Current systems rely on preset thresholds (e.g., if SPL > 210 dB, sound alarm). Researchers are training convolutional neural networks (CNNs) on spectrograms of deflagration vs. detonation events to automatically classify disposal events in real-time, distinguishing them from background noise (shipping, seismic surveys, biological sounds).
Even in a pure deflagration, hot gases expand rapidly. As the bubble collapses, it can generate secondary pressure pulses that, while far weaker than a detonation bubble pulse, may still be measurable. Characterisation must distinguish between the primary combustion source and secondary cavitation collapse. Current systems rely on preset thresholds (e
is a subsonic combustion process (flame front velocity typically < 1000 m/s, often < 100 m/s). In a controlled deflagration (used in commercial "low-order" systems like the RA-9 or EOD Robot Deflagration Systems), the energetic material is heated rapidly but not shocked into detonation. The burn propagates through the explosive filler via thermal conduction. The resulting gas release is relatively slow, generating a pressure pulse that lacks a distinct shock front. Underwater, this manifests as a longer-duration, lower-amplitude "thump" rather than a crack. Even in a pure deflagration, hot gases expand rapidly