文章摘要信息

[an error occurred while processing this directive]

世界地质 2023, 42(3) 577-587 DOI: ISSN: 1004-5589 CN: 22-1111/P

本期目录 | 下期目录 | 过刊浏览 | 高级检索 [打印本页] [关闭]

基础地质

	扩展功能

	本文信息
	Supporting info
	PDF(3761KB)
	[HTML全文]
	参考文献[PDF]
	参考文献
	服务与反馈
	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	引用本文
	Email Alert
	文章反馈
	浏览反馈信息
	本文关键词相关文章

	LASSO 算法
	BP 神经网络模型
	滆湖组黏性土
	工程地质特性
	抗剪强度
	本文作者相关文章
	顾春生
	许书刚
	杨鹏
	唐鑫
	张其琪
	李浩民
	PubMed
	Article by Gu C
	Article by Xu S
	Article by Yang P
	Article by Tang X
	Article by Zhang Q
	Article by Li H

基于LASSO??BP 神经网络模型的滆湖组黏性土抗剪强度预测

顾春生1,2 , 许书刚1,2 , 杨鹏3 , 唐鑫1,2 , 张其琪1,2 , 李浩民1,2

1. 江苏省地质调查研究院, 南京 210080;

2. 自然资源部地裂缝地质灾害重点实验室, 南京 210080;

3. 上海市城市建设设计研究总院 (集团) 有限公司, 上海 200125

摘要：为了研究苏锡常地区滆湖组黏性土工程地质特性, 建立能够描述其抗剪强度特性的预测模型, 笔者选取 727 组土工试验数据作为研究样本, 运用 LASSO 算法筛选出对抗剪强度影响显著的参数, 然后结合 BP 神经网络算法建立抗剪强度预测模型? 结果表明: ①滆湖组黏性土的黏聚力? 内摩擦角与液性指数? 压缩系数呈反相关, 与塑性指数? 饱和重度呈正相关; ②确定液性指数? 塑性指数? 压缩系数? 颗粒比重? 饱和重度等 5 个参数为抗剪强度的预测参数, 建立了基于 LASSO??BP 神经网络算法的滆湖组黏性土抗剪强度预测模型; ③黏聚力预测模型的拟合优度 R = 0. 92, 平均绝对误差 < 5. 4 kPa; 内摩擦角预测模型的拟合优度 R = 0. 75, 平均绝对误差 < 1. 7°? 模型预测精度及可靠性总体较高, 能够对滆湖组黏性土抗剪强度进行有效预测? 预测模型引入 LASSO 算法, 为运用数学方法研究土体参数间关系提供参考?

关键词： LASSO 算法 BP 神经网络模型滆湖组黏性土工程地质特性抗剪强度

Prediction of shear strength of cohesive soil in Gehu Formation based on LASSO??BP neural network model

GU Chunsheng 1,2 , XU Shugang 1,2 , YANG Peng 3 , TANG Xin 1,2 , ZHANG Qiqi 1,2 , LI Haomin 1,2

1. Geological Survey of Jiangsu Province, Nanjing 210080, China;

2. Key Laboratory of Earth Fissures Geological Disaster, Ministry of Natural Resources, Nanjing 210080, China;

3. Shanghai Urban Construction Design & Research Institute (Group) Co., Ltd., Shanghai 200125, China

Abstract: In order to study the engineering geological characteristics of cohesive soil in Gehu Formation in Suzhou??Wuxi??Changzhou area, and establish a prediction model of its shear strength, LASSO (The Least Absolute Shrinkage and Selection Operator) algorithm is used to identify the physical parameters that have great influence on shear strength, based on 727 sets of geotechnical test data. Then, a prediction model of shear strength of cohesive soil is established by the BP neural network model. Finally, the main results are as follows: ①The cohesion and internal friction angle of cohesive soil in Gehu Formation are negatively correlated with liquid index and compression coefficient, but positively correlated with plastic index and saturated unit weight. ②The following five parameters, liquid index, plastic index, compression coefficient, particle specific gravity and saturated unit weight, are identi? fied as the prediction parameters. Then a prediction model of shear strength of cohesive soil in Gehu Formation has been built based on the LASSO ??BP neural network algorithm. ③The goodness of fit of the prediction model for co? hesion is 0. 92, and the mean absolute error is less than 5. 4 kPa. The goodness of fit of the internal friction angle of prediction model is R = 0. 75, and the mean absolute error is less than 1. 7°. These results mean that the model has high accuracy and stability. In summary, the prediction model introduces LASSO algorithm, which can provide a reference for using mathematical methods to analyze the relationship among geological parameters.

Keywords: LASSO algorithm BP neural network model cohesive soil in Gehu Formation engineering geo? logical characteristics shear strength

收稿日期 修回日期 网络版发布日期

DOI:

基金项目:

通讯作者:

作者简介:

作者Email:

参考文献：

本刊中的类似文章

1．刘元, 李尚启, 张旗, 魏博, 李保辉, 常昊. 西昆仑北缘西河休地区中奥陶世中酸性岩体成因及其形成构造环境[J]. 世界地质, 2024,42(4): 595-610

2．吴海波,侯艳平,王璞珺,陈鸿平,申文静,彭威,陈崇阳. 海拉尔盆地红旗凹陷侏罗系油气来源及其勘探意义 [J]. 世界地质, 2024,42(4): 676-690

3．罗冀,张延军,刘强斌,李晓光,娄俊庆. 螺旋叶片强化中深层地热同轴换热器换热性能数值模拟 [J]. 世界地质, 2024,42(4): 740-748