PDF全文阅读

仪晓立1 , 王振军2, 侯向阳1 ,  惠冰2, 孙巍1 , 张旭2, 苗鑫2

(1. 中铁一局集团建设安装工程有限公司，陕西 西安 710000 ;

2. 山东省交通科学研究院， 山东 济南 250104)

摘 要：砂土液化导致地基承载力下降，合理判别砂土液化程度对防治地基下沉等灾害具有重要意义。借鉴机器学习方法，选取30组砂土液化数据样本，建立粒子群算法改进最小二乘支持向量机砂土液化判别模型，并与SVM砂土液化判别模型和BP砂土液化判别模型进行了对比分析。结果表明：LSSVM 模型通过PSO算法优化后确定正则化参数为323 . 125 247 535 、核参数为 1 . 015 053 246 5。对于15组训练样本，PSO-LSSVM砂土液化判别模型和SVM 砂土液化判别模型回判准确率为100% , BP砂土液化判别模型回判准确率为93.3%; 对于5组测试样本，PSO - LSSVM砂土液化判别模型预测准确率为100% , 而SVM砂土液化判别模型和BP 砂土液化判别模型预测准确率为80%; 在黄河流域砂土液化预测中，PSO - LSSVM 砂土液化判别模型具有更高的预测精度，可指导工程技术人员预测砂土状态制定防治措施。

关键词：粒子群算法；支持向量机；砂土液化；判别预测

中图分类号：TU435 

文献标志码：A 

文章编号：1005- 8249 (2024) 03- 0050- 05 

DOI：10.19860/j.cnki.issn1005-8249.2024.03.011

YI Xiaoli1 , WANG Zhenjun2, HOU Xiangyang1, HUI Bing2, SUN Wei 1, ZHANG Xu2, MIAO Xin2

1. China Railway First Group Building & Installation Engineering Co. ，LTD. ，Xi ’an 710000 , China;

2. Shandong Transportation Institute, Jinan 250104, China

Abstract : The liquefaction of sand soil causes the bearing capacity of the foundation to decrease. Reasonable judgment of the degree of sand liquefaction is of great significance to prevent and control disasters such as foundation subsidence. Drawing on machine learning methods , 30 sets of sand liquefaction data samples were selected to establish a sand liquefaction discrimination model improved by particle swarm optimization algorithm , and compared with SVM and BP sand liquefaction discrimination model. The results show that the LSSVM model is optimized by the PSO algorithm and determines the regularization parameter to be 323 . 125 247 535 and the kernel parameter to be 1 . 015 053 246 5 . For 15 sets of training samples , the PSO - LSSVM sand liquefaction discrimination model and SVM sand liquefaction discrimination model have a return accuracy of 100% , and the BP sand liquefaction discrimination model has a return accuracy of 93. 3% . For the five groups of test samples , the prediction accuracy of the PSO - LSSVM sand liquefaction discrimination model was 100% , while the prediction accuracy of the SVM sand liquefaction discrimination model and the BP sand liquefaction discrimination model was 80% . In the prediction of sand liquefaction in the Yellow River Basin , the PSO - LSSVM sand liquefaction discrimination model has higher prediction accuracy and can guide engineering and technical personnel to predict the state of sand and formulate prevention and control measures.

Keywords : particle swarm algorithm; support vector machine; sand liquefaction; discriminant prediction

*基金项目： 山东省交通运输科技计划项目 (2023B46) 。

作者简介：仪晓立 (1975—) , 男，本科，高级工程师，研究方向：土木工程。 

收稿日期：2023-10-06