摘要
将离散余弦变换(DCT)用于直径为φ26mm的变形高温合金棒材超声检测信号的特征分析。提取缺陷回波的离散会弦变换幅度谱作为特征矢量,并对提取的特征利用人工神经网络进行训练和分类。实验和距离可分性测度计算结果表明,与离散傅里叶变换提取的谱类别特征相比,离散余弦变换使缺陷信号的谱类别特征有明显增强。
Discrete cosine transform was used for extracting features of ultrasonic testing signals from the high temperature deformed alloy bars with diameter of 26mm. Then the flaw signal feature value based on discrete cosine transform spectrum was trained and classified by artificial neural networks. Theoretical and experimental results showed that discrete cosine transform could obviously enhance the spectrum features of flaw signals compared with Fourier transform.
出处
《无损检测》
2003年第3期124-127,401,共5页
Nondestructive Testing
基金
航空高校科学基金(EC99810915)
江西省教育厅科学基金(4962076)
