Machine Learning-Assisted Optimization of Modular Neutron Shielding Based on Monte Carlo Simulations

Li-Fang Chen · 2025

Read Machine Learning-Assisted Optimization of Modular Neutron Shielding Based on Monte Carlo Simulations by Li-Fang Chen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-17319.txt.

Introduction

Li-Fang Chen wrote a paper in two thousand twenty-five about designing neutron shielding with help from machine learning. Neutron beams serve science but pose radiation risks, so shutters must block them during maintenance. Testing each shield design with full physics simulation is slow and costly.

Chen simulated two hundred shield layouts with varied material order using a standard Monte Carlo code. A neural network learned to predict the neutron flow past each shield. The network then ranked one thousand fresh random layouts in a flash, and the twenty best were checked with full simulation. The top design cut neutron flow by about ten thousand times from entrance to exit. The study shows how learning models can speed shield design.

Catalog record

CREATORLi-Fang Chen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-17319

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Machine Learning-Assisted Optimization of Modular Neutron Shielding Based on Monte Carlo Simulations Li-Fang Chen Abstract This study proposes a novel design methodology for neutron beam shutters that integrates Monte Carlo simulations (MCNP) with machine learning techniques to enhance shielding performance and accelerate the design process. The target facility is a compact neutron science platform where neutrons are produced by proton beams from a cyclotron striking a neutron production target. A beam shutter is installed on the thermal neutron line to reduce occupational radiation exposure during maintenance activities. In this work, 200 neutron shutter configurations with varying material sequences were simulated using MCNP. The resulting dataset was used to train a fully connected neural network to predict the neutron flux downstream of the shielding. The trained model was subsequently applied to 1,000 randomly generated shielding configurations for rapid flux prediction and performance ranking. The 20 designs with the lowest predicted flux were selected and further validated via MCNP simulations. Results show that the optimal design reduces the neutron flux from 5.61 × 10 9 5.61\times 10^{9} n/cm 2 ⋅ {}^{2}{\cdot} s at the shutter entrance to 4.96 × 10 5 4.96\times 10^{5} n/cm 2 ⋅ {}^{2}{\cdot} s at the exit, achieving a reduction of four orders of magnitude. These findings confirm that the integration of machine learning techniques can effectively reduce simulation costs and assist in identifying high-performance shielding configurations, demonstrating the strong potential of data-driven approaches in neutron system design. Keywords: neutron beam shutter, MCNP simulation, machine learning, shielding design, cyclotron-based neutron source 1 Introduction A compact neutron source system based on a cyclotron accelerator has been established in Taiwan, where high-energy protons impinge on a neutron-generating target to produce neutrons. Originally developed for radiopharmaceutical production, the facility has gradually expanded into a multifunctional platform for neutron-based research and applications. The design and operation of such facilities require comprehensive simulations—ranging from proton beam transport and target configuration to neutron propagation and radiation shielding—to ensure performance and safety in compliance with international standards. A representative simulation study previously evaluated the cadmium ratio in a thermal neutron radiography system at this facility [1]. The facility currently operates two major neutron beamlines: one for thermal neutrons and one for fast neutrons. The latter has recently incorporated a quasi-monoenergetic neutron (QMN) source driven by a 30 MeV cyclotron. Utilizing a beryllium target and supported by Monte Carlo simulations[2, 3], this system has demonstrated promising neutron yield and engineering feasibility through detailed thermal and flux modeling[4]. To ensure radiation safety during maintenance or beam-off conditions, a beam shutter is currently under planning and design for the thermal neutron beamline to block neutron transmission and reduce occupational exposure. As a critical component of the neutron beam control system, the beam shutter must meet several engineering requirements, including high shielding efficiency, mechanical modularity, and structural replaceability. During shutdown states, the shutter must provide reliable neutron attenuation to safeguard downstream instruments and maintenance personnel. However, the configuration and layering of shielding materials exhibit highly nonlinear effects on both neutron flux and energy spectra. These challenges are exacerbated in mixed radiation fields, where both neutrons and photons are present. Given the significant differences in attenuation behavior and interaction mechanisms across particle types, shielding design must account for spectral characteristics of all radiation species. As such, relying solely on empirical approaches is insufficient for identifying optimal shielding configurations. Recent studies have investigated beam shutter design strategies in various applications [5–7], offering useful guidance for future development. To date, most shutter designs rely on Monte Carlo methods, such as MCNP, for evaluating shielding performance and conducting parameter sweeps. For instance, Didi, et al. designed a multilayered beam shutter for the Prompt Gamma Neutron Activation Analysis (PGAA) system at the TRIGA Mark II reactor and employed MCNP simulations to model the neutron attenuation and energy spectral changes through the shutter, highlighting its importance for radiation safety during system maintenance [8]. Although such simulations are highly accurate, their computational cost and time requirements become substantial when exploring high-dimensional design spaces, thereby limiting optimization efficiency. In recent years, machine learning (ML) has emerged as a powerful tool in scientific computing, particularly in applications requiring the extraction of complex patterns from large datasets. In their landmark review, LeCun et al. introduced deep learning as a multi-layered nonlinear transformation framework capable of learning abstract feature representations from high-dimensional data, thus providing a theoretical foundation for its application in engineering and physical modeling [9]. Alzubaidi et al. summarized the evolution of convolutional neural network (CNN) architectures in image recognition and emphasized their strong generalization capability in predictive modeling tasks [10]. Uddin et al. provided a taxonomy of supervised, unsupervised, and hybrid learning models, and reviewed their applications in medicine, materials, and cybersecurity [11]. Leveraging ML models enables the prediction of shielding performance from limited simulation data, facilitating rapid screening and accelerated optimization of neutron shielding designs. This study proposes a machine learning-assisted framework for neutron beam shutter design, integrating MCNP simulation data with a fully connected neural network (FCNN). We first generated a training dataset comprising 200 randomly assembled multilayer shielding configurations. The trained model was then applied to predict neutron flux performance for 1,000 candidate designs, from which the best-performing cases were selected and re-validated through MCNP simulation. The results demonstrate that data-driven ML models can significantly reduce simulation cost and human effort, thereby offering high potential for efficient and intelligent shielding design. 2 Methods 2.1 Design Parameters of the Neutron Beam Shutter The neutron beam shutter is positioned 170 cm downstream from the neutron target. The shielding layers are composed of polyethylene, borated polyethylene, iron, and lead, designed to attenuate neutron flux and radiation dose effectively. Following the concept of using standardized modular components, each shielding block was fabricated with dimensions of 7 cm in length, 13 cm in height, and 2 cm in width. A total of 25 blocks were assembled sequentially, with the final layer fixed as a 1 cm thick iron plate. This final iron layer not only contributes to neutron attenuation but also serves as an activation target for neutron spectrum and flux characterization through neutron activation analysis (NAA). The overall shielding section of the shutter measures 49 cm in total length and is housed within a removable stainless steel frame. This modular frame allows for rapid disassembly and relocation to a high-radiation laboratory when necessary, enabling testing and replacement of shielding materials. This design supports both spectral shaping of the neutron beam and replacement of radiation-degraded shielding components. Figure 1 presents a simplified schematic of the beam shutter assembly. The design includes a modular shielding cartridge mounted inside a removable steel frame, with an access panel for maintenance and replacement. The vertical actuator mechanism (component 1) controls shutter movement, while the shielding cartridge (component 2) holds the 25-layer arrangement. The transparent plate (component 3) illustrates the rear access interface for shutter replacement and testing. Figure 1: Schematic diagram of the modular beam shutter design. The figure shows the mechanical layout of the beam shutter system, including the actuator assembly (1), modular shielding block array (2), and detachable rear interface plate (3). The design supports modular maintenance and material replacement under high-radiation operation conditions. 2.2 MCNP Simulation Design and Data Generation

Read the complete text