Abstract

This study presents the newly developed “Antriebsbatterie-Inspektion mittels Röntgen” (AIR) system, a 2D X-ray imaging technology designed for radiation-safe scanning of entire vehicles. Utilizing a 360 kV X-ray source, AIR acquires high-resolution images from a bird’s-eye view, allowing for comprehensive inspection of battery packs, including their modules and cells. The aim of this study is to evaluate the mechanical integrity of various battery types, including prismatic, cylindrical, and pouch cells, by examining the visibility of cell and gap deformations. For this purpose, radiographic X-ray images of electric vehicles (EVs) with their built-in traction batteries are examined manually for the visibility and separability of individual cells and intercell gaps. Complementary measurements on reference cells with defined defects demonstrate that AIR can reveal typical mechanical anomalies such as layer misalignment, local deformations, and swelling based on cell and gap visibility.

Keywords: conditions and safety assessment, X-ray inspection, electric vehicles, battery packs, AIR

Materials Evaluation 84 (5): 30–36 | https://doi.org/10.32548/2026.me-04556 | ©2026 American Society for Nondestructive Testing

Introduction

The popularity of electric vehicles (EVs) is growing rapidly worldwide: in 2023, over 14 million electric cars were sold, representing a market share of around 18% of new registrations worldwide (IEA 2024). By 2030, it is anticipated that EVs will constitute over 70% of global vehicle production, with one in every four new passenger cars sold being electric (Sakno et al. 2024). This development poses new challenges in terms of the safety and reliability of cells built into battery stacks. Due to the high energy density of modern batteries, it is crucial to examine the mechanical integrity of battery modules, especially after accidents or other events that could cause structural damage (Fransson et al. 2023). Safety incidents, such as fires or explosive battery failures, have highlighted the urgency of these investigations and require proactive analysis of battery conditions.

A promising approach to noninvasive testing is the use of X-ray technology, which allows detailed images of battery modules to be created without compromising their integrity. A newly developed system for inspecting battery packs installed in vehicles using X-rays, known as “Antriebsbatterie-Inspektion mittels Röntgen” (AIR), offers real-time analysis that enables mechanical examination of batteries.

This study presents the investigation of various battery types—including prismatic, cylindrical, and pouch cells—using the AIR system to identify potential safety risks during their lifetime. The results of a comprehensive EV measurement campaign are used to evaluate the visibility of cell and gap deformations and to assess the mechanical condition of the battery modules, thereby demonstrating the capability of AIR to reliably detect defects in EV battery architectures.

Testing Methods

To evaluate the performance of the AIR system, the test setups and the measurement and comparison methods used are described in the following sections.

AIR System

The AIR system (see Figure 1) uses an X-ray inspection method to rapidly assess the mechanical integrity of battery modules in EVs. The abbreviation “AIR” refers to the bird’s-eye view from which the inspection is carried out. The concept of AIR originated from XXL-CT scans, in which EVs are scanned at many different angles to create a 3D digital twin of the vehicle. In contrast, AIR uses only a single scanning orientation, which allows operation with lower X-ray energy, faster scanning times, and simpler handling equipment. As a result, the system has a compact design that is only slightly larger than a car wash and can potentially be installed inside a standard garage instead of a heavily shielded bunker with walls several meters thick.

The X-ray images acquired with this method allow inspection of the interfaces between battery cells, as well as the shape and position of the cells in most regions of the battery system. Unlike in vehicles with an internal combustion engine (ICE), this approach benefits from the distributed arrangement of the cells across the EV underbody, resulting in a near-ideal information distribution for 2D X-ray imaging. Areas obscured by structures such as the steering wheel, seat frame, and similar components are subject to limitations in the analysis.

The main components of AIR include an X-ray source positioned above the vehicle that emits a narrow X-ray beam along the vehicle’s longitudinal axis onto a line detector. The source provides up to 360 kV and 5 mA, which is sufficient to penetrate more than 50 mm of steel or 200 mm of aluminum. Traction batteries typically have a shield consisting of 3 to 6 mm of aluminum or steel. The average dose rate at a distance of 1 m from the X-ray source is ~5 Sv/h.

The detector is an 800 mm long line detector with a pixel size of 83 μm, enabling pixel-accurate scanning of objects. For images requiring a field of view wider than 800 mm, such as entire vehicles, the X-ray components can be shifted horizontally. The system moves on a frame and inspects the region between the front and rear wheels; optionally, the vehicle can be positioned on steel beams so that the entire vehicle can be scanned.

In contrast to many existing X-ray systems, AIR features a cost-effective, simple design based on standard off-the-shelf components. The source and line detector are integrated into a compact top-down line-scan portal. The system is designed for rapid scanning of entire vehicles: between the axles, three to four partial images are acquired in a single longitudinal pass using a narrow, quasi-parallel fan beam and subsequently stitched into a 2D panorama. This enables AIR to achieve throughput on the order of minutes per vehicle at a moderate dose and without any disassembly.

Portable CT System

The small portable CT system (shown in Figure 2) is a computed tomography (CT) system that was also developed at the Fraunhofer IIS. It enables CT measurement of objects with a maximum height of 180 mm and a diameter of 80 mm. The X-ray system is equipped with a 90 kV X-ray source, which allows penetration of up to 80 mm of plastic and 15 mm of aluminum. For objects with higher absorption, a more powerful X-ray source can also be installed. The area detector has 2304 × 1300 pixels with a pixel pitch of 49.5 μm, resulting in a maximal achievable voxel size of 2.5 μm when applying magnification. The objects are positioned on a height-adjustable turntable that allows 360° rotation during scanning.

Experimental Design and Data Collection

Over 100 different EVs were scanned as part of a 10-week measurement campaign. Basic information was recorded for each vehicle, including the date of manufacture, mileage, and any damage or accidents. The scanned vehicles covered a wide range of mileage and model years (2013 to 2024), with Tesla being the most common manufacturer.

Each vehicle underwent inspection using standardized scanning parameters, with the X-ray source operating at an acceleration voltage of 320 kV, an amperage of 5 mA, and an exposure time of 40 ms per line. The images were acquired with an averaging over 8 pixels, corresponding to an effective pixel size of 664 μm. The scanning process was limited to the area between the vehicle axles. Due to the restricted detector size, a total of four images per vehicle were taken and then combined into a single panoramic image for analysis.

For each vehicle, a representative region was selected to quantitatively evaluate how well the individual cells in the battery module could be visualized. The analysis examined the consistency of cell visibility across different car brands, the detectability of potential defects, and the overall visibility of the cells and their gaps, with a specific focus on how easily these features could be recognized with the naked eye.

The visibility of the cells was assessed based on the visibility of the cell edges and the gaps in between. Similarly, gap visibility was assessed based on whether gaps could be clearly distinguished from neighboring cells. The manually determined counts were compared with the total number of cells installed according to the manufacturer's specifications to estimate the visibility of the cells and gaps as a percentage. In cases where the manufacturer did not specify the total number of installed cells or no reliable documentation was available, the cell count was determined directly from the AIR radiographs. Individual cells were manually counted by visual inspection of the images. In regions with overlapping or stacked cells, where single cells could not be clearly separated, the local cell density was estimated by interpolation from adjacent, well-resolved areas and extrapolated to the occluded regions. All cells were included in the total count, irrespective of their degree of visibility.

Visibility was calculated using the following formula: Cell or gap visibility (%) = (Visible cells or gaps / Total number of cells) × 100%

Figure 3 shows the classification of gap and cell visibility using the passenger-side portion of a Tesla Model S. The green outline corresponds to an area with complete visibility of both individual cylindrical cells and intercell gaps. In the yellow area, the cells can be partially detected but with limited discernibility of the intercell gaps. The area under the passenger seat rail is shown in red; here, the individual cylindrical cells and the spaces between them are not visible to the naked eye.

Results and Discussion

Based on the described measurement methodology, the results obtained are presented below and discussed with regard to visibility, defect detection and limitations of the AIR system.

Visibility Performance of Vehicle Models Under AIR

A total of 43 different vehicle models were examined as part of the measurement campaign and are included in the following analysis. For models that were scanned multiple times, one representative scan set was used. The analysis considered the visibility of the cells as well as the gap visibility.

The vehicles examined show a wide variation in the visibility of cells and gaps, ranging from ~95% to 0%. The bar chart in Figure 4 provides an overview of the vehicle models, categorized by visibility intervals. The Tesla Model S shows the best cell visibility at around 94.51%, whereas the Nissan Leaf ranks last with a cell visibility of 16.6% and a gap visibility of 0%. Both extreme cases are discussed in detail below.

Tesla models exhibit the highest cell and gap visibility, averaging around 90% or higher. This is mainly due to the distinct cylindrical battery shape. The Tesla Model S in Figure 5a has the best cell visibility at 94.51%. This image illustrates the visibility achieved with AIR and highlights the clear demarcation of the individual cells. The arrangement of these cylindrical cells allows for effective scanning without significant overlap, resulting in high visibility percentages. This visibility is crucial for assessing the structural integrity of the battery module and identifying potential safety risks. Only the steering axle, steering wheel, and seat rails negatively affect cell visibility. Almost all detectable cells can be clearly distinguished from neighboring cells, even with a gap. This provides the basis for investigating the mechanical integrity of the cells.

In contrast, the situation is fundamentally different for the first-generation Nissan Leaf (Figure 5b). The battery pack consists of eight encased modules arranged in two rows (highlighted in red and orange). Within these modules, the pouch-based cells are stacked horizontally—i.e., the cell layers lie flat on top of each other. As a result, individual cells—and particularly the gaps between them—are not distinguishable. This architecture differs markedly from most modern EV designs, in which cells are typically arranged in a predominantly vertical orientation. In the AIR image, the modules themselves can still be clearly localized, and in the four central modules, a residual X-ray signal is visible. In these regions, fewer cell layers are stacked along the penetration path, so that limited contrast between the cell stack and the surroundings remains. In the lower four modules (red), virtually no X-ray signal reaches the detector; here, multiple cells are stacked directly above one another in the vertical projection, resulting in a very large effective material thickness. Consequently, the image appears almost uniformly dark in these areas (Figure 5b).

In the upper battery module (yellow), the pouch cells are oriented along the scanning direction. In this configuration, the cells are arranged side by side rather than stacked, so that their electrical contacts are clearly visible as dark, well-defined features in the AIR image. Due to the opening angle of the beam, only a few narrow intercellular spaces are irradiated in parallel. These spaces cannot be reliably resolved, and the individual cell bodies appear as a largely contiguous structure without discernible gaps.

This example illustrates that, for the first-generation Nissan Leaf, the combination of horizontally stacked pouch cells and multiple cells stacked in projection severely limits the applicability of AIR: cell visibility is reduced to 16.6%, and gap visibility drops to 0%. The Leaf thus represents the worst-case scenario in the dataset and highlights how strongly cell orientation, stacking direction, and module architecture influence the achievable information content in 2D full-vehicle radiography.

Another limitation arises from layered battery modules arranged vertically, as implemented in the Peugeot e-208 and the similarly designed Opel Corsa (Figure 5c). In this case, the modules are not directly above each other but slightly offset (red in Figure 5c), which nevertheless limits subsequent evaluation. These arrangements increase the effective material thickness encountered by the X-ray beam, exceeding its penetration capability.

This observation demonstrates that the AIR with 360 kV is inherently constrained to single-layer battery systems. In the other areas of the vehicle where no second layer is present, inspection is still possible (green in Figure 5c). Vehicles originally powered by gasoline or diesel that were later converted to electric drive often exhibit this two-layer modular construction, as available space is used to maximize cell count and driving range. Vehicles designed to be electrically powered from the outset usually have only a single-layer modular design. However, with the growing trend toward cell-to-pack technology without modular construction, the applicability of the AIR method is expected to increase further.

Overall, areas with low visibility of gaps and cells were similar across all models. Components that are part of the seats, steering wheel, cooling system, and module mounts typically overlap the cells, resulting in reduced visibility.

This raises the question of how effectively X-rays can detect deformations in individual cells. The measurement campaign shows that a large proportion of the pouch, prismatic, and cylindrical cells can be imaged using AIR. No mechanical anomalies were detected in the scanned vehicles during the measurement campaign. Therefore, artificial samples were selected to assess the sensitivity of AIR. Standard lithium-ion batteries with known deformations were used for this verification. The cylindrical cells in the vehicles are comparable to standard 18650 lithium-ion batteries.

Validation of AIR for Detecting Internal Deformation in Cells

Four cells were selected for further tests to assess the quality of images generated by AIR when compared with computed tomography (CT). Three of these cells were morphologically indistinguishable 18650 3.7 V lithium-ion batteries, while the fourth cell had an open configuration allowing a direct view of the internal components. To verify the accuracy of the scanning techniques in detecting anomalies, defects were deliberately introduced into the open cell structure.

The CT was performed using a small portable CT to obtain 3D images of the cell interior. The distance between the source and the detector (source–detector distance, SDD) was 623 mm, while the distance between the source and the object (source–object distance, SOD) was 169.3 mm. The pixel resolution was 49.5 µm. The voltage and current values used for the radiation unit were 130 kV and 13 mA. The scans were performed individually and took about 5 minutes each. After the scans were completed, the raw data was exported for reconstruction.

Radiographic imaging was then performed using the AIR system, which produced high-resolution 2D images of the samples from a vertical perspective. The voltage and current values of the radiation unit were 200 kV and 4 mA, with a pixel size of 83 µm. For this purpose, a lower acceleration voltage was chosen compared to the vehicle measurement campaign, as only individual batteries were examined rather than entire vehicle structures.

Comparison of the CT and radiographic X-ray images of the first and second cells revealed no noticeable differences. Therefore, only the second cell was used for subsequent comparison with the other cells.

Figure 7 illustrates how well AIR can map internal structures and misalignments in cylindrical cells. The CT image (Figure 7: Intact, CT) of the intact cell shows regularly arranged, parallel anode and cathode layers with uniform layer spacing. This orderly internal structure means that the individual layers appear as clearly separated light and dark bands in the corresponding AIR image (Figure 7: Intact, AIR). The CT thus confirms that the layer structure visible in the AIR image corresponds to the actual electrode geometry.

In the third cell, the situation is reversed: the CT image (Figure 7: Deformed, CT) shows that the electrode layers are no longer parallel to the cell housing but are tilted or locally compressed. In the corresponding AIR image (Figure 7: Deformed, AIR), this misalignment causes the individual layers to overlap in the projection and appear only as a homogeneous dark area without any discernible layer separation; also, the internal winding appears more elliptical. The direct comparison of CT and AIR in Figure 7 demonstrates that layer parallelism and uniform spacing produce a clearly structured AIR signal, while deviations from this—such as inclination or local deformation—lead to a characteristic loss of layer resolution. This underscores that AIR is fundamentally suitable for the detection of internal malposition and structural abnormalities in cylindrical cells, even though, unlike CT, it does not give in-depth information.

In the open battery cell, the upper edge was mechanically deformed using a screwdriver to simulate degradation of the cell layers, as described in Blažek et al. 2025 (Figure 8a). The corresponding CT scan (Figure 8b) allows precise localization of these deformations in the upper part of the cell and clearly resolves the individual cathode–anode layers in this region. In the AIR radiograph (Figure 8c), however, the local indentations at the top of the cell cause the layers to overlap in projection, so that they can no longer be separated individually. Instead, this region appears as a compact, dark area with reduced internal structure.

In contrast, the deformation in the central region of the cell is much more prominent in the AIR image, appearing as a brighter zone relative to the neighboring layers, indicating a locally reduced effective density along the beam path. This suggests that the mechanical disturbance is confined to a limited axial section of the cell rather than extending over its full length, although the exact axial position cannot be determined from the radiograph alone. The CT data resolve this ambiguity, showing that the main deformation is restricted to the upper segment of the battery.

Taken together, Figure 8 demonstrates that AIR is sensitive to localized mechanical damage and density changes in the electrode stack, while CT provides complementary 3D information needed for precise spatial localization and detailed structural analysis.

Detection of Degradation in Cells Using AIR

Figure 9 shows how AIR can be used to detect degradation processes in pouch-based notebook batteries, allowing the findings to be transferred to pouch and prismatic cells in traction batteries. Figure 9a shows an example image of a notebook battery. Figure 9b shows three identical notebook batteries in different states of aging scanned in an upright orientation using AIR. The image was captured with an accelerating voltage of 200 kV, a current of 5 mA, an exposure time of 195 ms, and a pixel size of 83 µm.

The bottom battery shows no abnormalities, either externally or in the X-ray AIR image; the six pouch cells appear compact and homogeneous. In the middle, significantly swollen battery, the casing is clearly bulged, and the central folded area of the electrodes, in particular, exhibits an increased gray value, indicating gas formation, increased internal pressure, and localized density loss (indicated by the red arrow). The top battery shows only slight external swelling and a correspondingly moderate change in its internal structure, revealing a gradual transition from a largely intact state (bottom) to a clearly aged, gas-filled cell (middle). The AIR images clearly show that the external swelling is not primarily caused by severe folding or creasing of the electrode layers, but mainly by gas formation inside the battery, which leads to increased internal pressure. At the same time, the internal structure of the severely swollen cell differs from that of the less-aged cells. Figure 9b thus demonstrates that chemically induced aging processes leading to mechanical stress, gas formation, and structural changes are visible in AIR images. Depending on the cell type, these effects manifest in different geometric signatures (e.g., changes in the folding zone of pouch cells), which can be used to assess the condition of the cells.

When comparing the two X-ray techniques, CT and the AIR system, significant differences are observed. CT provides high 3D spatial resolution and enables volumetric imaging with substantial depth information, whereas the AIR system relies on radiography, which inherently limits its ability to produce fully resolved 3D images comparable to CT. Despite this limitation, the AIR system offers a considerable advantage in terms of scanning time. A comparison of vehicle-level inspections shows that AIR completes scans substantially faster than XXL-CT (Salamon et al. 2019).

The inspection of individual battery cells benefits from the superior performance of CT in terms of 3D volume assessment. With the AIR system, some of the relevant diagnostic tasks can be addressed with much lower effort and without disassembling the cell from the module.

The AIR system successfully identified structural deformations in lithium-ion cells (Figure 7), open-cell configurations (Figure 8), and revealed the mechanisms responsible for swelling in pouch cells, thereby underscoring its practical applicability.

Conclusion

This study introduces the AIR system as a novel, full-vehicle X-ray imaging approach for the safety assessment of traction batteries in electric vehicles. By systematically evaluating cylindrical, prismatic, and pouch cells in over 40 vehicle models, it was demonstrated that AIR can visualize a large portion of the installed cells and the gaps between them, and reveal mechanically relevant features such as misaligned layers, local density changes, and swelling-related degradation.

The results show strong variations in detectability between vehicle designs, with cell visibility ranging from about 95% in the best case (Tesla Model S) down to 16.6% in the most challenging configuration (Nissan Leaf, with 0% gap visibility). Single-layer arrangements of cylindrical cells are particularly well suited for AIR, whereas stacked pouch or prismatic modules and vertically arranged multilayer packs significantly reduce visibility.

Validation experiments with 18650 cells and notebook pouch cells confirmed that radiographic imaging using the AIR approach can reliably detect internal structural changes, such as nonparallel electrode layers, localized deformations, and gas-induced swelling. Compared with computed tomography (CT), AIR does not provide 3D information at comparable spatial resolution, but offers a rapid, cost-effective method for 2D screening of entire vehicles without dismantling.

Several limitations remain. The current evaluation relies partly on subjective assessment of visibility, and incomplete information about the exact number, layout, and orientation of cells in some vehicles restricts quantitative analysis. In addition, the penetration capability of the 360 kV setup inherently limits AIR to single-layer battery systems or regions with moderate material thickness.

Future work should therefore focus on two main directions. First, the diagnostic capability of AIR needs to be validated on traction battery modules and full packs with well-defined, safety-relevant damage—for example, from controlled crash tests, mechanical abuse, or thermal events. This would allow systematic assessment of detection limits, as well as false-positive and false-negative rates for typical defect scenarios at the pack level. Second, the largely manual and subjective visibility assessment used in this study should be replaced by automated evaluation pipelines. Possible approaches include segmentation of cell rows and module structures, quantitative analysis of gap contrast, and detection of characteristic anomalies such as layer misalignment, local density changes, or swelling signatures. In combination, these steps would transform AIR from an exploratory imaging tool into a quantitatively validated, partially automated screening method for traction battery safety assessment in real-world applications.

Acknowledgments

We would like to express our sincere gratitude to all those who contributed to the development and realization of this study. Special thanks are extended to our colleagues at the Fraunhofer Development Center for X-ray Technology for their support and expertise in X-ray imaging technologies. A big thank-you also goes to all private individuals who took part in a measurement campaign with their vehicles.

References

Blažek, P., et al. 2025. “Comprehensive Analysis of Degradation Mechanisms in 18650 Li-Ion Cells Under Prolonged Cycling Conditions.” Journal of Energy Storage 130: 117436. https://doi.org/10.1016/j.est.2025.117436

Fischer, R., et al. 2023. “A Demonstrator for Threat Detection in Volumetric CT Scans.” 12th iCT Conference. https://doi.org/10.58286/27701

Fransson, M., et al. 2023. “Sidewall Breach During Lithium-Ion Battery Thermal Runaway.” Journal of Energy Storage 71: 108088. https://doi.org/10.1016/j.est.2023.108088

International Energy Agency (IEA). 2024. “Global EV Outlook 2024.” Accessed 24 March 2026. https://www.iea.org/reports/global-ev-outlook-2024

Sakno, O., K. M. Bas, and I. Medvediev. 2024. “Trends in Vehicle Electrification.” Avtoshliakhovyk Ukrayiny 3 (280): 50–56. https://doi.org/10.33868/0365-8392-2024-3-280-50-56

Salamon, M., et al. 2019. “XXL-CT Capabilities for the Inspection of Modern Electric Vehicles.” International Symposium on Digital Industrial Radiology and Computed Tomography.

Salamon, M., et al. 2024. “A Novel Test Method for Rapid Evaluation of the Mechanical Integrity.” 32nd Annual Congress of the European Association for Accident Research and Analysis.

Schafer, D., and R. D. Swift. 1994. “AS&E Cargo Vehicle X-Ray Inspection System.” Proc. SPIE 2093: 472–482. https://doi.org/10.1117/12.172526