Sebastien Robert
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sebastien Robert has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Claire Prada
Institut Langevin, ESPCI Paris, Université PSL, CNRS, 75005 Paris, France
3 shared publications Claire Prada
Institut Langevin, ESPCI Paris, CNRS, PSL Research University, 1 rue Jussieu, 75238 Paris, France
2 shared publications- Dominique Lesselier
Laboratoire des Signaux et Systmes, CNRS, CentraleSuplec, Universit Paris-Saclay, Gif-sur-Yvette, France
2 shared publications - Lucas Merabet
Department of Imaging and Simulation for NDT, CEA List, 91191 Gif-sur-Yvette, France
2 shared publications - Roberto Miorelli
Université Paris-Saclay, CEA, List, F-91120 Palaiseau, France
2 shared publications - Corentin Menard
Paris-Saclay University, CEA, List, 91191 Palaiseau, France
1 shared publication
Latest Publications
- Open Access
- 1 September 2026
- Published by NDT.net GmbH & Co. KG in e-Journal of Nondestructive Testing
- Vol. 31 (9)
Etude sur la détection automatique d’indications à l’aide des méthodes d’apprentissage supervisé par réseaux de neurones pour le contrôle ultrasonore de soudures en acier austénitique
Lors d’un contrôle de soudures par techniques ultrasons multiéléments, l’acquisition des données est rapide du fait des nombreux progrès réalisés par les fabricants d’équipements pour simplifier leurs mises en oeuvre et démocratiser leurs usages. Cependant, la dernière étape du contrôle qui consiste à analyser les données en vue d’établir un rapport d’examen est souvent fastidieuse et nécessite des opérateurs expérimentés en particulier dans le cas de soudures en acier austénitique. Afin d’automatiser certaines étapes d’analyse, le CEA a travaillé sur la détection automatique d’indications à partir de la segmentation des signaux A-SCAN en se basant sur des critères en amplitude et en temps de vol. Toutefois, cette segmentation, disponible dans le logiciel CIVA depuis de nombreuses années, peut présenter certaines limitations dans le cas d’images avec un faible rapport signal sur bruit ou présentant un écho systématique proche des indications. Dans ce contexte et à travers un programme de R&D commun entre EDF, FRAMATOME et le CEA, le CEA a réalisé une étude sur l’apport des méthodes d’apprentissage supervisée par réseaux de neurones pour la détection automatique d’indications dans des soudures en acier austénitique. L’analyse porte sur les acquisitions ultrasonores du procédé END développé et qualifié par EDF pour la recherche de fissures de corrosion sous contrainte. A cette fin, EDF a mis à disposition du CEA des données ultrasonores acquises suivant le procédé qualifié. Cette communication présente la démarche adoptée pour la mise au point de la détection automatique d’indications à l’aide des méthodes d’apprentissages supervisées et les résultats obtenues sur des soudures présentant des fissures de corrosion sous-contrainte.
- 1 September 2026
- Published by NDT.net GmbH & Co. KG in e-Journal of Nondestructive Testing
- Vol. 31 (9)
Open AccessEtude sur la détection automatique d’indications à l’aide des méthodes d’apprentissage supervisé par réseaux de neurones pour le contrôle ultrasonore de soudures en acier austénitique
Lors d’un contrôle de soudures par techniques ultrasons multiéléments, l’acquisition des données est rapide du fait des nombreux progrès réalisés par les fabricants d’équipements pour simplifier leurs mises en oeuvre et démocratiser leurs usages. Cependant, la dernière étape du contrôle qui consiste à analyser les données en vue d’établir un rapport d’examen est souvent fastidieuse et nécessite des opérateurs expérimentés en particulier dans le cas de soudures en acier austénitique. Afin d’automatiser certaines étapes d’analyse, le CEA a travaillé sur la détection automatique d’indications à partir de la segmentation des signaux A-SCAN en se basant sur des critères en amplitude et en temps de vol. Toutefois, cette segmentation, disponible dans le logiciel CIVA depuis de nombreuses années, peut présenter certaines limitations dans le cas d’images avec un faible rapport signal sur bruit ou présentant un écho systématique proche des indications. Dans ce contexte et à travers un programme de R&D commun entre EDF, FRAMATOME et le CEA, le CEA a réalisé une étude sur l’apport des méthodes d’apprentissage supervisée par réseaux de neurones pour la détection automatique d’indications dans des soudures en acier austénitique. L’analyse porte sur les acquisitions ultrasonores du procédé END développé et qualifié par EDF pour la recherche de fissures de corrosion sous contrainte. A cette fin, EDF a mis à disposition du CEA des données ultrasonores acquises suivant le procédé qualifié. Cette communication présente la démarche adoptée pour la mise au point de la détection automatique d’indications à l’aide des méthodes d’apprentissages supervisées et les résultats obtenues sur des soudures présentant des fissures de corrosion sous-contrainte.
- 26 February 2022
- Published by Elsevier in Ultrasonics
- Vol. 124, 106708
Real-time 3D imaging with Fourier-domain algorithms and matrix arrays applied to non-destructive testing
- 26 February 2022
- Published by Elsevier in Ultrasonics
- Vol. 124, 106708
Real-time 3D imaging with Fourier-domain algorithms and matrix arrays applied to non-destructive testing
- Open Access
- 15 July 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (14) , 6505
Ultrasonic Array Imaging of Nuclear Austenitic V-Shape Welds with Inhomogeneous and Unknown Anisotropic Properties
The quality of ultrasound images of welds is highly penalized by the dendritic structure of the material that forms during cooling. The image of a flaw is all the more degraded as a reliable description of such a medium is most of the time not possible, due to the poor knowledge on the weld at time of inspection. In a previous work, we demonstrated the efficiency of an optimization procedure to correct a degraded Total Focusing Method (TFM) image of a point-like reflector inside a homogeneous weld, with uniaxial grain orientation. In the present contribution, this procedure is extended to more realistic welds with a varying grain orientation, and it is evaluated with a 64-element array emitting at 5 MHz on stainless steel weld samples featuring side-drilled holes of 1 mm diameter. A first proof of concept with simulated echoes, and then two experiments show that defects that were hardly visible on TFM images before the optimization are now well reconstructed and with positioning errors inferior to 1 mm.
- 15 July 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (14) , 6505
Open AccessUltrasonic Array Imaging of Nuclear Austenitic V-Shape Welds with Inhomogeneous and Unknown Anisotropic Properties
The quality of ultrasound images of welds is highly penalized by the dendritic structure of the material that forms during cooling. The image of a flaw is all the more degraded as a reliable description of such a medium is most of the time not possible, due to the poor knowledge on the weld at time of inspection. In a previous work, we demonstrated the efficiency of an optimization procedure to correct a degraded Total Focusing Method (TFM) image of a point-like reflector inside a homogeneous weld, with uniaxial grain orientation. In the present contribution, this procedure is extended to more realistic welds with a varying grain orientation, and it is evaluated with a 64-element array emitting at 5 MHz on stainless steel weld samples featuring side-drilled holes of 1 mm diameter. A first proof of concept with simulated echoes, and then two experiments show that defects that were hardly visible on TFM images before the optimization are now well reconstructed and with positioning errors inferior to 1 mm.
- Open Access
- 1 December 2020
- Published by Elsevier in NDT & E International
- Vol. 116, 102327
Optimization algorithms for ultrasonic array imaging in homogeneous anisotropic steel components with unknown properties
- 1 December 2020
- Published by Elsevier in NDT & E International
- Vol. 116, 102327
Open AccessOptimization algorithms for ultrasonic array imaging in homogeneous anisotropic steel components with unknown properties
- 29 January 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 66 (4) , 772-788
2-D and 3-D Reconstruction Algorithms in the Fourier Domain for Plane-Wave Imaging in Nondestructive Testing
Time-domain Plane Wave Imaging (PWI) has recently emerged in medical imaging and is now taking to Non- Destructive Testing (NDT) due to its ability to provide images of good resolution and contrast with only a few steered plane waves. Insonifying a medium with plane waves is a particularly interesting approach in 3D imaging with matrix arrays because it allows to tremendously reduce the volume of data to be stored and processed as well as the acquisition time. However, even if the data volume is reduced with plane wave emissions, the image reconstruction in the time domain with a delay-and-sum algorithm is not sufficient to achieve low computations time in 3D due to the number of voxels. Other reconstruction algorithms take place in the f-k wavenumber-frequency domain and have been shown to accelerate computation times in seismic imaging and in Synthetic Aperture Radar (SAR). In this paper, we start from time-domain PWI in 2D and compare it to two algorithms in the wavenumberfrequency (f-k) domain, coming from the Stolt migration in seismic imaging and the Lu theory of limited diffraction beams in medical imaging. We then extend them to immersion testing configurations where a linear array is facing a plane water-steel interface. Finally, the reconstruction algorithms are generalized to 3D imaging with matrix arrays. A comparison dwelling on image quality and algorithmic complexities is provided, as well as a theoretical analysis of the image amplitudes and the limits of each method. We show that the reconstruction schemes in the wavenumber-frequency domain allow to improve the lateral resolution and offer a theoretical and numerical computation gain of up to 36 in 3D imaging in a realistic NDT configuration.
- 29 January 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 66 (4) , 772-788
2-D and 3-D Reconstruction Algorithms in the Fourier Domain for Plane-Wave Imaging in Nondestructive Testing
Time-domain Plane Wave Imaging (PWI) has recently emerged in medical imaging and is now taking to Non- Destructive Testing (NDT) due to its ability to provide images of good resolution and contrast with only a few steered plane waves. Insonifying a medium with plane waves is a particularly interesting approach in 3D imaging with matrix arrays because it allows to tremendously reduce the volume of data to be stored and processed as well as the acquisition time. However, even if the data volume is reduced with plane wave emissions, the image reconstruction in the time domain with a delay-and-sum algorithm is not sufficient to achieve low computations time in 3D due to the number of voxels. Other reconstruction algorithms take place in the f-k wavenumber-frequency domain and have been shown to accelerate computation times in seismic imaging and in Synthetic Aperture Radar (SAR). In this paper, we start from time-domain PWI in 2D and compare it to two algorithms in the wavenumberfrequency (f-k) domain, coming from the Stolt migration in seismic imaging and the Lu theory of limited diffraction beams in medical imaging. We then extend them to immersion testing configurations where a linear array is facing a plane water-steel interface. Finally, the reconstruction algorithms are generalized to 3D imaging with matrix arrays. A comparison dwelling on image quality and algorithmic complexities is provided, as well as a theoretical analysis of the image amplitudes and the limits of each method. We show that the reconstruction schemes in the wavenumber-frequency domain allow to improve the lateral resolution and offer a theoretical and numerical computation gain of up to 36 in 3D imaging in a realistic NDT configuration.
- Open Access
- 1 January 2018
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1949 (1) , 230007
Comparative study of 2D ultrasound imaging methods in the f-k domain and evaluation of their performances in a realistic NDT configuration
In this paper, we present two frequency-domain algorithms for 2D imaging with plane wave emissions, namely Stolt’s migration and Lu’s method. The theoretical background is first presented, followed by an analysis of the algorithm complexities. The frequency-domain methods are then compared to the time-domain plane wave imaging in a realistic inspection configuration where the array elements are not in contact with the specimen. Imaging defects located far away from the array aperture is assessed and computation times for the three methods are presented as a function of the number of pixels of the reconstructed image. We show that Lu’s method provides a time gain of up to 33 compared to the time-domain algorithm, and demonstrate the limitations of Stolt’s migration for defects far away from the aperture.
- 1 January 2018
- Published by AIP Publishing in AIP Conference Proceedings
- Vol. 1949 (1) , 230007
Open AccessComparative study of 2D ultrasound imaging methods in the f-k domain and evaluation of their performances in a realistic NDT configuration
In this paper, we present two frequency-domain algorithms for 2D imaging with plane wave emissions, namely Stolt’s migration and Lu’s method. The theoretical background is first presented, followed by an analysis of the algorithm complexities. The frequency-domain methods are then compared to the time-domain plane wave imaging in a realistic inspection configuration where the array elements are not in contact with the specimen. Imaging defects located far away from the array aperture is assessed and computation times for the three methods are presented as a function of the number of pixels of the reconstructed image. We show that Lu’s method provides a time gain of up to 33 compared to the time-domain algorithm, and demonstrate the limitations of Stolt’s migration for defects far away from the aperture.
- 3 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 64 (9) , 1336-1344
Ultrasonic Imaging in Highly Attenuating Materials With Hadamard Codes and the Decomposition of the Time Reversal Operator
In this paper, defects in a high density polyethylene pipe are imaged with the total focusing method. The viscoelastic attenuation of this material greatly reduces the signal level and leads to a poor signal-to-noise ratio (SNR) due to electronic noise. To improve the image quality, the decomposition of the time reversal operator method is combined with the spatial Hadamard coded transmissions before calculating images in the time domain. Because the Hadamard coding is not compatible with conventional imaging systems, this paper proposes two modified coding methods based on sparse Hadamard matrices with +1/0 coefficients. The SNRs expected with the different spatial codes are demonstrated, and then validated on both simulated and experimental data. Experiments are performed with a transducer array in contact with the base material of a polyethylene pipe. In order to improve the noise filtering procedure, the singular values associated with electronic noise are expressed on the basis of the random matrix theory. This model of noise singular values allows a better identification of the defect response in noisy experimental data. Finally, the imaging method is evaluated in a more industrial inspection configuration, where an immersion array probe is used to image defects in a butt fusion weld with a complex geometry.
- 3 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 64 (9) , 1336-1344
Ultrasonic Imaging in Highly Attenuating Materials With Hadamard Codes and the Decomposition of the Time Reversal Operator
In this paper, defects in a high density polyethylene pipe are imaged with the total focusing method. The viscoelastic attenuation of this material greatly reduces the signal level and leads to a poor signal-to-noise ratio (SNR) due to electronic noise. To improve the image quality, the decomposition of the time reversal operator method is combined with the spatial Hadamard coded transmissions before calculating images in the time domain. Because the Hadamard coding is not compatible with conventional imaging systems, this paper proposes two modified coding methods based on sparse Hadamard matrices with +1/0 coefficients. The SNRs expected with the different spatial codes are demonstrated, and then validated on both simulated and experimental data. Experiments are performed with a transducer array in contact with the base material of a polyethylene pipe. In order to improve the noise filtering procedure, the singular values associated with electronic noise are expressed on the basis of the random matrix theory. This model of noise singular values allows a better identification of the defect response in noisy experimental data. Finally, the imaging method is evaluated in a more industrial inspection configuration, where an immersion array probe is used to image defects in a butt fusion weld with a complex geometry.
- Open Access
- 19 September 2015
- Published by Elsevier in Physics Procedia
- Vol. 70, 562-565
Noise Filtering in the Synthetic Transmit Aperture Imaging by Decomposition of the Time Reversal Operator: Application to Flaw Detection in Coarse-grained Stainless Steels
- 19 September 2015
- Published by Elsevier in Physics Procedia
- Vol. 70, 562-565
Open AccessNoise Filtering in the Synthetic Transmit Aperture Imaging by Decomposition of the Time Reversal Operator: Application to Flaw Detection in Coarse-grained Stainless Steels
- 1 July 2012
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15582809,p. 207-212
Real time nondestrutive testing of composite aeronautical structures with a self-adaptive ultrasonic technique
In aeronautical industry, composite structures under testing often have complex and variable geometries. In this case, an optimal use of ultrasonic transducer arrays requires specific algorithms in electronic systems in order to achieve rapid and reliable inspections. To fulfil such requirements, a new real time and adaptive technique is presented. This technique is based on an iterative algorithm that does not require a prior knowledge of the geometrical and acoustical properties of the component undergoing inspection. All different parts of one given component (with flat, concave, convex parts...) can be controlled using a same transducer array, such as a conventional linear array with a flat shape. In this paper, the adaptive processing is demonstrated and illustrated through acquisitions performed with different typical aircraft composite structures.
- 1 July 2012
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15582809,p. 207-212
Real time nondestrutive testing of composite aeronautical structures with a self-adaptive ultrasonic technique
In aeronautical industry, composite structures under testing often have complex and variable geometries. In this case, an optimal use of ultrasonic transducer arrays requires specific algorithms in electronic systems in order to achieve rapid and reliable inspections. To fulfil such requirements, a new real time and adaptive technique is presented. This technique is based on an iterative algorithm that does not require a prior knowledge of the geometrical and acoustical properties of the component undergoing inspection. All different parts of one given component (with flat, concave, convex parts...) can be controlled using a same transducer array, such as a conventional linear array with a flat shape. In this paper, the adaptive processing is demonstrated and illustrated through acquisitions performed with different typical aircraft composite structures.
- 16 April 2008
- Published by Taylor & Francis in Waves in Random and Complex Media
- Vol. 18 (2) , 241-254
Coherent waves in a multiply scattering poro-elastic medium obeying Biot's theory
Twersky's theory is generalized to multiple scattering by a uniform random distribution of cylinders in a poro-elastic medium. The high-frequency regime only, where no dispersion effects occur in the absence of scatterers, is investigated in the frame of Biot's theory. The scatterers lie within a slab of the host medium, and an incident wave gives rise to a fast longitudinal coherent wave, a slow longitudinal one, as well as a shear one in the slab. The dispersion equations of those three coherent waves are derived. The shear coherent wave propagates independently of the other two, while the longitudinal coherent waves obey a coupled dispersion equation involving conversion terms. Numerically speaking, coupling effects are significant only when forward scattering by a single cylinder of the fast wave into the slow one (or the slow wave into the fast) is larger than forward scattering with no conversion.
- 16 April 2008
- Published by Taylor & Francis in Waves in Random and Complex Media
- Vol. 18 (2) , 241-254
Coherent waves in a multiply scattering poro-elastic medium obeying Biot's theory
Twersky's theory is generalized to multiple scattering by a uniform random distribution of cylinders in a poro-elastic medium. The high-frequency regime only, where no dispersion effects occur in the absence of scatterers, is investigated in the frame of Biot's theory. The scatterers lie within a slab of the host medium, and an incident wave gives rise to a fast longitudinal coherent wave, a slow longitudinal one, as well as a shear one in the slab. The dispersion equations of those three coherent waves are derived. The shear coherent wave propagates independently of the other two, while the longitudinal coherent waves obey a coupled dispersion equation involving conversion terms. Numerically speaking, coupling effects are significant only when forward scattering by a single cylinder of the fast wave into the slow one (or the slow wave into the fast) is larger than forward scattering with no conversion.