基于CEEMDAN-VMD-PSO-LSTM模型的桥梁挠度预测
CEEMDAN-VMD-PSO-LSTM model for bridge deflection prediction
郭永刚 1张美霞 2王凯 1刘立明 1陈卫明2
作者信息
- 1. 安徽建工建设投资集团有限公司,安徽 合肥 230031
- 2. 中国地质大学(武汉)工程学院,湖北 武汉 430074
- 折叠
摘要
针对桥梁运行阶段的健康状态监测,构建了 CEEMDAN-VMD-PSO-LSTM模型对桥梁挠度进行预测.该模型主要分为二次模态分解平稳化、粒子群优化(PSO)算法和长短期记忆(LSTM)网络预测三大模块,共有5个步骤:①利用自适应噪声完备集合经验模态分解(CEEMDAN)算法对桥梁原始挠度序列进行初次模态分解,分解为若干本征模态分解函数(IMF);②使用样本熵(SampEn/SE)计算各IMF分量的复杂度,并通过K-means聚类为高频、中频和低频3个IMF分量;③通过变分模态分解(VMD)算法对高频IMF分量进行二次模态分解;④分别对各个IMF分量通过PSO算法得出LSTM最优超参数组合;⑤将各最优超参数分别代入LSTM模型进行训练,并将各预测结果融合为最终的预测结果.结果表明:该预测方法具有最高的预测精度,为智慧桥梁的安全监测监控提供了新的技术方法.
Abstract
In order to forecast bridge deflection for the operational phase of monitoring the health status of bridges,a CEEMDAN-VMD-PSO-LSTM model was built,including three main modules,namely,Quad-ratic modal decomposition smoothing,the PSO algorithm,and LSTM prediction.There are five steps.First,the original deflection sequence of a bridge is decomposed into several intrinsic mode decomposition func-tions(IMFs)using CEEMDAN.Second,the complexity of each IMF component is determined using sam-ple entropy(SampEn/SE),and three IMF components of high,medium,and low frequencies are clustered by K-means.Thirdly,the VMD is used to determine the original deflection sequence and high-frequency IMF components are then subjected to quadratic mode decomposition by VMD.Fourthly,the PSO optimiza-tion algorithm is used to derive the optimal LSTM hyperparameters for each IMF component.Fifthly,the LSTM model is trained using each of the optimal hyperparameters,and finally,the prediction results are combined to produce the final prediction result.The findings indicate that the prediction method offers a fresh technical strategy for the safety monitoring of smart bridges and has the highest prediction accuracy.
关键词
桥梁挠度预测/自适应噪声完备集合经验模态分解/变分模态分解/样本熵/K-means聚类/粒子群优化/长短期记忆网络Key words
bridge deflection prediction/complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)/variational mode decomposition(VMD)/SampEn/SE/K-means/particle swarm optimi-zation(PSO)/long short-time memory network(LSTM)引用本文复制引用
基金项目
国家重点研发计划(2022YFC3005900)
陕西省煤矿水害防治技术重点实验室开放基金(2021SKMS07)
中央高校基本科研业务费专项(CUG2642022006)
湖北省高等学校省级教学研究项目(2022)(2022143)
安徽省住房城乡建设科学技术计划(2022)(2022-YF104)
出版年
2024
国家科技期刊平台
NSTL
万方数据