MFL signal inversion using the finite element network

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Scopus citations

Abstract

Iterative algorithms that incorporate a numerical forward model have been typically used to solve the problem of flaw profile estimation from measured magnetic flux leakage (MFL) data. These approaches use the forward model to determine the measurement signal for a given defect profile, and obtain the desired profile by iteratively minimizing a cost function. The use of numerical models is computationally expensive, and alternative forward models are needed. This paper presents a finite element neural network (FEN) obtained by embedding a finite element model in a parallel neural network architecture that enables fast and accurate solution of the forward problem. Previous results have indicated that the FEN performance as a forward model is comparable to that of the conventional finite element method. In this paper, we investigate the applicability of the FEN to determining flaw profiles from MFL data in pipeline inspection and present results on synthetic MFL data.

Original languageEnglish
Title of host publicationReview of Progress in QuantitativeNondestructive Evaluation
Pages633-640
Number of pages8
DOIs
StatePublished - 2008
Externally publishedYes
Event34th Annual Review of Progress in Quantitative Nondestructive Evaluation - Golden, CO, United States
Duration: Jul 22 2007Jul 27 2007

Publication series

NameAIP Conference Proceedings
Volume975
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

Conference34th Annual Review of Progress in Quantitative Nondestructive Evaluation
Country/TerritoryUnited States
CityGolden, CO
Period07/22/0707/27/07

Keywords

  • Finite element models
  • Inverse problems
  • MFL
  • Neural networks

Fingerprint

Dive into the research topics of 'MFL signal inversion using the finite element network'. Together they form a unique fingerprint.

Cite this