1 College of Civil Engineering, Tongji University, Shanghai 200092, China
2 Key Laboratory of Geotechnical and Underground Engineering of the Ministry of Education, Tongji University, Shanghai 200092, China
3 Guangxi Communication Design Group Co., Ltd., Nanning 530029, China
4 Zhejiang Jiaogong Group Co., Ltd., Hangzhou 310052, China
5 Shanghai Baoye Group Co., Ltd., Shanghai 200941, China
Show less
文章历史+
Received
Published
Revised
2025-09-20
2026-09-08
2026-01-19
Online Date
2026-09-18
PDF (14911K)
摘要
为提升交通基础设施的耐久性及安全性,有必要对路基变形进行精准预测。提出了一种融合自注意力机制(self-attention,SA)的双向长短期记忆网络(bidirectional long short-term memory,Bi-LSTM)预测方法,采用改进粒子群优化算法(improved particle swarm optimization,IPSO)进行了超参数优化,进而建立了一种路基变形(沉降和水平位移)的智能预测模型I-B-L-S(IPSO-Bi-LSTM-SA),并结合现场实测数据进行了对比验证。结果表明:在相同实测数据输入条件下,I-B-L-S模型的变形预测误差比BP(Back Propagation)模型、B-L(Bi-LSTM)模型及B-L-S(Bi-LSTM-SA)模型更小,预测结果变化趋势与实测数据更接近。基于上述预测结果分别计算了路基变形的性能评价指标:平均绝对误差、均方根误差、平均绝对百分误差和决定系数。所建立的I-B-L-S模型在4个预测性能评价指标上均表现出显著优势,说明其具有更好的准确性和鲁棒性,可为路基变形预测提供理论参考。
Abstract
To enhance the durability and safety of transportation infrastructure, accurate prediction of subgrade deformation is essential. A predictive method fusing a bidirectional long short-term memory (Bi-LSTM) network with a self-attention (SA) mechanism was proposed. An improved particle swarm optimization (IPSO) algorithm was employed for hyperparameter optimization to develop an intelligent prediction model for subgrade deformation (settlement and horizontal displacement), designated as IPSO-Bi-LSTM-SA (I-B-L-S). The established model was validated using field-measured data. The results indicate that under identical input conditions, the I-B-L-S model exhibited smaller prediction errors compared to the BP, Bi-LSTM (B-L), and Bi-LSTM-SA (B-L-S) models, and the predicted varying trend of results is closer to the measured data. Based on the predicted results, performance evaluation metrics for subgrade displacement, including mean absolute error, root mean square error, mean absolute percentage error, and the coefficient of determination, were calculated. The established I-B-L-S model demonstrated significant advantages across all four metrics, indicating that superior accuracy and robustness are achieved, and can provide a theoretical reference for subgrade deformation prediction.
YUAN Tian-gang, LÜ Xi-lin, LIU Xian-lin, MA Yi-yue, SHAO Yu, XU Ke-feng.
Intelligent Prediction Model for Subgrade Deformation Driven by Measured Data[J].
Science Technology and Engineering, 2026, 26(25): 11044-11056 DOI:10.12404/j.issn.1671-1815.2506768
传统LSTM模型为单向循环网络,其基本结构决定了信息流的单向性,只能接收前一时刻的记忆细胞状态信息,但未考虑下一时刻记忆细胞状态信息。双向长短期记忆网络(bidirectional long short-term memory,Bi-LSTM)通过建立两个独立的LSTM层,即向前传递层和向后传递层,可将两个方向的隐藏状态拼接和融合,如图2所示。其特征在于可获取双向输入信息,从而更好地获取序列数据信息的全局特征并保持长期记忆[32]。在向前传递阶段,网络沿时间正向序列(从时刻1到时刻t)计算,在向后传递阶段,网络则沿时间逆向序列(从时刻t到时刻1)计算,并保存两个阶段每个时刻的隐藏层输出。最后结合前向层和后向层的相应时刻输出结果得到最终结果。该过程可表示如下。
LiuXianlin, LüXilin, LanRiyan, et al. Identification-early warning-control integrated treatment method of landslide disasters along mountainous highways[J]. China Journal of Highway and Transport, 2025, 38(10): 21-35.
YangZhou, ChengXiaohui, GuanWen, et al. High fill embankment instability induced by groundwater seepage: a case study and finite element limit analysis[J]. China Journal of Highway and Transport, 2025, 38(5): 1-10.
ZhangPingsong, YuHongqing, XuShiang, et al. Review of highway subgrade deformation and testing technology[J]. Science Technology and Engineering, 2024, 24(15): 6134-6145.
XieYongli, DongGengzhi, YueXiabing, et al. Review and perspective on the state characteristics and engineering issues of fill subgrades[J]. China Journal of Highway and Transport, 2025, 38(8): 228-249.
WangJinchang, LongYuhang, FanZhihong, et al. Influence of post-construction settlement and countermeasures on soft foundation of Zhejiang highways[J]. Journal of Ground Improvement, 2025, 7(6): 574-583.
LiangBin, BuYiming, DengHonglong, et al. Analysis of the mechanism of settlement and slip of weak subgrade slope of expressway and micro-pile composite reinforcement[J]. Journal of North China University of Water Resources and Electric Power (Natural Science Edition), 2024, 45(3): 67-76.
YueZhongqi. Study on the instability condition and landslide mechanism of subgrade slope in Mei-Da Expressway[J]. The Chinese Journal of Geological Hazard and Control, 2024, 35(4): 1-12.
[15]
DuY, HeX J, WuC, et al. Long-term monitoring and analysis of the longitudinal differential settlement of an expressway bridge-subgrade transition section in a loess area[J]. Scientific Reports, 2022, 12(1): 19327.
YuYongtang, ZhengJianguo, SunMo, et al. Evaluation methods for performance of post-construction settlement prediction models in thick loess filled ground[J]. The Chinese Journal of Geological Hazard and Control, 2023, 34(4): 39-48.
CaoWengui, LiPeng, TangYini, et al. Layer-wise summation analysis method for settlement of ground foundation based on porous medium features[J]. China Journal of Highway and Transport, 2016, 29(3): 1-8, 65.
LüXilin, FangHang, ZhangJiafeng. A theoretical solution for long-term settlement of soft subgrade induced by traffic loading[J]. Rock and Soil Mechanics, 2016, 37(S1): 435-440.
HuJing, TangYue, ZhangJiakang, et al. Influences of water level rise on dynamic responses and long-term settlement of high-speed railway subgrade[J]. Journal of Traffic and Transportation Engineering, 2023, 23(4): 75-91.
ZhaoXiushao, ChengAn, ZhaoLinhao, et al. Mechanical properties of composite improved phyllite soil under uniaxial compression[J]. Journal of Traffic and Transportation Engineering, 2024, 24(3): 171-180.
XieZhiwei, MaYiyue, ZhangYan, et al. Experimental study on accumulative deformation characteristics of tunnel slag under cyclic dynamic loading considering particle breakage effect[J]. China Journal of Highway and Transport, 2024, 37(6): 144-156.
ChenTao, ChenXinguo, WuHuiming, et al. A case study on mechanism analysis and treatment measures of subgrade instability in soft soil area[J]. Journal of Ground Improvement, 2024, 6(2): 178-184.
WuJiangbin, WangWeidong, HuYun, et al. Design and implementation of high fill foundation treatment for the terminal zone in the north area expansion and renovation project of Urumqi International Airport[J]. Journal of Ground Improvement, 2025, 7(1): 1-9, 35.
ChaiShaobo, YueShanqiu, WangMingyi, et al. Settlement prediction of high fill subgrade in loess areas based on variational mode decomposition and particle swarm optimization of long short-term memory networks[J]. Journal of Taiyuan University of Technology, 2025, 56(5): 907-915.
[36]
HuangH, ChangJ, ZhangD, et al. Machine learning-based automatic control of tunneling posture of shield machine[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2022, 14(4): 1153-1164.
[37]
YeX, JinT, ChenY. Machine learning-based forecasting of soil settlement induced by shield tunneling construction[J]. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research, 2022, 124: 104452.
WangYaowen, YangWeifeng, ZhengXinyuan, et al. Optimization model for parameter value of Ningbo silt clay based on machine learning[J]. Journal of Engineering Geology, 2025, 33(1): 38-46.
LuWenxin, ZhangLei, LiuBo, et al. Prediction model of dam settlement based on improved EMD-MGWO-LSTM[J]. Science Technology and Engineering, 2024, 24(35): 15224-15232.
[44]
RongM, FengC, WangH, et al. Dynamic prediction of slope displacement using VMD decomposition with collaborative LSSVM-LSTM optimization[J]. Scientific Reports, 2025, 15(1): 35176.
[45]
DengZ, HeY, LiY, et al. Monitoring and prediction of surface deformation along railways lines in karst areas using multi-source data: a case study of the Beijing-Guangzhou Railway in China[J]. Transportation Geotechnics, 2025, 52: 101564.
ChenWeihang, LuoQiang, WangTengfei, et al. Bi-LSTM based rolling forecast of subgrade post-construction settlement with unevenly spaced time series[J]. Journal of Zhejiang University (Engineering Science), 2022, 56(4): 683-691.
XuJiangbo, HouXinmin, WuXiong, et al. Slope displacement prediction using MIC-XGBoost-LSTM model[J]. China Journal of Highway and Transport, 2024, 37(10): 38-48.
GeQi, WangDongfei, SunHongyue, et al. Displacement prediction of filling road bed slope surface based on dynamic spatial panels model[J]. Journal of Engineering Geology, 2019, 27(2): 367-375.
[54]
HochreiterS, SchmidhuberJ. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
LiBo, ZhangChuanhuang, YuXiao. Short-term traffic flow prediction based on VMD-CNN-BiLSTM model[J]. Journal of Wuhan University of Technology (Transportation Science & Engineering), 2026, 50(2): 250-257.
[59]
VaswaniA, ShazeerN, ParmarN, et al. Attention is all you need[C]// Advances in Neural Information Processing Systems 30 (NIPS 2017). Long Beach: ACM, 2017: 6000-6010.
[60]
EberhartR C, KennedyJ. A new optimizer using particle swarm theory[C]// Proceedings of the Sixth International Symposium on Micro Machine and Human Science. Nagoya: IEEE, 1995: 39-43.
SongMuyuan, YangMinghui, ChenWei, et al. Prediction of shield tunneling-induced soil settlement based on self-attention recurrent neural network model[J]. Rock and Soil Mechanics, 2025, 46(8): 2613-2625.
[65]
ArsalanM, RezaH N, MokhtarM, et al. Forecasting tunnel boring machine penetration rate using LSTM deep neural network optimized by grey wolf optimization algorithm[J]. Expert Systems with Applications, 2022, 209: 118303.
[66]
HamedF, BehrouzA, SahandH, et al. A data-driven structural damage identification approach using deep convolutional-attention-recurrent neural architecture under temperature variations[J]. Engineering Structures, 2023, 276: 115311.
[67]
WangZ, XiaoC, GohS H, et al. Metamodel-based reliability analysis in spatially variable soils using convolutional neural networks[J]. Journal of Geotechnical and Geoenvironmental Engineering, 2021, 147(3). DOI: 10.1061/(ASCE)GT.1943-5606.0002486.