A physics-guided digital twin framework for synthetic gamma-ray spectrum generation and validation in radiation monitoring applications

Nosenko, Tetiana та Zabulonov, Yuii (2026) A physics-guided digital twin framework for synthetic gamma-ray spectrum generation and validation in radiation monitoring applications Simulation Modelling Practice and Theory (153). ISSN 1569190X

Текст Nosenko_Zabulonov_Digital_Twin_AAM_2026.pdf Download (2MB)

Анотація

The limited availability of labelled gamma-radiation data constrains the development and evaluation of machine-learning methods for radiation-monitoring applications, particularly under shielded or otherwise difficult-to-reproduce conditions. This study presents a physics-guided Digital Twin framework integrating Geant4-based Monte Carlo radiation transport, automated scenario orchestration, detector-response aggregation, structured synthetic-data generation, and multiple explicitly separated levels of assessment. Eight radionuclide-labelled source classes were represented by characteristic monoenergetic gamma emissions ranging from 0.051 to 2.610 MeV. Simulations covered four shielding materials-lead, stainless steel, concrete, and polyethylene-at thicknesses of 0, 1, 2, 5, and 10 cm, resulting in 160 source-material-thickness configurations. The simulated detector responses reproduced the expected dependence of photon transmission on energy, shielding material, and absorber thickness. After incorporating counting-statistical uncertainties, the fitted linear attenuation coefficient for 1.25-MeV 60Co-labelled photons in lead was 0.618 ± 0.042 cm-1, with a 95% confidence interval of 0.535 - 0.700 cm-1. The corresponding NIST XCOM reference value of 0.680 cm-1 lies within this interval. A limited detector-level experimental assessment was performed using independently acquired NaI(Tl) spectra. A multi-peak blind spectrum was assigned to material containing the 232Th decay chain before disclosure of the source identity, and the assignment was subsequently confirmed by the data provider. This result is reported as a proof-of-concept case rather than a statistical estimate of identification accuracy. An independent UAV radiation-survey dataset containing 3,722 observations across ten flight transects was additionally analysed. It demonstrated a strong and spatially stable relationship between integral gamma activity and estimated 137Cs activity r=0.990, R2=0.980. These UAV results provide an application-level reference but are not interpreted as direct validation of Geant4-generated spectra. The proposed framework provides a transparent and extensible basis for physics-guided synthetic-data generation while explicitly distinguishing attenuation verification, limited detector-level spectral assessment, and application-level consistency analysis.

Тип елементу :	Стаття
Ключові слова:	Digital twin; Geant4; Monte Carlo simulation; Synthetic data generation; Gamma-ray spectra; Radiation monitoring; Verification and validation; Physics-guided modelling
Типологія:	Статті у базах даних > Scopus > У виданнях Q1 Scopus
Підрозділи:	Факультет інформаційних технологій та математики > Кафедра комп'ютерних наук
Користувач, що депонує:	доцент Тетяна Іванівна Носенко
Дата внесення:	24 Жов 2026 19:18
Останні зміни:	24 Жов 2026 19:18
URI:	https://elibrary.kubg.edu.ua/id/eprint/59782

Actions (login required)

Перегляд елементу