Demet Sariyer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Demet Sariyer has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Mustafa Emre Erbil
Isparta University of Applied Sciences
4 shared publications - Elif Yıldırım
Department of Computer Engineering, İstanbul Technical University Faculty of Computer and Informatics, İstanbul, Türkiye
2 shared publications N. Küçer
Celal Bayar University, Faculty of Arts and Sciences, Department of Physics, 45140, Muradiye, Manisa, Turkey
2 shared publicationsR. Küçer
Celal Bayar University, Faculty of Arts and Sciences, Department of Physics, 45140, Muradiye, Manisa, Turkey
2 shared publications- Rahmi Küçer2 shared publications
- Ali Osman Gokcan
Manisa Celal Bayar University
1 shared publication
Latest Publications
- 23 October 2026
- Published by Duzce Universitesi Bilim ve Teknoloji Dergisi in Düzce Üniversitesi Bilim ve Teknoloji Dergisi
- Vol. 14 (4) , 1020-1039
Comprehensive Analysis of Particle Production, Transport, and Spatial Dose Distribution in Two Multilayer Shielding Designs for a 50 MeV Proton Accelerator Using FLUKA Monte Carlo Simulations
This study investigates secondary particle production and transport processes in two different multilayer shielding configurations (Design 1 and Design 2) for a 50 MeV proton accelerator using the FLUKA Monte Carlo (MC) code. A monoenergetic proton beam interacting with a copper target generates a complex radiation field composed of photons, neutrons, protons, and light nuclear particles. Analysis of inelastic interactions shows that photon production is dominant in both designs at approximately 50%, followed by neutrons (~27%) and protons (~17%), indicating the significant role of electromagnetic processes in the low-energy regime. Stopping particle statistics reveal that nearly 99% of primary protons are absorbed within the shielding structures, while the contribution of light ions remains negligible. The investigation of low-energy neutron interactions demonstrates a strong neutron multiplication behavior, with more than 97% of the generated secondary particles being neutrons. Although both shielding designs exhibit similar overall particle production characteristics, Design 1 yields a slightly higher neutron multiplication factor (~4.16 particles per proton) compared to Design 2 (~3.74 particles per proton), suggesting that Design 2 provides relatively more effective moderation for low-energy neutrons. These findings indicate that while initial particle production is governed by electromagnetic interactions, the evolution of the radiation field is primarily controlled by neutron transport and multiplication mechanisms. Furthermore, one-dimensional ambient dose equivalent H*(10) profiles along the x and y directions clearly demonstrate the directional attenuation performance of the shielding configurations. In addition, relative error distributions obtained from MC simulations were analyzed to assess computational reliability, confirming that the majority of dose results fall within acceptable uncertainty limits (generally below 10%). Overall, the study provides a detailed physical interpretation of particle production mechanisms and dose distribution characteristics, offering valuable insights for the optimization of multilayer shielding designs in low-energy proton accelerator environments.
- 23 October 2026
- Published by Duzce Universitesi Bilim ve Teknoloji Dergisi in Düzce Üniversitesi Bilim ve Teknoloji Dergisi
- Vol. 14 (4) , 1020-1039
Comprehensive Analysis of Particle Production, Transport, and Spatial Dose Distribution in Two Multilayer Shielding Designs for a 50 MeV Proton Accelerator Using FLUKA Monte Carlo Simulations
This study investigates secondary particle production and transport processes in two different multilayer shielding configurations (Design 1 and Design 2) for a 50 MeV proton accelerator using the FLUKA Monte Carlo (MC) code. A monoenergetic proton beam interacting with a copper target generates a complex radiation field composed of photons, neutrons, protons, and light nuclear particles. Analysis of inelastic interactions shows that photon production is dominant in both designs at approximately 50%, followed by neutrons (~27%) and protons (~17%), indicating the significant role of electromagnetic processes in the low-energy regime. Stopping particle statistics reveal that nearly 99% of primary protons are absorbed within the shielding structures, while the contribution of light ions remains negligible. The investigation of low-energy neutron interactions demonstrates a strong neutron multiplication behavior, with more than 97% of the generated secondary particles being neutrons. Although both shielding designs exhibit similar overall particle production characteristics, Design 1 yields a slightly higher neutron multiplication factor (~4.16 particles per proton) compared to Design 2 (~3.74 particles per proton), suggesting that Design 2 provides relatively more effective moderation for low-energy neutrons. These findings indicate that while initial particle production is governed by electromagnetic interactions, the evolution of the radiation field is primarily controlled by neutron transport and multiplication mechanisms. Furthermore, one-dimensional ambient dose equivalent H*(10) profiles along the x and y directions clearly demonstrate the directional attenuation performance of the shielding configurations. In addition, relative error distributions obtained from MC simulations were analyzed to assess computational reliability, confirming that the majority of dose results fall within acceptable uncertainty limits (generally below 10%). Overall, the study provides a detailed physical interpretation of particle production mechanisms and dose distribution characteristics, offering valuable insights for the optimization of multilayer shielding designs in low-energy proton accelerator environments.
- Open Access
- 25 September 2026
- Published by Elsevier in Nuclear Engineering and Technology
- Vol. 59 (2)
Data-driven surrogate modeling of FLUKA-generated ambient dose equivalent distributions in an asymmetric proton accelerator shielding system
- 25 September 2026
- Published by Elsevier in Nuclear Engineering and Technology
- Vol. 59 (2)
Open AccessData-driven surrogate modeling of FLUKA-generated ambient dose equivalent distributions in an asymmetric proton accelerator shielding system
- Open Access
- 16 September 2026
- Published by Elsevier in SSRN Electronic Journal
Hybrid MC-ML Modeling of Neutron Spectra and Multidirectional Dose Distributions in Multilayer Fe-Concrete Shielding for a 50 MeV Proton Accelerator
In this study, the multidirectional behavior of secondary neutron fields generated by the interaction of 50 MeV protons with a copper target was comprehensively investigated in a multilayer Fe-concrete shielding geometry using a hybrid framework combining Monte Carlo (MC) simulations and machine learning (ML). The proton accelerator tunnel was modeled as a three-dimensional structure consisting of symmetric 2 m Fe + 4 m concrete side shields, a concrete roof, a concrete floor, and the surrounding soil environment. FLUKA MC simulations were performed to analyze secondary particle production, neutron energy spectra, and spatial ambient dose equivalent H*(10) distributions. Neutron energy spectra were evaluated using the USRTRACK scoring estimator, providing insight into the energy-dependent transport, moderation, and attenuation mechanisms within the multilayer shielding system. The results demonstrated that the iron layer effectively suppresses the fast-neutron component through inelastic scattering, whereas the surrounding concrete layers provide efficient neutron moderation and absorption, producing a progressive spectral softening together with significant reductions in H*(10). Using the FLUKA-generated H*(10) datasets, in which the one-dimensional spatial position along each axis (x or y) was used as the single input feature and the corresponding H*(10) value as the output target, Linear Regression (LR), Random Forest (RF), XGBoost, Support Vector Regression (SVR), and LightGBM models were developed and compared. Following log(1+y) transformation, min-max normalization, an 80%/20% train-test split, 5-fold cross-validation, and GridSearchCV-based hyperparameter optimization, the Random Forest model achieved the highest predictive accuracy (X-axis test R² = 0.9994; Y-axis test R² = 0.9991), while XGBoost and LightGBM exhibited comparable performance. Computational efficiency analysis showed that high-statistics FLUKA simulations require several hours on the TRUBA/TR-Grid HPC infrastructure, whereas the trained ML models generate predictions within milliseconds, corresponding to computational speedups of approximately 10⁶–10⁹ during the prediction stage. The combined analyses of neutron spectra, multidirectional dose distributions, and surrogate modeling demonstrate that the proposed hybrid MC-ML framework provides both a physically interpretable and computationally efficient methodology for rapid radiation shielding optimization and accelerator shielding design.
- 16 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHybrid MC-ML Modeling of Neutron Spectra and Multidirectional Dose Distributions in Multilayer Fe-Concrete Shielding for a 50 MeV Proton Accelerator
In this study, the multidirectional behavior of secondary neutron fields generated by the interaction of 50 MeV protons with a copper target was comprehensively investigated in a multilayer Fe-concrete shielding geometry using a hybrid framework combining Monte Carlo (MC) simulations and machine learning (ML). The proton accelerator tunnel was modeled as a three-dimensional structure consisting of symmetric 2 m Fe + 4 m concrete side shields, a concrete roof, a concrete floor, and the surrounding soil environment. FLUKA MC simulations were performed to analyze secondary particle production, neutron energy spectra, and spatial ambient dose equivalent H*(10) distributions. Neutron energy spectra were evaluated using the USRTRACK scoring estimator, providing insight into the energy-dependent transport, moderation, and attenuation mechanisms within the multilayer shielding system. The results demonstrated that the iron layer effectively suppresses the fast-neutron component through inelastic scattering, whereas the surrounding concrete layers provide efficient neutron moderation and absorption, producing a progressive spectral softening together with significant reductions in H*(10). Using the FLUKA-generated H*(10) datasets, in which the one-dimensional spatial position along each axis (x or y) was used as the single input feature and the corresponding H*(10) value as the output target, Linear Regression (LR), Random Forest (RF), XGBoost, Support Vector Regression (SVR), and LightGBM models were developed and compared. Following log(1+y) transformation, min-max normalization, an 80%/20% train-test split, 5-fold cross-validation, and GridSearchCV-based hyperparameter optimization, the Random Forest model achieved the highest predictive accuracy (X-axis test R² = 0.9994; Y-axis test R² = 0.9991), while XGBoost and LightGBM exhibited comparable performance. Computational efficiency analysis showed that high-statistics FLUKA simulations require several hours on the TRUBA/TR-Grid HPC infrastructure, whereas the trained ML models generate predictions within milliseconds, corresponding to computational speedups of approximately 10⁶–10⁹ during the prediction stage. The combined analyses of neutron spectra, multidirectional dose distributions, and surrogate modeling demonstrate that the proposed hybrid MC-ML framework provides both a physically interpretable and computationally efficient methodology for rapid radiation shielding optimization and accelerator shielding design.
- Open Access
- 31 August 2026
- Published by International Journal of 3D Printing Technologies and Digital Industry in International Journal of 3D Printing Technologies and Digital Industry
- Vol. 10 (2) , 402-424
MULTIDIRECTIONAL NEUTRON TRANSPORT AND DOSE COMPLIANCE IN MULTILAYER IRON-CONCRETE SHIELDING FOR A 50 MEV PROTON ACCELERATOR: A MONTE CARLO-MACHINE LEARNING APPROACH
In this study, the multidirectional transport and dose compliance of secondary neutron fields generated by the interaction of 50 MeV protons with a copper target were investigated within a multilayer iron-concrete hybrid shielding system using a hybrid Monte Carlo (MC)-Machine Learning (ML) approach. High-accuracy FLUKA MC simulations were employed to compute the ambient dose equivalent H*(10) distributions along both horizontal (±x) and vertical (±y) directions in a realistic tunnel-type accelerator geometry. Based on the resulting dose-distance dataset, Linear Regression (LR), Random Forest (RF), and Gradient Boosting Regressor (GBR) models were developed. During the data preprocessing stage, log(1+y) transformation and min-max normalization were applied. Model performances were evaluated using an 80% training – 20% testing split, 5-fold cross-validation, and GridSearch-based hyperparameter optimization. The results demonstrate that tree-based ensemble models can accurately capture nonlinear dose distributions spanning a wide dynamic range (RF: R² = 0.9999 / 0.9992; GBR: R² = 0.9997 / 0.9988), whereas the performance of the LR model remains limited. From a radiation protection perspective, the analysis indicates that dose levels along the horizontal direction and particularly in the upward (+y) region exceed international regulatory limits. In contrast, a significant attenuation is observed in the downward direction due to the combined moderating and absorbing effects of the concrete floor and surrounding soil. In conclusion, the proposed MC-ML framework provides a reliable and scalable methodology for the rapid analysis and optimization of multilayer shielding systems by significantly reducing the computational cost associated with conventional MC-based dose predictions.
- 31 August 2026
- Published by International Journal of 3D Printing Technologies and Digital Industry in International Journal of 3D Printing Technologies and Digital Industry
- Vol. 10 (2) , 402-424
Open AccessMULTIDIRECTIONAL NEUTRON TRANSPORT AND DOSE COMPLIANCE IN MULTILAYER IRON-CONCRETE SHIELDING FOR A 50 MEV PROTON ACCELERATOR: A MONTE CARLO-MACHINE LEARNING APPROACH
In this study, the multidirectional transport and dose compliance of secondary neutron fields generated by the interaction of 50 MeV protons with a copper target were investigated within a multilayer iron-concrete hybrid shielding system using a hybrid Monte Carlo (MC)-Machine Learning (ML) approach. High-accuracy FLUKA MC simulations were employed to compute the ambient dose equivalent H*(10) distributions along both horizontal (±x) and vertical (±y) directions in a realistic tunnel-type accelerator geometry. Based on the resulting dose-distance dataset, Linear Regression (LR), Random Forest (RF), and Gradient Boosting Regressor (GBR) models were developed. During the data preprocessing stage, log(1+y) transformation and min-max normalization were applied. Model performances were evaluated using an 80% training – 20% testing split, 5-fold cross-validation, and GridSearch-based hyperparameter optimization. The results demonstrate that tree-based ensemble models can accurately capture nonlinear dose distributions spanning a wide dynamic range (RF: R² = 0.9999 / 0.9992; GBR: R² = 0.9997 / 0.9988), whereas the performance of the LR model remains limited. From a radiation protection perspective, the analysis indicates that dose levels along the horizontal direction and particularly in the upward (+y) region exceed international regulatory limits. In contrast, a significant attenuation is observed in the downward direction due to the combined moderating and absorbing effects of the concrete floor and surrounding soil. In conclusion, the proposed MC-ML framework provides a reliable and scalable methodology for the rapid analysis and optimization of multilayer shielding systems by significantly reducing the computational cost associated with conventional MC-based dose predictions.
- 28 May 2026
- Published by Suleyman Demirel Universitesi Fen Edebiyat Fakultesi Fen Dergisi in Süleyman Demirel Üniversitesi Fen Edebiyat Fakültesi Fen Dergisi
- Vol. 21 (1) , 1-18
Rapid Secondary Neutron Dose Estimation in Proton Accelerators Using FLUKA and Machine Learning
Proton accelerators are facilities of critical importance for radiation safety, as interactions of high-energy proton beams with matter produce secondary neutrons. In this study, neutron dose distributions were calculated using FLUKA Monte Carlo simulations for three energy levels (50, 100 and 250 MeV) across different shielding and environmental configurations (soil-air and soil-shield; along the x- and y-axes). The resulting data were predicted using machine learning models, namely Linear Regression (LR), Random Forest (RF), Gradient Boosting Regressor (GBR) and K-Nearest Neighbors (KNN). Model performance was evaluated based on the coefficient of determination (R²) and normalized root mean square error (NRMSE).The results demonstrate that along the x-axis in the soil–air configuration, GBR (R² ≈ 0.906-0.987) and RF (R² ≈ 0.833-0.988) exhibited strong performance, while in the soil-shield x-axis configuration, GBR (R² ≈ 0.762-0.797) and RF (R² ≈ 0.810-0.896) also achieved reliable predictions. Along the y-axis, GBR and RF models showed high accuracy, with R² ≈ 0.843-0.912 and R² ≈ 0.823-0.853 for soil–air, and R² ≈ 0.844-0.981 and R² ≈ 0.887-0.972 for soil-shield, respectively. These findings confirm that RF and GBR models can rapidly and reliably predict secondary neutron doses under varying energy levels and environmental conditions. The proposed hybrid Monte Carlo-Machine learning approach reduces simulation times and provides a fast and reliable method for shielding design, emerging as an effective tool for radiation safety and shield optimization in proton accelerators.
- 28 May 2026
- Published by Suleyman Demirel Universitesi Fen Edebiyat Fakultesi Fen Dergisi in Süleyman Demirel Üniversitesi Fen Edebiyat Fakültesi Fen Dergisi
- Vol. 21 (1) , 1-18
Rapid Secondary Neutron Dose Estimation in Proton Accelerators Using FLUKA and Machine Learning
Proton accelerators are facilities of critical importance for radiation safety, as interactions of high-energy proton beams with matter produce secondary neutrons. In this study, neutron dose distributions were calculated using FLUKA Monte Carlo simulations for three energy levels (50, 100 and 250 MeV) across different shielding and environmental configurations (soil-air and soil-shield; along the x- and y-axes). The resulting data were predicted using machine learning models, namely Linear Regression (LR), Random Forest (RF), Gradient Boosting Regressor (GBR) and K-Nearest Neighbors (KNN). Model performance was evaluated based on the coefficient of determination (R²) and normalized root mean square error (NRMSE).The results demonstrate that along the x-axis in the soil–air configuration, GBR (R² ≈ 0.906-0.987) and RF (R² ≈ 0.833-0.988) exhibited strong performance, while in the soil-shield x-axis configuration, GBR (R² ≈ 0.762-0.797) and RF (R² ≈ 0.810-0.896) also achieved reliable predictions. Along the y-axis, GBR and RF models showed high accuracy, with R² ≈ 0.843-0.912 and R² ≈ 0.823-0.853 for soil–air, and R² ≈ 0.844-0.981 and R² ≈ 0.887-0.972 for soil-shield, respectively. These findings confirm that RF and GBR models can rapidly and reliably predict secondary neutron doses under varying energy levels and environmental conditions. The proposed hybrid Monte Carlo-Machine learning approach reduces simulation times and provides a fast and reliable method for shielding design, emerging as an effective tool for radiation safety and shield optimization in proton accelerators.
- Open Access
- 20 May 2026
- Published by Elsevier in SSRN Electronic Journal
Millisecond-Level Surrogate Prediction of Neutron Dose Distributions in Multilayer Fe-Concrete Shielding Using Hybrid MC-ML Modeling
In this study, the spatial ambient dose equivalent H*(10) distributions of secondary neutron fields produced by the interaction of 50 MeV protons with a copper target were investigated in a multilayer iron-concrete shielding geometry, using a hybrid approach combining Monte Carlo (MC) and machine learning (ML) methods. In the simulations performed with the FLUKA MC code, the proton accelerator tunnel was modeled as a three-dimensional structure including symmetric 2 m Fe + 1 m concrete side shields, a concrete roof, a concrete floor, and the surrounding soil environment. Using the H*(10) datasets obtained from the FLUKA simulations, Linear Regression, Random Forest, XGBoost, Support Vector Regression, and LightGBM models were developed and compared. In the data preprocessing stage, log(1+y) transformation and min-max normalization were applied; the models were trained under an 80%/20% train-test split with 5-fold cross-validation and GridSearchCV-based hyperparameter optimization. The results showed that the Random Forest model provided the highest predictive performance along both axes (X-axis test R2 = 0.9994; Y-axis test R2 = 0.9991). XGBoost and LightGBM models also exhibited high accuracy. The computational efficiency analysis revealed that high-statistics FLUKA simulations require run times on the order of hours on the TRUBA/TR-Grid HPC infrastructure, whereas the trained ML models can be trained within seconds and produce predictions at the millisecond level. A computational speed-up of approximately 106-109 was achieved with ML-based predictions compared to MC simulations. The results demonstrate that the proposed hybrid MC-ML approach provides an effective computational framework for rapid parametric screening and radiation shielding optimization.
- 20 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMillisecond-Level Surrogate Prediction of Neutron Dose Distributions in Multilayer Fe-Concrete Shielding Using Hybrid MC-ML Modeling
In this study, the spatial ambient dose equivalent H*(10) distributions of secondary neutron fields produced by the interaction of 50 MeV protons with a copper target were investigated in a multilayer iron-concrete shielding geometry, using a hybrid approach combining Monte Carlo (MC) and machine learning (ML) methods. In the simulations performed with the FLUKA MC code, the proton accelerator tunnel was modeled as a three-dimensional structure including symmetric 2 m Fe + 1 m concrete side shields, a concrete roof, a concrete floor, and the surrounding soil environment. Using the H*(10) datasets obtained from the FLUKA simulations, Linear Regression, Random Forest, XGBoost, Support Vector Regression, and LightGBM models were developed and compared. In the data preprocessing stage, log(1+y) transformation and min-max normalization were applied; the models were trained under an 80%/20% train-test split with 5-fold cross-validation and GridSearchCV-based hyperparameter optimization. The results showed that the Random Forest model provided the highest predictive performance along both axes (X-axis test R2 = 0.9994; Y-axis test R2 = 0.9991). XGBoost and LightGBM models also exhibited high accuracy. The computational efficiency analysis revealed that high-statistics FLUKA simulations require run times on the order of hours on the TRUBA/TR-Grid HPC infrastructure, whereas the trained ML models can be trained within seconds and produce predictions at the millisecond level. A computational speed-up of approximately 106-109 was achieved with ML-based predictions compared to MC simulations. The results demonstrate that the proposed hybrid MC-ML approach provides an effective computational framework for rapid parametric screening and radiation shielding optimization.
- Open Access
- 6 May 2026
- Published by Elsevier in SSRN Electronic Journal
Uncertainty-Aware Hybrid Monte Carlo–Machine Learning Surrogate Modeling of Neutron Dose Attenuation in Ferroboron–Concrete Multilayer Shielding
This study investigates the spatial distribution of secondary neutron-induced dose fields generated by 250 MeV proton–copper interactions within a multilayer ferroboron (FeB)-concrete shielding configuration using a hybrid Monte Carlo (MC) and machine learning (ML) framework. High-fidelity simulations were performed using the FLUKA code to obtain three-dimensional ambient dose equivalent H*(10) [µSv·h−1] distributions in a realistic accelerator tunnel geometry consisting of symmetric FeB-concrete lateral shielding, a concrete floor and roof, and surrounding soil.Based on the MC-generated dataset, four regression models—Gaussian Process Regression (GPR), Multilayer Perceptron (MLP), Kernel Ridge Regression (KRR), and CatBoost (CBR)—were trained and systematically evaluated. Data preprocessing included log(1+y) transformation and min–max normalization, while model development employed an 80/20 train–test split, 5-fold cross-validation, and GridSearchCV-based hyperparameter optimization. The results indicate that the GPR model with a Matérn-1.5 kernel achieved the highest prediction accuracy along both spatial axes (CV R2 = 0.9989 for x-axis and 0.9995 for y-axis), while CatBoost also demonstrated strong performance (CV R2 = 0.9991 along the y-axis). Furthermore, epistemic uncertainty estimates obtained from the GPR model were compared with statistical uncertainties from FLUKA simulations, revealing a statistically significant correlation (Pearson r = 0.541 for x-axis and r = 0.701 for y-axis, p < 10−40). This finding demonstrates that ML-based uncertainty quantification is physically consistent with the stochastic nature of MC simulations.The proposed hybrid MC–ML framework significantly reduces computational cost from hours to milliseconds while preserving high accuracy, enabling rapid and reliable evaluation of dose attenuation in multilayer shielding systems.
- 6 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessUncertainty-Aware Hybrid Monte Carlo–Machine Learning Surrogate Modeling of Neutron Dose Attenuation in Ferroboron–Concrete Multilayer Shielding
This study investigates the spatial distribution of secondary neutron-induced dose fields generated by 250 MeV proton–copper interactions within a multilayer ferroboron (FeB)-concrete shielding configuration using a hybrid Monte Carlo (MC) and machine learning (ML) framework. High-fidelity simulations were performed using the FLUKA code to obtain three-dimensional ambient dose equivalent H*(10) [µSv·h−1] distributions in a realistic accelerator tunnel geometry consisting of symmetric FeB-concrete lateral shielding, a concrete floor and roof, and surrounding soil.Based on the MC-generated dataset, four regression models—Gaussian Process Regression (GPR), Multilayer Perceptron (MLP), Kernel Ridge Regression (KRR), and CatBoost (CBR)—were trained and systematically evaluated. Data preprocessing included log(1+y) transformation and min–max normalization, while model development employed an 80/20 train–test split, 5-fold cross-validation, and GridSearchCV-based hyperparameter optimization. The results indicate that the GPR model with a Matérn-1.5 kernel achieved the highest prediction accuracy along both spatial axes (CV R2 = 0.9989 for x-axis and 0.9995 for y-axis), while CatBoost also demonstrated strong performance (CV R2 = 0.9991 along the y-axis). Furthermore, epistemic uncertainty estimates obtained from the GPR model were compared with statistical uncertainties from FLUKA simulations, revealing a statistically significant correlation (Pearson r = 0.541 for x-axis and r = 0.701 for y-axis, p < 10−40). This finding demonstrates that ML-based uncertainty quantification is physically consistent with the stochastic nature of MC simulations.The proposed hybrid MC–ML framework significantly reduces computational cost from hours to milliseconds while preserving high accuracy, enabling rapid and reliable evaluation of dose attenuation in multilayer shielding systems.
- Open Access
- 6 May 2026
- Published by Elsevier in SSRN Electronic Journal
Multidirectional Neutron Shielding Analysis and Hybrid MC-ML Surrogate Modeling of Multilayer Concrete-FeB-Concrete Systems in a 250 MeV Proton Accelerator Tunnel
In high-energy proton accelerators, neutron-dominated secondary radiation fields from proton–target interactions critically govern shielding design. In this study, the transport and attenuation of secondary particles from 250 MeV proton–copper interactions were investigated within a multilayer concrete–ferroboron (FeB)–concrete shielding system using the FLUKA Monte Carlo (MC) code. A three-dimensional tunnel geometry was employed to capture the anisotropic nature of radiation transport, and ambient dose equivalent H*(10) distributions were obtained using high-resolution USRBIN scoring.The results indicate strongly direction-dependent attenuation behavior. The concrete–FeB system provides effective neutron reduction through combined moderation and absorption mechanisms, while the concrete–soil configuration in the downward direction yields multi-order dose attenuation. In contrast, elevated dose levels persist in lateral and upward directions due to high-energy neutron components and skyshine effects. Dose values were assessed against International Commission on Radiological Protection (ICRP) limits, showing compliance in certain regions but highlighting the need for further optimization in lateral and roof shielding.To mitigate the computational cost of MC simulations, machine learning (ML)-based surrogate models were developed using MC-generated datasets. Ensemble methods, particularly Extra Trees and Random Forest, achieved high predictive accuracy (R² > 0.99) and effectively captured nonlinear, spatially varying dose behavior. The proposed hybrid MC–ML framework significantly reduces computation time while maintaining high accuracy, providing an efficient tool for the analysis and optimization of multilayer shielding systems.
- 6 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMultidirectional Neutron Shielding Analysis and Hybrid MC-ML Surrogate Modeling of Multilayer Concrete-FeB-Concrete Systems in a 250 MeV Proton Accelerator Tunnel
In high-energy proton accelerators, neutron-dominated secondary radiation fields from proton–target interactions critically govern shielding design. In this study, the transport and attenuation of secondary particles from 250 MeV proton–copper interactions were investigated within a multilayer concrete–ferroboron (FeB)–concrete shielding system using the FLUKA Monte Carlo (MC) code. A three-dimensional tunnel geometry was employed to capture the anisotropic nature of radiation transport, and ambient dose equivalent H*(10) distributions were obtained using high-resolution USRBIN scoring.The results indicate strongly direction-dependent attenuation behavior. The concrete–FeB system provides effective neutron reduction through combined moderation and absorption mechanisms, while the concrete–soil configuration in the downward direction yields multi-order dose attenuation. In contrast, elevated dose levels persist in lateral and upward directions due to high-energy neutron components and skyshine effects. Dose values were assessed against International Commission on Radiological Protection (ICRP) limits, showing compliance in certain regions but highlighting the need for further optimization in lateral and roof shielding.To mitigate the computational cost of MC simulations, machine learning (ML)-based surrogate models were developed using MC-generated datasets. Ensemble methods, particularly Extra Trees and Random Forest, achieved high predictive accuracy (R² > 0.99) and effectively captured nonlinear, spatially varying dose behavior. The proposed hybrid MC–ML framework significantly reduces computation time while maintaining high accuracy, providing an efficient tool for the analysis and optimization of multilayer shielding systems.
- Open Access
- 6 April 2026
- Published by Elsevier in SSRN Electronic Journal
A Hybrid Monte Carlo-Machine Learning Framework for Neutron Dose Attenuation Modeling in Layered Iron-Concrete Shields of a 1000 MeV Proton Accelerator
In this study, a hybrid Monte Carlo–Machine Learning (MC-ML) framework was developed to model neutron dose attenuation in a multilayer iron–concrete shielding configuration representing a 1000 MeV proton accelerator environment. High-fidelity reference dose distributions obtained from FLUKA Monte Carlo simulations were used to train and evaluate Random Forest (RF), Extra Trees (ETR), XGBoost (XGBR), Hist Gradient Boosting (HGBR), and Support Vector Regression (SVR) models independently along both the x and y directions.The results demonstrate that tree-based ensemble models achieve high predictive accuracy over a wide dynamic range. In the x-direction, the HGBR model exhibited the best performance with a test coefficient of determination of R2 = 0.996 and RMSE = 0.061. In the y-direction, RF, ETR, and XGBR produced consistent results with test R2 values of approximately 0.985. The strong agreement between the model predictions and the MC reference dose profiles indicates that the proposed hybrid approach successfully captures the complex and nonlinear characteristics of dose attenuation.From a physical standpoint, the multilayer iron-concrete structure provides a multi-stage attenuation mechanism for high-energy neutrons, leading to pronounced directional (anisotropic) behavior in the dose distributions. The proposed hybrid framework effectively reduces computational cost to a one-time MC simulation phase and enables rapid and reliable dose prediction through trained surrogate models. This approach offers a practical and efficient tool for the design and optimization of multilayer shielding systems in high-energy accelerator environments.
- 6 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Hybrid Monte Carlo-Machine Learning Framework for Neutron Dose Attenuation Modeling in Layered Iron-Concrete Shields of a 1000 MeV Proton Accelerator
In this study, a hybrid Monte Carlo–Machine Learning (MC-ML) framework was developed to model neutron dose attenuation in a multilayer iron–concrete shielding configuration representing a 1000 MeV proton accelerator environment. High-fidelity reference dose distributions obtained from FLUKA Monte Carlo simulations were used to train and evaluate Random Forest (RF), Extra Trees (ETR), XGBoost (XGBR), Hist Gradient Boosting (HGBR), and Support Vector Regression (SVR) models independently along both the x and y directions.The results demonstrate that tree-based ensemble models achieve high predictive accuracy over a wide dynamic range. In the x-direction, the HGBR model exhibited the best performance with a test coefficient of determination of R2 = 0.996 and RMSE = 0.061. In the y-direction, RF, ETR, and XGBR produced consistent results with test R2 values of approximately 0.985. The strong agreement between the model predictions and the MC reference dose profiles indicates that the proposed hybrid approach successfully captures the complex and nonlinear characteristics of dose attenuation.From a physical standpoint, the multilayer iron-concrete structure provides a multi-stage attenuation mechanism for high-energy neutrons, leading to pronounced directional (anisotropic) behavior in the dose distributions. The proposed hybrid framework effectively reduces computational cost to a one-time MC simulation phase and enables rapid and reliable dose prediction through trained surrogate models. This approach offers a practical and efficient tool for the design and optimization of multilayer shielding systems in high-energy accelerator environments.