郑跃滨,雷振坤*,徐浩,周凯,张佳奇,马书义,武湛君*.基于数据驱动的无基准导波损伤诊断成像方法研究[J].实验力学,2021,36(4):458~470 |
基于数据驱动的无基准导波损伤诊断成像方法研究 |
Research on a data-driven reference-free probability diagnostic imaging method using guided waves |
投稿时间:2020-12-11 修订日期:2021-01-20 |
DOI:10.7520/1001-4888-20-239 |
中文关键词: 损伤诊断 超声导波 无基准 数据驱动 字典学习 |
英文关键词:damage diagnostic ultrasonic guided waves reference-free data-driven dictionary learning |
基金项目:国家重点研发计划(2018YFA0702800),国家自然科学基金(11772081, 11972106, 12072056)资助 |
|
摘要点击次数: 807 |
全文下载次数: 1701 |
中文摘要: |
在超声导波方法中,通常需要比较损伤前后(基准信号与当前测试信号)的数据以实现损伤诊断。然而实际情况下,基准信号往往不可靠或未知。本研究提出一种基于数据驱动的无基准损伤诊断成像(DRPDI)方法。DRPDI方法基于损伤稀疏假设,将聚类分析后的核心路径集作为训练字典库,利用字典学习算法重构信号,代替传统基准信号,从而实现无基准的损伤诊断概率成像。本方法以聚类数据作驱动,无需结构的物理模型和材料属性等先验知识。这一特性使得该方法在复杂加筋结构上尤其适用。通过加筋薄壁结构的损伤检测实验,验证了DRPDI方法在定位精度上与传统有基准方法接近。 |
英文摘要: |
In the ultrasonic guided waves-based method, it is usually necessary to compare the data before and after damage (reference signal vs. current test signal) for damage diagnostic. However, in practical scenarios, the reference signal is often unreliable or unavailable. A data-driven reference-free probability diagnostic imaging (DRPDI) method is proposed in this research. Based on the sparse hypothesis of damage, the DRPDI method reconstructs signals by the dictionary learning algorithm with core path setting as the training dictionary, instead of traditional reference signal, and realizes the reference-free probability diagnostic imaging. This method is driven by the clustering data and does not require prior knowledge of physical models of structure and material properties. This characteristics makes the method particularly suitable for complex stiffened structures. Through a damage detection experiment carried out on a stiffened thin-walled structure, it is verified that the localization accuracy of DRPDI method is close to the traditional reference method. |
查看全文 下载PDF阅读器 |
关闭 |