De-Noising of Ultrasonic Signals Based on Entropy Theory and Recognition of Defect in Material.

In order to eliminate the noise which exists in ultrasonic detection signal and improve the diagnostic accuracy of defects inside the material, the method of de-noising ultrasonic signal by applying adaptive threshold(EAT) based on entropy theory is put forward in this paper. The searching algorithm of best wavelet packet basis adopting Shannon entropy is analyzed. The state of signal with noise is characterized by entropy and the threshold of wavelet packet decomposed in different scales is determined according to the entropy of the wavelet packet energy spectrum. Experiment of processing ultrasonic signal which comes from Steyr engine crankshaft with flaws has been implemented. Information that characterizes defect size and location could be extracted accurately from the processing results. The result indicates that the proposed EAT method has better de-noising performance and it has benefit to enhancing the degree of accuracy for quantitatively analyzing the defect inside material. [ABSTRACT FROM AUTHOR]