Skip to content

Source Catalog

The supplied bundle contains exactly two original PDFs. Citations use PDF page numbers checked against the page-separated local text.

Peri and Orion (2017)

Eyal Peri and Itzhak Orion, “Shielding calculations for industrial 5/7.5 MeV electron accelerators using the MCNP Monte Carlo Code,” EPJ Web of Conferences 153, 03011 (2017). DOI 10.1051/epjconf/201715303011. Six pages. File peri-orion-2017.pdf; SHA-256 473285f378ccb089e8616b9f520ef0e1cad35615049a8ddb7d6fa9dd04a9b8aa.

EXTERNAL LITERATURE — supports:

  • Industrial converted-X-ray context and the importance of energy-, converter-, and angle-dependent shielding information at 5 and 7.5 MeV [Peri and Orion 2017, pp. 1-2].
  • MCNP 4C2 calculations of angular photon dose, bremsstrahlung spectra, and concrete attenuation for specified converters, energies, and angles [Peri and Orion 2017, pp. 2-5].
  • Pencil-beam converter calculations, ordinary concrete at 2.3 g/cm³, angular detector sectors, spherical concrete thicknesses from 10 to 180 cm, and geometry-splitting variance reduction for thick shielding [Peri and Orion 2017, pp. 2-3].
  • Target dimensions of 1.5 mm Ta at 5 MeV and 2.3 mm Ta at 7.5 MeV; these describe the authors' isolated targets selected with reference to electron range, not universal optima [Peri and Orion 2017, p. 2].
  • Photon angular-dose calculations at one metre, energy spectra at selected angles, and concrete attenuation at 0°, 90°, 135°, and 180° [Peri and Orion 2017, pp. 2-5].
  • Energy-angle dependence, harder forward spectra, higher dose at 7.5 MeV, and more effective attenuation of backward radiation in thick concrete [Peri and Orion 2017, pp. 4-6].
  • Comparisons with earlier calculations and a measured bremsstrahlung spectrum as limited checks of the authors' configurations [Peri and Orion 2017, pp. 2-4].
  • The authors' reported history counts and relative statistical uncertainties for their specific angular, spectral, and attenuation calculations [Peri and Orion 2017, pp. 4-5].

Does not support:

  • A broad scanned plate, water/SS304 backing, finite-edge treatment, complete rectangular room, PHITS implementation, RAMAL-EBX, or a three-dimensional surrogate.
  • Treating its F6:p KERMA-approximation dose as H*(10) [Peri and Orion 2017, p. 2].
  • Copying its isolated-target thicknesses as an approved D3 construction.
  • Approval of D5 scan details, D7 criteria, D9 validation, safe D4 dimensions, Dataset inclusion, surrogate acceptance, or regulatory compliance.
  • Treating the paper's reported histories, variance reduction, or uncertainty levels as Paper 1 numerical criteria.

Chowdhury et al. (2024)

Rajarshi Pal Chowdhury, Juan Zamora, Tom Ginter, and Georg Bollen, “Machine Learning in Accelerator Shielding: Application Using Convolutional Neural Networks,” RPSD 2024 proceedings. DOI 10.13182/T131-45920. Three pages. File chowdhury-et-al-2024.pdf; SHA-256 6f67d6f4a1ee0d09c740e50038762864918d845b2924e3e648d1ee8d9af82b3a.

EXTERNAL LITERATURE — supports:

  • Motivation for faster shielding calculations where Monte Carlo neutron transport may require hours to weeks [Chowdhury et al. 2024, p. 1].
  • A PHITS-trained one-dimensional CNN mapping incident neutron spectra and concrete shielding conditions to downstream differential neutron flux [Chowdhury et al. 2024, pp. 1-2].
  • PHITS 3.29 calculations for 250 quasi-monoenergetic neutron inputs from 1-250 MeV, followed by 10,000 superposed datasets and a reported 70:30 training/testing split [Chowdhury et al. 2024, p. 2].
  • Normally incident narrow neutron beams, vacuum upstream of the shield, and an air tally region downstream in the authors' simplified transport setup [Chowdhury et al. 2024, p. 2].
  • Testing with a realistic Ca-48-induced neutron spectrum distinct from the training spectra while retaining the training-data simulation configuration [Chowdhury et al. 2024, pp. 2-3].
  • Reported effective-dose agreement within approximately 10% for the demonstrated case after ICRP-74 folding and millisecond-scale model computation [Chowdhury et al. 2024, pp. 1, 3].
  • The authors' conclusion that the method succeeded for one material and thickness, with extension to trained thicknesses and future multiple-material work [Chowdhury et al. 2024, p. 3].

Does not support:

  • Conventional industrial 5-7.5 MeV electron-to-bremsstrahlung conversion, a finite scanned head, complete shielding rooms, or three-dimensional H*(10) fields.
  • D3-D6 source, converter, room, material, photonuclear-library, or normalization decisions.
  • Whole-configuration separation, untouched final-test governance, local underprediction limits, uncertainty-aware voxel inclusion, or independent PHITS verification of complete-room Predictions.
  • Physical validation of RAMAL-EBX or transfer of the approximately 10% demonstrated agreement to Paper 1.
  • A claim that a spectrum distinct from training is equivalent to holding out a complete physical configuration.

Bundle boundary

UNRESOLVED: No paper in this original supplied bundle establishes the Ta–water–SS304 industrial configuration, approved Paper 1 room family, D7–D11 criteria, or full claimed evidence gap. The separate D3/D5 primary-source review is complete and owns the broader source-head evidence; neither review supports a universal novelty, completeness, or priority claim.