Program

Tomography in Aerospace: Safran’s Journey, Current Challenges, and Future Needs
Frédéric JENSON, SAFRAN Tech
Over the past decade, Safran has made significant strides in the application of tomography within the aerospace and aircraft industry. This plenary talk will highlight major achievements, current challenges, and future needs, all within the context of Safran’s experience.
Key accomplishments include the deployment of CT inspection in the production of highly critical components for our flagship LEAP engine. This involved the development of automated data analysis algorithms to assist inspectors, ensuring robust and efficient quality control.
Additionally, we have pioneered AI-based algorithms for Assisted Defect Recognition (ADR) and the characterization of material properties in advanced composite materials, addressing both NDT-related challenges and design needs.Efforts also encompass the upgrade of off-the-shelf CT scanners to mitigate specific artefacts and enhance overall performance. Furthermore, studies have been initited on high-energy CT for complex and thick specimens, pushing the boundaries of what is possible with current technology. Looking ahead, we identify strong needs, particularly in inspecting complex-shaped components from additive manufacturing processes.
There is also a significant opportunity to leverage CT for multiple tasks with a single system, integrating volumetric non-destructive inspections and dimensional measurements. This all-in-one approach will ensures a well-integrated system, producing 3D data central to the concepts of the digital twin and numerical continuity.
In this presentation we explore Safran’s journey in tomography, the challenges we face, and our vision for the future of aerospace innovation.

9:45 – WE1A1
CyXTraX: Object Placement Optimization for Industrial CT
Simon WITTL¹, Anton WEISS¹, Jonas SCHNITZER², Gabriel HERL¹ | 1.TC Plattling, Deggendorf Institute of Technology, Germany ; 2.Pattern Recognition Lab, University of Erlangen–Nuremberg, Germany
This study addresses the challenge of optimal object placement by formulating it as an optimization problem for a circular CT trajectory around a known object. To enhance image quality, a localized approach to modeling voxel signals in CT scanning is introduced. Additionally, a novel simulation workflow is developed, incorporating voxelbased signal maps to improve computational efficiency. Validation conducted using a twin robotic CT system and simulated data demonstrates notable improvements in image quality and computational efficiency

10:05 – WE1A2
RadalyX: Portable Multimodal Robotic Scanner
Josef UHER¹, Jana BOHÁČOVÁ¹, Richard KADEŘÁBĚK¹, Jakub VESELݹ | 1.Radalytica a.s., U Pergamenky, Prague, Czech Republic
Robotic Computed Tomography (CT) is establishing itself as a versatile and transformative tool in non-destructive inspection. Robots not only facilitate upscaling CT measurements for large objects but also offer the flexibility of diverse scanning trajectories, making advanced inspection of regions of interest possible. Our portable scanner RadalyX design eliminates the need for objects to “go to the CT machine”; instead, the CT machine “comes to the object,” significantly enhancing usability in aerospace, space, and other demanding fields [1]. This mobility broadens the spectrum of object sizes that can be inspected while enabling the use of smaller, more precise robots even for scans of regions-of-interest of very large structures. This paper presents such capability. and case studies demonstrating the portability of the system, including the inspection of a damaged aircraft, and discuss installation procedures and new robot position calibration methods. We will also showcase recent improvements in scanning strategies, image reconstruction, and the integration of multimodal imaging techniques, providing a glimpse into the future of portable robotic inspection.

10:25 – WE1A3
Masked X-ray Tomographic Reconstruction on a Dual-Arm Robotic Cell
Victor BUSSY¹, Marius COSTIN¹, Jitendra RATHORE¹ | 1.Université Paris Saclay, CEA, List, Palaiseau, France
Tomographic reconstruction is typically performed on a voxel grid that fully covers the field of view and usually the object under study fits entirely within this region. However, many industrial samples are parts or assemblies with a lot of void and discontinuous surfaces and therefore a reduced number of voxels contain the relevant information. By applying a mask for the reconstruction algorithm, it is possible to restrict the process only to the voxels of interest, which has two advantages: an improvement of the reconstruction quality through a reduction of artefacts and a reduction of the reconstruction time. This approach was proven to be effective, but it assumes a prior alignment of the real object with the used mask. In this article, we propose a novel method that uses a mask generated with the help of a distance sensor fixed on the X-ray source and therefore specific to any sample and without a prior alignment.

9:45 – WE1B1
Reflection about the interest to create a Certification for the NDT operators dedicated to X-ray Computed Tomography
Lionel GAY¹, Andre BAILLARD², Vivian DIDIER¹ | 1.COFREND, France ; 2.FrANDTB (French Aerospace Non Destructive Testing Board), France ;
X-ray Computed Tomography (XCT) is increasingly used to get expertise diagnostic for new kinds of components, such as the additive manufacturing parts or the battery cells. 3D volumes generated by XCT are used to detect flaws in the material (usual NDT) but also to measure geometrical characteristics (metrology) when these measures are not achievable with usual metrology means (as some internal dimensions into complex geometry or wall thicknesses). Nevertheless, we have to admit that this use of XCT stays currently rather focused for R&D than for production inspection. As an explanation, we have to consider that these new kinds of components as well as XCT are still involving relatively new technologies in their respective domains. Consequently, still in 2025, 3D XCT stays relatively new for NDT specialists in Radiographic Testing (RT), mostly focused on current 2D Techniques: usual X-Ray films or digital detectors (CR as Computed Radiography and DR as Digital Radiology with flat panels). These 2D Techniques are giving images quite exclusively dedicated to flaw detection in the material. And so, unless exception, these current 2D Techniques are not capable to perform dimensional measurement, such as XCT. Last point to take into account: no certification for the NDT personnel is required when inspection results are not used to attest conformity before delivering the part to the customer. Nevertheless, new kinds of components as above could require the use of XCT, as a new NDT technique, for production inspection and conformity statement (with a Certificate of Conformity or else).

10:05 – WE1B2
Creation of Digital Radiography Certification
Arnaud BAILLY¹, Patrick BOUVET², Benoît CHASTAIN³, Mehdi EL OUARDANI⁴, Raphaël GABEN⁵, Romain JONCHIERE⁶, Olga JOULIE⁷, Charles JUMEAUX⁸, Bruno PAPIN⁹, Yves ROUFFIANDIS¹⁰, Tony SCHROYERS¹¹, Pierrette VAXELAIRE¹² | 1.Mistras, France ; 2.Cetim, France ; 3.SGS, France ; 4.Teneo, France; 5.KNDS, France ; 6.EDF ; 7.Framatome, France ; 8.Wortest, France ; 9.Institut de Soudure, France ; 10.Cap Ain, France ; 11.Technisonic, France ; 12.Self-Employee, France
The place of digital and computed radiography in the industry has increased for the last years. Standardization of these techniques is now enough complete to perform them on welding and casting parts. The last step of this deployment concerns the Non-destructive testing (NDT) personal certification through the last evolution of ISO9712. Cofrend, the French national certification body decided to put in place certification in 2025. Working group started in Feb 2024 with 12 digital radiography French experts to create certification conform to ISO9712 and coherent with the film certification as a new technique in the radiographic testing (RT) method, considering computed radiography (CR) and digital radiography (DR) in the same certification. In the paper, we’ll present the ways chosen by France to create this certification as a new technique.

10:25 – WE1B3
ASME B89.4.23 Performance Evaluation and Interim Testing of X-Ray CT Systems
Joe SCHLECHT¹, Eric FERLEY¹ | 1.VJ Technologies, Inc., USA
ASME recently published the B89.4.23 standard on performance evaluation of X-Ray CT systems. This standard has several advantages over other published standards and guides that benefit users of CT systems. The requirements for measurement direction and length exceed those of other standards to ensure error sensitivity. Material influence and the effects of penetration length are also a required part of the performance evaluation. While this test does not reveal all errors during measurement of an arbitrary workpiece, it does improve over other standards. One drawback of this performance evaluation is the time and cost of testing. To address this concern, the B89.4.23 project team is working towards including a recommendation for interim testing that significantly reduces the time of testing, while maintaining a high likelihood of adherence to specification.

11:15 – WE2A1
On the joint use of simulations and experimental RX images for generating very large datasets with application to object detection tasks
Anthony TOURON¹, Roberto MIORELLI¹, Julie ESCODA¹, Edouard DEMALDENT¹, Souad BANNOUF² | 1.Université Paris-Saclay, CEA, List, France ; 2.IRSN, Fontenay-aux-Roses, France
Machine Learning (ML) and Deep Learning (DL) algorithms are increasingly used in industrial domains to enhance accuracy and efficiency in error-prone tasks. Their performance hinges on data quantity and quality. In Nondestructive Testing and Evaluation (NdT&E), data scarcity, often due to confidentiality and rare events, poses a challenge. This is particularly true for Radiographic Testing (RT) data. To address this, we propose a systematic approach to combine simulated data with experimental radiographic (RX) images, creating large datasets with common defects (e.g., porosity, flaws) found in nuclear welding inspections. This process, called in this work Virtual Flaws (VFs) insertion, enables training ML algorithms without manual labelling or extensive experimental datasets. We evaluate the effectiveness of VFs-based datasets in object detection tasks using state-of-the-art algorithms.

11:35 – WE2A2
Hybrid Modeling of Material Imperfection Indications in X-ray Computed Tomography Datasets
Erik LINDGREN¹, Fabian HANNING¹, Linda SQUILLACI^1-2 | 1.Department of Engineering Science, University West, Trollhättan, Sweden ; 2.GKN Aerospace Sweden AB, Trollhättan, Sweden
Machine learning, more specifically Deep Learning models, have shown remarkable potential for analyzing 3D CT data within both material science and NDT applications. However, the models require large amounts of training data, especially Deep Learning models. Because high quality real experimental data is expensive to derive, humans annotating, synthetic training data is often utilized. In this work we have explored if training data can be modeled with a hybrid approach; where material imperfection indications are transformed directly into real experimental 3D CT reconstructions from 2D and 3D binary representations at higher contrast and resolutions, skipping the usual synthetic dataset derivation steps of reconstructing from simulated 2D projections. Such representations are often abundant from process-development and can be re-used this way for NDT activities. The simple transformations are derived using existing X-ray imaging models (gVXR) and reconstruction algorithms, the FDK in ASTRA Toolbox. We have limited our study to Metal Additively Manufactured samples, Ti-6Al-4V built using Power Bed Fusion Laser Beam process, and superalloy Hastelloy X built with Directed Energy Deposition Laser Beam powder. We demonstrate realistic looking training data sampled with transformation operations low on computational resources. The high-resolution imperfection data is derived both from real experimental data (lack-of-fusion) as well synthetically from other unrelated imaging modalities (lack-of-fusion) as well as from parametric statistical models of superalloy cracks. We believe that the proposed simple and convenient dataset derivation approach, for both Machine Learning training as well as conventional segmentation algorithm parameter optimization, can be valuable to the X-ray NDT community.

11:55 – WE2A3
3D Augmentation of Raw CT Data for Improved Automated Defect Inspections
Nina LASSALLE-ASTIS^1-2, Pascal DESBARATS², Fabien BALDACCI², Romain BRAULT¹ | 1.Cetim, Senlis, France ; 2.Univ. Bordeaux, CNRS, INP, LaBRI, UMR, France
The identification of defects in metallic parts from additive manufacturing is crucial. In many industrial applications, this task is time-consuming and repetitive, prone to errors and high costs. In non-destructive testing, defect inspection is enabled by X-ray computed tomography. This study introduces a novel methodology for future massive defect detection using deep neural networks. Our in-house dataset needs to be augmented with thousands of additional data to ensure the effectiveness and reliability of their training process. Two interconnected modules have been implemented for defect extraction and defect augmentation from projection samples (or sinograms), bypassing the need for reconstruction. The 2D slice-by-slice extractor framework was integrated into a 3D approach to ensure realistic defect representations following augmentation. Geometric transformations with varying parameters within the limits of additive manufacturing observations in the industry were applied to introduce variability into our dataset. Defect (or porosity) sinograms were then merged with the ones of acquired parts in the Radon space to generate realistic volumetric data. This hybrid 2D and 3D processing enhances the quality and realism of the training datasets, enabling robust neural network development. The proposed methodology supports accurate and efficient defect detection in industrial additive manufacturing.

12:15 – WE2A4
X-rays: using simulation to enrich data
Stéphane AMIEL¹, Sofia MARINO¹, Jean RINKEL¹, Julian BETANCUR¹, Jennifer VANDONI¹, Benoît GERARDIN¹ , Frédéric JENSON¹ | ¹Safran Tech, Digital Sciences & Technologies Department, Magny-Les-Hameaux, France
During the manufacture of metal components for aero-engines, a number of non-destructive tests are carried out using digital X-rays to detect potential volume indications. These checks, carried out by certified inspectors, represent time-consuming and tedious work. The use of detection algorithms is therefore envisaged to speed up the inspection process, improve repeatability, decrease fatigue and assist inspectors in their decision-making. Training these algorithms requires a large database of annotated images. Work is therefore being carried out to generate synthetic X-ray data in the form of realistic simulated images, annotated with the presence of defects.

11:15 – WE2B1
New Concept for Detail Sensitivity Monitoring in industrial Computed Tomography – the EURAMET Project SensMonCT
Uwe EWERT¹, Holger ROTH², Simon BURKHARD³, Josephine GUTEKUNST⁴, Virpi KORPELAINEN⁵, Anne-Françoise OBATON⁶, Ulrich NEUSCHAEFER-RUBE⁷, Thomas BLUMENSATH⁹, Marko KATIC⁸ | 1.KOWOTEST GmbH, Germany ; 2.Waygate Technologies, Baker Hughes Digital Solutions GmbH, Germany ; 3.METAS, Federal Institute of Metrology METAS, Switzerland ; 4.Microworks GmbH, Germany ; 5.National Metrology Institute VTT MIKES, Finland ; 6.Laboratoire National de Métrologie et d’Essais (LNE), France ; 7.PTB, Physikalisch-Technische Bundesanstalt, Germany ; 8.FSB, Fakultet Strojarstva I Brodogradnje, Croatia ; 9.University of Southampton, Institute of Sound and Vibration Research (ISVR), UK
Digital X-Ray Imaging and Computed Tomography (CT) are applied in industry for flaw detection, flaw evaluation and dimensional measurement. This requires correct experimental system settings for sufficient visibility and detectability of flaws and structure elements. A new parameter, the Detail Detection Sensitivity (DDS), is introduced. A related standard draft (WK84836-24) has been submitted to the ASTM E07 committee. The ASTM guide E 1441 [1] describes three essential functions for the characterisation of iCT systems. These are the Contrast Discrimination Function (CDF), the MTF, and the Contrast Detail Diagram (CDD). The related procedures and formulas for the determination of these functions and the DDS will be discussed, based on measurements of newly developed test gauges with holes of different diameters with Industrial Computed Tomography systems (iCT) and by modelling. Currently, the DDS of iCT systems is evaluated by human operators which is unreliable and costly. Therefore, within the EURAMET project “SensMonCT” new test gauges and traceable automated measurement and monitoring methods will be developed and procedures to evaluate the DDS of iCT systems and its standardisation.

11:35 – WE2B2
X-ray tomography as a tool for dimensional measurement: challenge of the surface determination
Malik ENNIAFA^1-2-3, Valerie KAFTANDJIAN², Anne-Françoise OBATON³, Sébastien BRZUCHACZ¹ | 1.CETIM, France ; 2.Univ Lyon, INSA Lyon, France ; 3.Laboratoire National de Métrologie et d’Essais (LNE), France
X-ray CT is now more and more used for dimensional measurements. However, the presence of artifacts can degrade image quality, and our aim is to assess their influence on the measurements. In order to carry out dimensional measurements, the first step is to create a surface from the gray-level CT reconstructed volumes. For this purpose we have implemented several surface determination based on different definitions. The 3 criteria defining the position of the surface considered are: a gray level between the one of air and of the part, an edge modeled by a sigmoid, and the maximal local gradient. The first one is computed on different zones: volume wise, region of measurand wise, and point wise. In order to evaluate different physical effects, in this work, we use the simulation software CIVA-RT/CT to create several virtual scans using different physical conditions, from the ideal case to the more realistic ones. The manifestations of physical effects (unsharpness, noise, scattering…) are thus studied for all the volume, all along the XCT measurement chain: projection, reconstructed volume, edge profile, and finally errors to the known true position. Through this study, we hope to bring the XCT operator a better understanding of the complete measuring chain. Moreover, because of its complexity and specificity to use case (part), we propose qualitative metrics and the use of visualisations to constrain and optimize the scan parameters for dimensional measurement

11:55 – WE2B3
Image Quality Analysis of a Laminographic Multiplicative Algebraic Reconstruction Algorithm compared to Filtered Shift Averaging
Weslley Carlos Dias DA SILVA¹, Uwe ZSCHERPEL², Uwe EWERT³ | 1Petrobras Petróleo Brasileiro, Brazil; 2.Bundesanstalt für Materialforschung und -prüfung (BAM), Germany ; 3.Radiology Committee of DGZfP, Germany
 Ionizing radiation imaging techniques are widely used in the industrial area for inspection of components and equipment. Among these techniques, Laminography has been used in situations where there is an asymmetry of the object of interest or in situations where it is not possible to perform a complete rotation of the object or of the X-ray tube and detector around the object for Computed Tomography (CT). Typically, Laminography uses reconstruction algorithms producing volumetric information (3D solid) from a certain number of projections. Several reconstruction algorithms are described in the literature, namely: Tomosynthesis, modified Filtered Backprojection, Algebraic Reconstruction Technique (ART), Simultaneous Iterative Reconstruction Technique (SIRT), Simultaneous Algebraic Reconstruction Technique (SART), Maximum Likelihood and Statistical Reconstruction. In this work, an analytical algorithm, called “Filtered Shifted Average” (FSA) and another iterative algorithm, called “Multiplicative Algebraic Reconstruction Technique” (MART) were optimized and used to perform coplanar translational laminographic reconstructions. Several aspects of image quality, such as analysis of Modulation Transfer Function (MTF), Contrast-to-Noise ratio (CNR), Noise and Detectability were measured to characterize and evaluate the relative performance between the FSA and MART techniques. All laminographic image volumes were obtained using the Coplanar Translational Laminography (CTL) arrangement. Several image acquisitions were made for an increasing number of projections (from 25 up to few thousands) in order to have a greater range of analysis and conclusions about the two algorithms used in this study. The results show that there is no significant dependence of the MTF10% value regarding the tested reconstruction algorithms and the number of projections. The MTF analysis is a suitable tool to describe the different filter properties of FSA and the optimized MART reconstruction. Furthermore, both CNR and the 1/Noise analysis showed higher values for all acquisition conditions with the MART algorithm in comparison with FSA. Regarding detectability, the differences between the algorithms were more accentuated for the condition with a low number of projections. The CTL was applied as one unidirectional scan and as a combination of two orthogonal planar scans, called cross coplanar translational laminography (C-CTL). Finally, there was a significant improvement in the contrast modulation for the images obtained with C-CTL, applying both algorithms for CTL. The choice and correct use of an iterative algorithm (MART) yields several advantages regarding the image quality, despite requiring more computational resources than the analytical algorithm (FSA).

12:15 – WE2B4
Use of Digital Focal Spot Measurement for Optimisation of X-ray Tubes in Development and Production
Klaus BAVENDIEK¹, Florian FELSKE², Alexander WARRIKHOFF² | 1.KOWOTEST GmbH, Germany ; 2.rtw RÖNTGEN-TECHNIK DR. WARRIKHOFF GmbH & Co. KG, Germany
For the imaging and standard-compliant measurement of X-ray tube focal spots, Kowotest GmbH launched the digital focal spot camera KowoSpot X in 2021, which uses a small digital detector to produce high resolution images of focal spots within seconds. In addition to the lifetime monitoring of focal spots in existing systems, the KowoSpot X also offers great advantages in the development and production monitoring of X-ray tubes. The KowoSpot system is used very successfully at rtw RÖNTGEN-TECHNIK DR. WARRIKHOFF GmbH & Co. KG in the development and production of MCTG rod anode X-ray tubes. These X-ray tubes have been specially developed for industrial use for the inspection of welds in heat exchangers and operate with voltages of up to 150kV with focal spot sizes of less than 200µm in some cases. As the focal spot is electromagnetically guided and focused by the anode rod, it is necessary to adjust the focal spot during the development and production process. For this purpose, rtw uses the KowoSpot X. The digital display and standard-compliant evaluation of the focal spot simplifies and speeds up this process. An exact and standard-compliant measurement of the focal spot and the resulting resolution is essential for the development and production of X-ray tubes at rtw, which is why the KowoSpot system is currently being integrated into the production process for other types of tube, e.g. panoramic tubes.

Potentials of Quantum Computing for Tomography
Theobald FUCHS (Fraunhofer EZRT)
We will discuss the current state of Quantum Computing and give a short overview of the theoretical background. The concept of qubits, superposition, entanglement and measurement of quantum states will be explained. Quantum gates and circuits built from these gates will be introduced. Next, we will report on the current state of our research on several applications of QC. These include an implementation of cross-sectional image reconstruction for CT imaging based on a QUBO-method (Quadratic Unconstrained Binary Optimization), used to find a mathematical minimum of a metric. Another topic, we are working on, is optimization of X-ray source trajectory in CT with respect to image quality achievable by a restricted number of angular positions. In addition, we evaluate common problems of image processing for instance noise reduction and search for ways to implement respective methods on the QC.

14:45 – WE3A1
Learning a projection term on a manifold to constrain tomographic reconstruction
Victor BUSSY¹, Julie ESCODA¹, Anthony TOURON¹ | 1.Université ParisSaclay, CEA, List, France
Inverse problems, such as signal recovery from noisy measurements, have traditionally been addressed through optimisation techniques that balance data fidelity with regularisation. Classical regularisation methods, including total variation (TV), promote sparsity in a predefined transform domain, such as the Fourier domain or learned dictionaries [1]. However, with the advances in deep learning and large-scale datasets, significantly more expressive regularisation techniques have emerged [2]. Deep neural network post-processors have gained popularity, yet they do not inherently ensure that the reconstructed signals adhere to the underlying physical constraints, which is unacceptable for non destructive applications [3]. More recent approaches, such as unrolled optimisation methods—including primal-dual hybrid gradient techniques with learned parameters—integrate data fidelity terms directly within the optimisation framework alongside regularisation [4]. Additionally, neural networks can serve as priors within variational inference frameworks, such as diffusion model based methods [5]. In this study, we focus on a family of priors using an adversarial regulariser. The core idea behind this latter approach is to construct a distance term capable of distinguishing between the distribution of ground-truth images and that of corrupted images by learning the manifold M on which ground-truth samples reside [6]. This class of models is generally simpler to implement than generative adversarial networks (GANs), as it does not require access to corrupted samples. Adversarial regularisers and GAN-based methods have already been employed in various ways for ill-posed problems, such as direct optimisation within a latent space [7] or through projection onto this space [8].

15:05 – WE3A2
An Investigation into Fully AI-Driven Reconstruction in X-Ray Computed Tomography
Robin TENSCHER-PHILIPP¹, Tim SCHANZ¹, Martin SIMON¹ | 1.Institute of Applied Research (IAF), University of Applied Sciences, Karlsruhe, Germany
Classical image reconstruction methods, based on analytical or iterative techniques, despite their maturity, still exhibit limitations in industrial applications. Notably, artifacts often arise when combining materials with high and low absorption properties. Recent studies have explored AI-assisted reconstruction methods, typically implemented as hybrid approaches, which also inherit some weaknesses of traditional techniques. In this work, we present the initial results of our investigation into fully AI-driven reconstruction methods. By utilizing end-to-end models, we optimize the entire reconstruction process, enhancing both image quality and efficiency. Adaptive methods are employed that dynamically adjust to varying scan conditions and application requirements. In the first phase, the system is trained and validated using synthetic data, starting with ideal datasets and progressively introducing real-world elements such as noise and artifacts. Our findings demonstrate that the convolutional neural network (CNN) is capable of learning from projection data and producing accurate reconstructions, both in terms of geometry and grayscale values.

15:25 – WE3A3
A Fast Deep Incremental Angle Refinement Model for Limited-Angle CT Reconstruction
Xingyu LIU¹, Guangpu YANG¹, Ammar ALSAFFAR¹, Faizan AHMAD¹, Steffen KIESS¹, Sven SIMON¹ | 1.Department of Computational Imaging Systems, ITI, University of Stuttgart, Germany
Industrial computed tomography (CT) plays a critical role in non-destructive testing and quality control across various industries. However, one of the key challenges in industrial CT lies in artifacts caused by limited-angle tomography, where geometric constraints prevent full rotation of the object. To mitigate these artifacts, we propose an innovative framework leveraging the diffusion model, a state-of-the-art generative model that has garnered significant attention for its powerful generation capabilities. Despite their impressive results, traditional diffusion models typically require upwards of 1,000 intermediate steps to achieve accurate output, leading to substantial computational demands and long inference time, which significantly limit their practical application. Our approach addresses these issues by integrating reconstructed data from different limited angles scans as intermediate steps, substituting the conventional steps of adding random Gaussian noise. This modification significantly reduces the number of intermediate steps necessary for both training and inference, thereby enhancing computational efficiency. Unlike other acceleration techniques that often compromise output quality for speed, our method achieves a significant reduction in computational burden while improving key performance metrics, such as structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). By aligning the training process more closely with real-world data degradation, our approach not only accelerates diffusion models but also enhances the fidelity and quality of the resulting reconstructions, making it a highly promising solution for industrial CT applications.

15:45 – WE3A4
Exact reconstruction for helical cone-beam X-ray CT using the ASTRA toolbox
Pavel PARAMONOV^1-2, Joaquim SANCTORUM^1-2, Femke DANCKAERS^1-2, Jan De BEENHOUWER^1-2, Jan SIJBERS^1-2 | 1.imec-Vision Lab, University of Antwerp, Belgium ; 2.DynXlab: Center for 4D Quantitative X-ray Imaging and Analysis, University of Antwerp, Belgium
The ASTRA toolbox is a popular open-source software package for X-ray CT (XCT), supporting CT reconstruction for a wide variety of scan types. However, helical CT, which is extensively applied in medical and industrial imaging, has so far been available only through iterative reconstruction algorithms. This paper introduces an open-source Python add-on to ASTRA, implementing the state-of-the-art algorithm for helical CT – Katsevich’s algorithm. Belonging to the filtered backprojection (FBP) family, it offers exact reconstruction with high computational efficiency. Our implementation leverages the GPU acceleration through ASTRA’s algorithms and dedicated CUDA kernels to provide efficient backprojection, significantly reducing computation time compared to iterative methods. The developed Python package is built on top of ASTRA and includes all routines needed for helical CT: definition of scanning geometry, projection data filtering, and backprojection. To evaluate the accuracy and performance of the proposed implementation, we simulated projections of a CAD model resembling an 18650 battery. The XCT image reconstructed with Katsevich’s algorithm is compared to the one reconstructed with SIRT. Our findings show that our implementation achieves high-quality image reconstruction with substantially faster processing times and reduced memory requirements.

14:45 – WE3B1
Application of radiographic criteria to Computed Tomography inspection in non-destructive testing
Sébastien Antoine BRZUCHACZ¹ | 1.MCO, CETIM, Senlis, France
Non-destructive testing of industrial parts using X-ray computed tomography is hampered by the absence of standards defining acceptance criteria for defects generated by the various manufacturing processes available, such as casting or additive manufacturing. To define acceptance criteria for a part inspected by tomography, it is currently necessary to use the standards used for X-ray inspection. A study has been launched at Cetim to test different tomographic image processing methods for applying the criteria defined in radiographic standards. The aim of this study is not to compare both inspection techniques, but to define defect sizing methods using voxel data processing software, and then to assess their relevance to the radiographic criteria applied. Our examination of existing standards for foundry, forging and welding has led us to consider two main types of criteria: acceptance threshold criteria, for which the dimensions of the defect are compared with those accepted by the standard, and visual comparison criteria, which involve comparing an area of the image under examination with a reference image of the same dimensions containing the same type of defect as the one whose harmfulness we are seeking to assess. For each of the two types of criteria, different sizing and comparison methods were tested on a scan of a casting and a scan of a weld. Their relevance was then assessed in terms of level of similarity with a radiographic examination, ease of implementation and level of performance in terms of dimensioning. It appears that not all criteria are systematically applicable in tomography and that they do not offer the possibility of exploiting the 3rd dimension offered by tomography. Finally, recommendations were formulated, as well as prospects for further development, which we feel are essential if we are to take greater advantage of the specific features of X-ray tomography and avoid further penalizing this type of inspection, whose acquisition times are long and already quite costly.

15:05 – WE3B2
New method for characterizing contrast sensitivity in radiography: application to a photon-counting detector
Jean RINKEL¹, Clément REMACHA¹, Alexiane ARNAUD¹, Benoît GERARDIN¹, Frédéric JENSON¹ | 1.Safran Tech, France
The contrast sensitivity of X-ray detectors is a key parameter for the quality of flaw detection and evaluation, as well as dimensional analysis of high-density turbine blades by radiography and tomography. Photon-counting detector technology is a promising solution for non-destructive testing, already commercially available for medical applications. In this paper, we propose a new method for determining the contrast sensitivity of X-ray detectors, which is both robust and easy to implement. We aim to compare the results obtained with a conventional energy-integrating detector and a photon-counting detector.

15:25 – WE3B3
3D resolution in microCT: a comparison study on visual resolution versus MTF measurement
Solène VALTON¹, Elona LIETAERT¹, Paul MOYAU¹ | 1.RX Solutions SAS, France
As any measurement instrument, X-ray microcomputed tomography (microCT) systems are concerned by a number of complex specifications like accuracy, maximum permissible error, or resolution. In optics, resolution refers to the ability of an imaging system to distinguish two closely spaced objects. This is commonly extended to 3D microCT images as the ability to visually distinguish features on a reconstructed slice. In ISO 15708 standard, resolution in computed tomography is quantified by the 3D modulation transfer function (MTF). In this work we have studied the correlation between a 3D MTF measurement method and the visual separation of features on a CT volume on microfocus and nanofocus tubes for voxel size ranging from 10 μm to 0.4 μm. We found a very good correlation down to 1 μm. Below 1 μm the comparison was not possible due to the lack of 3D sample for the visual feature observation but the MTF values were in accordance to the expected resolution.

15:45 – WE3B4
Evaluating Image Quality: A Comparative Study of Computed Radiography in Aerospace Applications
Muzibur KHAN¹, and Trent GILLIS¹ | 1.Aerospace Research Centre, National Research Council Canada, Montreal, Canada
The aerospace industry enforces rigorous product quality standards to guarantee the structural integrity of essential components, ensuring the safety and airworthiness of aircraft. Non-destructive inspections (NDI) are regularly conducted to ensure product quality and detect problems before they reach critical dimensions. Various NDI techniques, such as industrial radiography, are essential in the inspection process, with industrial radiography being the predominant technique for identifying volumetric problems. Industrial radiography offers excellent efficiency in non-destructive testing, especially for aircraft structural components, making it a crucial tool for aviation maintenance. Film-based radiography requires consumables (films, hazardous chemicals, and appropriate disposal of chemical waste), a darkroom facility, and manual processing. This process is not only time-consuming but also entails greater radiation exposure compared to digital systems. Digital radiography eliminates the need for films and processing chemicals. Industrial radiographic inspection is currently evolving towards two types of digital technologies: (1) Digital Detector Arrays (DDA), sometimes referred to as flat panel detectors in digital radiography (DR), and (2) Computed Radiography (CR). The digital radiographic images possess common attributes and characteristics, regardless of the technology used for their acquisition. Both CR and DDA technologies have their own advantages and disadvantages, and the specific requirements of the application play a pivotal role in the selection. This paper provides a review of the key parameters that characterize the image quality of Digital Radiography (DR) and Computed Radiography (CR), explores the available evaluation methods used to measure these parameters, ensuring a thorough understanding of how to assess and enhance image quality in both DR and CR systems. It examines the various factors that influence each parameter of image quality. In addition, this article presents a detailed comparison of the image quality produced by deploying both Computed Radiography (CR) and Digital Radiography (DR) on the same aerospace components. By using identical components for both imaging techniques, the study aims to highlight the differences in image clarity, resolution, and diagnostic effectiveness. This comparative analysis will provide insights into the strengths and limitations of each method, offering a comprehensive evaluation of their performance in aerospace applications.

16:35 – WE4A1
Inspection of Composite Pipelines Using X-ray Imaging 
Weslley Carlos Dias da SILVA¹, Uwe ZSCHERPEL², Uwe EWERT³ | 1.Petrobras Petróleo Brasileiro, Brazil ; 2.Bundesanstalt für Materialforschung und -prüfung (BAM), Germany ; 3.Radiology Committee of DGZfP, Germany
The integrity of subsea equipment is of paramount importance for the operational continuity of oil and gas production. Due to recent problems with stress crack corrosion (SCC) corrosion in flexible metallic pipelines, Oil & gas operators and pipelines suppliers are developing flexible pipelines made of composite materials, as these are not susceptible to classic corrosion mechanisms. This work presents the results obtained with three radiographic techniques (digital radiography with digital detector array (DDA), computed laminography (CL) and computed tomography (CT)) applied to TCP (Thermoplastic Composite Pipe) pipelines. Artificial discontinuities were inserted in these samples by crushing test experiments. In sequence, the above X-ray imaging techniques were used to detect and size the discontinuities generated in the samples (typically: cracks and delaminations). The results demonstrate the technical viability of digital radiography using DDAs as well as CL and CT for 3D reconstructions of the inspections of TCP pipelines. Notably, CT showed the best results. It was possible to detect and size delaminations with openings larger than 96 µm with the used system. Using DDA and CL it was possible to detect and size other discontinuities in the samples. It is important to mention the CL and CT 3D volume reconstruction capability. Additionally, CT in a fast scan mode showed satisfactory results compared to a slower scanning CT. Based on the results obtained and the operational characteristics of each technique, it is vertified that radiography can be used as a standard testing method, recommending the use of CL, and best CT, especially at points of the TCP pipelines which require more accurate and detailed information on the discontinuities present.

16:55 – WE4A2
Radiographic testing of prestressed concrete bridges: investigation of alternatives to gamma sources
James GILBERT¹, Sylvie ARNAUD¹, Florent PLASSARD1 | 1.Cerema, France
Prestressed concrete bridges in France require regular detailed inspection, with one focus being the evaluation of steel tendon conditions. The aging infrastructure and recent bridge failures underscore the importance of structural investigations. Corrosion of steel tendons, often caused by grouting voids, is a significant risk. Nondestructive testing (NDT) is preferred, with radiographic testing (RT) being the primary acceptable technique for evaluating tendon grouting. Cerema operates the only bridge radiography unit in France capable of investigating prestressed concrete elements up to 60 cm thick. Due to stricter safety regulations, Cerema is exploring alternative methods to replace sealed radioactive sources. The project aims to investigate high-energy X-ray generators, particularly particle accelerators, as potential replacements. The methodology includes comparing NDT methods, conducting market studies, and evaluating particle accelerators. Criteria for comparison include X-ray energy, dose rate, size, and robustness. Two types of accelerators, Betatron and miniaturized LINAC, are being studied further. The project also explores alternative detection techniques and has developed test concrete blocks for repeatable comparisons. Support systems for positioning accelerators are being investigated, to access underneath bridges. The research prioritizes circular and linear particle accelerators as X-ray sources, with ongoing experimentation to optimize source-detector pairing.

17:15 – WE4A3
Comparison of CT Imaging Methods for Defect Detection in a Multi-Material 7-Pin Power Connector
Jochen BUTZER¹, Thomas RIEDEL¹, Céline WOELK1, Uta PROETTEL¹, Annika DOERING¹, Philipp KURTH², Fabian RÜCKERT² , Darius RÜCKERT² | 1.Corporate Research, Robert Bosch GmbH, Germany ; 2.Voxray GmbH, Germany
This paper explores industrial computed tomography (CT) for detection of anomalies such as voids and gaps in power connectors for automotive application. Such defects can lead to leakage paths which can cause electrical malfunction or short circuits. A primary challenge in CT analysis arises from the proximity of dense metal parts next to plastic material, which induces strong imaging artifacts that hinder the evaluation of such defects. This study evaluates three imaging methods – standard Filtered Back Projection reconstruction (FBP), Quantum Reconstruction Technique (QRT), and dual-energy combination (DE) – to mitigate these artifacts and improve defect visualization. Each method’s efficacy is assessed by analyzing the accuracy in detecting and characterizing gaps and voids. Results are compared to a target preparation and cross-section analysis with optical microscopy, providing a benchmark for reliability and accuracy. Full experimental findings and comparisons are detailed in the complete paper.

16:35 – WE4B1
Characterization of the damage mechanisms of aeronautical composite materials via synchrotron in-situ tensile testing
Georges GIAKOUMAKIS¹, Juan-Manuel GARCIA¹, Henry PROUDHON², Cedric HUCHETTE¹ , Frédéric LAURIN¹ | 1.Université Paris-Saclay, ONERA, France ; 2.Centre des Matériaux – Mines Paris, CNRS UMR, France
For the past few decades, the aerospace industry has been increasingly using carbon-fiber-reinforced polymers in the production of its aircraft, thanks to their excellent specific properties. In particular, novel composite materials made of carbon fibers embedded in a thermoplastic matrix are being considered by aircraft manufacturers for future generations of civil aircraft fuselages. These high-performance materials offer numerous advantages compared with standard composite materials, in particular new possibilities for welded assemblies, higher toughness and improved recyclability. Prior to their integration into aeronautical applications, their specific features, particularly in terms of damage, need to be studied experimentally. For this purpose, we present a fine characterization and quantification of the damage behaviour during synchrotron in situ tensile tests of two carbon fiber reinforced polymer matrices. The first type consists of a thermoplastic matrix (CFRM-TP), while the second type consists of a thermoset matrix (CFRM-TS). The motivation of this study is to : 1) Understand the differences between the damage mechanisms of the CFRM-TP and the CFRM-TS composites; 2) Extract quantitative data (average crack length, Crack Opening Displacement (COD), etc.) to develop a micromechanical discrete damage model [1]; 3) Generate a data basis to identify and validate an efficient continuum damage model [2,3] at the mesoscale.

16:55 – WE4B2
Non-destructive testing of electronic components using X-ray computed laminography and tomography
Sascha SENCK¹, Boris KOMLJEN², Sarah HEUPL¹, Patrick WEINBERGER², Simon ZABLER², Till HUESGEN³, Stephan KASEMANN⁴, and Otto KREUTZER² | 1.University of Applied Sciences Upper Austria, Austria ; 2.Technologie Campus Plattling, Technische Hochschule Deggendorf, Germany ; 3.Hochschule Kempten, Germany ; 4.Vorarlberg University of Applied Sciences, Austria
Ensuring the reliability of electronic components is critical in industry 4.0. Non-destructive testing (NDT) methods, such as X-ray based laminography and microcomputed tomography (μCT), are powerful tools for the detailed inspection of printed circuit boards (PCBs) and other complex electronic assemblies. While μCT provides three-dimensional volume data, its utility for large, planar samples like PCBs is constrained by geometrical limitations. Laminography is a complementary technique that enables layer-specific imaging in high-resolution, overcoming constraints associated with traditional μCT. This study investigates the integration of these imaging modalities to characterize internal defects, such as bond failures and microcracks in multi-material PCBs assemblies. Scanning and reconstruction parameters were optimized to reduce image artifacts and improve defect detectability. The focus is on laminography to inspect large, flat assemblies that were loaded under realistic operation conditions. The method was applied to operational PCBs, demonstrating its effectiveness in isolating defects within specific layers while minimizing edge distortion. For example, scans of fully assembled PCBs with a 10 cm diameter achieved voxel resolutions of 5.5 x 5.5 x 19.3 μm. By combining laminography and μCT, this work provides a comprehensive analysis of their respective advantages and limitations for NDT of electronic components. Insights gained from these imaging techniques are crucial for improving the design, manufacturing, and maintenance of high-reliability electronic devices, ensuring their long-term performance and safety in critical applications, e.g. in electric vehicles.

17:15 – WE4B3
In-situ observation of electrolyte movement in commercial 18650 Li-ion battery cells under various temperature conditions by Industrial X-ray Computed Tomography
Zuzana STRAVOVÁ¹, Tomáš ZIKMUND¹, Jozef KAISER¹, Jakub ŠALPLACHTA³ , Gergő BALLAI², Dániel SEBOK², Ákos KUKOVECZ² | 1.Central European Institute of Technology, Brno University of Technology, Czech Republic ; 2.Department of Applied and Environmental Chemistry, University of Szeged, Hungary ; 3.CactuX s.r.o., Brno, Czech Republic
Industrial X-ray Computed Tomography (CT) is widely used for non-destructive imaging but traditionally relies on ex-situ experiments. In contrast, in-situ CT enables real-time observation of structural changes under operational conditions, offering valuable insights into dynamic processes. This study presents a novel in-situ approach to visualising dynamic electrolyte movement in the commercial 18650 Li-ion cell during charge-discharge cycles. This critical yet underexplored phenomenon is linked to pressure fluctuations and potential safety risks. The in-situ approach advances LIB research by combining real-time imaging of battery ageing and environmental temperature simulation, offering deeper insights into structural behaviour during operation.

High energy laser-based X-ray source
Cédric Thaury, ENSTA – Paris Tech /LOA
Thanks to accelerator fields that are 3 to 4 orders of magnitude stronger than those in conventional electron accelerators, laser-plasma accelerators promise a dramatic reduction in the footprint of these instruments. Although the average current produced by this emerging technology is still much lower compared to established technologies, it is nevertheless strong enough to consider medium-term industrial applications. The first targeted applications leverage the strengths of laser-plasma accelerators, such as the extremely high peak current, which is crucial for flash radiotherapy, or the micrometric size of the electron beam source, which is a key factor for high-resolution industrial tomography. In this presentation, we will introduce the principle of laser-plasma acceleration, highlighting its strengths and current limitations. We will then discuss the ongoing efforts to demonstrate its potential for radiography.

9:45 – THU1A1
Tomography strategies dedicated to laser-based Kα X-ray source
Adrien STOLIDI¹, Victor BUSSY¹, Raphaël CLADY², Olivier UTÉZA² , Amélie FERRɲ | 1.Université Paris-Saclay, CEA List, France ; 2.Aix-Marseille Université, CNRS, LP3 UMR, France
Laser-based Kα X-ray sources offer compact and cost-effective alternatives to large-scale facilities like synchrotrons, providing high photon flux and ultrashort pulses ideal for advanced imaging. However, challenges such as flux variability, target refresh limitations and acquisition constraints hinder their use in tomography. This study presents tomographic strategies leveraging a priori knowledge of the sample and advanced denoising techniques, tested on simulated data.

10:05 – THU1A2
Thermal neutron imaging in the local lab
Serge DUARTE PINTO¹, Olivier BONNET¹, Julie DALLEAU¹, Sean GARDELL¹,Olivier MERLIN¹, David PASQUALE¹, Steve RITZAU¹, Axel RIZZO¹, Dmitriy TIPIKIN¹, Alexis TOUSSAINT¹, and Vincent VAN STEENBERGEN² | 1.Exosens, Paris, France ; 2.ThinkDeep AI, Mérignac, France
We have started an effort to bring the benefits of neutron radiography to the local laboratory. We have started imaging experiments with a fusion neutron generator as a source. A generator is a commercially available, compact neutron source. With a flux that is orders of magnitude weaker than a reactor, ingenuity is required to make good-quality images. The sensitivity of the detector is of vital importance. We have developed the most sensitive neutron imaging detector on the market, based on our patented neutron-sensitive microchannel plates. We will show the principles and design of the detector and demonstrate some results. We reveal the first imaging results using our DD fusion generator as a source, proving the feasibility of this type of instrument. And we outline our plans and ambitions for the immediate future. This development will democratize access to neutron imaging and make it available to the non-destructive testing community.

10:25 – THU1A3
Localized Phase Retrieval Application for Analyzing Multi-Material Samples
Lukáš MALEČEK¹, Marek ZEMEK¹, Yoshihiro TAKEDA², Kazuhiko OMOTE², Tomáš ZIKMUND¹, Jozef KAISER¹ | 1.CEITEC, University of Technology, Brno, Czech Republic ; 2.Rigaku Corporation – Tokyo, Japan
X-ray computed tomography (CT) is a powerful imaging technique that provides three-dimensional insights into the internal structures of objects without mechanical intervention. Initially developed for medical diagnostics, CT has evolved into a versatile tool in various scientific and industrial fields. However, standard absorption-based CT has limitations when imaging materials with low X-ray attenuation or multi-material samples with similar attenuation values, often resulting in insufficient contrast for detailed analysis. To overcome these challenges, phase-contrast CT has emerged as a valuable approach, leveraging changes in the phase of X-rays as they interact with the sample. One of the most accessible phase contrast imaging methods is the propagation-based approach, which typically requires a phase retrieval application. This application increases the contrast of individual structures, but on the other hand, it leads to degradation of the spatial resolution in heterogeneous samples, particularly when high-density particles are present. This paper presents so-called localized phase retrieval, which offers a novel solution to this limitation. Instead of applying the phase retrieval algorithm globally, this method focuses on specific regions within reconstructed tomographic slices, preserving spatial resolution and improving material distinction. This localized approach holds promise for more accurate and efficient analysis of complex multi-material samples, paving the way for advancements in computed tomography applications.

9:45 – THU1B1
Crack segmentation in radiographic images of additively manufactured parts 
Tebogo LEDWABA¹, Christine STEENKAMP¹, and Anton DU PLESSIS^1-2 | 1.Department of Physics, Stellenbosch University, South Africa ; 2.Comet Technologies Canada Inc
Cracking constitutes a critical defect in additive manufacturing (AM) and mechanical testing. X-ray imaging and processing allows for non-destructive evaluation of cracks and is well suited for AM. Research has been carried out on deep learning-based image segmentation in X-ray computed tomography (XCT) data [1–3], and specifically also focusing on AM parts [4,5]. However, radiographic X-ray inspection is popular in industry, not so much in research, because it is affordable and allows for quick visualization. Cracks in radiographic images are hard to detect and may be missed because they are often thin and small. Therefore, there is a need to enhance visualization of cracks in X-ray radiographs to assure accurate quality inspection and evaluations. This paper presents preliminary work towards the development of a generic deep learning-based network that successfully segment cracks in radiographs of AM parts. We used an XCT system to acquire radiographic images of AM hourglass samples made from Ti6Al4V from different angles. These samples were subjected to fatigue testing which resulted in various kinds of internal cracks. We trained and validated different deep learning architectures on thousands of manually segmented and augmented datasets. The resulting models were evaluated and employed to unseen images. We present the inspection results as well as a comparison of the trained models to conventional segmentation methods. Future work involves extending the network by adding datasets of various sample geometries and image qualities to make it more robust and widely applicable. The project aims to develop a scientifically effective non-destructive testing method that allows automation, quick crack detection and quantitative analysis in metal AM.

10:05 – THU1B2
A projective digital volume correlation approach for wall thickness measurement by multiple view X-ray imaging of metallic turbine blades
Julian BETANCUR¹, Gautier LOKOU², Gael NUE², Clément REMACHA³ | 1.Safran Tech, Digital Sciences & Technologies Department, France ; 2.CEI Aubes de turbines, Safran Aircraft Engines, France ; 3.SafranTech, Plateforme Aubes de Turbines Avancées, France
Projective digital volume correlation (P-DVC) enables to estimate geometric indicators of superalloys turbine blades with complex geometry from few X-ray projections. During P-DVC, the geometry of the part is estimated by optimizing the difference between simulated and acquired X-ray projections. In this work, we study the use of P-DVC from initial geometries by 3D fringe scanners. We present the comparison of wall thicknesses measured from surfaces estimated by P-DVC with wall thicknesses measured in the metal volumes from X-ray computed tomography (CT). Results showed improved concordance of wall thicknesses by P-DVC with wall thicknesses by X-ray CT than the concordance of wall thicknesses from initial geometries and wall thicknesses by X-ray CT. The P-DVC approach has the potential to decrease the cost of wall thickness inspection compared to their inspection by X-ray CT alone.

10:25 – THU1B3
Image Super-resolution Algorithm for Fan Beam CT Based on Denoising Diffusion Probabilistic Models 
Han YU¹, Yingchi LIU¹, Xinyue LI¹, Xingjie LI¹ | 1.China Academy of Machinery Shenyang Research Institute of Foundry
With the development of advanced manufacturing industries, such as the emergence of 3D printing technology, the structure of metal parts is becoming increasingly complex. However, traditional non-destructive testing techniques are powerless for measuring internal dimensions with cavity structures. Fan beam CT technology collects multi angle projections of the object to be measured through a linear detector, and combines reconstruction algorithms to observe a certain section, thereby achieving size measurement and defect detection. The micro-motion technology of detectors aims to merge multiple discrete data points that are less than one pixel apart by making small movements of the detector in the array direction, in order to obtain images with higher resolution than the detector. However, the micro-motion technology of the detector will prolong the CT detection time. To address this issue, this paper proposes a fan beam CT super-resolution technique based on conditional denoising diffusion model[1]. This method uses high-resolution images obtained by detector micro-motion technology as the starting point of the diffusion model, and low resolution images obtained by detector micro motion technology as conditions. Through forward add-noising and backward denoising processes, higher resolution fan beam CT images can be obtained without the need for detector micro-motion technology. This method overcomes the problem of excessive smoothness in images caused by regression-based deep learning methods, and has good preservation of small defects in images. In human eye evaluation tests, the image effects generated by this method far exceed those based on CNN[2], Transformer[3], GAN[4], which has a great promoting effect on improving CT detection efficiency.

11:15 – THU2A1
X-ray Computed Tomography of Polymers exposed to High-pressure Hydrogen Gas
Johann KASTNER¹, Julia MAURER¹, Bernice MILLS², Nalini MENON², and Michael SCHÖBEL³ | 1.University of Applied Sciences Upper Austria, | 2.Sandia National Laboratories, Livermore, CA, USA ; 3.Leobersdorfer Maschinenfabrik GmbH, Austria
Hydrogen (H2) is a highly versatile energy carrier with growing significance in energy storage and transportation. As a clean alternative to fossil fuels, its use is rapidly expanding, particularly in the mobility sector, where compact storage solutions are essential. However, hydrogen’s low volumetric energy density necessitates compression to pressures exceeding 350 bar, creating unique engineering challenges. While polymers, often combined with metals, are widely used in dry-running high-pressure hydrogen compressors to avoid contamination of the process gas with oil lubricants, they are subject to thermo-mechanical stresses that can lead to fatigue and degradation. This study leverages X-ray computed tomography (CT) to characterize such degradation in polymers, providing insights into their performance and long-term viability in hydrogen applications. The first part of the study focuses on elastomeric O-rings, commonly used for sealing in hydrogen infrastructure. These components operate under demanding cyclic pressure conditions, ranging from 0.9 MPa to 87 MPa. Initial defects in the O-rings were identified using CT imaging before subjecting the samples to repeated pressure cycling. The progression of these defects was tracked over hundreds of cycles to assess both their growth and the emergence of new damage sites. By employing model compounds of known compositions, the study evaluates the role of fillers and additives in the formation of voids and other defects. These insights are critical for understanding how operational stresses influence damage evolution and for improving the design of materials resistant to hydrogen-induced wear and fatigue. The second focus of the research is on Polyphenylene sulfide (PPS) composites, which are increasingly used as pistons and packing rings in high-pressure hydrogen compressors. PPS is a high-performance polymer known for its thermal stability, chemical resistance, and high mechanical strength, making it suitable for extreme operating environments like reciprocating piston compressors. However, pure PPS suffers from high friction and wear, limiting its application in non-lubricated use. To address these limitations, the study investigates the addition of carbon fibers and polytetrafluoroethylene (PTFE) to the PPS matrix. The inclusion of carbon fibers enhances the material’s mechanical properties, while PTFE serves as a solid lubricant, forming tribo-films that reduce friction and improve wear resistance. The modified PPS formulations from this study will be exposed to select hydrogen environments to evaluate for friction properties which can be further co-related to polymer-filler distribution studied using X-ray CT towards understanding the influence of filler type and loading in the PPS matrix as necessary to applications such as pistons and packing rings.

11:35 – THU2A2
Holistic Quality Assessment of Additively Manufactured Heat Exchangers Using Robotic CT
Jitendra-Singh RATHORE¹, Marius COSTIN¹, Quynh trang PHAM², Pierre COSTE², Gilles-Charles GAILLARD² , Vincent BONNEFOY² | 1.Université Paris-Saclay, CEA, List, France ; 2.Université Grenoble Alpes, CEA LITEN, France
The increasing demand for higher heat transfer efficiency has driven the integration of compact heat exchangers with additive manufacturing (AM). The ability to fabricate intricate geometries using diverse materials allows for optimized trade-offs between heat transfer and pressure drop. This work demonstrate a holistic quality inspection approach utilizing a robotic Computed Tomography (CT) setup for the comprehensive evaluation of various types of AM heat exchangers. Three distinct geometric designs—OSF, WAVY, and TPMS—were analyzed using representative mock-ups. For global inspection, full-part scanning was performed to assess geometrical deviations and conduct statistical wall thickness analyses. As depicted in Figure 1(a), some regions exhibited unmelted powder, impacting their efficiency; additionally, various small porosities were observed in another design type, as shown in Figure 1(b). A holistic wall thickness analysis on the fins was conducted (Figure 1(c)), enabling statistical evaluations of uniformity.

11:55 – THU2A3
X-ray tomo-ptychography of single micrometic carbon and basalt fibres
Mathieu DUCOUSSO^1-2, Jean RINKEL¹, Willem BOUTU^3-4, Frédéric JENSON¹, Javier PÉREZ⁵, Pierre MARGERIT², Nicolas QUAGLIA¹, Eva HERIPR˲, Juan-Pablo MARQUEZ COSTA², Léonard COURAPIED¹ | 1.Safran Tech, France ; 2.Laboratoire PIMM, Arts et Metiers Institute of Technology, CNRS, Cnam, HESAM Université, France ; 3.Université Paris-Saclay, CEA, LIDYL, France ; 4.CY Cergy Paris Université, CEA, LIDYL, France ; 5Synchrotron Soleil, France
Carbon fibre-based composite materials are essential for emissions reduction in transportation. However, the physical properties of their main component, i.e. the micro-size carbon fibres, are still poorly documented because of experimental difficulties. Such material is however known for presenting high anisotropy between off- and on-axis properties. Here we have used Ptychography X-ray Computed Tomography to probe quantitatively the bulk electronic density and the morphology of micrometric carbon and basalt fibres. The carbon fibre exhibits a density variation as function of its radius, with a maximum at its center and a minimum around half of the radius, while the basalt fibre is homogeneous. Morphology is investigated at the fibre scale to quantify its deviations from a perfect cylinder and at the nanoscale to evaluate the texture of the surface. Resolutions are around 200 nm. This pioneering work should open new understandings and bulk or surface experimentations on single micron-size fibres at the nanoscale.

12:15 – THU2A4
Quatifying the Impact of XCT Data Compression on Defect Shapes
Thomas LANG1, Christoph HEINZL1-2 | 1.Fraunhofer Institute of Integrated Circuits IIS, Division Development Center X-ray Technology, Germany ; 2.University of Passau, Germany
X-ray computed tomography represents an essential tool in non-destructive analysis of industrial components in a wide variety of applications, ranging from fast inline inspection, to imaging of very large structures such as cars, and micro-scale characterization of novel material candidates. Due to the high demand regarding detailed analyses, both hardware and software components are improving continuously, yielding larger quantities of three-dimensional datasets. Data compression techniques aim at reducing storage requirements, but yet there is a risk of losing valuable information if a compression is too strong. In this work, we thus investigate a three-dimensional lossy compression technique tailor-made for XCT datasets and evaluate the deterioration of defect representations due to compression. For this purpose, a virtual XCT dataset is generated by modeling defects of randomized shapes and using X-ray simulation to generate realistic XCT data, whereas the pipeline will be applied on real XCT data as well in the full paper. While varying the maximum admissible compression error, an image processing pipeline will segment the defects and evaluate a selection of shape factors. When increasing the compression factor, the quantitative results indicate a continuous drop in shape factors, i.e., a moderate loss in shape is encountered at high compression rates.

11:15 – THU2B1
Active Contour for Applying CT Segmentation on a Single Image to Surrounding Multiple Images
Takashi KAWAMOTO¹, Yutaka OHTAKE² | 1.Tokyo Boeki Techno-System Ltd, Japan ; 2.University of Tokyo, Japan
There is a growing interest in using X-ray CT scanners to measure industrial products, as CT scan data holds significant potential for reverse engineering and model-based design. To fully utilize CT scan data from assembled parts, it is essential to segment each individual component. The commonly used thresholding method is ineffective when there is no luminance difference at the boundaries between parts. Consequently, segmentation becomes particularly challenging when adjacent parts are made of the same material (i.e., have the same density). When threshold-based segmentation fails, manual interventions such as outlining objects with mouse clicks is often required. However, manual segmentation on a single image may lead to misalignments in surrounding slices, even though contour shapes generally remain similar within a certain region. If artificial intelligence can correct these misalignments and propagate contours across multiple slices, it could enable segmentation in many cases that are currently unfeasible with existing technologies. In this study, we propose a deep learning-based active contour algorithm for applying the manual segmentation performed on a single image to its surrounding slices.

11:35 – THU2B2
Semi-Automatic Approach for Parts Segmentation of CT Scanned Assembled Car
Naoki MURAKAMI¹ and Yutaka OHTAKE¹ | 1.The University of Tokyo, Japan
In recent years, large-scale X-ray CT scanners have made it possible to perform CT scans on objects which are the size of automobiles. The resulting CT volume can be utilized for reverse engineering. However, handling them is difficult due to their large data sizes. Furthermore, the segmentation required for reverse engineering is currently performed manually by experts, which is time-consuming. In this study, we propose a semi-automatic method for parts segmentation of the CT scanned assembled car. The method involves automatically segmenting the CT scanned assembled car by thickness and material and then allowing the user to manually select from the automatically segmented parts. The segmentation based on thickness is carried out by the disappearing thin objects when the isosurface is offset. The segmentation based on material is performed by referencing CT values of the input CT volume. As a result, the parts composed of iron and aluminum, such as motors, were properly segmented.

11:55 – THU2B3
Distribution analysis in xCT data for identifying factors influencing uncertainty of repeated
measurements
Lionel KIELHÖFER¹, Dominik WOLFSCHLÄGER¹, Robert H. SCHMITT^1-2 | 1.IQS Laboratory for Machine Tools and Production Engineering WZL, RWTH Aachen University, Germany ; 2.Fraunhofer Institute for Production Technology IPT,  Germany
X-ray computed tomography (xCT) enables the measurement of inner and outer dimensional quantities by reconstructing the 3D model of an object. Due to limited understanding of the xCT acquisition and reconstruction processes modern guidelines typically estimate the measurement uncertainty using a standardized procedure based on repeated measurements and basic statistical analysis. In this work, we analyze voxel-wise distributions in the reconstructed volume and identifyclusters that reveal factors influencing measurement uncertainty in repeated measurements, with the eventual goal of reducing the number of CT scans required for accurate uncertainty estimation.

12:15 – THU2B4
A novel scatter correction method for dual-layer flat-panel detector based CBCT imaging
Yongshuai GE^1-2, Xin ZHANG¹, Jixiong XIE¹, Ting SU¹ | 1.Research Center for Advanced Detection Materials and Medical Imaging Devices, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China ; 2.Paul C Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China
In recent years, research on dual-energy cone-beam CT (DE-CBCT) imaging with dual-layer flat-panel detectors (DL-FPD) has steadily increased. However, the image quality of DE-CBCT is still affected by Compton scatter. This study introduced an energy-modulated scatter correction method for industrial CBCT imaging based on DL-FPD. By establishing a signal distribution model for the two layers and pre-calibrating relevant parameters, the X-ray scatter signals obtained from both detector layers can be analytically extracted. The method was validated through phantom experiments. Results showed that this approach significantly reduces scatter artifacts in both low-energy and high-energy CBCT images obtained from DL-FPD. In conclusion, the newly proposed DL-FPD based CBCT scatter correction method can effectively reduce the shading artifacts, yielding significantly improved CBCT images.

X-ray micro-tomography for non-destructive study of natural history specimens: a review from the MNHN
Patricia WILLS, MNHN Paris
Since 2011, the Muséum national d’Histoire naturelle has been equipped with X-ray micro-tomography systems, to study its natural history collections. The main purpose of this 3D digitization activity is to support scientific research on natural heritage collections. Access to the micro-tomography technique also contributes to providing information on the conservation of specimens, enabling them to be enriched, as well as providing data for teaching, dissemination and expertise. Looking back over more than a decade of activity, we can propose a review of how academics from different disciplines have appropriated the technique to develop new methods and new knowledge and address the ongoing challenges.

14:45 – THU3A1
X-Ray imaging for cultural heritage preservation in C2RMF
Elsa LAMBERT¹, Clotilde BOUST¹ | 1.C2RMF, France
In 1931, the Louvre museum set up a laboratory dedicated to the scientific study of paintings, in which radiography played a major role as it is non invasive. In 1995, this laboratory became the Centre de Recherche et de Restauration des Musées de France (C2RMF). The C2RMF is a national department of the French Ministry of Culture, dedicated to the study, preservation and restoration of art objects from the 1200 French national museums. Conservation workshops and laboratories are situated in the Musée du Louvre and in the Château de Versailles. C2RMF has numerous scientific instrumentation, from chemistry to optical analysis, and is hosting a unique particule accelerator dedicated to art works called AGLAE – Accélérateur Grand Louvre d’Analyse Elémentaire. Every year, several hundred works of art come to the C2RMF for analysis and conservation, around 300 of them are going through radiography. Each one is coming to the C2RMF with a question related to his preservation. Among all analysis, radiography see inside the artworks and this technique is invaluable when seeking information on the state of conservation and the techniques used to produce them. It is of interest not only for our knowledge of art and practices, but also for our knowledge of materials. Indeed, the C2RMF has 4 X-ray rooms (160 Kv, 420 Kv, 100 Kv, 200 Kv) and one mobile equipment for missions in museums (200 Kv). The diversity of the art pieces studied in terms of dimensions (from 1 cm to 7 meters), materials (wood, plaster, marble, metal or mixed media…), fragility (pieces are often several century old) and preciousness (unique pieces) are all constraints that need to be taken into account during radiographic examination. In particular, C2RMF is specialised in bronze statues up to 2 meters with the 420 Kv X-Rays room. Each work of art is absolutely unique, so C2RMF radiologists have developed the know-how to meet these technical challenges and obtain the most accurate images answering the conservation question. Several examples of X-rays imaging will be presented, in particular taken on large statues, both in the C2RMF and on in-situ missions.

15:05 – THU3A2
Parametric study of lightning induced damage in carbon composite using X-ray phase contrast imaging 
Adrien STOLIDI¹, Laureen GUITARD^1-2, Loïc TOULEMONDE¹, Olivier GHIBAUDO¹, Jerôme PRIMOT², Rafael SOUSA MARTINS³ , Amélie JARNAC³ | 1.Université Paris-Saclay, CEA List, France ; 2.DOTA, ONERA, Université Paris-Saclay, France ; 3.DPHY, ONERA, Université Paris-Saclay, France
This study presents a parametric analysis of lightning-induced damage in Carbon Fibre Reinforced Polymers (CFRP) using X-ray Phase Contrast Imaging (XPCI) using MultiLateral Shearing Interferometry technique. Damage severity was evaluated at different electrical energies deposited in the CFRP and cross-validated with other non-destructive techniques. The results demonstrate the potential of XPCI for precise characterization and improved damage modeling in aerospace applications.

15:25 – THU3A3
Comparison of porosity rate calculation implementing different software
Janka WILBIG¹, Anne-Françoise OBATON¹, Alexander WILSON-HEID², Tanguy LAURENCIN³, Oliver GUIRAUD³, Joseph BAPTISTA⁴, Laurent BERNARD⁴ | 1.Laboratoire National de Métrologie et d’Essais (LNE), France ; 2.Sandia National Laboratories, USA ; 3.Rubis Control, Switzerland ; 4.Reactiv’IP, France
Additive manufacturing (AM) processes come with the benefit to produce parts with highly complex geometries, and internal features easily. Porosity, however, is a typical flaw correlated to the layer-wise manufacturing approach, which can be critical for parts in operation. Thus, the porosity rate represents an important factor for the acceptability or rejection of parts within the frame of quality assurance. The porosity can be viewed on X-ray computed tomography (XCT) 3D volumetric images and then quantified. Hence, porosity analysis is a common analysis asked by customers to tomography service providers. First, the parts need to be scanned by XCT, then the 3D volume reconstructed, and finally a dedicated software is used to evaluate the porosity rate on the 3D volume. Several analysis software exist, that allow calculating the porosity rate. Six of them were identified, but the list is not exhaustive: IPSDK Explorer from Reactiv’IP [1], Zeiss inspect from Zeiss, Avizo from Thermo Fischer Scientific [2], VG Studio Max from Hexagon, Dragonfly from Comet Technologies Canada Inc. [3], and ImageJ as a software in the public domain, developed by the National Institutes of Health. It has been observed, that the results of porosity analysis processed by different operators using different software lead to differences. The customers should be aware of that.

15:45 – THU3A4
Development of a 20 kN Tension/Compression testing system for In-Situ microtomography of lattice structures under compression
Guillaume BRAVAIS^¹, Salaheddine MADI², Jean BOURGEAS^¹, Romain JARNIAS^¹, Sofiane TERZI^¹ | 1.Novitom, France ; 2.KU Leuven, Leuven (Arenberg), Belgium
Microtomography, particularly in-situ microtomography, provides a non-destructive means to visualize and monitor the microstructural evolution of lattice structures under mechanical loading, offering 4D insights into failure mechanisms, such as localized strut failure, node deformation, and strain redistribution. However, the size-dependent behavior of lattice structures complicates material testing, as small-scale specimens may not accurately represent the performance of larger components. This necessitates testing at larger scales to capture critical phenomena like strain redistribution and failure mechanisms that occur at full scale. To address this, a 20kN tension/compression device has been designed and developed for in situ microtomography, offering high capacity, autonomy, and precise data collection. In situ microtomography experiments conducted on lattice structures at laboratory sources and synchrotron facilities provided critical data for optimizing design and improving performance in high-stress applications.

14:45 – THU3B1
A Hybrid X-ray Computed Tomography System for Education and Synthetic Data Generation for AI Model Training
Martin SIMON¹, Robin TENSCHER-PHILIPP¹, Tim SCHANZ¹ | 1.University of Applied Sciences Karlsruhe, Faculty of Mechanical Engineering and Mechatronics, Germany
We present a hybrid X-ray Computed Tomography (CT) system designed for both education and AI data generation. It upgrades a legacy X-Ray system with digital detectors, precise mechanics, and modern controls. The system produces real CT scans and synthetic data through a specialized pipeline. Its dual purpose is to teach students about X-ray imaging and CT, and to create data for training AI models. Synthetic datasets simulate various scanning conditions, helping develop AI, especially when real data is difficult or expensive to obtain. The system generates surface and voxel-level data using modeling tools and AI techniques, enabling the creation of large volumes of realistic CT data. Additionally, AI methods automate the extraction of key features from CT images, aiding in tasks like anomaly detection and material classification. This innovative system combines real data with synthetic data generation, benefiting both educational goals and AI research. It provides hands-on experience for students while supporting the advancement of AI applications in industrial CT.

15:05 – THU3B2
DnCNN based Compton Backscatter Imaging Denoising Algorithm
Peiyuan MA^1-2, Yongshun XIAO^1-2, Changrong SHI^1-2 | 1.Department of Engeering Physics, Tsinghua University, Beijing, China ; 2.Key Laboratory of Particle & Radiation Imaging (Tsinghua University), Ministry of Education, Beijing, China
Compton Backscatter Imaging (CBST) is well-suited for single-sided non-destructive testing, particularly for low atomic number materials like carbon columns. However, traditional CBST imaging faces challenges such as attenuation correction and multiple scattering, leading to low signal-to-noise ratios (SNR) and reduced image quality. To address this, we propose a Denoising Convolutional Neural Network (DnCNN)-based approach for CBST image denoising. The model enhances image quality by applying deep learning-based post-processing to traditional point-by-point reconstructions. A specialized carbon column dataset was constructed for training and evaluation, with Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) used as performance metrics. Experimental results show that the DnCNN-based method significantly improves image quality compared to Total Variation (TV) minimization. Additional comparisons with FFDNet and U-Net reveal that while these models achieve higher PSNR and SSIM in most cases, they suffer from severe degradation in regions affected by multiple scattering. In contrast, the DnCNN-based approach demonstrates greater stability across different noise conditions, making it a promising solution for CBST image denoising.

15:25 – THU3B3
Filling quantification of capilary target using X-ray radiography and X-ray fluorescence
Alexandre ROYER¹, Lena RIGNY¹ , Romain CAPIAU¹ | 1.CEA, DAM, Valduc, France
As part of the simulation program, the Laser Mégajoule (LMJ) was created to allow the reproduction of the physical conditions of nuclear fusion by heating and compressing matter with laser. This laser is focused on a millimeter-sized target. Among the variety of targets available are the tube filled targets, consisting of a tube filled line composed of a spherical plastic capsule of approximately 2 mm diameter, glued to a glass tube. In the context of this study, the glass capillary has an external diameter of 10 μm. After manufacturing, the first step is to ensure that it is not clogged or cracked, so that the gas filling of the capsule can take place. To do this characterization X-ray radiography and X-ray fluorescence are used. Moreover as X-ray radiography and X-ray fluorescence are two non-destructive characterization which are influenced by the quantity of gas inside the capsule, they can be used to determine the pressure of gas inside the capsule and also the time of filling [1][2]. This is the main purpose of this study.

15:45 – THU3B4
Unveilling degradation mechanisms: Morphological Analysis of Li-ion battery electrodes using X-ray nano-computed tomography (nXRCT)
S. Mitra¹, F. Monaco¹, G. Oney¹, D. Karpov², Q. Jacquet¹, S. Tardif⁴, Q. Arnoux³, S. Lyonnard¹ | 1.Univ. Grenoble, Alpes, CEA, CNRS, IRIG, SyMMES, France ; 2.European Synchrotron Radiation Facility, France ; 3.TotalEnergies OneTech, France ; 4.Univ. Grenoble Alpes, CEA, IRIG, MEM, France
Lithium-ion batteries (LIBs) play a pivotal role in various technological applications, including portable electronics, electric vehicles, and grid storage systems. However, their performance and lifespan are critically influenced by morphological degradation mechanisms occurring within the cathode and anode materials during cycling. Understanding these degradation processes is essential for enhancing battery performance and durability. Battery ageing is strongly impacted by morphological evolution at different scale: microscale (electrode swelling and cell defects), nanoscale (electrode porosity changes, particle cracking), atomic scale (crystal defects). Nano X-ray computed tomography (nXRCT) offer the advantage of high resolution ranging from 25 – 150 nm and good statistics with large field of view of several tenth of microns, hence complementary to FIB/SEM or micro X-ray CT [1]. This study investigates SEI formation and degradation mechanism due to aging of commercial Li-ion batteries with porous graphite electrodes (anode) and Lithium Iron Phosphate (LFP)/ Nickel Cobalt Aluminum Oxide (NCA) electrodes (cathode) using ex-situ post-mortem synchrotron nXRCT at the ID16a beamline of the ESRF. In this work, we demonstrated that a minimum volume of 20 × 20 × 20 μm3 must be used to be representative of the total electrode volume. On the cathode side, most NCA particle are cracked after ageing even though the cutoff voltage was 3.8 V – hence with little (de)lithiation of NCA particles. Aged graphite particles appear swollen, leading to a reduction in electrode porosity for aged electrode (~25%) as compared to pristine electrode (~35%) with this reduction being more pronounced towards the separator side. The study emphasizes the significance of high-resolution nano-CT in revealing volume-specific micro/macro pore size distribution within graphite after long-term ageing.

16:35 – THU4A1
Experimental and simulation results for inherent unsharpness estimation of cassette contribution 
Anthony TOURON¹, Romain JONCHIERE², Andreas SCHUMM³, Uwe ZSCHERPEL⁴ | 1.Université Paris-Saclay, CEA List, France ; 2.EDF, Direction of Industrial Quality, France ; 3.EDF Lab les Renardières, France ; 4.BAM, Berlin, Germany
In the absence of motion unsharpness, which can usually be neglected in industrial radiography, the total unsharpness is defined as the square root of the sum of the squared geometrical and inherent unsharpness. While the concept of geometrical unsharpness is well understood, and to a certain extent can be controlled by the radiographer with a careful consideration of the distance between the source, the object, and the film, inherent unsharpness is directly related to the radiation energy. Inherent unsharpness is produced by scattered secondary radiation within the cassette containing the photographic emulsion, which in turn renders silver halide grains within a certain vicinity of the primary radiation’s absorption developable. This vicinity increases with the energy of the generated electrons, which in turn depends on the energy of the incident radiation. With digital radiography, it is relatively straightforward to determine inherent unsharpness. With film radiography, this is not the case. However, a detailed comparison of the values of this unsharpness between these two techniques is important for transitioning from one to the other, with implications for industrial implementation or regulatory requirements. This study presents the results obtained through simulation and experiments on measuring this unsharpness in film radiography using different methods, including the Klasens method, which involves evaluation of the profile of a particular edge image.

16:55 – THU4A2
High resolution CBCT imaging with a dual-layer detector
Ting SU¹, Xin ZHANG¹, Jiongtao ZHU², Yongshuai GE¹ | 1.Research Center for Advanced Detection Materials and Medical Imaging Devices, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China ; 2.College of Physics and Optoelectronic Engineering, Shenzhen University, China
This study presents an innovative super-resolution dual-energy cone-beam CT (CBCT) imaging technique utilizing a dual-layer flat-panel detector (DL-FPD). The proposed method, termed suRi, employs a half-pixel shifted binning approach across the two detector layers, effectively doubling the spatial sampling rate during signal acquisition. This method enables high spatial resolution CBCT imaging while maintaining high signal readout speeds through large detector binning rates. Furthermore, we implement a penalized likelihood material decomposition algorithm to directly reconstruct high-resolution material bases from the acquired dual-energy CBCT projections. Experimental validation using the Catphan phantom demonstrates that suRi achieves a 23% improvement in image spatial resolution compared to conventional DL-FPD-based CBCT imaging at equivalent signal readout speeds. This method effectively resolves the inherent trade-off between spatial resolution and signal readout speed in traditional CBCT systems. The suRi technique demonstrates notable improvements in DL-FPD-based dual-energy CBCT imaging, showing potential for enhancing imaging performance in clinical applications.

17:15 – THU4A3
Hardware and software developments on ESRF-EBS BM05
Fabien LEONARD¹, Florian JÜRRIES¹, Elodie BOLLER¹ | 1.The European Synchrotron Radiation Facility (ESRF), France
X-ray computed tomography and radiography imaging are currently the second most exploited technique at the European Synchrotron Radiation Facility (ESRF) for commercial services. To maximise the growth, the beamlines must be flexible enough to accommodate a wide range of materials and sample sizes, whilst conserving operational efficiency and ensuring reliable and high-quality imaging outcomes. In this paper, we present some of the latest developments on the BM05 tomography beamline from the ESRF in terms of automatisation, processes, and procedures that ensure an efficient and robust operation of the beamline. For all users, the automatisation developments are focused on being able to set-up the beamline in a pre-defined state as easily and quickly as possible, so that time consuming user interventions (and potential mistakes) are minimised. For complex structured samples, a Graphical User Interface (GUI) called Daiquiri tomo has been developed as a simple tool for non-experts to be able to perform most of their scanning operations in a user-friendly and efficient fashion, with minimum support from beamline scientists. For large series of samples, a robotic sample changer has been assembled to be able to scan series of up to 45 samples, and maximise usage of night shifts. Overall, those improvements are key in reaching operational excellence and X-ray computed tomography and radiography imaging are projected to rival macromolecular crystallography (MX) as the first technique for commercial services at ESRF by the end of 2025.

17:35 – THU4A4
New generation of high-resolution 225 kV transmission X-ray tubes with LaB6 cathodes
Vladimir BURLAKA¹, Alexander DASCHKEWITSCH¹, Oliver ABE¹, Jens Peter STEFFEN¹, Thorsten FRÖBA¹ | 1.X-RAY WorX GmbH, Germany
High-resolution microfocus X-ray tubes with a transmission target offer the possibility of investigating the smallest structures in the submicrometer range. The desired ultimate spatial resolution requires precise geometry of the electron beam and the use of two-stage electromagnetic focusing optics, including condenser and objective lenses (1). The main challenges are long filament lifetime, long-term thermal stability, cathode-dependent beam geometry and target performance. In our new generation of transmission X-Ray tubes, we use LaB6 cathodes to achieve the highest brightness (1, 2), exceptional beam precision, an improved target-to-emission current ratio and optimal target performance without overheating the cathode. LaB6 cathodes are known for their long lifetime compared to standard tungsten cathodes (1, 2, 3). Thanks to the precision of the primary electron beam, only one objective lens is required to reduce the beam spot, which positively impacts the efficiency and performance of the X-ray tube. The new technical solution provides an ultimate JIMA resolution of 0.5 micrometers (μm) at 225 kV (Fig.1), an extended filament lifetime of approximately 1000+ hours and excellent thermal stability during long-term measurements (72 hours). Furthermore, it generates a focal spot with a perfectly round shape (Fig. 2). All these characteristics are unattainable with a conventional single-stage microfocus X-ray tube that uses a tungsten cathode (typical resolution: 2 μm–3 μm at 225 kV, 3 W; typical filament lifetime: 200–300 hours). To achieve this outstanding performance, we developed a specialized cathode arrangement optimized for specific LaB6 cathodes, created a new vacuum system, applied actively cooled electromagnetic optics, and utilized a special transmission target with a diamond substrate and an active tungsten layer.

16:35 – THU4B1
Resolution improvement by Speckle-based learning 
Yukie NAGAI¹ , Yutaka OHTAKE¹ | 1.School of Engineering, The University of Tokyo, Japan
Focal spot size is a factor that determines the resolution of industrial X-ray CT and radiography. High voltage and current values for high penetration measurement will increase the focal spot size and make it difficult to measure the details of fine features. Some objects have fine details that are easily lost in the transmission images due to the large focal spot size that is unavoidable for appropriate parameter settings for sufficient penetration. Talbot-Lau interferometers and specklebased imaging, which requires a detector with very small sensors, allow such detailed observations, but are not generally applicable to common laboratory CT scanners. Inspired by speckle-based imaging, we propose a detailed transmission image generation by machine learning using speckle patterns as visual cuea. Our system is trained by pairs of a transmission image with speckle patterns scanned with a large focal spot size and a transmission image with a small focal spot size, both of which are easily obtained with lab-based CT scanners, and it generates transmission images with fine features.

16:55 – THU4B2
A Fast Metric Based on Marching Cubes and Hausdorff Distance for the Quality Assessment of 3D Volume Data for Computed Tomography Measurement Technology 
Ahmed BARAKA¹, Faizan AHMAD², Kilian GEIGER¹, Dominik WOLFSCHLÄGER¹, Sven SIMON² , Robert H. SCHMITT^1-3 |1.WZL, RWTH Aachen University, Germany ; 2.Department of Computational Imaging Systems, ITI, University of Stuttgart, Germany ; 3.Fraunhofer Institute for Production Technology IPT, Aachen, Germany
Inline X-ray Computed Tomography (xCT) offers a promising approach for the quality assurance and control of various workpieces during production. This is due to its capability to non-destructively inspect both internal and external structures. However, the typical duration of measurements, which extends to several hours, remains a limiting factor in practical applications. The sparse view xCT method, which limits the number of necessary projection images, can reduce this duration to a few minutes. Nonetheless, this reduction in time increases the overall measurement uncertainty, primarily due to streak artifacts. Therefore, the ability to efficiently quantify a quality measure sensitive to the number of projections captured during xCT measurement is crucial for determining the reliability of conducted measurements. This paper introduces a quality metric sensitive to streak artifacts, which uses a rapid surface determination algorithm known as marching cubes, followed by the computation of the Hausdorff distance between a reference surface and the extracted surface. An experimental study correlating the Hausdorff distance to the measurement uncertainty as a function of the utilized number of projections is conducted. The findings show that the Hausdorff distance increases for surfaces extracted from measurements taken with fewer projections. Additionally, by analyzing the correlation between the Hausdorff distance and the measurement uncertainty, quantified through the standard deviation of multiple extracted geometrical features from the measured test workpiece, a strong positive correlation is observed.

17:15 – THU4B3
Study of parameter influency and measurement uncertainties with X-ray computed tomography on a weld
Alexandre CHOUX¹, Alexis GUINARD¹, Cyril HERMEREL¹, Bastien VAN VLIERBERGHE¹ | 1.Université Paris-Saclay, CEA List, France
Computed Tomography could be a powerful instrument of control and acquisition of information for manufactured products. This technology can be useful not only during the production control, but also in the validation of geometric parameters, which are difficult to obtain with classical measurement. It can be useful in the volume and geometry measurement of the internal defects or specific dimensional measurement. The obtained tomographic volume can be analysed with a wide range of post processing visualization tools. Thus, the characterization of elementary parts of complex geometry is carried out by X-ray tomography. This control takes place in three stages (acquisition, reconstruction and analysis), each involving different parameters. To test the influence of these parameters, an experimental design is set up. Repeatability tests, foreseen in the experimental design, have been implemented. Thanks to the obtained results and to additional manipulations highlighting each parameter determined in the experimental design, the influence of the latter on the measurements is carried out. The repeatability tests are also used to determine the limit of detection of defects within the parts. Finally, the means and standard deviations of these tests are used to determine the measurement uncertainties associated with this control. This works details the method and all the results obtained are given here.

8:30 a.m.: Shuttles departure for Saclay
9:30 a.m.: Arrival on site (check-in at the reception pavilion) and coffee reception
10:00 a.m. – 10:30 a.m.: Presentation in the room of the Synchrotron SOLEIL and its areas of application for research and industry

10:30 a.m. – 12:15 p.m.: Tour of the synchrotron facilities and beamlines (maximum 48 people):

• 2 groups of 12 people each will visit the ANATOMIX (https://www.synchrotron-soleil.fr/fr/lignes-de-lumiere/anatomix) and ROCK (https://www.synchrotron-soleil.fr/fr/lignes-de-lumiere/rock) beamlines
• 2 groups of 12 people each will visit the PSICHE (https://www.synchrotron-soleil.fr/fr/lignes-de-lumiere/psiche) and CRISTAL beamlines (https://www.synchrotron-soleil.fr/fr/lignes-de-lumiere/cristal)

12:30 p.m.: Shuttle departure for Paris
2:00 p.m. approx. arrival

8:30 a.m.: Shuttle departure
9:30 a.m.: Coffee reception
9:45 a.m. – 10:00 a.m.: Presentation of the CEA List and the site
10:00 a.m. – 12:30 p.m.: Tours in groups of the following facilities:

1. Robotic CT
2. X-ray inspection with automatic decision
3. Laboratory phase contrast imaging setup

12:30 p.m. Shuttle transfer to Paris
2:00 p.m. approx. arrival

National Museum of Natural History – Paris 5ème
10h00 – 12h00 : 4 groups of 12 participants. 
The AST-RX platform is the micro-tomography facility at the MNHN, dedicated to the 3D non-destructive imaging of specimens of natural history from the institution or collaborators. More on

Research and Restoration Center of the Museums of France – Paris 1er
10h15 – 11h45: 3 groups of 15 persons will visit 3 platforms
– Tomography platform
– AGLAE platform (Grand Louvre Elementary Analysis Accelerator)
– VISHNU studies