Frequency diverse Bayesian ultrasonic flaw detection

When the range cell of an ultrasonic system contains many unresolved reflectors such as grains, fibers, etc. the detection of flaws embedded in the range cell is obscured by echoes scattered from the microstructure. In general, microstructure echoes (clutter) are random, and are more sensitive to changes in frequency than flaws, which are generally larger in size. Through split-spectrum processing, frequency-diverse scattering information of both microstructure and flaw can be obtained. The statistical difference between clutter and flaw echoes has led to the development of a Bayes classifier that is quadratic and can incorporate the correlation properties of scattered echoes. The performance of the Bayes classifier has been examined for both experimental and computer-simulated data, and is compared to other commonly used techniques such as mean, minimum, median, and polarity detectors