Physical and Mechanical Properties and BPNN Prediction Model of Silty Soil in Coastal Area of Jiangsu Province
-
摘要: 江苏沿海地层以第四系海相或海陆相交互沉积成因的粉土和淤泥质粉土为主，地下水位高，工程性质较差，直接影响电网基础设施工程建设与后期运营安全。根据现场原位测试和室内试验数据，分析了江苏沿海地区粉土物理力学指标的变化范围、变异性以及指标之间的相互关系，土体黏聚力等力学指标的变异系数值在0.5附近，明显大于土体密度、液塑限等物性指标的变异系数；应用BPNN神经网络方法建立了粉土土性指标的预测模型，结果表明该模型预测效果良好，能反映粉土物理与力学性质指标之间的非线性关系，采用均方根误差（RMSE）和平均绝对误差（MAE）两个指标评价，土体压缩指标的预测精度要优于剪切强度指标。Abstract: Shallow soil deposits in coastal area of Jiangsu Province are mainly composed of silty soil and muddy silty soil that formed by Quaternary marine sediment or marine and terrestrial sediment. High groundwater level and poor geotechnical properties of the soil will directly affect the infrastructure construction and later operation safety. Based on the in-situ and lab test data of several new substation projects in coastal area of Jiangsu, the change ranges, variability and correlations of physical and mechanical characteristics of the typical silt soil were summarized. It shows that coefficient of variation of mechanical indexes (such as soil cohesion etc.) is about 0.5, which is significantly greater than that of physical indexes (such as soil density and liquid plastic limit etc.). A BPNN model of silty soil was constructed by using neural network method and the predicted result is good for reflecting the complex non-linear relationship between silty soil parameters. The prediction accuracy of soil compression indexes are better than that of shear strength indexes by using RMSE and MAE.
-
表 2 粉土土性指标概率分布特征
概率分布 天然密度
/(g·cm−3)含水率
/%饱和度
/%孔隙比 液限
/%塑限
/%液性
指数塑性
指数黏聚力
/kPa内摩擦角
/(°)压缩系数
/MPa−1压缩模量
/MPa${\chi ^{\text{2}}}$ 8.0 3.9 24.5 9.2 23.2 0.8 34.5 14.1 15.7 13.6 183.7 35.3 偏度 −0.52 0.05 −80.99 −0.33 −0.01 −0.15 0.47 0.44 0 0 1.68 0.04 峰度 0.52 −0.08 4.88 −2.44 −0.82 −0.27 0.99 −0.46 −0.5 −0.5 −1.33 −3.01 正态检验 接受 接受 拒绝 接受 拒绝 接受 拒绝 拒绝 拒绝 拒绝 拒绝 拒绝 
下载: 导出CSV
表 3 BPNN模型预测结果汇总
评估指标 压缩系数
/MPa−1压缩模量
/MPa黏聚力
/kPa内摩擦角
/(°)训练集 RMSE 0.05 1.91 9.17 4.94 MAE 0.03 1.37 6.61 4.00 测试集 RMSE 0.09 2.17 9.71 6.53 MAE 0.07 1.56 7.60 5.58
下载: 导出CSV
-
[1] 沈珠江. 软土工程特性和软土地基设计[J]. 岩土工程学报,1998,20(1):100-111. doi: 10.3321/j.issn:1000-4548.1998.01.025 [2] 梅芹芹,龚绪龙,史雅栋,等. 江苏沿海盐渍土盐渍化程度及其影响因素分析[J]. 工程地质学报,2020,28(5):959-965. doi: 10.13544/j.cnki.jeg.2020-247 [3] 朱志铎,刘松玉,孙海军. 江苏徐宿地区粉土的基本特性及加固方法研究[J]. 岩土力学,2004,25(7):1155-1158. doi: 10.3969/j.issn.1000-7598.2004.07.033 [4] 王文军,刘用海,朱向荣. 宁波海相软土工程特性研究[J]. 工程勘察,2008,(10):19-24. [5] 董冬冬. 江苏沿海地区土性参数统计分析与随机特性研究[D]. 南京: 东南大学, 2017. [6] 汪名鹏. 江苏泗阳城区浅层粉土工程地质特性分析[J]. 地下空间与工程学报,2018,14(2):452-460. [7] 宋许根,王志勇,柏威伟,等. 珠海西部中心城区大面积深厚软土工程特性研究[J]. 岩石力学与工程学报,2019,38(7):1434-1451. doi: 10.13722/j.cnki.jrme.2018.1331 [8] TIZPA P,CHENARI R J,FARD M K,et al. ANN prediction of some geotechnical properties of soil from their index parameters[J]. Arabian Journal of Geosciences,2014,8(5):1-10. [9] LIU L,MOAYEDI H,RASHID A S A,et al. Optimizing an ANN model with genetic algorithm (GA) predicting load-settlement behaviours of eco-friendly raft-pile foundation (ERP) system[J]. Engineering with Computers,2020,36(1):421-433. doi: 10.1007/s00366-019-00767-4 [10] 李旭昶,孙德安,徐永福. 扬州土物理与力学性质指标的相关性分析[J]. 工程地质学报,2019,27(2):333-340. [11] 王才进,张 涛,骆俊晖,等. 神经网络反馈分析方法预测土体热阻系数研究[J]. 岩土工程学报,2019,41(S2):109-112. [12] 郭子奇,杨双锁,李彦斌,等. 基于PSO-BP神经网络的地铁盾构场地土体参数反演[J]. 太原理工大学学报,2020,51(2):171-176. [13] 章定文,刘松玉,于新豹. 连云港海相软土工程特性及处治方法探讨[J]. 工程地质学报,2003,11(3):250-257. doi: 10.3969/j.issn.1004-9665.2003.03.005 [14] 冯秀丽,沈渭铨,杨荣民,等. 现代黄河水下三角洲软土沉积物工程地质特性[J]. 中国海洋大学学报,1994,(S3):132-137. [15] 杜 星,孙永福,宋玉鹏,等. 基于MPL神经网络的地震作用下砂土液化评估及预测[J]. 工程地质学报,2020,28(6):1458-1465. doi: 10.13544/j.cnki.jeg.2019-321 -
图(7) / 表(3)
计量
- 文章访问数: 106
- HTML全文浏览量: 20
- PDF下载量: 22
- 被引次数: 0
