Neutron Yield Predictions with Artificial Neural Networks: A Predictive Modeling Approach

TL;DR

This paper presents a deep neural network model trained on Monte Carlo simulations to rapidly predict neutron yields from ion-based sources, streamlining the comparison and optimization of different neutron production concepts.

Contribution

It introduces a novel predictive modeling approach using deep neural networks trained on extensive simulations to replace time-consuming Monte Carlo calculations.

Findings

Neural network model accurately predicts neutron yields.

Model significantly reduces computation time.

Addresses and mitigates model shortcomings.

Abstract

The development of compact neutron sources for applications is extensive and features many approaches. Let alone ion-based approaches, several projects with different parameters exist. This article focuses on ion-based neutron production below the spallation barrier for arbitrary light ion beams. With this model, it is possible to compare different ion-based neutron source concepts against each other quickly. This contribution derives a predictive model using Monte Carlo simulations (50k simulations) and deep neural networks. This model can skip the necessary Monte Carlo simulations, which individually take a long time to complete, increasing the effort for optimization and predictions. The models' shortcomings are addressed, and mitigation strategies are proposed.