By Topic

Flaw identification from time and frequency features of ultrasonic waveforms

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
T. J. Case ; Dept. of Electr. Eng., Rochester Univ., NY, USA ; R. C. Waag

Time and frequency features have been used with classification algorithms to distinguish between ultrasonic echoes from flaws in pipes and ultrasonic echoes from various geometric configurations of weld root and counterbore. Waveforms containing reflections from known geometries and from flaws were obtained and sets of features were defined using a k-nearest neighbor approach to separate waveforms into classes. Two independent databases containing various flaws and pipe geometries were used to determine these feature sets. From these databases, optimal feature sets were found to separate counterbore waveforms from crack waveforms. Optimal feature sets were also found to distinguish between waveforms from counterbore, waveforms containing both counterbore and root echoes, and waveforms from flaws. The best feature sets used with the classifier algorithms could separate waveforms from the same database with accuracy in the 92-97% range and with high confidence., Another database was obtained from pipe structures in a nuclear power plant to provide a field test of the method. When applied to this database, the same classifier algorithms and feature sets used with the other databases either resulted in a comparable percentage of correct decisions, but with low confidence, or could not classify anywhere from 79 to 88% of the waveforms. Spatial parameters based on averaging feature vectors in axial and circumferential directions were also defined and used for classification. These classifications had higher accuracy but lower confidence levels than the classifications based on individual waveforms.

Published in:

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control  (Volume:43 ,  Issue: 4 )