AI predicts future slip physics in laboratory earthquakes


The seismogram is recorded by a seismograph at the Weston Observatory in Massachusetts, USA. Credit: Wikipedia

An AI approach borrowed from natural language processing — such as language translation and text autofilling on your smartphone — can predict future fault friction and the next failure time with high accuracy in laboratory earthquakes. The technology, which applies artificial intelligence to the audio signals of a fault, advances past work and goes beyond by predicting aspects of the future state of the fault’s physical system.

Chris Johnson, co-lead author of a research paper on the findings, said: Geophysical Research Letters.

Paul Johnson, corresponding author of the paper, geophysicist and lab fellow at Los Alamos National Laboratory, leads a team that has made steady progress in applying many of the machine learning Techniques for challenging earthquake prediction in the laboratory and in the field.

“The acoustic signals emitted by the laboratory error contain predictive information about the future fundamental physics of the system by Earthquake “We’ve never seen that before,” Paul Johnson said.

In a new approach, the Los Alamos team applied a deep learning transducer model of acoustic emissions beamed from a laboratory error to predict the state of friction.

“The deep learning converter model we used is synonymous with a language translation model, like Google Translate, using a code book to translate a sentence into a different language,” said Chris Johnson. “You can think of this as writing an email in English and having the AI ​​translate from English to Japanese while also anticipating your words and autofilling them at the end of the sentence.”

Chris Johnson said AI “takes data on what’s happening now and says what happens next on the wrong.”

The Los Alamos team had previously predicted the timing of failures in laboratory earthquakes and in historical slow-slip Earth data using a number of machine learning techniques. Applying machine learning to data from laboratory shear experiments showed that fault emissions are imprinted with information regarding their current state and place in the slip cycle.

In fact, the statistical properties of the fault’s continuous seismic signal identified through machine learning allowed Los Alamos researchers to predict instantaneous developing—not future—friction, friction and other features, along with the timing of the next laboratory earthquake.

In this previous work, waveform data (or acoustic emission) is fed into a model to predict the current state of a fault system. This forecast includes an estimate of the countdown, or time of failure, of the next slip event, with some degree of uncertainty, which is not a future prediction but a description of the current state of the system.

“Now we are making future prediction from past data, which goes beyond describing the instantaneous state of the system. The model learns from waveforms To predict future error friction and when the next slip will occur using only past information, without using any data from the significant future time step,” said Chris Johnson.

“The model is not constrained by physics, but it does predict the physics, and the actual behavior of the system,” Chris Johnson said.

“The next challenge is whether we can do this on Earth to predict the future Error Offset, for example, “Paul Johnson said.” This is an open question, because we don’t have such long datasets to train the model as we do in laboratory. “

The method can be applied to other disciplines, such as testing for non-destructive materials, where it can provide information about gradual damage and imminent damage to a bridge, for example.

Using scattered data to predict laboratory earthquakes

more information:
Kun Wang et al, Predicting future laboratory error friction with deep learning transformer models, Geophysical Research Letters (2022). DOI: 10.1029/2022GL098233

the quote: AI predicts future slip physics in laboratory earthquakes (2022, Oct 10) Retrieved Oct 11, 2022 from

This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.

Leave a Comment