Neural network based geomagnetic estimation for multi-site observation system

• Katori Yuta

  Graduate School of System Design, Tokyo Metropolitan University

• Okubo Kan

  Graduate School of System Design, Tokyo Metropolitan University

Abstract

For years, researchers have been studying the relation between earthquakes and geomagnetic field signals using observation results. Following our continuous geomagnetic observation, this report describes geomagnetic signal changes generated by earthquakes and tsunami waves. Results show that detection of their occurrence using geomagnetic field measurement is effective for providing an early alarm system for disaster mitigation. Robust detection requires the robust estimation of a geomagnetic field using multiple observation results. This study introduced a deep neural network (DNN) to estimate geomagnetic fields. Results clarified that the proposed DNN model using data of multiple axes at multiple observation points as input data provides efficient solutions for geomagnetic estimation.

Journal

• IEICE Communications Express

  IEICE Communications Express 7 (10), 352-357, 2018

  The Institute of Electronics, Information and Communication Engineers

Keywords

• time-series geomagnetic estimation
• anomaly detection
• neural network
• autoencoder
• multi-site observation

Details

• CRID: 1390845713003823360
• NII Article ID: 130007494880
• DOI: 10.1587/comex.2018xbl0017
• ISSN: 21870136
• Web Site: https://www.jstage.jst.go.jp/article/comex/7/10/7_2018XBL0017/_pdf
• Text Lang: en
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed