应用双树复小波包和NCA-LSSVM检测磁瓦内部缺陷

Application of DTCWPT and NCA-LSSVM to inspect internal defects of magnetic tile

提出结合双树复小波包变换(DTCWPT)、邻域成分分析法(NCA)、最小二乘支持向量机(LSSVM)的磁瓦内部缺陷检测方法.通过双树复小波包将采集的声音信号分解为6层,得到64个不同频带的子信号;求取特定频带信号的能量、偏度、峭度、模糊熵,并将能量、偏度、峭度、模糊熵作为分类特征;利用邻域成分分析法对分类特征降维;将降维构造的新特征集输入到最小二乘支持向量机,判断磁瓦是否含有内部缺陷.通过实验验证,对提出的检测方法进行可行性分析.3种不同类型磁瓦的内部缺陷识别率均可以达到99%,与以往双谱切片方法相比,提高了检测识别率.试验结果表明,提出的方法具有检测速度快、可靠性高、适应性强等特点,为高效、准确地进行磁瓦内部缺陷检测提供了有效的技术手段.

A novel method was proposed to detect magnetic tile internal defects based on dual-tree complex wavelet packet transform（DTCWPT）,neighborhood component analysis（NCA）,least squares support vector machines（LSSVM）.The impact sound was decomposed up to 6levels,resulting in 64sub-signals.Then four statistical features：energy,skewness,kurtosis and fuzzy entropy were calculated to construct feature set.The dimension of feature set was reduced and optimized by NCA.The new feature set was input to LSSVM to judge whether the magnetic tile is with internal defects.The reliability of the proposed method was verified by the experimental results.The classification results showed that the accuracy rates of three kinds of magnetic tiles could reach 99%,higher than the bispectrum method.The experimental results demonstrate that the proposed method is fast,adaptable,efficient and reliable,providing a technical support to detect the internal defects of magnetic tile.