Development of a machine learning model for neutron spectrum unfolding

Area of Science:

• Nuclear physics
• Computational science

Background:

• Neutron spectrometry is crucial for nuclear applications but faces challenges due to the ill-posed unfolding problem.
• Traditional methods like Monte Carlo simulations and iterative techniques have limitations including computational cost and sensitivity to noise.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for neutron spectral unfolding.
• To assess the accuracy and generalization capabilities of ML-based unfolding compared to conventional methods.

Main Methods:

• A ML model was developed and trained using the International Atomic Energy Agency (IAEA) neutron spectra compendium.
• The model was used to perform spectral unfolding from Bonner sphere spectrometer count rates.

Main Results:

• The ML-based unfolding approach achieved high accuracy in reconstructing neutron energy spectra.
• Simulation results showed strong generalization capabilities, with reconstructed spectra closely matching reference benchmarks.

Conclusions:

• Machine learning presents a robust and efficient alternative to traditional neutron spectrum unfolding techniques.
• The developed ML model demonstrates significant potential for accurate and reliable neutron spectrum analysis.