基于残差修正的混凝土坝位移双层阶预测模型

Dual-layer Prediction Model of Concrete Dam Displacement Based on Residual Amendment

传统的混凝土坝位移监测模型往往忽略残差序列中的有效信息,导致对大坝位移的预测效果不佳。旨在提高预测模型的精度。针对这一问题,提出一种考虑信号残差修正的混凝土坝位移双层阶预测模型。首先根据传统混凝土坝位移预测的统计模型选取大坝位移的影响因子;其次通过麻雀搜索算法(SSA)确定极限学习机(ELM)中的超参数,建立单层阶预测模型SSA-ELM,进而得到大坝的单层阶模型位移预测值;再次,基于最小样本熵(SE)和相关分析法,通过变分模态分解法(VMD)对残差序列分解重构;最后利用SSA-ELM对重构后序列进行修正,并将修正值与单层阶模型位移预测值进行叠加,构建双层阶预测模型SSA-ELM-VMDρ+,进而得到最终的位移预测值。工程实例验证表明,与其他模型相比,该双层阶模型预测精度高,泛化能力强,有效的挖掘了残差中的有效信息,并克服了噪声干扰。本研究为大坝安全监控、健康服役诊断与管理运行提供了新的参考。

The traditional monitoring model for concrete dam displacement often ignores analyses of pertinent information within the residual sequence,leading to suboptimal predictive performance for dam displacement.The purpose of this paper is to improve the accuracy of prediction models.To address this issue,a dual-layer displacement prediction model for concrete dam,which considers signal residual amendment,is proposed.Firstly,influential factors affecting dam displacement are selected based on the statistical model of traditional concrete dam displacement.Secondly,the sparrow search algorithm(SSA)is employed to determine the hyperparameters of extreme learning machines(ELM)in order to establish the single-layer prediction model SSA-ELM,which yields single-layer model displacement prediction values for dams.Next,the Variational Mode Decomposition(VMD)technique along with the Minimum Sample Entropy(SE)and correlation analysis is utilized,the residual sequence is decomposed and reconstructed.Finally,by using the SSA-ELM model to correct the reconstructed sequence and combining the corrected values with the single-layer model displacement prediction values,the dual-layer prediction model SSA-ELM-VMDr+is constructed,allowing for the final displacement prediction values.The engineering validation of the model demonstrates that the dual-layer prediction model exhibits higher prediction accuracy and stronger generalization ability compared with other models.Moreover,the model effectively extracts valuable information from the residual and overcomes noise interference.This paper provides a new reference for dam safety monitoring,health service diagnosis and management operations.