|
| 基于全导波场图像识别的损伤检测研究 |
| Research on damage detection based on guided wave field image recognition |
| 投稿时间：2023-06-13 修订日期：2023-08-20 |
| DOI： |
| 中文关键词: 损伤检测 全导波场识别 深度学习 YOLOv5s目标检测算法 |
| 英文关键词:damage detection full-field guided-wave recognition deep learning YOLOv5s target detection algorithm |
| 基金项目:国家青年科学基金项目（11702118） |
|
| 摘要点击次数: 500 |
| 全文下载次数: 0 |
| 中文摘要: |
| 本文提出了一种基于深度学习目标检测算法的导波损伤识别方法。该方法针对工程中典型的板状结构Lamb导波检测技术，根据结构中的损伤将导致局域波数变化的特性，利用YOLOv5s的目标检测功能，对结构瞬态波场图像进行检测识别。为了提高样本生成效率，本研究构建含盲孔损伤铝板的数值模型，以多频率激励下结构的稳态响应波场图像为样本，建立神经网络模型训练的数据库。为了验证训练后网络模型对瞬态导波场的检测效果，首先对数值模拟含盲孔结构的时域导波场进行分析。结果表明，当导波传播经过损伤后，导波场中将存在损伤局部畸变，利用所训练的网络可以准确的识别检测。在实际铝板结构检测中，采用压电片激励，扫描式激光测振仪检测的方式，获得结构的全场时域导波图像。由检测结果可见，基于深度学习目标检测的全场波场损伤识别算法，能够快速准确识别结构中损伤的位置和尺寸。本文提出的方法检测效率高，不依赖人工经验，为工程结构的智能化损伤检测提供了新的研究思路和检测方案。 |
| 英文摘要: |
| In this paper, a method of guided wave damage identification based on deep learning target detection algorithm is proposed. The method targets the Lamb guided-wave detection technique for typical plate structures in engineering, and uses the target detection function of YOLOv5s to detect and identify transient guided-wave field images of structures based on the property that damage in the structure will lead to changes in the local wave number. In order to improve the efficiency of sample generation, this study constructs a numerical model of an aluminum plate containing blind hole damage, and builds a database of neural network models trained on the steady-state response guided-wave field images of the structure under multi-frequency excitation as samples. In order to verify the detection effect of the trained network model on the transient guided-wave field, the time domain guided-wave field of the numerically simulated structure containing blind holes is first analyzed. The results show that when the guided-wave propagates through the damage, there will be local distortion of the damage in the guided-wave field, which can be accurately identified and detected using the trained network. In the actual aluminum plate structure inspection, a piezoelectric sheet excitation and scanning laser vibrometer detection is used to obtain a full-field time-domain guided-wave image of the structure. The detection results show that the full-field guided-wave field damage identification algorithm based on deep learning target detection can quickly and accurately identify the location and size of damage in the structure. The proposed method is highly efficient and does not rely on human experience, providing a new research idea and detection solution for intelligent damage detection of engineering structures. |
| 下载PDF阅读器 |
| 关闭 |
