小波分析和人工神经网络在金属超声无损检测缺陷分类中的应用

Application of Wavelet Analysis and Artificial Neutral Networks to Flaw Classification in Ultrasonic Non-destructive Testing

摘要: 基于金属超声检测中的缺陷脉冲回波为非稳态信号的特点,对高温合金材料超声检测信号的小波变换进行了特征分析,提取了各级小波分解信号的能量分布特征,最后将这些特征输入人工神经网络进行训练和分类,实验表明,这种方法具有良好效果.

Abstract: As the flaw pulse echo signals were non-stationary in ultrasonic testing,wavelet transform was used for analyzing feature of the flaw signals in ultrasonic testing of high temperature alloy.Features based on power distribution of the decomposed signals were extracted.Finally,a artificial neutral networks classifier was used for the features.Experimental results show that the method is effective.