Jump to: navigation, search

Youtube search... ...Google search ..Google News ... mapped

Daniel Trugman: Characterizing Earthquake Hazards and Source Dynamics Using Machine Learning
Dr. Daniel Trugman, Feynman Postdoctoral Fellow at Los Alamos National Laboratory, presents "Characterizing Earthquake Hazards and Source Dynamics Using Machine Learning" at the MIT Earth Resources Laboratory on May 11, 2018. "Observational seismology is an increasingly data-rich field in which data-driven machine learning techniques show significant potential. In this seminar, I focus on three specific examples from my research where simple algorithms and concepts from machine learning prove useful in characterizing earthquake source properties and hazard. First, I discuss how clustering and graph theoretical techniques can effectively be applied to large-scale differential travel time datasets to provide precise earthquake hypocentral relocations. The open-source software package GrowClust that incorporates these concepts is currently being used for large-scale relocation problems in study areas ranging from California, Nevada, and Kansas to Japan and Costa Rica. Next, describe a new framework for empirical Ground Motion Prediction Equations (GMPEs) based on a supervised learning algorithm known as a Random Forest. I then use the Random Forest GMPE to measure the influence of dynamic stress drop on the measured peak ground accelerations of moderate earthquakes in the San Francisco Bay Area. Finally, I discuss ongoing collaborative efforts to systematically analyze the P-waveform features of large magnitude earthquakes for potential signatures of nucleation and rupture onset characteristics that correlate with event size. Results from this study may yield insight into the physics of earthquake rupture and have practical implications for real-time earthquake early warning algorithms."

08FORCE Raknes Automated seismic interpretation using machine learning and field interpretations
Espen B. Raknes, Aker BP, June 13, 2018 Seismic interpretation is dependent on the experience of the interpreter and the quality of the seismic data. Therefore, interpretations include uncertainties, in particular in areas with complex geology. Today, interpretation is a task that is performed mainly by manual work resulting in a tedious work flow, often consisting of repeated tasks. With the increase in computational power over the recent decade, methods involving artificial intelligence (AI)/machine learning (ML) have been introduced with success in different disciplines for automating methods and processes. Seismic interpretation is a relatively new area for ML methods. Therefore, theory and methodology must be developed as well as tested on real world data applications. This is not a straightforward task as seismic data typically are very large data volumes, representing physical measurements significantly influenced by noise and the convolution of many different types of signals. Moreover, using human interpretations in supervised machine learning is challenging due to human errors, individual differences when interpreting, and a difficulty finding large enough training sets with few or no false negatives due to under-interpretation. We present a proof-of-concept study which purpose was to develop a machine learning work flow using human-made interpretations from fields in both exploration and production phases. The goal of this work was to create a supervised deep learning method using convolutional and fully convolutional neural networks that could automatically classify faults and horizons from 2D and 3D seismic data samples. Most of the work was performed in designing the neural networks as well as an efficient work flow starting with the interpretations that was used as input in the training of neural networks, to a result that the interpreter could use in daily work. The results show that our image recognition method is able to detect and classify faults and horizons accurately. However, when training using data from a single field, the model shows limited ability to generalize to other fields. Using real-time image augmentation when training helps reduce the over-fitting. In the continuation of this work, we propose using much larger datasets drawn from fields with different characteristics combined with data augmentation

Machine Learning in Seismic Data Processing and Interpretation - Maxim Ryabinskiy
Yandex School of Data Analysis Conference Machine Learning: Prospects and Applications There are several most frequent challenges in modern seismic explora- tion that demand the use of machine learning techniques: adjusting the forward and inverse filtering parameters during seismic data processing, pattern recognition during geological strata tracking, the data regular- ization problem, fields interpolation during geological interpretation and geological structures classification at the final interpretation stage. Solving these problems involves a great variety of artificial intelligence techniques: different modifications of regression in first-type problems, the kriging method and artificial neural networks and fuzzy logic systems in second- and third-type problems accordingly. The parameters adjusting problem demands the construction of learning sets from the modeled data, then we use iterative routines to realize the learning procedure and try to use the established parameters on real data. Commonly we use different regression models, especially the power regression technique. Machine learning is more popular in geological interpretation of seismic surveys data. Geological interpretation is an attempt to solve the inverse geophysical problem using all the available a priori information: well-log data, regional geological maps, etc. The a priori information is always insufficient, so the inverse problem is poorly conditioned and so we have to use either geologists’ experience or artificial intelligence. The main aim of the oil geologist is locating hydrocarbon traps; this problem seems like facial recognition and involves the same kits to solve it, but otherwise, the number of parameters in geological interpretation reaches several hundred. Therefore, the dimension reduction of the feature space also is an actual problem. So, applying machine learning techniques in seismic data processing and interpretation is a modern and actual branch of geology.

EAGE E-Lecture: Seismic interpretation with deep learning by Anders U. Waldeland
To learn more about what EAGE has to offer on Machine Learning, please visit the EAGE calendar of events here: How and why can Deep Learning be used for seismic interpretation? The machine learning technique called deep learning is revolutionizing the field of computer vision. A central part of deep learning is convolutional neural networks (CNN). This E-lecture gives a simple and intuitive introduction to CNNs in the context of seismic interpretation. Github link: Paper link

Phoebe Robinson DeVries: Deep learning for aftershock location patterns and the earthquake cycle
Over the past few years, deep learning has led to rapid advances in applied computer science, from machine vision to natural language processing. These methods are now accessible to scientists across all disciplines due to the availability of easy-to-use APIs and affordable GPU acceleration. We demonstrate two specific applications of deep learning within earthquake science. In the first, we train a deep neural network to learn computationally efficient representations of viscoelastic solutions, across large ranges of times, locations, and rheological structures. Once found, these efficient neural network representations may accelerate computationally intensive viscoelastic calculations by a factor of 500. In the second, we focus on aftershock location patterns and find that a fully connected neural network trained on 131,000+ mainshock-aftershock pairs can explain aftershock locations in an independent testing data set of 30,000+ mainshock aftershock pairs more accurately than static elastic Coulomb failure stress change. In contrast to the common assertion that deep learning produces “black box” results, the trained neural networks can provide some interesting physical insights. Acoustic emissions (AE) were continuously recorded and ultrasonic velocities were monitored in 20-40 s intervals using up to 16 P-wave sensors attached directly to the sample surface. Full waveforms of AEs were stored in a 16-channel transient recording system (Proekel, Germany) with a bandwidth of 16 bit and 10 MHz sampling rate. Event location is based on automatic P-wave picking and updated time-varying velocity models. We calculated full moment tensors (FMT) using P-wave amplitudes corrected for incidence angle and coupling. Spatio-temporal changes in AE event densities, fault mechanisms, local damage and stress distribution, FMT components, and magnitude-frequency distributions (b-values) from fracture tests and multiple stick-slip events were analyzed and compared to post-mortem fault structures using optical microscopy and Xray tomography. In general, we observe a correlation of fault roughness with AE hypocenter patterns indicating fault asperities. Spatio-temporal changes in AE activity and b-value distributions correlate with changing AE source mechanisms, in particular varying contributions from non-double components (NDC). Major stick-slip events associated with pronounced stress drops are spatially correlated with large AE events possibly indicating shearing of asperities along combined R- and P-type shears. Post-slip increase in b-values and increasing NDC contributions possibly indicate gouge compaction by grain crushing and sliding. Our studies suggest that the observed seismic characteristics are controlled by boundary conditions (confining pressure), sample material (porosity), and spatio-temporal changes in fault zone structure and stress heterogeneities.

Earthquake Simulator Finds Tremor Triggers
Using a novel device that simulates earthquakes in a laboratory setting, a Los Alamos researcher has found that seismic waves-the sounds radiated from earthquakes-can induce earthquake aftershocks, often long after a quake has subsided. The research provides insight into how earthquakes may be triggered and how they recur. Los Alamos researcher Paul Johnson and colleague Chris Marone at Penn State have discovered how wave energy can be stored in certain types of granular materials-like the type found along certain fault lines across the globe-and how this stored energy can suddenly be released as an earthquake when hit by relatively small seismic waves far beyond the traditional “aftershock zone” of a main quake. Perhaps most surprising, researchers have found that the release of energy can occur minutes, hours, or even days after the sound waves pass; the cause of the delay remains a tantalizing mystery.

AI Can Save You From An Earthquake
Physicist and author Louis A. Del Monte discusses a recently published finding that indicates artificial intelligence can save you from an earthquake. The new finding, recently published in Science Advances, indicates that AI is capable of detecting 17 times more earthquakes than older methods in a fraction of the time. There’s even some hope that it could predict earthquakes before they occur. This could be done by looking for patterns in the data; for example, finding times when a number of small earthquakes have happened in quick succession, triggering a bigger, potentially damaging quake. In addition to describing the new finding, Lou explains the Richter magnitude scale (often shortened to Richter scale), which is the most common standard of measurement for earthquakes. It was invented in 1935 by Charles F. Richter of the California Institute of Technology as a mathematical device to compare the size of earthquakes. The Richter scale is used to rate the magnitude of an earthquake, that is the amount of energy released during an earthquake.

Earthquake Prediction

Geophysics and Earthquake Prediction
Hank and Dr. Rebecca Bendick talk about her work in the science of earthquake forecasting, and then Jessi joins the show to show off Sandy the sand boa!

Global AI On Tour, Pune 2020: Predicting earthquake damages with Azure ML Workspace
Eva Pardi delivered a sesion on "Predicting earthquake damages with Microsoft Azure ML Workspace"

StStW Ep 5: Can You Help Seismic Researchers and AI Predict Earthquakes?
This episode explores the possibility of using machine learning to predict earthquakes more accurately. See how you can help by classifying seismic data!

Project to predict Earthquake using ML (RNN,LSTM) Microsoft CodeFunDo++
Sourav Kumar Dash In this video we are going to explain how our project will work and how much it will effect, what will be the limitation's with that.

Machine learning accurately predicts slow slip in earthquakes
Machine-learning research has detected seismic signals accurately predicting the slow slipping of the Cascadia fault. It has also found similar signals predicting slow slip failure in Chile and New Zealand. Los Alamos National Laboratory researchers applied machine learning to analyze 12 years of historic Cascadia data. Cascadia’s constant tremors produce an acoustic signal, like sound waves. The key to Cascadia’s behavior was buried in that acoustic data

LANL seismologist doing breakthrough earthquake research
Research being done in New Mexico is showing promise when it comes to addressing a worldwide concern

Why are earthquakes so hard to predict? - Jean-Baptiste P. Koehl
Take a look at the theories behind why earthquakes occur, what makes them so hard to predict and the warning system technologies we rely on today. In 132 CE, Zhang Heng presented his latest invention: a large vase he claimed could tell them whenever an earthquake occurred for hundreds of miles. Today, we no longer rely on pots as warning systems, but earthquakes still offer challenges to those trying to track them. Why are earthquakes so hard to anticipate, and how could we get better at predicting them? Jean-Baptiste P. Koehl investigates. Lesson by Jean-Baptiste P. Koehl, directed by Cabong Studios.

Predicting earthquakes via machine learning by Bikas K Chakrabarti

Data mining digs up hidden clues to major California earthquake triggers
A powerful computational study of southern California identified 10 times more earthquakes occurred, and 10 times more frequently than previously measured. Researchers were able to detect, understand, and locate quakes more precisely, and they created the most comprehensive earthquake catalog to date. The big data analysis of the entire region also revealed new information, such as triggering mechanisms and hidden foreshocks, which may help predict larger quakes.

Earthquake Prediction System
Part 2 Will Explain the Earthquake Factors In Depth, and Hypothesize the Mechanisms

Richter's predictor Modelling Earthquake Damage
In this research the basic motto is “Prediction of earthquake damage using Machine Learning” and the idea is to use existing data set of seismic activity for training and then to predict, when an earthquake will happen and how much damage will be use through it. In this study, we are using Random forest classifier for the implementation.

Kaggle: Earthquake Prediction Challenge

Forecasting earthquakes is one of the most important problems in Earth science because of their devastating consequences. Current scientific studies related to earthquake forecasting focus on three key points: when the event will occur, where it will occur, and how large it will be. In this competition, you will address when the earthquake will take place. Specifically, you’ll predict the time remaining before laboratory earthquakes occur from real-time seismic data. If this challenge is solved and the physics are ultimately shown to scale from the laboratory to the field, researchers will have the potential to improve earthquake hazard assessments that could save lives and billions of dollars in infrastructure. This challenge is hosted by Los Alamos National Laboratory which enhances national security by ensuring the safety of the U.S. nuclear stockpile, developing technologies to reduce threats from weapons of mass destruction, and solving problems related to energy, environment, infrastructure, health, and global security concerns.

MSDS696 Kaggle LANL Earthquake Prediction
Masters Degree final project report for Data Science at Regis University in Denver, CO

(  Please see the GitHub repository at: or look for my postings on the Kaggle LANL Earthquake Prediction Challenge under the user name Vettejeep. 

What is LNAL earthquake Predictions
What is LANL Earthquake Predictions..

Kaggle NYC meetup 2019 02 26 Neural Networks and earthquake detection | Jacob Peters

LANL-Earthquake-Prediction XGBoost Model Visualize
Too large to plot at one shot

Siraj Raval's

Data Science coding challenge time! The popular Data Science competition website Kaggle has an ongoing competition to solve the problem of earthquake prediction. Given a dataset of seismographic activity from a laboratory simulation, participants are asked to create a predictive model for earthquakes. In this video, I'll attempt the challenge as a way to teach 3 concepts; the Data Science mindset, Categorical Boosting, and Support Vector Regression (SVR) models. I'll be coding this using Python from start to finish in the online Google Colaboratory environment. The credits are from a mix of different Kaggle Kernels that I liked (lots), as well as my own code + explanations.This is the code for the Kaggle Earthquake Challenge videp | Siraj Raval on YouTube...

Kaggle Earthquake Prediction Challenge
Data Science coding challenge time! The popular Data Science competition website Kaggle has an ongoing competition to solve the problem of earthquake prediction. Given a dataset of seismographic activity from a laboratory simulation, participants are asked to create a predictive model for earthquakes. In this video, I'll attempt the challenge as a way to teach 3 concepts; the Data Science mindset, Categorical Boosting, and Support Vector Regression models. I'll be coding this using python from start to finish in the online Google colab environment. Enjoy!