Computers can be trained to better detect distant nuclear explosions, chemical explosions and volcanic eruptions by learning from artificial explosion signals, a new study has found.
The study was published in the journal ‘Geophysical Research Letters’.
Witsil, at the Geophysical Institute’s Wilson Alaska Technical Center, and his colleagues created a library of synthetic infrasound burst signals to train computers in recognizing the source of the infrasound signal. Infrasound has a very low frequency that cannot be heard by humans and travels much longer than high-frequency audible waves.
“We used modeling software to generate 28,000 synthetic infrasound signals, which are computer generated but can hypothetically be recorded by infrasound microphones positioned hundreds of kilometers from a large explosion,” said Witsil. “
Artificial signals reflect variations in atmospheric conditions, which can change the signal of an explosion regionally or globally as sound waves propagate. Those changes can make it difficult to trace the origin and type of an explosion very far away.
Why create artificial sounds of explosions instead of using real-world examples? Since explosions have not occurred everywhere on the planet and the atmosphere is constantly changing, there are not enough real-world examples to train generalized machine-learning detection algorithms.
“We decided to use synthetics because we can model many different types of atmospheres through which signals can propagate,” Vitsil said. “So even though we don’t have access to any eruptions that happened in North Carolina, for example, I can use my computer to model the North Carolina eruptions and use machine-learning algorithms to detect explosion signals there.” I can build.”
Today, detection algorithms typically rely on infrasound arrays consisting of several microphones placed close to each other. For example, the International Comprehensive Test Ban Treaty Organization, which monitors nuclear explosions, has deployed infrasound arrays around the world.
“It’s expensive, difficult to maintain, and can break a lot,” Vitsil said. Vitsil’s method improves detection, using hundreds of single-element infrasound microphones already in place around the world. This makes detection more cost-effective.
The machine-learning method expands the usefulness of single-element infrasound microphones, enabling them to detect more subtle burst signals in near real time. Single-element microphones are currently only useful for retrospective analysis of known and generally high-amplitude signals, as they did with the massive eruption of Tonga volcano in January.
Vitsil’s method can be deployed in an operational setting for national defense or mitigation of natural hazards.