Automatic Ultrasonic Image Analysis Method for Defect Detection

Ultrasonic examination of seams raises well known signals perturbed by this austenitic steel weld problems of interpreting type of material. The JUKEBOX ultrasonic imaging system developed at the Cadarache Nuclear Research Center provides a malor improvement in the general area of defect localization and characterization, based on processing overall images obtained by iX,Y) scanning. (X,timel images are formed by ~uxtaposing input signals. A series of parallel images shifted on the Y-axis is also available. The authors present a novel deCect detection method based on analysing the timeline positions of the maxima and minima recorded on (X,time) images. This position is statistically stable when a defect is encoutered, and is random enough under spurious noise conditions to constitute a discriminating parameter. The investigation involves calculating the trace variance: this parameter is then taken into account for detection purposes. Correlation with parallel images enhances detection reliability. A significant increase in the signal-to-noise ratio during tests on artificial defects is shown.