针对孔隙率接近0的小孔隙率碳纤维复合材料(Carbon Fiber Reinforced Composite, CFRP)的富树脂检测需求, 提出富树脂超声检测技术。对超声检测信号中的噪声消除方法、衰减抑制方法和富树脂检测的多视图成像技术进行研究, 并开发小孔隙率CFRP富树脂超声检测软件。首先提出共振频率估计方法, 通过低通滤波抑制高频随机噪声。其次根据频率差异, 应用变分模态分解算法分离并消除共振结构噪声, 提取低频成分。该低频成分包括表面回波、底面回波、富树脂反射信号和由层间反射信号、材料散射噪声等构成的相干噪声。再次, 引入瞬时幅值比修正低频成分的幅值衰减并描述被检测小孔隙率CFRP的局部反射能力。最后, 应用Otsu多阈值方法自适应获得富树脂识别的阈值, 消除相干噪声的影响, 完成富树脂识别。进一步对小孔隙率CFRP的超声检测结果进行多视图成像, 在三维视图、C扫描视图和B扫描视图内识别富树脂。结果表明: 变分模态分解的分量数为2, Otsu多阈值的类别数为3时, 能够准确识别小孔隙率CFRP超声检测信号中的富树脂反射信号; 采用0.15作为多视图成像的阈值, 可简洁有效地描述富树脂在小孔隙率CFRP中的分布。

To satisfy the demand of rich-resin defect detection in the so-called low-porosity carbon fiber reinforced composite (CFRP) with porosity close to zero, an ultrasonic testing methodology was proposed in this article. The denoising methods, attenuation suppression method, and 3D imaging technology for rich-resin identification are investigated, and low-porosity CFRP rich-resin detection software was developed. The rich resin was detected in four steps. First, the resonant frequency was estimated, and the high-frequency stochastic noise was suppressed. Second, variational mode decomposition (VMD) was used to separate the resonant structure noise and extract the low-frequency component. The low-frequency component consisted of the front-wall echo, back-wall echo, rich-resin reflection signal, and remaining coherent noise made up of the interlayer reflection signals and material scattering noise. Third, the instant amplitude ratio was introduced to correct the envelop attenuation of the low-frequency component and describe the local reflectivity of the low-porosity CFRP. Finally, the multi-threshold Otsu method was used to search the threshold of the rich-resin detection, resulting in the elimination of interference and finishing the detection of rich resin. Further, multi-view imaging was performed on the test results, and the rich resin was identified in the 3D, C-scan, and B-scan imaging processes. The experimental results show that when the VMD mode was set to two and the classes in the multi-threshold Otsu method are set to three, a rich-resin reflection signal can be detected. When the threshold in the multi-view imaging is set to 0.15, the rich resin can be effectively characterized.