SONG Weihua^1,2 WANG Xiaomin¹  LI Mingxuan¹

(1 Institute of Acoustics, The Chinese Academy of Sciences  Beijing 100190)

(2 Graduate School of the Chinese Academy of Sciences  Beijing 100029)

Received Feb.5, 2007

Revised Mar.21, 2007

Abstract  The proper frequency is experimentally chosen to be the operation frequency of the electromagnetic acoustic transducer. The instantaneous amplitude, phase and frequency of the detected ultrasonic echoes from a multilayer adhesive sample of steel and rubber materials are calculated and composed to form three-dimensional instantaneous spectrum which is successful to distinguish the testing signals from different adhesive states qualitatively. Then, average instantaneous parameters in sensitive time window are picked up and used as the input eigenvectors for the BP artificial neural network. Identified results in both training and testing volumes demonstrate that the detected electromagnetic ultrasonic interfacial echoes can be identified and classified automatically with the correctness ratio larger than 95%.

PACS numbers: 43.35, 43.60