Investigation of block filtering and deconvolution for the improvement of lateral resolution and flaw sizing accuracy in ultrasonic testing

In ultrasonic testing, the accuracy of flaw sizing is normally low because of the use of relatively long wavelengths. In this paper, we investigate two signal processing methods: block filtering and block filtering with deconvolution for the improvement of lateral resolution and flaw sizing accuracy...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Ultrasonics 1995-05, Vol.33 (3), p.187-194
Hauptverfasser: Zhao, J., Gaydecki, P.A., Burdekin, F.M.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In ultrasonic testing, the accuracy of flaw sizing is normally low because of the use of relatively long wavelengths. In this paper, we investigate two signal processing methods: block filtering and block filtering with deconvolution for the improvement of lateral resolution and flaw sizing accuracy in ultrasonic testing. In block filtering, each returned echo is Fourier transformed, and a band of frequency components is used to build a C-scan image that is based on the estimation of flaw size. In deconvolution, a block-filtered C-scan image is deconvolved with the point spread function of a transducer by a frequency-domain Weiner deconvolution algorithm. The deconvolved C-scan image is then used for the determination of flaw size. A unique feature in our implementation of the deconvolution technique is that the point spread function is obtained based on an ultrasonic scattering model 1. By doing so, a separate experiment needed for measuring the point spread function is avoided 2. Experiments were conducted on several artificial flaws with different sizes (flat-bottom holes). It was found that the C-scan images, after processing, become sharper; flaw sizing errors were reduced after block filtering and could be further reduced by deconvolution if proper parameters in the deconvolution were selected.
ISSN:0041-624X
1874-9968
DOI:10.1016/0041-624X(94)00024-J