Operationalizing nuclear safety: A remote α/β detection system integrating CNN and radioluminescence for emergency...

Area of Science:

• Nuclear Physics
• Radiation Detection
• Artificial Intelligence

Background:

• Discriminating alpha/beta radioactive materials is critical for public health and environmental safety during nuclear emergencies.
• Conventional detection methods face limitations like signal interference and short detection ranges, hindering rapid and precise identification.

Purpose of the Study:

• To develop an innovative remote alpha/beta discrimination technique using Convolutional Neural Networks (CNN).
• To improve the efficiency and reliability of radioactive material detection in nuclear accident scenarios.

Main Methods:

• A multilayer CNN model was developed, accepting 2D alpha/beta imaging maps as input.
• Particle transport simulations generated 100,000 samples for model training and testing.
• The model was trained on 80,000 datasets and tested on 20,000 datasets.

Main Results:

• The CNN model achieved a 99.1% accuracy rate in discriminating alpha/beta radioactive materials.
• Position coordinate predictions showed a relative error of less than 10% for 72.4% of samples.
• Experimental validation confirmed the model's efficacy and reliability.

Conclusions:

• The proposed CNN-based remote sensing technique offers a significant advancement for radioactive substance safety monitoring.
• This technology enhances detection efficiency and has the potential to mitigate hazards associated with nuclear accidents.