查看更多>>摘要：By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Current study results on Machine Learn ing have been published. According to news reporting out of Stanford, California , by NewsRx editors, research stated, "During the 3 month long eruption of Kilau ea volcano, Hawaii in 2018, the pre-existing summit caldera collapsed in over 60 quasiperiodic failure events. The last 40 of these events, which generated Mw > 5 very long period (VLP) earthquakes, had inter-event times between 0.8 and 2.2 days." Financial support for this research came from National Science Foundation (NSF). Our news journalists obtained a quote from the research from Stanford University , "These failure events offer a unique data set for testing methods for predicti ng earthquake recurrence based on locAlly recorded GPS, tilt, and seismicity dat a. In this work, we train a deep learning graph neural network (GNN) to predict the time-to-failure of the caldera collapse events using only a fraction of the data recorded at the start of each cycle. We find that the GNN generalizes to un seen data and can predict the timeto- failure to within a few hours using only 0 .5 days of data, substantiAlly improving upon a null model based only on inter-e vent statistics. Predictions improve with increasing input data length, and are most accurate when using high-SNR tilt-meter data. Applying the trained GNN to s ynthetic data with different magma-chamber pressure decay times predicts failure at a nearly constant stress threshold, revealing that the GNN is sensing the un derling physics of caldera collapse. These findings demonstrate the predictabili ty of caldera collapse sequences under well monitored conditions, and highlight the potential of machine learning methods for forecasting real world catastrophi c events with limited training data. Plain Language Summary During the summer of 2018, Kilauea volcano, Hawaii, experienced a dramatic series of large earthquak es, coinciding with the collapse of the summit caldera in a series of repeated f ailure events. These collapse events occurred periodicAlly, with inter-event tim ings between 0.8 and 2.2 days. Because of the significance of this event, there is interest to understand more about the dynamics of this collapse sequence. We study whether observational measurements of deformation recorded on the surface of the volcano carry signatures that indicate the timing of the upcoming collaps e events. By using machine learning, we train a series of models that aim to pre dict the time-to-failure of each cycle based on the observed deformation data, a nd we experiment with using different combinations of input data sets. We find o ur models can accurately predict the timing of most collapse events to within a few hours, including for events that the models were never trained on and that h ave longer durations than the training events."
NETL
NSTL