Conveners
T3.6 e-poster session: e-poster session - T3.6 - Artificial Intelligence and Machine Learning
- Vera Miljanovic Tamarit (CTBTO Preparatory Commission, Vienna, Austria)
T3.6 e-poster session: T3.6 - Artificial Intelligence and Machine Learning
- Wolfgang Sommerer (CTBTO Preparatory Commission, Vienna, Austria)
- Vera Miljanovic Tamarit (CTBTO Preparatory Commission, Vienna, Austria)
Description
T3.6
Recent advances in convolutional neural networks (CNNs) have brought impressive detection capabilities to one- and three-component seismic stations. Still, the highest sensitivity to repeating events is obtained by beamforming signals over a seismic array. We propose a new neural network architecture that combines the two, by introducing a two-dimensional convolutional layer that encodes the...
According to OSI Operation Manual, IT/ISP living and working areas should be well-protected. Scenarios like the management of the different living and working areas for IT and ISP, require entry permission granted separately to either IT or ISP members. This work would provide a customized management supporting system solution to the above mentioned scenario. The system is based on Artificial...
Advancements in AI/ML are creating a paradigm shift in virtually every sector of the tech industry. Among the endless applications, AI vision technology based on Deep Neural Network, finds its strength at image processing, pattern recognition and image interpretation, which can be utilized for manufacturing, medical diagnosis, and OSI. Current OSI search logic relies on finding and identifying...
In this study, we aim to develop a new approach using machine learning and data mining algorithms to estimate the activity concentration of radioxenon isotopes of any unknown sample without extensive mathematical calculations from calibrated raw spectra. So far, several methods have been applied such as the region-of-interest (ROI) and the simultaneous decomposition analysis tool (SDAT) to...
Large aftershock sequences cause problems for the International Data Centre (IDC) because the seismic event rate increases dramatically during an aftershock sequence, making correct association of arrivals difficult for the automated pipeline. Aftershock sequences can continue for days or even months after a large earthquake and although aftershocks aren’t events of interest for treaty...
Automatic recognition of seismic event source has been a primordial task since the introduction of digital seismic networks. Nowadays, this task becomes more important due to the huge quantity of data recorded continually and the need for real time results. Different approaches have been addressed in the literature. Currently, artificial intelligence techniques have attracted increasing...
Phase arrival time estimation for tele-seismic signals is a critical and fundamental step in the detection and localization of nuclear explosions and seismic study in general. Typically, this process involves heavy human interaction with more than half of all automatically detected arrivals being manually re-timed by a human analyst. Developments in Artificial Intelligence and specifically in...
One of the implementations to support the CTBT instrument measurement is radionuclide identification. An automatic real-time identification radionuclide can be an option for some applications, including monitoring of environmental contamination and prevention of nuclear terrorism. This research is about the automatic algorithms that provide feedback about the presence of any radiations...
Three-component stations traditionally rely on polarization analysis to estimate the backazimuth of each arriving wave. Unfortunately, these polarization estimates suffer from both high error and low confidence, and contribute very little to the downstream association algorithms at the IDC. Here, we present BazNet, a deep neural-network-based backazimuth predictor for three-component stations....
Seismic waveform data are generally contaminated by noise from various sources. To date, the most common noise suppression methods have been based on frequency filtering. These methods, however, are less effective when the signal of interest and noise share similar frequency bands. We implemented a seismic denoising method that uses a trained deep convolutional neural network (CNN) model. In...
In this work, a committee machine was used to combine supervised and unsupervised artificial neural networks to discriminate between Earthquakes and quarries blasts. The unsupervised network is used as a measure of accuracy for the results of the supervised neural network. The unsupervised Self-Organized Map (SOM) and the k-means clustering algorithms are used to estimate support and...
Typically, data-driven learning works best when we can exploit expectations from our data domain. For example, the development of recurrent neural network architectures to deal with the temporal dependence in language, geometric deep learning for 3-D problems, and physics-constrained Bayesian learning for more interpretable dependencies. Yet it can be unclear how to interject expectations, and...
Discriminating between explosions and earthquakes is necessary for building hazard maps and monitoring applications. Previous studies have used classical ML techniques based on the amplitudes of various phases. More recent methods based on Deep Learning use the full seismic waveform; however, they rely on detections made by nearby stations. These methods are inapplicable for global-scale...
Discrimination between earthquakes and explosions is an essential component of nuclear test monitoring. Discrimination methods currently used by the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) are often ineffective for regional events, particularly in Israel’s region.
For instance, five seismic events whose epicenters lie near the Sea of Galilee (Lake Kinneret) were reported...
The environment of infrasound stations is characterized by mesoscale wind speed and temperature fluctuations that affect the temporal variability of the Atmospheric Boundary Layer (ABL). While the statistical characteristics of turbulence are poorly constrained, modeling such statistics appears to be crucial since each sensor of infrasound stations is subject to this local noise that may mask...
The International Monitoring System (IMS) includes waveform sensor stations connected to a centralized processing system in the International Data Center (IDC) in Vienna. While the performance of the IMS is known to be related to atmospheric properties, the usual approach at the IDC still relies on expert judgments and simple models to incorporate the environmental knowledge. In this work, we...
An event detection method based on deep neural network combined with the average wave speed ratio of multiple stations is proposed for detecting local events under the global sparse seismic network. Firstly, the method uses multi-task convolution neural network to detect and identify the direct P and S phases, as well as estimate their arrival time. Then a joint network of GAN and LSTM is used...
NET-VISA has benefitted tremendously from the interaction between its developers and CTBT State Signatories experts. One way that this interaction has taken place is through the delivery of data sets processed with successively enhanced versions of the software, with feedback from the experts. We present the results of a full simulation of operational results conducted offline with a recent...
Time-frequency distributions (TFDs) are useful tools for nonstationary signals analysis. Due to the presence of unwanted cross-terms, useful information extraction from TFDs has proven to be a challenging task, in particular when analysing noisy real-life signals.
One way to suppress the cross-terms is by employing compressive sensing methods that enforce sparsity in the resulting TFD. In...
Machine learning has advanced radically over the past 10 years, and machine learning algorithms now achieve human-level performance or better on a number of tasks. Machine learning techniques have been extensively deployed for a variety of applications in different areas of life. The success of machine learning algorithms has led to an explosion in demand.
Machine learning models are also...
Manual identification of seismic events in long and signal-rich records is a challenging and time-consuming task. Power detectors for single stations or array beams are widely just but often provide a vast number of ungrouped events. The need for screening these events arises for example when no a priori information about expected events is available, precise locations cannot be obtained, or...
