基于LSTM的泵闸工程混凝土施工期温度场预测.pdfVIP

第43卷第2期 水 利 水 电科 技 进 展 2023年3月 Vol.43 No.2 Advances in Science and Technology of Water Resources Mar. 2023 DOI:10.3880/ j.issn.1006 7647.2023.02.012 基于 LSTM的泵闸工程混凝土施工期温度场预测 程摇 井,孔垂穗,邹科辉 (河海大学水利水电学院,江苏 南京摇 210098) 摘要:为了快速准确地预测混凝土施工期温度过程线,结合主成分分析,提出了一种基于长短期记 忆网络(LSTM)算法的预测模型。 以上海崇明岛四滧港和八滧港水闸底板为例,采用主成分分析 对混凝土温度场的可能影响因素进行降维,建立以四滧港水闸底板温度数据为基础的LSTM 温度 过程线预测模型并对输入主成分进行训练,将训练后的模型用于八滧港温度过程线的拟合和预测 并与实测结果进行对比。 结果表明,该模型预测温度过程线与实际测点温度过程线拟合良好,均方 根误差在2益以内,判定系数接近1,预测结果符合工程精度要求。 该预测模型可部分替代有限元 反馈分析,从而提高泵闸混凝土温度场预测的效率。 关键词:泵闸结构;温控防裂;温度预测;深度学习;主成分分析;长短期记忆网络 中图分类号:TV331摇 摇 摇 文献标志码:A摇 摇 摇 文章编号:1006 7647(2023)02 0076 06 Temperature field prediction during concrete construction period of pump and sluice project based on LSTM/ / CEHNGJing,KONG Chuisui,ZOU Kehui(College of Water Conservancy and Hydropower Engineering,Hohai University, Nanjing 210098,China) Abstract:To quickly and accurately predict the temperature history during the concrete construction period,a prediction model based on the long short鄄term memory (LSTM) network algorithm was proposed combined with the principal component analysis (PCA).Taking the baseboard of the Siyaogang Sluice and Bayaogang Sluice in Shanghai Chongming Island as an example,the PCA method was used to reduce the number of the possible influencing factors of the concrete temperaturefield, and then the LSTM prediction model of temperature history based on the temperature data of the Siyaogang Sluice baseboardwas established to train the input principal components. Then the trained model was used for thefitting and prediction of the temperature history of the Bayaogang Sluice and was compared with the measured results. The results show that the predicted valuefitswellwi