| 张思远,于可新.基于CNN-LSTM-Attention的桥梁结构异常预测模型研究[J].安徽建筑大学学报,2026,34(2):68-76 |
| 基于CNN-LSTM-Attention的桥梁结构异常预测模型研究 |
| Research on a Bridge Structure Anomaly Prediction Model Based on CNN-LSTM-Attention |
| |
| DOI： |
| 中文关键词: 结构健康监测 异常预测 CNN-LSTM-Attention 斜拉桥 斜拉索振动信号 |
| 英文关键词: structural health monitoring anomaly prediction CNN-LSTM-Attention cable-stayed bridge stay-cable vibration signals |
| 基金项目:国家自然科学基金项目（52308129）；安徽省自然科学基金项目（2308085QE185） |
|
| 摘要点击次数: 160 |
| 全文下载次数: 0 |
| 中文摘要: |
| 针对斜拉桥结构振动信号具有非平稳性、高噪声及强时序依赖特征，传统异常检测方法难以实现早期精准识别的问题，提出一种融合卷积神经网络（CNN）、长短期记忆网络（LSTM）与自注意力机制（Attention）的桥梁结构异常预测模型，即 CNN-LSTM-Attention。该模型通过 CNN 提取局部空间特征，LSTM 捕捉长时序依赖关系，Attention 机制自适应强化关键时间步权重分配，实现对结构动态响应中潜在异常的高效识别。以芜湖长江公路二桥斜拉索加速度监测数据为案例，结果表明，该模型在预测精度与稳健性方面均优于传统时间序列预测方法。 |
| 英文摘要: |
| The structural vibration signals of cable-stayed bridges are characterized by non-stationarity, high noise, and strong temporal dependence, making it difficult for traditional anomaly detection methods to achieve early and accurate identification. Hence, a bridge structure anomaly prediction model is proposed, which integrates Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) network, and Self-Attention mechanism, namely CNN-LSTM-Attention. The model extracts local spatial features through CNN, captures long-term temporal dependencies via LSTM, and uses the Attention mechanism to adaptively enhance the weight allocation of key time steps, thereby realizing efficient identification of potential abnormal patterns in the structural dynamic response. Taking the acceleration monitoring data of the stay cables of the Second Wuhu Yangtze River Highway Bridge as a case study, the results show that the proposed model outperforms traditional time-series prediction methods in terms of prediction accuracy and robustness. It provides strong technical support for intelligent early warning and operation-maintenance decision-making of bridge health monitoring systems. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
