摘要
【目的】为实现黄土地区高填方路基沉降趋势的准确预测。【方法】通过建立基于变分模态分解(VMD)和粒子群算法(PSO)优化的长短期记忆网络(LSTM)预测模型VMD-PSO-LSTM,学习高填方路基沉降数据的高层次特征并预测其发展变化趋势。通过工程实例,对所建立的预测模型进行验证。【结果】结果表明:VMD-PSO-LSTM模型对高填方路基沉降曲线的预测效果良好,且精度均比反向传播神经网络模型(BP)、LSTM模型与只用PSO优化的LSTM模型(PSO-LSTM)有了进一步的提高,具有更高的鲁棒性和适用性。
【Purposes】This research is conducted to achieve accurate prediction of settlement trends for high fill subgrades in loess areas.【Methods】In this work,a Variational Mode Decomposition(VMD)and Particle Swarm Optimization(PSO)optimized Long Short-Term Memory(LSTM)network prediction model was proposed,referred to as VMD-PSO-LSTM model.This model was designed to learn high-level features of the settlement data for high fill subgrade and predict their developmental trends.To validate the effectiveness of the proposed model,an engineering case study was conducted.【Results】The results clearly indicate that the VMD-PSO-LSTM model performs well in predicting the settlement curves of high fill subgrade.Moreover,its accuracy surpasses that of the Back Propagation Neural Network(BP)model,the standard LSTM model,and the LSTM model optimized solely by PSO(PSO-LSTM),suggestings that the proposed VMD-PSO-LSTM model not only provides enhanced predictive accuracy but also demonstrates increased robustness and wider applicability.
作者
柴少波
岳山丘
王铭一
吕龙龙
范康凯
CHAI Shaobo;YUE Shanqiu;WANG Mingyi;LYU Longlong;FAN Kangkai(School of Civil Engineering,Chang’an University,Xi’an,Shaanxi,China;School of Civil and Hydraulic Engineering,Ningxia University,Yinchuan,Ningxia,China)
出处
《太原理工大学学报》
北大核心
2025年第5期907-915,共9页
Journal of Taiyuan University of Technology
基金
国家自然科学基金项目(41902277)
宁夏教育厅高等学校科学研究项目青年支持项目(NYG2024051)。
关键词
黄土
高填方路基
沉降预测
变分模态分解
长短期记忆网络
loess
high fill subgrade
settlement prediction
variational mode decomposition
long short-term memory network
