Explosives detection using prompt-gamma neutron activation and neural networks

Area of Science:

• Nuclear Physics and Spectroscopy
• Artificial Intelligence and Machine Learning
• Materials Science and Security

Background:

• Accurate detection of explosives is critical for security applications.
• Prompt gamma-ray spectroscopy offers a non-destructive method for material identification.
• Challenges exist in detecting explosives when they are occluded by other materials.

Purpose of the Study:

• To investigate the application of neural networks for explosive detection using neutron capture prompt gamma-ray spectra.
• To evaluate the capability of neural networks to identify explosives despite occluding materials.
• To assess the generalization ability of neural networks in explosive pattern recognition.

Main Methods:

• Simulating neutron capture prompt gamma-ray spectra using the Monte Carlo N-particle transport code (MCNP4B).
• Training neural networks on simulated gamma-ray spectra to recognize explosive patterns.
• Testing the trained neural network's performance in identifying C-4 explosives under various occlusion scenarios.

Main Results:

• The trained neural network successfully identified the presence of C-4 explosive.
• Detection remained effective even when the explosive was occluded by several materials.
• The neural network demonstrated generalization capabilities, identifying explosives in untrained scenarios.

Conclusions:

• Neural networks are powerful tools for recognizing prompt gamma-ray explosive patterns, even with occluding materials.
• The developed approach shows potential for in situ inspection systems for explosive detection.
• The ability to generalize highlights the robustness of the neural network for security applications.