基于IPSO-Bi-LSTM的油页岩粉煤灰改良土路基抗变形预测研究  

Research on Deformation Resistance Prediction of Oil Shale Fly Ash Improved Soil Roadbed Based on IPSO-Bi-LSTM

作　　者：陈汉 CHEN Han(China Railway 22nd Bureau Group Co.Ltd.,Beijing 100043,China)

机构地区：[1]中铁二十二局集团有限公司,北京100043

出　　处：《铁道建筑技术》2024年第12期55-58,79,共5页Railway Construction Technology

基　　金：中铁二十二局集团有限公司科技研究开发计划项目(2021-01A)。

摘　　要：油页岩粉煤灰改良土(OFMS)路基填料具有良好的抗冻性。为探究季冻区服役中其强度及抗变形变化规律,以长春地铁6号线地铁车站市政道路采用OFMS填料回填工程为依托,结合OFMS动态回弹模量试验结果,综合不同条件因素耦合作用,建立基于IPSO-Bi-LSTM的动态回弹模量预测模型,并与PSO-Bi-LSTM、PSO-SVR结果进行预测对比。研究表明:OFMS动态回弹模量满足土质路基规范要求,三种模型测试集平均相对误差分别为5.477%、6.097%与10.209%,在OFMS动态回弹模量预测中均有较好表现,其中IPSO-Bi-LSTM测试集MAE为5.057,RMSE为6.008,R^(2)为0.984,表现相对优异。研究中含水率、压实度、围压与应力条件等因素作为影响参数较为合理,能够有效反映复杂环境影响下的变化情况。综上,IPSO-Bi-LSTM对于多因素耦合的OFMS动态回弹模量预测具有较高精度,可为OFMS路基寿命预测和评估提供可靠依据,并拓展其工程应用。Oil shale fly ash improved soil(OFMS)roadbed filler has good frost resistance.In order to explore the change rule of strength and deformation resistance during service in seasonal freezing area,this study is based on the backfill project of OFMS filler in the municipal road of Changchun Metro Line 6.In combination with the results of OFMS dynamic rebound modulus tests,and taking into account the coupling effects of different conditions and factors,a dynamic rebound modulus prediction model based on IPSO-Bi-LSTM(improved particle swarm optimization-bidirectional long-short term memory neural network)was established,and compared with PSO-Bi-LSTM and PSO-SVR for prediction.The results show that:OFMS dynamic resilience modulus meets the specification requirements of soil roadbed.The average relative errors of the three model test sets are 5.477%,6.097%and 10.209%,respectively,showing good performance in the prediction of OFMS dynamic resilience modulus.Among them,IPSO-Bi-LSTM test set MAE is 5.057,RMSE is 6.008,and R^(2) is 0.984,which is relatively excellent.In the study,water content,compaction degree,confining pressure and stress conditions are more reasonable as influencing parameters,which can effectively reflect the changes of complex environmental influences.In conclusion,IPSO-Bi-LSTM can predict the dynamic resilience modulus of OFMS with high accuracy,and can provide a reliable basis for predicting and evaluating the life of OFMS roadbed,and expand its engineering application.

关 键 词：道路工程 油页岩粉煤灰改良土 动态回弹模量 抗变形预测 双向长短时记忆神经网络 

分 类 号：U416.1[交通运输工程—道路与铁道工程]