Self-Organizing Map (SOM) of Space Acceleration Measurement System (SAMS) data

Alok Sinha, A. D. Smith

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

In this paper, space acceleration measurement system (SAMS) data have been classified using self-organizing map (SOM) networks without any supervision; i.e., no a priori knowledge is assumed regarding input patterns belonging to a certain class. Input patterns are created an the basis of power spectral densities of SAMS data. Results for SAMS data from STS-50 and STS-57 missions are presented. Following issues are discussed in details: impact of number of neurons, global ordering of SOM weight vectors, effectiveness of a SOM in data classification, and effects of shifting time windows in the generation of input patterns. The concept of 'cascade of SOM networks' is also developed and tested. It has been found that a SOM network can successfully classify SAMS data obtained during STS-50 and STS-57 missions.

Original languageEnglish (US)
Pages (from-to)78-87
Number of pages10
JournalMicrogravity Science and Technology
Volume12
Issue number2
StatePublished - Jan 1 1999

All Science Journal Classification (ASJC) codes

  • Modeling and Simulation
  • Engineering(all)
  • Physics and Astronomy(all)
  • Applied Mathematics

Fingerprint Dive into the research topics of 'Self-Organizing Map (SOM) of Space Acceleration Measurement System (SAMS) data'. Together they form a unique fingerprint.

Cite this