Abstract
BACKGROUND: Designing shields for gamma radiation sources is particularly important due to their extensive use in medical, industrial, and research studies. OBJECTIVE: This study aimed to explore the ability of an Artificial Neural Network (ANN) to identify the optimized shield for a typical gamma source. Despite the effectiveness of Monte Carlo simulations in determining optimal shielding materials and geometries, they are time-consuming and require numerous simulations for each configuration. MATERIAL AND METHODS: In this simulating study, the MCNPX Monte Carlo code was utilized to conduct simulations using a previously proposed shielding material. After validating the simulation accuracy, a large dataset was generated to serve as input and target data for the machine learning process. The method's precision was assessed by comparing the results of the ANN with those of Monte Carlo simulations. Dose calculations were performed using a water phantom. RESULTS: The deviation of less than 1% was computed between the simulation and the ANN. The network also exhibited satisfactory predictions for unknown data. Additionally, the dose was evaluated using a water phantom to assess further and optimize the selected shielding material. CONCLUSION: The ANNs are widespread and significant in radiation shielding studies. The developed network can accurately predict unknown weight fraction combinations. The designed network can effectively predict unknown weight fraction combinations.