数据驱动的特大跨桥梁全生命周期服役性能-荷载作用耦合分析模型数据

该数据集编号2022YFB2602700-003，汇集了桥梁结构健康监测系统采集的原始加速度时程数据及其预处理可视化样本，涵盖中国某大跨度斜拉桥梁的实际服役状态。数据时间范围与时间精度足以捕捉桥梁在长期荷载作用下的振动特征，空间分布覆盖关键结构节点，空间精度满足局部与整体动态响应分析需求。数据经过严格的预处理与可视化处理，以保证训练和分析的可靠性。基于此数据集，构建了卷积神经网络（CNN）模型，用于桥梁健康状态分类与异常检测，模型输入为可视化图像数据，经训练与多次迁移学习后，分类准确率达到95%，并实现了模型的稳定收敛。模型评估通过混淆矩阵计算召回率、准确率和F1指标，验证了方法在结构异常识别与状态监测中的有效性。该数据集与方法为桥梁全生命周期服役性能分析、荷载作用耦合研究及智能健康监测提供了高质量基础数据与可复用的模型框架，具有重要应用价值和推广潜力。

This dataset, numbered 2022YFB2602700-003, compiles raw acceleration time-history data collected by bridge structural health monitoring systems and preprocessed visualized samples thereof, covering the actual service states of a long-span cable-stayed bridge in China. The temporal range and temporal resolution of the data are sufficient to capture the vibration characteristics of the bridge under long-term loads, while the spatial distribution covers key structural nodes, and the spatial resolution meets the requirements of local and global dynamic response analysis. The data have undergone strict preprocessing and visualization processing to ensure the reliability of training and analysis. Based on this dataset, a convolutional neural network (CNN) model was constructed for bridge health state classification and anomaly detection. The model takes visualized image data as input; after training and multiple transfer learning sessions, the classification accuracy reached 95%, and stable convergence of the model was achieved. Model evaluation was conducted by calculating recall, accuracy, and F1 scores via a confusion matrix, which verified the effectiveness of the method in structural anomaly recognition and condition monitoring. This dataset and the corresponding method provide high-quality basic data and a reusable model framework for bridge life-cycle service performance analysis, load coupling research, and intelligent health monitoring, possessing important application value and promotion potential.