all considered velocities remain within this range, specifcally between 0.9 m/s and 2.8 m/s.In practical scenarios, sandwich composite structures can be subjected to low-velocity impacts in numerous ways, such as tool drops during servicing or routine maintenance, hailstorms, bird strikes for moving objects in space, during transportation and handling prior to installation, and part installation whether standalone or in assembly [3,4].For this study, low-velocity impact damage was induced on the sandwich composite structures by utilising the Instron Dynatup Impact Testing Machine.Notably, recent research regarding sandwich composite structures primarily concentrated on understanding the efects of extremely low temperatures on carbon fbre reinforced polymers [5].Tis highlights the many factors that can infuence the impact behaviour of these complex structural materials.

Damage Assessment Methods.
Damage assessment methods are techniques that aim to detect, locate, quantify, and characterise damage in structures.Tere are diferent types of damage assessment methods, such as nondestructive testing (NDT), structural health monitoring (SHM), and numerical modelling and simulation.Te application and the available resources determine the advantages and disadvantages of each method.

Nondestructive Testing (NDT).
Tese techniques assess damage without causing further harm or altering the structure being inspected.Examples include ultrasound, radiography, X-ray micro-computed tomography (X-ray μ-CT), acoustic emission, shearography, and thermography [6].Each of the mentioned technique has its own advantages and disadvantages that are not covered in this work except X-ray μ-CT that is the focus of this study and presented in detail as an isolated subsection.Some advantages of NDT methods are as follows: (i) Tey are fast and reliable, as they can provide immediate results without afecting the structure's performance or service life.(ii) Tey are cost-efective, as they can reduce the need for destructive testing, repair, or replacement of the structure or its parts.(iii) Tey are versatile, as they can be applied to various materials, shapes, sizes, and conditions of the structure.

Structural Health Monitoring (SHM).
Tese are monitoring techniques that continuously or periodically measure and analyse the structural responses to external stimuli, such as loads, vibrations, temperature, and strain [6].Tey can provide information about the structural behaviour, performance, and condition of the structure.Te advantages include the following: (i) Tey are proactive, as they can detect and identify damage before it becomes critical or catastrophic and provide early warning and diagnosis.(ii) Tey are adaptive, as they can adjust to the changing environmental and operational conditions of the structure and provide updated and accurate information.(iii) Tey are comprehensive, as they can provide a global and holistic view of the structure and capture the interactions and interdependencies among the structural components and systems.
Eforts from researchers are currently underway to create hybrid systems that combine the strengths of nondestructive testing (NDT) and structural health monitoring (SMH) approaches, aiming to enhance the monitoring and assessment of damage.

Numerical Modelling and Simulation.
Tese are computational techniques that use mathematical models and algorithms to simulate the physical phenomena and processes that afect the structure.Tey can provide information about the structural response, damage evolution, and failure mechanisms of the structure [7].Although not discussed in the current work, this approach is slated for exploration in subsequent studies.Some advantages are as follows: (i) Tey are predictive, as they can forecast the future behaviour and performance of the structure and provide optimal design and maintenance strategies.(ii) Tey are fexible, as they can incorporate various parameters, scenarios, and uncertainties of the structure and provide sensitivity and robustness analysis.(iii) Tey are innovative, as they can explore new and complex phenomena and mechanisms that are difcult or impossible to observe or measure experimentally and provide novel and creative solutions.

X-Ray Micro-Computed Tomography (X-Ray μ-CT).
Initially, the primary utilisation of X-ray μ-CT was in the realm of medical applications.However, over time, enhancements in the technique have broadened its application to nonmedical felds.X-ray μ-CT introduces a novel possibility of measuring internal geometries that surpasses the capabilities of optical measuring techniques [8].In their study on the CT Scanner Facility at Stellenbosch, the researchers highlighted emerging applications in material science, wood science, and industrial applications, owing to the developments in computer power, hardware, and software advancements [9].Evidently, there is a vibrant research efort to leverage the capabilities of X-ray μ-CT, with explorations extending to areas such as industrial nondestructive testing [10], dimensional metrology [11], additive manufacturing [12], food science [13], and biological sciences [14].Within the scope of this study, the limitations of optical methods in generating reliable qualitative and quantitative data on internal damage led to the choice of X-ray μ-CT as the preferred technique.X-ray μ-CT technique uses X-ray radiation to capture 2D images of an object in many positions about the axis of rotation on which the platform is mounted.In this work, the specimen underwent a complete 360 °revolution while the Xray generator emitted X-ray beams that passed through the specimen and refected on the digital detector in the form of radiographs (projected 2D images) as shown in Figure 1.Te 2D images were subsequently transformed into 3D images using specifc algorithms included in the Volume Graphics software.Tis conversion was done to provide qualitative and quantitative data for more in-depth analyses.
Te calibration of the X-ray μ-CTsystem for dimensional metrology entails optimising key geometric parameters, including the source-to-detector distance, source-to-object distance, detector resolution, detector tilt angle, and rotation centre ofset.Tese parameters are critical for ensuring the accuracy and quality of the reconstructed images and must be meticulously established.However, the focus of this study is not to explore the calibration process in depth or to examine all potential sources of uncertainty that may arise during the measurement process.Te system resolution for this study was estimated from the formula provided in [15].
where r s is the system resolution; M g is the geometrical magnifcation; D r is the resolution of the detector; S is the size of the X-ray focal size; d OD is the distance of the object to the detector; and d SO is the distance of the source to the object.Te geometrical magnifcation: Te typical pixel resolutions of commercial fat panel detectors are within the range of 75-200 μm [16].Te detector resolution is primarily determined by the pixel size of the detector.A detector with a smaller pixel size will have a higher resolution, as it can capture more detail.Conversely, a detector with a larger pixel size will have a lower resolution.
Te smallest focal spot sizes for X-ray μ-CT system using a microfocus and nanofocus tubes are 7 μm and 3 μm, respectively [17].Te focal spot size in an X-ray μ-CT system plays a vital role in determining the resolution and contrast of the X-ray beam.Te smaller the focal spot size, the greater the resolution and contrast of the X-ray beam.Tis high resolution and contrast would allow for the detection and quantifcation of the elemental composition and distributions in the sample with greater detail.
Te detector resolution and focal spot size are not the only relevant parameters in estimating the overall system resolution, but they are among the most important ones.

Experimental Methodology
Tis section deals with the steps and procedures that were followed to carry out the diferent tests.Te construction of the sandwich panel involved an autoclave process using preimpregnated resin (prepregs) with glass fbers for the outer layers, and a foam material at its core.Te arrangement of the prepregs followed a symmetrical and balanced pattern, with the sequence being [(0/90), (+45/-45)] for the top face sheet, followed by the foam core, and then mirrored by [(+45/-45), (0/90)] for the bottom face sheet.A crosslinked and closed-cell polyvinyl chloride foam of density 80 kg/m 3 was used in the construction of the panel.Te planar dimensions of the specimens were 150 mm × 100 mm, selected in accordance with the ASTM D7136/7136M-15 standard for drop weight impact testing.
3.1.Impact Tests.Low-velocity impact tests were conducted on an Instron Dynatup Impact Testing Machine equipped with a hemispherical impactor of diameter 12.5 mm.Various levels of impact energy were applied to the specimens, ranging from barely visible impact damage (BVID) to the point of achieving complete penetration.Te total impact mass of 5.61 kg was used.All tests were conducted at a room temperature of 21 °C.Te testing machine was equipped with a data acquisition device as depicted in Figure 2. Te specifcations of the testing machine are provided in Table 1.
Impact damage was induced by varying the drop height of the impactor so that the desired impact energy levels were reached.In this study, the selected energy levels do not necessarily represent specifc industrial applications except to understand the impact responses aimed at optimising the use of sandwich structures in identifed industrial applications.Te impacted specimens were then taken for damage analysis using X-ray μ-CT.

X-Ray Experimental Setup.
Te setup of an X-ray microcomputed tomography (μ-CT) scanner is essential in ensuring accurate and precise measurements.Te micro-CT scans were conducted on the General Electric Phoenix V-Tome-X L240 scanner with the following specifcations as presented in Table 2.
Te setup steps are summarised as follows: (1) Positioning of the X-ray source, the detector, and the specimen: Te X-ray source, the detector, and the specimen were positioned as illustrated in Figure 3, to optimise the geometric parameters.Te detector was positioned at an angle perpendicular to the X-ray beam to prevent image distortion.(2) Specimen mounting: Each specimen was mounted on a rotating stage located between the X-ray source and detector.It was placed in such a way that the centre of rotation was aligned with the axis of the Xray beam to avoid reconstruction errors.Each specimen was secured on a low-density foral foam Journal of Engineering so that its image would not interfere with that of the tested specimen.Because the sizes of the tested specimens were too long in relation to the detection spectrum, part of either side of the specimen length from the centre was covered.Te covering was aided by a sticky stuf that can be observed on the scan.Tus, the focus of inspection was concentrated on the impacted zone around the centre.
(3) Selection of scanning parameters: Important scanning parameters that were selected include the X-ray tube voltage and current, the exposure time, the It is worth noting that the setup and calibration of the X-ray μ-CT scanner are critical steps that directly infuence the quality and precision of the measurements, thereby necessitating meticulous attention.
Te μ-CT scanner was operated in a temperaturecontrolled environment to prevent thermal drifts that could afect the accuracy of the scans.Te scanner boasts robust shielding designed to impede radiation leakage effectively.Tis includes enclosures fortifed with lead lining, complemented by observation windows crafted from leaded glass, which allow for secure monitoring of the scanning procedure.As an additional safety measure, the computer utilised for data capture is stationed externally to the scanner chamber, as depicted in Figure 4.
Te scanning time lasted for about 2 hours per specimen.Important parameters for the X-ray settings are provided in Table 3.
Te ratio of the distance of the detector from the X-ray source to the distance of the sample or specimen stage from the X-ray source determines the magnifcation, which in this case was 8 (refer Table 1).Voxel size was set at 25 μm.Voltage and current were 130 kV and 150 μA, respectively.Te acquisition time per image was 333 ms carried out during the full rotation of the specimen on which 3000 images were captured.
Te 2D images acquired through the detector were stored in the memory of the personal computer desktop stationed outside the X-ray room.Te CT scans were reconstructed and visualised in a 3D graphic software called Volumetric Graphics VGStudio Max.Complimentary viewer program myVGL 3.5 was used to interactively view the analysis results and 3D visualisations of the impacted zones of specimens.Te 3D reconstruction allowed for the impacted zones to be examined without morphological ambiguities.
Te resolution of the cone-beam X-ray μ-CT system depends on the geometrical magnifcation as illustrated in Figure 5.
Te graph shows a steep reduction of system resolution up to a geometrical magnifcation of 5 for the range of commercial fat panel detectors.Te system resolution curves then decreased gradually as the geometrical   magnifcation increased, approaching the minimum system resolution.Given a focal spot size of 7 μm and a detector resolution range of 75-200 μm, the system produced a resolution range of 11-26 μm.Tis suggests that the system's resolution is more infuenced by the detector resolution than the focal spot size or any other parameters.When the focal spot size was changed to 4 μm, the system resolution was estimated between 10 and 25 μm for the same range of fat panel detector resolution.In other words, even with the very small focal spot size (4 μm), it is the detector's resolution that primarily determines the system resolution.It is also worth noting that the system resolution is much smaller than the detector resolution, which indicates that the reconstruction algorithm and system geometry successfully resolved details smaller than a single detector pixel.Tis setup, with a highresolution detector and small focal spot size, provides the capacity to image fne details within the specimens being examined.

Results and Discussion
4.1.Impact Test Results.Te specimens subjected to impact were examined to assess their responses to such events.Barely visible impact damage was scrutinised using X-ray micro-computed tomography.Following each impact occurrence, notes were recorded regarding the condition of each specimen, as detailed in Table 4. Te accepted margin of error for both velocity and impact energy measurements is within ± 0.1%.Te impact responses were observed on load-time curves.Te area under each curve is representative of the total energy absorbed during the impact event.For all curves, the load increased linearly with the increase in time before the peak load was reached.Te time to reach the peak load depended on the impact energy values.Tus, the higher the impact energy, the shorter the time to reach peak value.For barely visible impact damage, two cases exist.(1) Te impact energy is completely dissipated throughout the specimen, as refected on a load-time graph with a smooth curve.For this loading profle, the peak load was reached at 6 ms.( 2) Te profle depicting a sharp drop in load by approximately 18% is a manifestation of damage initiation as well as a reduction of stifness in the structure.In this case, the peak load was reached at only 3 ms.From the sharp bottom end, the load increased nonlinearly to below the initial peak load.Te second peak load was followed by a smooth curve to zero load at the end of the impact event that lasted for a total 13 ms, as shown in Figure 6.Te frst notable damage mechanism to manifest was delamination (interfacial debonding or debonding of the face sheet and the foam core).However, delamination is preceded by matrix cracking caused by shear and tensile stresses during impact.
Te reduction in stifness continued until the top face sheet was completely perforated with a sharp drop of the load to approximately 75% of the peak load.Tis drop in load suggests a minimum contact force recorded at the impactor at the time of perforation.When the impactor regained full contact with the foam core, it occurred at a lower load in comparison to the peak load before another slight sharp increase.Te load then gradually decreased from a lower peak load to zero and remained constant, as illustrated in Figure 7.
In the near penetrated specimen, the foam core was crushed immediately after the top face sheet was penetrated.Tis resulted in a decrease in load before the densifcation of the foam core due to load increase as shown in Figure 8. Te second sharp curve shows that the back face sheet was nearly perforated because it did not reach the zero load.At this stage, the perforated foam core remained in contact with the impactor.Te load started to decrease gradually from a point above the end point to indicate a complete perforation of the foam core.Te decrease reached the zero load, remained constant for a while before a gradual increase, and then became constant again.Te gradual increase is perhaps an indication of residual energy before full penetration.
It was important to carry out visual inspections for the impacted sandwich composite structures for the diferent impact energies.However, visual inspection failed to identify the delamination inficted on the specimen, as shown in Figure 9.
Near penetration and its corresponding damage morphology suggested a dent diameter to impactor diameter ratio of 1 : 2 as depicted in Figure 10.Te fully perforated damage morphology served as an indicator of the specimen's impact response for the utilised stacking sequence, as demonstrated in Figure 11.

6
Journal of Engineering 4.2.Damage Assessment on the Impacted Specimens Using X-Ray Micro-Computed Tomography. Figure 12 shows the position of impact with respect to the orthogonal views, while Figures 13 and 14 are surface and profle plots, respectively, for the 2 J impact energy.
Te plot was generated using data from the impacted zone and its peripheries.It reveals that within the region of interest (the impacted zone), there is a notable and consistent decline in gray values when contrasted with the adjacent areas.
Te frst impact energy induced an indentation that was only visible through a CT scan.Te second impact energy (3 J) resulted in further stretching of fbres and propagation of crack tips in the matrix along diferent stacking sequence orientations.An indentation was observed at the impacted zone that resulted in initiation of delamination that could not be visually noticed, as illustrated in Figure 15.Te respective surface and profle plots for the 3 J impact energy are depicted in Figures 16 and 17.
Te intensity-distance plot shows a drop in intensity as the depth of penetration increases.Tus, the lower gray values were identifed at the impacted zone.
Te observed consistent reduction in intensity towards the centre of the impacted area suggests the presence of valuable data that could be extracted.Tis fnding necessitates additional research to ascertain whether the grayscale profle within the impacted zone could yield insights into the penetration depth and the characteristics of the impactor employed when inducing the damage.

Validation of Quality of Reconstructed Images.
Te reconstructed images were further refned using ImageJ 2.9.0/1.54 h software, a widely used tool for image processing.Te objective was to enhance the quality of the images.An iterative process was employed, removing outliers and applying various flters to fnd the ones that ofered the best improvements.Interestingly, by visual means, the median flter was found to deliver images that closely resembled the original ones, albeit with a slight reduction in quality.Tis outcome supports the quality and accuracy of the images produced by the X-ray μ-CT.Te images revealed both internal and external features with precision, demonstrating the efectiveness of the adjustments made to the   Journal of Engineering X-ray μ-CT system.Figure 18 shows the comparison of reconstructed (raw) and processed images for 2 J and 3 J impact energies.
Barely visible impact damage was observed at low-level impact energies.Progressive damage was attained through increasing the drop heights and observing the damaged areas until full perforation.Te area of complete penetration displayed a diamond shape, as shown in Figure 11.Tis shape could have been infuenced by the stacking sequence which featured 0/90 °at the outer surface.In a nearly perforated specimen, all failure damage mechanisms were vividly visible using X-ray microcomputed tomography, as featured in Figure 19.Journal of Engineering compression from the top face sheet to the bottom face sheet due to the impactor.Te back face sheet experienced delamination as the primary failure damage mechanism.Fibre breakages occurred at the point of impact, as there was no complete penetration of the impactor at the bottom face sheet.

Conclusions
Te X-ray micro-computed tomography system with an estimated resolution of up to 26 μm successfully revealed the severity of damage induced on sandwich composite structures.Te fndings indicate an immediate reduction in strength and load-bearing capacity upon the onset of damage.Even low-level impact energies can cause barely visible yet structurally signifcant damage, including matrix cracking, delamination, and fbre breakage in the top face sheet, stemming from shear and tensile stresses.Te foam core experienced shear and crushing from the impactor's penetration.Delamination was especially pronounced at the bottom interface, highlighting the vulnerability of sandwich composites to impact damage and their potential for premature failure.Te X-ray micro-computed tomography technique ofers potential opportunities for use in various felds such as engineering, manufacturing, medicine, and biology due to its nondestructive nature and its ability to provide highly detailed internal images of objects without the need for disassembly or destruction.Despite the benefts that this technique ofers, there is need for further resolution improvements.Improving the resolution and sensitivity of micro-CT systems can be accomplished by using novel X-ray sources, detectors, algorithms, and reconstruction techniques.Tis could lead to better characterisation of microstructures in materials science or more detailed inspections of small components.Furthermore, enhancing the speed at which micro-CT scans are conducted would make the technology more suitable for high-volume production.Another potential area of research involves developing hybrid techniques through combining micro-CT with other nondestructive testing methods such as ultrasonics, acoustic emission, or thermography, to provide complementary data, leading to a more comprehensive analysis of the specimen under review.Detailed analysis could also be aided by coupling micro-CT systems with mechanical testing rigs to allow for the observation of material behaviour under diferent loads or environmental conditions, thereby ofering insights into material properties and failure mechanisms.In a broader sense, tailoring micro-CT systems to meet the unique needs of specifc industries, such as electronics, automotive, marine, or aerospace, could lead to custom solutions that address industry-specifc challenges.

Figure 3 :
Figure 3: Experimental setup for the X-ray micro-computed scanner.

Figure 4 :
Figure 4: Personal computer desktop for data acquisition installed outside the scanner room.

FDDFigure
Figure System resolution versus geometrical magnifcation for spot size of 7 μm.

Figure 6 :Figure 7 :
Figure 6: Damage initiation at the second lowest drop height.

Figure 8 :
Figure 8: Damage evolution for a nearly perforated specimen.

Figure 9 :
Figure 9: Front face sheet of impacted specimen with delamination BVID.

Figure 10 :
Figure 10: (a) Specimen showing the state of closer to full perforation.(b) Specimen illustrating the relative damage morphology tomogram.

Figure 11 :
Figure 11: Fully penetrated bottom side of the impacted specimen.

Nomenclature
ASTM: American Society for Testing and Materials BVID: Barely visible impact damage CT:Computed tomography FDD: Focus-to-detector distance FOD: Focus-to-object distance GFRP: Glass fbre reinforced polymer LVID: Low-velocity impact damage NDT: Nondestructive testing PVC: Polyvinyl chloride SHM: Structural health monitoring VID: Visible impact damage μ-CT: Micro-computed tomography.

Table 3 :
X-ray machine test settings.

Table 4 :
Impact energies and statuses of impacted specimens.