Automation of Electric Current Injection NDE by Neural Networks

Thumbnail Image
Barbosa, C. Hall
Bruno, A.
Vellasco, M.
Pacheco, M.
Camerini, C.
Major Professor
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Is Version Of
Review of Progress in Quantitative Nondestructive Evaluation
Center for Nondestructive Evaluation

Begun in 1973, the Review of Progress in Quantitative Nondestructive Evaluation (QNDE) is the premier international NDE meeting designed to provide an interface between research and early engineering through the presentation of current ideas and results focused on facilitating a rapid transfer to engineering development.

This site provides free, public access to papers presented at the annual QNDE conference between 1983 and 1999, and abstracts for papers presented at the conference since 2001.


The Electric Current Injection (ECI) method of nondestructive evaluation is applied to materials that are electrically conductive but not magnetically permeable, as aluminum, magnesium, and titanium. It consists of detecting current-flow anomalies due to voids, nonmetallic inclusions and open cracks in the sample, through distortions introduced in the magnetic field generated by the plate [1]. We have applied an ECI method, with low dc current levels, to aluminum plates with circular voids in the millimeter range. These plates have already been measured with a SQUID magnetometer, with large lift-off distances [2], being necessary the use of image enhancement techniques, in order to improve the visual detection of the flaws [3]. In this paper a small lift-off distance was used, so it was possible to detect the associated magnetic field with a fluxgate magnetometer without further processing. It is proposed a method for the automation of magnetic signal analysis using Time-Delay Neural Networks (TDNN), which exempts the need of a trained technician, necessary to most of the usual NDE methods. In this method a TDNN is trained to dynamically scan the magnetic signals, automatically locating the voids and classifying their diameters. Next section depicts the experimental setup, and presents some of the magnetic signals obtained. After that, the neural network used to detect and classify the defects are explained, and the results of this technique are shown.

Subject Categories
Wed Jan 01 00:00:00 UTC 1997