|
| |
环境温度影响下基于LSTM神经网络识别结构损伤 |
Structural Damage Identification Based on LSTM Neural Networks Under Ambient Temperature Variations |
投稿时间:2022-07-26 修订日期:2022-09-12 |
DOI: |
中文关键词: LSTM神经网络 结构健康监测 温度 模态频率 变分模态分解 |
英文关键词:LSTM neural network structural health monitoring temperature modal frequency variational mode decomposition |
基金项目:国家自然科学(51908395);江苏省高等学校自然科学研究项目(19KJB580004);江苏省研究生科研与实践创新计划项目(SJCX22_1569) |
|
摘要点击次数: 248 |
全文下载次数: 0 |
中文摘要: |
结构损伤会导致模态参数的改变,因而,模态参数的改变是结构损伤识别的依据。环境温度变化也会引起模态参数的变化,其变化程度有时会更大,掩盖或部分掩盖结构损伤,导致结构健康监测系统发出“假阳性”或“假阴性”的误判。因此,消除温度对结构损伤识别是提高损伤识别精度的关键。基于LSTM神经网络的非线性映射优势,文章提出了采用LSTM剔除环境温度影响,并以频率为损伤指纹判断结构损伤状况的方法。通过建立频率与环境温度的神经网络模型消除温度的影响,并在此基础上采用控制图判断频率异常变化以确定损伤状况。最后将该方法在数值模型和实际桥梁中加以应用,结果表明:该方法能够有效识别损伤时刻,并具有一定的抗噪性。 |
英文摘要: |
Structural damage will lead to the change of modal parameters. Therefore, the change of modal parameters is the basis of structural damage identification. The change of ambient temperature will also cause the change of modal parameters, and the degree of change is sometimes greater. It covers up or partially covers up the structural damage, resulting in the misjudgment of "false positive" or "false negative" issued by the structural health monitoring system. Therefore, eliminating the influence of temperature on structural damage identification is the key to improve the accuracy of damage identification. Based on the nonlinear mapping advantage of LSTM neural network, a method of using LSTM to eliminate the influence of ambient temperature and using frequency as damage fingerprint to judge structural damage is proposed in this paper. The neural network model of frequency and ambient temperature is established to eliminate the influence of temperature, and on this basis, the control chart is used to judge the abnormal change of frequency to determine the damage condition. Finally, the method is applied to numerical models and practical bridges. The results show that the method can effectively identify the damage time and has certain noise resistance. |
查看/发表评论 下载PDF阅读器 |
|
|
|
|