Fractal Approach to the Regional Seismic Event Discrimination Problem

Abstract

In the framework of the Comprehensive Test Ban Treaty, development of reliable methods to discriminate between underground nuclear explosions and earthquakes at regional distances (less than 2500km) continues to be very important especially in connection with the last (in May, 1998) nuclear explosions conducted at Indian and Pakistan test sites. Since the lithosphere is a fractal, we suppose the signals, which propagate through the media, inherit its self similar' (scaling) features. We assumed that these features of explosions and earthquakes or their topological reconstructions (embeddings) have to be different. Scaling reflects correlations of more high order then it is possible to estimate by linear discriminating methods and can be used as base of non-linear discrimination. We propose to build a universal geometrical model of a seismic signal using the canon algorithm of F. Takens and to estimate scaling of the model. The scaling features were used as patterns of seismic signals for entering them into an artificial neural network. Records of nuclear explosions and earthquakes from different regions were included into the training set. The net was trained to classify types of seismic events. Results have shown 80% correct classification of the unknown signals. As additional tools for distinguishing between nuclear explosions and earthquakes we propose to use Hurst's method and the cross correlation method. Results of using these methods are demonstrated on examples of some explosions and earthquakes.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2000
Accession Number
ADP010917

Entities

People

  • D. N. Belyashov
  • I. V. Emelyanova
  • L. M. Karimova
  • M. M. Novak
  • N. G. Makarenko

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Complex Systems
  • Cross Correlation
  • Discrimination
  • Earthquakes
  • Electronic Mail
  • Embedding
  • Explosions
  • Integrals
  • Kazakhstan
  • Mathematics
  • Neural Networks
  • Nuclear Explosions
  • Pattern Recognition
  • Real Numbers
  • Technical Information Centers
  • Training

Readers

  • Educational Psychology
  • Neural Network Machine Learning.
  • Seismology

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference