Generalized neural network approach to eddy current inversion for real cracks

Noritaka Yusa, Weiying Cheng, Zhenmao Chen, Kenzo Miya

Research output: Contribution to journalArticlepeer-review

26 Citations (Scopus)

Abstract

This paper proposes a generalization for the solutions of an inversion method developed by the authors. The method is based upon an artificial neural network that simulates mapping between eddy current signals and crack profiles. One of the biggest advantages of the approach is that it can deal with conductive cracks, which is necessary to reconstruct natural cracks. However, it has one significant disadvantage: the reliability of reconstructed profiles was unknown. This paper also proposes a novel parameter that provides an index for assessment of the crack profile and overcomes this disadvantage. After the parameter is validated by reconstruction of simulated cracks, it is applied to reconstruction of natural cracks that occurred in steam generator tubes of a pressurized water reactor. It is shown that the parameter is applicable to not only simulated cracks but also real (natural) ones.

Original languageEnglish
Pages (from-to)609-614
Number of pages6
JournalNDT and E International
Volume35
Issue number8
DOIs
Publication statusPublished - 2002 Dec 1
Externally publishedYes

Keywords

  • Eddy currents
  • Inverse problem
  • Natural crack
  • Neural network
  • Steam generator

ASJC Scopus subject areas

  • Materials Science(all)
  • Condensed Matter Physics
  • Mechanical Engineering

Fingerprint Dive into the research topics of 'Generalized neural network approach to eddy current inversion for real cracks'. Together they form a unique fingerprint.

Cite this