市政技术2025，Vol.43Issue(8)：63-69,76,8.DOI:10.19922/j.1009-7767.2025.08.063
基于麻雀优化与LSTM基坑顶部变形预测
Deformation Prediction at the Top of Foundation Pit by SSA and LSTM
摘要
Abstract
In order to improve the accuracy of deformation prediction at the top of foundation pit,a deformation pre-diction model based on sparrow search algorithm(SSA),variational mode decomposition(VMD)and long-term and short-term memory network(LSTM)is proposed.The deformation monitoring data of a foundation pit in Guangzhou is taken as an example.Firstly,the control parameters of VMD are optimized by SSA to reduce the interference of noise components and the time series data of the deformation at the top of the foundation pit are decomposed into different intrinsic mode function(IMF).Acording to different frequency characteristics,the decomposition results are divided into trend items,seasonal items and random items;The polynomial fitting method is used to predict the trend term.The combination of LSTM with self attention mechanism is used to model the seasonal term and random term.The deformation prediction value is obtained by superimposing the prediction results of each sub item.The results show thatthe mean square error(MSE)of SSA-VMD-LSTM model is 0.48,and the coefficient of determination(R2)is 0.971 when the input step is 3.Compared with the original sequence model without decomposition,MSE is re-duced by 0.99,which verifies the significant improvement of the prediction accuracy of the improved model.The re-search results provide an important reference for the deformation prediction at the top of similar foundation pits.关键词
基坑顶部变形/麻雀搜索算法/变分模态分解/本征模态函数/长短期记忆网络Key words
deformation at the top of the foundation pit/sparrow search algorithm(SSA)/variational mode decom-position(VMD)/intrinsic mode function(IMF)/long short-term memory networks(LSTM)分类
建筑与水利引用本文复制引用
聂彦冲,周明文,祝敏刚,段平,王以姣..基于麻雀优化与LSTM基坑顶部变形预测[J].市政技术,2025,43(8):63-69,76,8.基金项目
广州市设计院集团有限公司重点科研项目(24RD22(B1) （24RD22(B1）
24RD24(B1)) （B1）
