基于CNN-LSTM的桥梁结构损伤诊断方法

Bridge structure damage diagnosis method based on CNN-LSTM

针对传统桥梁结构损伤诊断方法在时间联合序列信号特征提取及损伤识别方面不理想的问题,提出一种基于联合卷积神经网络(convolutional neural network, CNN)和长短时记忆神经网络(long short-term memory, LSTM)模型的桥梁结构损伤诊断方法,通过CNN对动挠度、动应变进行传感器拓扑相关特征提取后,再利用LSTM网络进一步提取时间维度相关性特征,实现了健康、损伤预警、一级损伤和二级损伤4种特定工况下桥梁损伤的识别。实验结果表明,该方法对桥梁结构损伤的诊断准确率高达87.6%,具有实际工程价值。

Aiming at the unsatisfactory problem of traditional methods in time joint sequence signal feature extraction and damage recognition, this paper proposes a method based on convolutional neural network(CNN) and long short-term memory(LSTM) model bridge structure damage diagnosis method. After CNN is used to extract the sensor topology-related features of dynamic deflection and dynamic strain, the LSTM network is used to further extract the time-dimension correlation features to realize the identification of bridge damage under several specific working conditions. The experimental results show that the accuracy of this method for the diagnosis of bridge structure damage is as high as 87.6%, which has practical engineering value.