Remote alpha/beta measurement system with support vector machine algorithm

Decommissioning nuclear reactor sites presents challenges due to the presence of various radionuclides, including alpha emitters (e.g., Pu, Am, Cm) and beta emitters (e.g., ¹³⁷Cs, ⁹⁰Sr–⁹⁰Y), which pose significant internal exposure risks to workers. Traditional measurement methods require multiple instruments and are time-consuming, particularly in high gamma-ray environments. To address these issues, we developed a remote alpha and beta discrimination measurement system that integrates a stilbene scintillator detector with a silicon photomultiplier, enabling simultaneous detection of both alpha and beta particles. This study further incorporates machine learning techniques, specifically Support Vector Machines (SVM), for automatic discrimination, eliminating the need for user-defined thresholds and ensuring consistent operational conditions. The system was tested with known radiation sources, demonstrating over 96 % classification accuracy for alpha and beta particles. Measurements conducted in motion effectively identified contamination sources, confirming the system's capability for real-time analysis. This innovative approach enhances radiation safety and efficiency in nuclear decommissioning operations, making it particularly beneficial in environments where human access is limited.