This research is supported with funding from Institutional Support (LDRD) at Los Alamos National Laboratory including funding via the Center for Nonlinear Studies.
\nAbstract
\nWe apply machine learning to data sets from shear laboratory experiments, with the goal of identifying hidden signals that precede earthquakes. Here we show that by listening to the acoustic signal emitted by a laboratory fault, machine learning can predict the time remaining before it fails with great accuracy. These predictions are based solely on the instantaneous physical characteristics of the acoustical signal and do not make use of its history. Surprisingly, machine learning identifies a signal emitted from the fault zone previously thought to be low-amplitude noise that enables failure forecasting throughout the laboratory quake cycle. We infer that this signal originates from continuous grain motions of the fault gouge as the fault blocks displace. We posit that applying this approach to continuous seismic data may lead to significant advances in identifying currently unknown signals, in providing new insights into fault physics, and in placing bounds on fault failure times.
\nPlain Language Summary
\nPredicting the timing and magnitude of an earthquake is a fundamental goal of geoscientists. In a laboratory setting, we show we can predict \u201clabquakes\u201d by applying new developments in machine learning (ML), which exploits computer programs that expand and revise themselves based on new data. We use ML to identify telltale sounds\u2014much like a squeaky door\u2014that predict when a quake will occur. The experiment closely mimics Earth faulting, so the same approach may work in predicting timing, but not size, of an earthquake. This approach could be applied to predict avalanches, landslides, failure of machine parts, and more.
\n1\u00a0Introduction
\nA classical approach to determining that an earthquake may be looming is based on the interevent time (recurrence interval) for characteristic earthquakes, earthquakes that repeat periodically (Schwartz & Coppersmith,\u00a01984). For instance, analysis of turbidite stratigraphy deposited during successive earthquakes dating back 10,000 years suggests that the Cascadia subduction zone is ripe for a megaquake (Goldfinger et al.,\u00a02017). The idea behind characteristic, repeating earthquakes was the basis of the well-known Parkfield prediction based strictly on seismic data. Similar earthquakes occurring between 1857 and 1966 suggested a recurrence interval of 21.9\u00a0\u00b1\u00a03.1\u00a0years, and thus, an earthquake was expected between 1988 and 1993 (Bakun & Lindh,\u00a01985), but ultimately took place in 2004.
\nWith this approach, as earthquake recurrence is not constant for a given fault, event occurrence can only be inferred within large error bounds. Over the last 15\u00a0years, there has been renewed hope that progress can be made regarding forecasting owing to tremendous advances in instrumentation quality and density. These advances have led to exciting discoveries of previously unidentified slip processes that include slow slip (Melbourne & Webb,\u00a02003), low frequency earthquakes and Earth tremor (Brown et al.,\u00a02009; Obara,\u00a02002; Shelly et al.,\u00a02007) that occur deep in faults. These discoveries inform a new understanding of fault slip and may well lead to advances in forecasting, impending fault failure if the coupling of deep faults to the seismogenic zone can be unraveled.
\nThe advances in instrumentation sensitivity and density also provide new means to record small events that may be precursors. Acoustic\/seismic precursors to failure appear to be a nearly universal phenomenon in materials. For instance, it is well established that failure in granular materials (Michlmayr et al.,\u00a02013) and in avalanche (Pradhan et al.,\u00a02006) is frequently accompanied by impulsive acoustic\/seismic precursors, many of them very small. Precursors are also routinely observed in brittle failure of a spectrum of industrial (Huang et al.,\u00a01998) and Earth materials (Jaeger et al.,\u00a02007; Schubnel et al.,\u00a02013). Precursors are observed in laboratory faults (Goebel et al.,\u00a02013; Johnson et al.,\u00a02013) and are widely but not systematically observed preceding earthquakes (Bouchon et al.,\u00a02013,\u00a02016; Geller,\u00a01997; McGuire et al.,\u00a02015; Mignan,\u00a02014; Wyss & Booth,\u00a01997).
\nSeismic swarm activity which exhibits very different statistical characteristics than classical impulsive precursors may or may not precede large earthquakes but can mask classical precursors (e.g., Ishibashi,\u00a01988).
\nThe International Commission on Earthquake Forecasting for Civil Protection concluded in 2011 that there was \u201cconsiderable room for methodological improvements in this type of (precursor-based failure forecasting) research\u201d (International Commission on Earthquake Forecasting for Civil Protection,\u00a02011: Jordan et al.,\u00a02011). The commission also concluded that published results may be biased toward positive observations.
\nWe hypothesize that precursors are a manifestation of critical stress conditions preceding shear failure. We posit that seismic precursor magnitudes can be very small and thus frequently go unrecorded or unidentified. As instrumentation improves, precursors may ultimately be found to exist for most or all earthquakes (Delorey et al.,\u00a02017). Furthermore, it is plausible that other signals exist that presage failure.
\nRead the source article at AGUPublications.com
\nSource: AI Trends<\/p>\n","protected":false},"excerpt":{"rendered":"This research is supported with funding from Institutional Support (LDRD) at Los Alamos National Laboratory including funding via…\n","protected":false},"author":1,"featured_media":3501,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","footnotes":""},"categories":[811],"tags":[812],"class_list":{"0":"post-3500","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai"},"yoast_head":"\n