For several years, CEA has been developing highenergy imaging and photon interrogation techniques in the CINPHONIE irradiation bunker (CHICADE facility, CEA IRESNE, Cadarache, France). A new linear electron accelerator (K15 LINAC from Varex) has recently been commissioned and its irradiation field requires precise characterization. The manufacturer data for the mean electron energy are 15 MeV, but we asked for a slightly higher dose rate, resulting in a higher end-energy that needs to be characterized. In addition, the knowledge of the beam current is essential to ensure the reliability of both experimental results and numerical simulations. For that purpose, dose rate measurements in a water tank and delayed gamma-ray spectroscopy of thin metal foils activated in the X-ray field were carried out. A numerical model of the LINAC internals was built with MCNP6.3 to calculate dose rates in water and reaction rates in irradiation samples. A thorough characterization campaign was carried out to validate and calibrate this MCNP model against experiments. Finally, a Bayesian framework was applied to fine-tune the electron source energy distribution, estimate the average LINAC current, and quantify the X-ray angular deviation. Finally, the best estimates are E = (17.8 +/- 0.2) MeV and I-beam = (107 +/- 6) & micro;A.

Industrial X-ray imaging exams based on linear accelerators (LINAC), while being a powerful tool for the physical inspection of large and/or thick objects, are difficult to implement due to the important radiological constraints associated to the photon spectrum and the irradiation dose rate. Indeed, as most of the interrogating photons have an energy higher than 0.5 MeV, thick concrete shielding (>1 m) is usually required, and the investment cost of such bunkers can be prohibitive: an industrial 9 MV LINAC would typically need 4 m thick concrete in the beam axis and 1.5 m everywhere else. Moreover, in case of large objects, some of the pieces to be inspected cannot, or only with difficulty, be transported. This is particularly true in the nuclear waste management field [1]. To overcome these constraints, CEA led, since 2017, the development of a high-energy low dosimetry impact transportable tomograph: TOMIS. Developed as part of an investment supported by the French Government, TOMIS aims at providing a Non Destructive physical characterization of large-volume packages (diameter < 140 cm, height < 130 cm, mass < 5 t) with millimeter spatial resolution in less than an hour directly on their production/storage site. The innovation of TOMIS is to integrate in a standard truck container all the elements of the tomograph: - the X-ray source is an industrial LINAC (Varex M9) offering a dose rate close to 30 Gy/min at 1 meter from the target in the beam-axis; - the linear and collimated detector which ensures a high efficiency of detection over 150-cm width; - the lifting unit to handle large objects. The truck container also brings most of the mandatory shielding for photons (lead) and neutrons (polyethylene) in order to provide the user with a mobile and autonomous tomograph (see Figure 1). After 6 years of development, TOMIS was commissioned in December 2023 (see Figure 2). In this paper, we report on the radiography and tomography performances of TOMIS and compare them to those achieved in a fixed irradiation cell with several materials (stainless steel, concrete, etc) and objects of various size [2], [3]. Performances are detailed in terms of Modulation Transfer Function, contrast-over-noise ratio, limit of detection. TOMIS is scheduled to be used over the decades for industrial measurement campaigns on alpha bearing-nuclear waste packages.

As part of its R&D programs on large objects characterization, the Nuclear Measurement Laboratory at the CEA-Cadarache center has equipped its high-energy tomograph with a new linear accelerator (linac): a VAREX K15 (9 to 15 MV range). This linac delivers a very high dose rate: up to 130 Gy/min at 1 m from the target. Combined with a mechanical bench and optimized detectors, this X-ray source allows handling very large objects for radiographies and tomographies, up to 1600 mm in diameter and 5 t in mass [1]. Compared with the kilovoltage range, MV energies offer two advantages: higher photon flux and deeper penetration capabilities (steel from 100 to 400 mm). The new X-ray source has been fully characterized in terms of dose rate, focal spot size and photon spectrum using water attenuation measurements. Depending on the geometry of the object to be scanned, two detectors can be used. The first is dedicated to larger objects and is a lens-based detector with different scintillator screens (Gadox or CsI, as described in [2]) specially designed for this configuration (with a screen size up to 800x600 mm2). The second, a commercial flat-panel with a small pixel pitch (0.1 mm) is used for the smallest but densest objects, where the spatial resolution is critical [3]. The performance of these detectors is characterized, compared and discussed. The tomograph set-up and its final performance at low (9 MV) and high (15 MV) energies are detailed in terms of MTF curves, and contrast-over-noise ratio obtained on specific mock-ups. Examples of tomography on real objects (industrial packages produced by metal additive manufacturing or radioactive waste drums) are also presented. Finally, the main drawbacks in this energy range are listed and detailed: 1) the scattering background caused by the Compton effect, 2) the thickness of the scintillator, which must be optimized according to the spatial resolution or expected efficiency and 3) the size of the X-ray source (limited to 1.5 mm). To overcome these limitations, various studies are currently underway, and the expected solutions are presented and discussed.

Usual computed tomography (CT) systems provide information on the layout and nature of materials composing an object. However, this information is limited to the apparent linear attenuation μ of the materials. To reach a more precise and accurate description, in the form of the effective atomic number Zeff and the electronic density ρe, dual-energy imaging can be used. Conventional dual-energy computed tomograohy (DECT) techniques are: (a) pre-processing dual-energy data sets and performing conventional CT reconstruction [1], (b) reconstruct dual-energy data sets and analyse the ratio of obtained linear attenuation coefficients [2, 3] and (c) reconstruct data sets after a decomposition on a material basis [4-6]. While the second technique is relatively convenient to set-up, it is not completely energy-independent. The third technique has proven rather efficient; however, it raises the question of the choice of material base used for decomposition. When inspecting complex objects composed of a large number of differents materials, this choice can be crucial. Therfore, this work focuses on extending the first technique to high energies, as it does not require any assumptions on the materials to be detected and takes into account beam-hardening effects through the system spectral response.

Electron linear accelerators (LINACs) are versatile and powerful X-ray sources, that can be used in medical radiotherapy as well as in various industrial applications including non-destructive testing, imaging and security inspection. LINACs accelerate electrons by passing them through a series of oscillating electric fields within a vacuum tube. These high-energy electrons are then directed towards a metallic target, producing X-rays (bremsstrahlung radiation) when they decelerate upon impact. In the field of non-destructive radioactive waste characterization, high-energy photon imaging (radiography, tomography) is used on large cemented radiological waste containers, with a volume of the order of 1 m3, to check their integrity and assess their content [1][2]. However, for such packages, passive gamma-ray spectroscopy, passive neutron counting and even active neutron interrogation fail in measuring nuclear materials, like plutonium and uranium. Therefore, high-energy photon interrogation techniques is under study to detect and quantify nuclear materials through the detection of induced-photofission particles. For the past years, CEA has been developing high-energy imaging [3] and photon interrogation techniques in CINPHONIE irradiation bunker (CHICADE facility, CEA IRESNE, Cadarache, France). CINPHONIE was recently upgraded with a K15 Varex accelerator that can reach a maximum dose rate of 130 Gy/min at 1 m from the X-ray target [4]. For advanced techniques (high-energy photon and photoneutron activations, photofission, bi-energy imaging), it is paramount to simulate precisely the irradiation field. For that purpose, a numerical model of the LINAC internals was built (with MCNP 6.3). It aims at simulating photon and neutron fields in view to calculate dose rates and reaction rates in irradiation samples, waste packages, but also in the whole casemate. A thorough characterization campaign was carried out to validate and calibrate this MCNP model against various experiments, including dose rate measurements in a water tank and delayed gamma-ray spectroscopy of thin metal foils activated in the X-ray field. These experimental results were used to fine-tune the electron source energy distribution as well as to estimate the average beam current. Its high-energy part is indeed particularly crucial for photofission and bi-energy studies.

L’identification des matériaux présents dans des objets volumineux comme des colis de déchets nucléaires nécessite la plupart du temps soit une opération destructive (carottage, découpe) pour les colis bétonnés, soit une intervention humaine (tri, inventaire) sur des colis de faible activité dont le déchet n’est pas bloqué. Dans le cadre de ses activités de R&D sur la gestion des colis de déchets, l’ONDRAF/NIRAS (Belgique) a sollicité le CEA (Institut IRESNE) pour évaluer les performances et les limites de détection et d’identification de matériaux dans des colis bétonnés étalons de 400 l/ 900 kg. La technique mise en oeuvre est la tomographie par transmission au moyen d’un accélérateur linéaire dont l’énergie est réglable de 9 à 21 MeV (avec un maximum de dose à 15 MeV). Des essais ont ainsi été réalisés sur cinq fantômes spécialement conçus par l’ONDRAF/NIRAS pour refléter la gamme de matériaux devant être identifiés (plexiglas, cellulose, bois, métal, eau, huile, filtres, déchets compactés, etc). Les premières séries de mesures ont consisté en des tomographies demi-champ Cone-Beam (3D) afin d’avoir une vue globale des colis et de leur contenu. La diffusion Compton étant majoritaire à ces énergies, il n’a pas été possible, à cause du rayonnement diffusé parasite, de réaliser une calibration en densité. Une seconde série de mesures a alors visé à comprendre les perturbations observées afin de les corriger et obtenir la meilleure précision possible sur la densité à l’aide d’une géométrie collimatée Fan-Beam (2D). Les précisions obtenues sur la densité sont de l’ordre de quelques pourcents pour la majorité des matériaux, la sensibilité variant suivant leur nature. Par exemple, l’eau et l’huile contenues dans un même récipient sont distinguées autant par la valeur de leur densité que par leur stratification alors même que la nature des métaux est plus difficile à extraire à cause du durcissement de spectre. La tomographie 3D et la tomodensitométrie 2D apparaissent donc comme des examens efficaces pour identifier des matériaux contenus dans des objets volumineux et lourds comme des colis de déchets nucléaires.

La fabrication additive, métallique en particulier, est en plein essor, mais les pièces ainsi produites peuvent présenter des défauts tels que des anomalies d'impression, de la rétention de poudre ou des fissures. Pour contrôler l'intégrité de ces pièces, la tomographie par transmission haute résolution reste la méthode de référence. Cependant, pour inspecter des pièces massives et fortement absorbantes, la tomographie haute énergie avec un accélérateur linéaire d'électrons est nécessaire. Le Laboratoire de Mesures Nucléaires du CEA IRESNE dispose d'un tomographe haute énergie et a souhaité améliorer la qualité des projections acquises en mettant en place des post-traitements numériques. Afin d’essayer de dépasser les performances des méthodes de restauration classiques, basées sur des algorithmes de déconvolution, une approche de post-traitement novatrice a été étudiée : la déconvolution de flou par réseaux de neurones convolutifs. Pour ce faire, un jeu de données d’images a tout d’abord été généré par simulation. Un réseau de neurones convolutifs, basé sur la structure du réseau SRCNN (Super-Resolution Convolutional Neural Network), a ensuite été adapté, entraîné et évalué. Chaque hyperparamètre du réseau a alors été spécialement optimisé. Enfin, ce réseau a été validé sur des tomographies à 9 MeV d’objets réels afin d’évaluer les performances finales obtenues, mais aussi comprendre les limitations de ce type d’approche. Le réseau de neurones convolutifs ainsi optimisé démontre de bonnes performances de défloutage tout en limitant l’amplification du bruit.

Le Laboratoire de Mesures Nucléaires de l’institut IRESNE du CEA (centre de Cadarache) réalise dans la cellule d’irradiation CINPHONIE (INB CHICADE) des expertises radiographiques et tomographiques sur des objets denses et/ou de grandes dimensions. Pour ce faire, la cellule est équipée d’un accélérateur linéaire de type SATURNE permettant la production de photons ayant une énergie allant de 9 à 21 MeV (débit de dose maximal à 1 m d’environ 62 Gy/mn). Un banc mécanique de haute précision et d’une capacité de 5 tonnes permet les mouvements en translation (horizontale et verticale) et en rotation d’objets ayant un diamètre maximal de 1,50 m. La détection est assurée au moyen d’un détecteur dit « Grand Champ », composé d’un écran scintillateur de Gadox monté sur un renforçateur en tantale (80 x 60 cm²) et filmé, via un renvoi d’angle, par une caméra sCMOS bas bruit. Compte-tenu des contraintes géométriques de la cellule, le champ de vue du tomographe permettait de mesurer au maximum des objets de diamètre 102 cm en demi-champ. En 2018, une nouvelle méthode de traitement des projections a permis d’étendre le champ de vue à des objets de toute taille. Cette extension du champ de vue est effectuée en enregistrant, pour une coupe tomographique donnée, plusieurs séries de rotation sur 2π obtenues en décalant le centre de rotation de l’objet. Les projections résultantes sont ensuite fusionnées puis utilisées comme données d’entrée des algorithmes de reconstruction. Celle-ci est effectuée au moyen du logiciel RTK (algorithme FDK). Après une phase de validation par simulation au moyen du logiciel MODHERATO, développé au sein du laboratoire, la méthode a été appliquée à divers objets de gros volumes et de densité variées tels que des coques de déchets nucléaires bétonnés (Ø110 cm × 130 cm, d = 2,5) ou encore le crâne du Mammouth de Durfort (110 cm × 110 cm, d = 1,5) appartenant au Museum National d’Histoire Naturelle. Les dimensions maximales des objets sont ainsi définies par les capacités physiques (et non algorithmiques) du système tomographique à disposer d’un rapport signal sur bruit suffisant. Ce rapport dépend uniquement des caractéristiques du faisceau ainsi que du diamètre et de la densité des objets tomographiés.

The Nuclear Measurement Laboratory (LMN) at CEA Cadarache in France is developing a high-energy tomograph reaching energies up to 21 MeV with high dose rates. It allows performing tomographies on massive objects (5 tons, 140 cm diameter) with a multimillimetric spatial resolution. For the control of absence of cracks, bubbles or defects in the concrete coating of some CEA waste drums, the laboratory needs a “high-resolution” version of this tomograph. The purpose of this study is to design an imaging detector able to provide a high spatial resolution on large objects with high dose resistance and scanning times of a few hours. To achieve this design, several steps are necessary and will be presented: – A first step consists in characterizing the current detection elements of the tomograph, such as the camera and its lens with experimental measurements. This enables to compare performance of these elements with that available on the market and to consider a replacement. – Then, the scintillator is selected according to experimental measurements on few scintillator types and a state of the art. – Next, the detector configuration is optimized to achieve the higher spatial resolution. Throughout these steps, we present the design of a highresolution detector with a spatial resolution around 300 µm for a 10%contrast.

The Nuclear Measurement Laboratory (LMN) at CEA Cadarache in France uses a high-energy electron linear accelerator, LINAC (9-21 MeV), to characterize nuclear waste drums. It enables to explore new examination modalities, such as active photon interrogation or dual-energy CT to scan large concrete objects with diameters up to 140 cm. These techniques require precise awareness of the photon spectrum emitted by the LINAC. However, direct measure of this photon energy spectrum cannot be achieved because of the accelerator pulses causing detector saturation. During the last few years, a large number of indirect methods has been developed. From an experimental point of view, the simplest indirect method for spectrum estimation method is ransmission measurements. Because it can be set up easily and accurately using an ionization chamber as well as an appropriate screen. The obtained transmission curve depends on the photon energy spectrum, which can be estimated using inverse models. In this paper, we present the development of a numerical model to determine the energy spectrum from an attenuation curve via transmission measurements which combines two types of inverse models: a continue model and a discrete model. We validate this tool using a test spectrum and its transmission curve obtained via Monte-Carlo simulation. This qualification allowed us to determine its sensitivity (signal-to-noise ratio, SNR) in order to have a good convergence. We show that if the SNR is less than 4%, we have a good estimation of the photon energy spectrum. Then, it was experimentally tested with a transmission curve obtained at the laboratory.

The characterization of radioactive waste packages is mandatory for their transport, interim storage and final disposal. In this framework, the Nuclear Measurement Laboratory of CEA DES IRESNE Institute, at Cadarache, France, uses a high-energy electron linear accelerator (LINAC) to produce an interrogating bremsstrahlung beam with endpoint energies ranging from 9 to 21 MeV to perform X-ray imaging and high-energy photon interrogation on large concrete packages. In particular, highenergy photon beam induces photofission reactions in both fissile ( 235 U, 239 Pu, 241 Pu) and fertile ( 238 U, 240 Pu, 232 Th, etc.) actinides possibly present in the radioactive waste. In order to assess their mass, we use delayed gamma rays emitted by their photofission products, which are measured with a 50 % relative efficiency High-Purity Germanium (HPGe) detector. Actinide differentiation, which is important for the fissile mass estimation, is based on the ratios of gamma rays emitted by different photofission products and requires appropriate corrections for the gamma attenuation in concrete. To this aim, we report here a localization method of point-like nuclear materials in the concrete matrix, based on the differential attenuation of several gamma rays emitted by a same photofission product. We use here the 1435.9 and 2639.6 keV lines of 138 Cs, with both experimental data and MCNP numerical simulations to determine the (r,θ) coordinates of nuclear materials. Then, the depth inside the concrete matrix, which is determined with a precision of a few percent, mainly depending on counting statistics on 1435.9 and 2639.6 keV net peak areas, is used to correct for the different gamma ratios used in the actinide identification method. Experimental tests with uranium samples have been performed to validate the localization method.

One of the main drawbacks of Cone Beam Computed Tomography (CBCT) is the contribution of the scattered photons due to the object and the detector. Scattered photons are deflected from their original path after their interaction with the object. This additional contribution of the scattered photons results in increased measured intensities, since the scattered intensity simply adds to the transmitted intensity. This effect is seen as an overestimation in the measured intensity thus corresponding to an underestimation of absorption. This results in artifacts like cupping, shading, streaks etc. on the reconstructed images. Moreover, the scattered radiation provides a bias for the quantitative tomography reconstruction (for example atomic number and volumic mass measurement with dual-energy technique). The effect can be significant and difficult in the range of MeV energy using large objects due to higher Scatter to Primary Ratio (SPR). Additionally, the incident high energy photons which are scattered by the Compton effect are more forward directed and hence more likely to reach the detector. Moreover, for MeV energy range, the contribution of the photons produced by pair production and Bremsstrahlung process also becomes important. We propose an evaluation of a scattering correction technique based on the method named Scatter Kernel Superposition (SKS). The algorithm uses a continuously thickness-adapted kernels method. The analytical parameterizations of the scatter kernels are derived in terms of material thickness, to form continuously thickness-adapted kernel maps in order to correct the projections. This approach has proved to be efficient in producing better sampling of the kernels with respect to the object thickness. This technique offers applicability over a wide range of imaging conditions and gives users an additional advantage. Moreover, since no extra hardware is required by this approach, it forms a major advantage especially in those cases where experimental complexities must be avoided. This approach has been previously tested successfully in the energy range of 100 keV – 6 MeV. In this paper, the kernels are simulated using MCNP in order to take into account both photons and electronic processes in scattering radiation contribution. We present scatter correction results on a large object scanned with a 9 MeV linear accelerator.

As part of its R&D programs on non-destructive testing of nuclear waste drums, CEA is commissioning an irradiation cell named CINPHONIE, at Cadarache. This cell allows high-energy imaging (radiography and tomography) on large volumes (up to 5 m 3 ) and heavy weights (up to 5 tons). A demonstrator has been finalized, based on existing components. The X-ray source is a 9 MeV LINAC which produces Bremsstrahlung X-rays (up to 23 Gy/min at 1 meter in the beam axis). The mechanical bench is digitally controlled on three axes (translation, rotation, elevation) and can handle objects up to 2 t. This bench performs trajectories necessary for acquisition of projections (sinograms) according to different geometries: Translation-Rotation, Fan-Beam and Cone-Beam. Two detection systems both developed by CEA-Leti are available. The first one is a large GADOX scintillating screen (800×600 mm 2 ) coupled to a low-noise pixelated camera. The second one is a multi-CdTe semiconductor detector, offering measurements up to 5 decades of attenuation (equivalent to 25 cm of lead or 180 cm of standard concrete). At the end of the acquisition, a Filtered Back Projection-based algorithm is performed. Then, a density slice (fan-beam tomography) or a density volume (cone-beam tomography or helical tomography) is produced and used to examine the waste. Characterization of LINAC, associated detectors as well as the full acquisition chain, are presented. Experimental performances on phantoms and real drum are discussed and expected limits on defect detectability are evaluated by simulation. The final system, designed to handle objects up to 5 tons is then presented.