Machine Learning Surrogate Modeling of Multilayer Concrete-FeB-Soil Shielding in a 250 MeV Proton Accelerator Tunnel Based on FLUKA Monte Carlo Simulations

Şaşal S., Sarıyer D.

RAP 26 International Conference on Radiation Applications,, Lisbon, Portekiz, 25 - 29 Mayıs 2026, ss.1-2, (Özet Bildiri)

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Lisbon
  • Basıldığı Ülke: Portekiz
  • Sayfa Sayıları: ss.1-2
  • Manisa Celal Bayar Üniversitesi Adresli: Evet

Özet

In proton accelerator environments, neutrons produced through proton–target interactions constitute the dominant secondary radiation component and play a critical role in shielding design due to their high penetration capability and broad energy spectrum. In this study, a hybrid Monte Carlo–Machine Learning (MC–ML) framework was developed to enable rapid and reliable prediction of ambient dose equivalent distributions in a multilayer concrete–FeB–soil shielding system for a 250 MeV proton accelerator tunnel.

The secondary neutron field was modeled in a three-dimensional tunnel geometry using the FLUKA Monte Carlo particle transport code for a monoenergetic proton beam incident on a copper target. Dose data obtained along horizontal and vertical directions were used to train surrogate models based on linear, distance-based, and tree-based ensemble learning algorithms, including Linear Regression, K-Nearest Neighbors, Random Forest, Gradient Boosting, HistGradientBoosting, and Extra Trees Regressors.

The results demonstrate that tree-based ensemble methods provide the highest prediction accuracy. In particular, the Extra Trees and Random Forest models achieved R² values of 0.9970 and 0.9964 along the x-direction, and 0.9832 and 0.9815 along the y-direction, respectively. The developed surrogate models successfully capture the highly nonlinear dose–distance relationship and offer significant computational efficiency compared to full Monte Carlo simulations.

These findings highlight the strong potential of hybrid MC–ML approaches for fast evaluation and optimization of complex multilayer shielding systems in proton accelerator applications.