环境温度影响下基于LSTM神经网络识别结构损伤

环境温度的改变会引起模态参数的变化，其变化程度会掩盖或部分掩盖损伤引起的变化量，导致结构健康监测系统发出假阳性或假阴性的误判，因此，消除温度效应是提高损伤识别精度的关键。本文基于LSTM神经网络提出了一种环境温度影响下识别结构损伤的方法。充分利用LSTM神经网络的非线性映射优势，建立多元温度-模态频率的相关模型，在此基础上采用数据标准化方法消除温度效应，并结合控制图判断模态频率异常变化以确定损伤状况。最后将所提方法在数值模型和实际桥梁中加以应用，结果表明，方法能够有效消除温度效应；结合控制图能识别损伤时刻，并具有一定的抗噪性；在实桥数据分析中仍能表现出较好的损伤敏感性。

Structural damage identification based on LSTM neural networks under ambient temperature variations

The change of ambient temperature will cause the change of modal parameters,and the degree of change will cover up or partially cover up the change caused by damage,resulting in the misjudgment of false positive or false negative issued by the structural health monitoring system.Therefore,eliminating the temperature effect is the key to improve the accuracy of damage identification.Based on LSTM neural network,this paper proposes a method to identify structural damage under the impact of ambient temperature.By making full use of the nonlinear mapping advantage of LSTM neural network,the correlation model of multivariate temperature and modal frequencies is established.On this basis,the data normalization method is used to eliminate the temperature effect,and the control chart is used to judge the abnormal change of modal frequencies to determine the damage condition.Finally,the proposed method is applied to a numerical model and an actual bridge.The results show that the method can effectively eliminate the temperature effect.Combined with the control chart method,it can identify the damage time and has a certain noise resistance.In the real bridge data analysis,it can still show good damage sensitivity.