Sparse Data Foreseeing Laboratory Quakes

Dmitry Spodarets
Dmitry Spodarets

A  special machine-learning concept created for sparse data reliably predicts fault slip in laboratory earthquakes. This can be essential for predicting fault slip and potentially earthquakes in the industry. The recent study by a Los Alamos National Laboratory team has evolved on their previous success using data-driven approaches that worked for slow-slip events in Earth but came up small on large-scale stick-slip faults that generated relatively little data and massive quakes.

The team trained a convolutional neural network on the output of numerical simulations of laboratory moves, as well as on a limited set of data from lab experiments. Therefore, they were able to foresee fault slips in the remaining unseen lab data.