基于超声导波和机器学习的蜂窝夹层结构脱黏诊断

Debonding diagnosis of honeycomb sandwich structures based on guided waves and machine learning

针对蜂窝夹层结构的脱黏损伤诊断,首先通过集成压电陶瓷传感器构建传感器网络,采用超声导波加权分布诊断成像方法对损伤进行平面内定位诊断;然后利用超声导波在结构厚度截面内对不同脱黏层的敏感度差异提取损伤特征;最后通过蜂窝夹层结构有限元模型进行大量的导波传播仿真,形成训练数据库,进而训练形成稳定的支持向量机(SVM)脱黏层分类机器学习模型,进行截面内脱黏层诊断。验证试验结果表明,该方法能够有效诊断出蜂窝夹层结构的脱黏损伤,平面内定位误差小于2 cm,截面内脱黏层诊断准确度为100%。

Aiming at the diagnosis of debonding damage of honeycomb sandwich structure, the sensor network was first constructed by integrating piezoelectric ceramic sensors, and the ultrasonic guided wave weighted distribution diagnostic imaging method was used to locate and diagnose the damage in the plane. The sensitivity difference of the debonding layer was used to extract the damage characteristics;finally, a large number of guided wave propagation simulations were carried out through the finite element model of the honeycomb sandwich structure to form a training database, and then a stable support vector machine(SVM) debonding layer classification machine learning model was formed. Diagnosis of intra-section debonding layer. The verification test results show that the method can effectively diagnose the debonding damage of the honeycomb sandwich structure, the positioning error in the plane is less than 2 cm, and the diagnostic accuracy of the debonding layer in the section was 100%.