Paul Johnson's research team at Los Alamos National Laboratory is applying artificial intelligence to earthquake prediction, using machine-learning algorithms, supercomputers, and big data storage and analysis.
The team is feeding machines vast datasets of measurements taken continuously before, during, and after lab-simulated earthquakes. The algorithm mines the data for patterns that reliably signal when an artificial quake will occur.
The team also has started applying the machine-learning analysis to raw seismic data from actual quakes. Via experiments run on Pennsylvania State University's earthquake simulator, the researchers found the algorithm detected reliable "creaking and grinding" noises that change in a very specific way as the artificial tectonic system gets closer to a simulated quake.
"Not only could the algorithm tell us when an event might take place within very fine time bounds--it actually told us about physics of the system that we were not paying attention to," Johnson says.
From Scientific American
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found