多层粘接结构中脱粘界面的人工神经网络余弦变换谱特征识别

Classification of de-bond in multi-layered steel-rubber adhesive structure with character of DCT spectra by artificial neural networks

针对钢－橡胶多层粘接结构中界面脱粘的超声检测难题，利用余弦变换（ＤＣＴ）提取的表征检测信号的模式特征矢量，通过人工神经网络模式识别方法对不同界面脱粘时实验检测信号的正确识别，实现了脱粘一、二、三和四界面的检测。本文脱粘界面信号模式的人工神经网络识别系统对现代工业中ＮＤＴ＆ＮＤＥ的自动化有着积极的意义。

The character of bond defects in the multi-layered steel-rubber adhesive structure was studied by DCT. Using artificial neural networks, the de-bond at interface Ⅰ, Ⅱ,Ⅲ and Ⅳ was classified. The result show that the features extracted by our algorithms is good pattern for recognition. This enable automatic detection and recognition of de-bond in industry application.