深圳土工参数数据库及基于生成对抗网络的多元参数分布预测模型研究  

Shenzhen geotechnical parameter database and multivariate parameter distribution prediction model based on generative adversarial network

作　　者：潘秋景[1] 孙广灿 蔡永敏 苏栋[3] 李凤伟 PAN Qiu-jing;SUN Guang-can;CAI Yong-min;SU Dong;LI Feng-wei(School of Civil Engineering,Central South University,Changsha,Hunan 410075,China;Department of Architecture and Sustainable Design,Singapore University of Technology and Design,Singapore 487372;College of Civil and Transportation Engineering,Shenzhen University,Shenzhen,Guangdong 518060,China;China Railway Construction Corporation Limited,Beijing 100855,China)

机构地区：[1]中南大学土木工程学院,湖南长沙410075 [2]新加坡科技设计大学建筑学与可持续设计系,新加坡487372 [3]深圳大学土木与交通工程学院,广东深圳518060 [4]中国铁建股份有限公司,北京100855

出　　处：《岩土力学》2025年第2期563-572,共10页Rock and Soil Mechanics

基　　金：国家自然科学基金(No.52108388,No.52378424);国家重点研发计划(No.2023YFC3009300);湖南省科技创新计划(No.2021RC3015);深圳大学2035追求卓越研究计划(No.2022B007)。

摘　　要：借鉴大数据思想,充分利用岩土工程勘察数据,实现岩土参数精细化表征和建模,是岩土工程数字孪生的重要组成部分。通过收集深圳市75个工程项目的岩土工程勘察报告,建立了深圳黏性土及风化残积土8个土工试验参数数据库SZ-SOIL/8/11369,分析了深圳黏性土及风化残积土土工试验参数的分布特征和规律。进一步利用该数据库,提出了基于生成对抗网络(generative adversarial network,简称GAN)的土工试验物理力学参数概率分布及预测模型,并将提出的方法应用于深圳某项目,针对单组土工试验样本利用物理参数成功预测了其力学参数,并利用少量样本正确预测了该工程场地的土工试验参数的分布。结果表明,所提方法能够对缺失参数样本进行合理预测,并实现了通过大范围地区勘察数据降低局部工程场地岩土参数不确定性的目的,可为深圳岩土与地下工程结构韧性设计和风险评价提供参数保障。Inspired by big data,fully utilizing geotechnical data for precise characterization and modeling of geotechnical parameters is critical for the digitalization of geotechnical engineering.This study collected geotechnical investigation reports from 75 engineering projects in Shenzhen,established a database containing 8 geotechnical parameters of clay and weathered residual soil(SZ-SOIL/8/11369),and thoroughly analyzed the distribution characteristics of geotechnical parameter data in Shenzhen.Subsequently,a model for predicting geotechnical parameters was developed using this database and a generative adversarial network(GAN).The proposed method was applied to a project in Shenzhen,successfully predicting mechanical parameters from known physical parameters and accurately forecasting the geotechnical parameter distribution of the project site using small samples.The results indicate that the proposed method can make reasonable predictions for samples with missing parameters,achieving the goal of reducing the uncertainty in geotechnical parameters at local engineering sites through extensive regional survey data.This provides parameter assurance for the resilience design and risk assessment of geotechnical and underground engineering structures in Shenzhen.

关 键 词：土工参数分布 数据库 预测 生成对抗网络 

分 类 号：TU413[建筑科学—岩土工程]