基于CNN-BiLSTM的特高拱坝变形预测模型  

CNN-BiLSTM-based deformation prediction model for extra-high arch dams

作　　者：欧斌 张才溢 傅蜀燕 杨霖 陈德辉 杨石勇 OU Bin;ZHANG Caiyi;FU Shuyan;YANG Lin;CHEN Dehui;YANG Shiyong(College of Water Conservancy,Yunnan Agricultural University,Kunming,Yunnan 650201,China;State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering,Hohai University,Nanjing,Jiangsu 210098,China;Yunnan Province Small and Medium-sized Water Conservancy Engineering Research Centre for Intelligent Management and Maintenance,Kunming,Yunnan 650201,China)

机构地区：[1]云南农业大学水利学院,云南昆明650201 [2]河海大学水文水资源与水利工程科学国家重点实验室,江苏南京210098 [3]云南省中小型水利工程智慧管养工程研究中心,云南昆明650201

出　　处：《排灌机械工程学报》2024年第10期1031-1035,1043,共6页Journal of Drainage and Irrigation Machinery Engineering

基　　金：国家自然科学基金资助项目(52069029,52369026);云南省教育厅科学研究基金资助项目(2023J0519)。

摘　　要：为提高特高拱坝的变形预测精度,提出了一种基于卷积神经网络(CNN)和双向长短期记忆网络(BiLSTM)的大坝变形预测模型.该模型利用CNN捕捉数据之间的空间关系,进行特征提取,再将其输入到BiLSTM中进行时间维度上的演变规律考虑.通过特征融合和全连接层的拼接,得到更丰富和综合的特征表示,最终映射到预测输出层进行拱坝变形预测.以某拱坝为例,验证了CNN-BiLSTM模型在RMSE等评价指标上具有高精度和稳定性,为混凝土拱坝结构的安全监测提供了新的思路.To improve the deformation prediction accuracy of extra-high arch dams,a dam deformation prediction model based on convolutional neural network(CNN)and bidirectional long and short-term memory network(BiLSTM)was proposed.The model used CNN to capture the spatial relationship between the data for feature extraction,which was then fed into BiLSTM for the consideration of evolutio-nary patterns in the time dimension.A richer and integrated feature representation was obtained through feature fusion and splicing of fully connected layers,which was finally mapped to the prediction output layer for arch dam deformation prediction.Taking an arch dam as an example,the CNN-BiLSTM mo-del was verified to have high accuracy and stability in evaluation indexes such as RMSE,providing a new idea for the safety monitoring of concrete arch dam structures.

关 键 词：混凝土拱坝 卷积神经网络 双向长短期记忆网络 预测模型 

分 类 号：S277.9[农业科学—农业水土工程]