【摘要】 本文在对各向同性均匀固体中横穿孔、平底孔和裂缝缺陷超声散射特性分析的基础上，分别用回波幅度话和去卷积幅度措作为特征量，利用人工神经网络对缺陷类型进行识别．结果表明，用去卷积幅度谱作为特征量时，利用人工神经网络对这三类缺陷的类型识别，可获得较理想的结果．更多还原

【Abstract】 In this paper, the scattering characteristics of an ultrasonic wave in a homegeneous solid by three types of flaw are considered the three types being the traversecylindrical cavily, the blat-bottom hole and the plane crack. The flaws are then classifiedby a neural network on using the respectively amplitude spectra of ultrasonic echoes andthose of deconvolved ultrasonic echoes as characteristic features. It is demonstrated thatthe latter improve greatly the classification accuracy.更多还原