设计、施工、监测

基于FCN的地层信息AI重构方法的实现及应用

  • 冯伟健 ,
  • 陆勇 ,
  • 顾琳琳 ,
  • 曹宇鹏 ,
  • 范存新
展开
  • 1.苏州科技大学 土木工程学院,江苏 苏州 215011;
    2.南京理工大学 土木工程系,南京 210094;
    3.日本国立京都大学 城市管理系, 东京 606-8511
冯伟健(2000—),男,江苏扬州人,主要从事人工智能方法在岩土工程中的应用研究工作。E-mail: wjfeng0422@163.com

收稿日期: 2025-06-18

  网络出版日期: 2026-04-28

基金资助

国家重点研发计划项目(2023***3106500);国家自然科学研究基金(52331010);江苏省高等学校基础科学(自然科学)研究项目资助(22KJB170020)

The Realization and Application of AI Reconstruction Method of Stratum Information Based on FCN

  • Feng Weijian ,
  • Lu Yong ,
  • Gu Linlin ,
  • Cao Yupeng ,
  • Fan Cunxin
Expand
  • 1. School of Civil Engineering, Suzhou University of Science and Technology, Suzhou, Jiangsu 215011, P.R China;
    2. Department of Civil Engineering, Nanjing University of Science and Technology, Nanjing 210094, P.R. China;
    3. The Department of Urban Management, Department of Urban Management, National Kyoto University of Japan, Kyoto 606-8511, Japan

Received date: 2025-06-18

  Online published: 2026-04-28

摘要

通过岩土工程勘察获取一定量的钻孔数据继而确定地层剖面信息,这对于岩土工程精细化设计、施工至关重要。然而,部分工程场地因所处区域的特殊性(既有老城区、文保建筑区等)往往存在钻孔数据难以获取的问题,使得相应的地层信息确定存在挑战。为此,基于全卷积网络(FCN)开发了用于地层信息重构的人工智能(AI)方法,该方法的核心思路是将区域既有钻孔数据作为学习样本,分析并提取样本的多维信息特征(竖向成层、水平延展),继而将该信息特征作为模板针对仅有少量钻孔数据的工程场地进行基于概率的地层剖面信息插值重构。通过对苏州古城区某隧道工程与基坑工程的地勘数据进行学习与重构,发现地层预测的准确度在模拟次数增加至30次以上后逐渐趋于稳定,并可达90%左右。这验证了所开发的地层信息AI重构方法的适用性,将能为相关工程的复杂地层信息预测提供有效选择。

本文引用格式

冯伟健 , 陆勇 , 顾琳琳 , 曹宇鹏 , 范存新 . 基于FCN的地层信息AI重构方法的实现及应用[J]. 地下空间与工程学报, 2026 , 22(2) : 653 -663 . DOI: 10.20174/j.JUSE.2026.02.27

Abstract

It is quite important for the refined design and construction of geotechnical engineering to obtain a certain amount of borehole data and then determine the stratum profile information through geotechnical investigation. However, due to the particularity of the area where some engineering sites are located (existing old urban areas, cultural Relic Protection Building, etc.), there is often a problem that the borehole data is difficult to obtain, which makes the corresponding stratum information determination challenging. To this end, an artificial intelligence (AI) method for stratum information reconstruction is developed based on the fully convolutional network (FCN). The core idea of this method is to use the existing borehole data in the region as a learning sample, analyze and extract the multi-dimensional information features of the sample (vertical stratification, horizontal extension), and then use this information feature as a template to perform probability-based stratum profile information interpolation reconstruction for engineering sites with only a small amount of borehole data. Through the study and reconstruction of the geological survey data of a tunnel project and foundation pit project in the ancient city of Suzhou, it is found that the accuracy of stratum prediction gradually tends to be stable after the number of simulations increases to more than 30 times, and can reach about 90%. This verifies the applicability of the developed AI reconstruction method of stratum information, which will provide an effective choice for the prediction of complex stratum information in related projects.

参考文献

[1] 安晓晓,李云燕.基于应用场景的城市深部地下空间规划模式研究[J].地下空间与工程学报,2024,20(3):701-709. (An Xiaoxiao,Li Yunyan. Research on urban deep underground space planning model based on application scenarios[J].Underground space and engineering,2024,20(3):701-709.(in Chinese))
[2 ] Giardina G, Hendriks M A N, Rots J G. Numerical analysis of tunnelling effects on masonry buildings: the influence of tunnel location on damage assessment[J]. Advanced Materials Research, 2010, 133: 289-294.
[3] 沈珠江. 土体结构性的数学模型──21世纪土力学的核心问题 [J]. 岩土工程学报,1996, 18(1): 95-97. (Shen Zhujiang. Mathematical model for soil structure—The core topic of soil mechanics in the 21st century[J]. Chinese Journal of Geotechnical Engineering, 1996, 18(1): 95-97. (in Chinese))
[4] 刘松玉,蔡国军,张炜,等.岩土工程勘察、测试与评价进展[J].土木工程学报,2024:1-18.(Liu Songyu, Cai Guojun, Zhang Wei, et al. Progress of geotechnical Investigation, Test and evaluation [J]. Chinese Journal of Civil Engineering, 2024:1-18. (in Chinese))
[5] Lu C, Wu X, Wang W. Modern drilling managementsystem based on field data monitoring[J]. Earth Science Informatics, 2018, 11: 403-412.
[6] 赵晨德,王心义,任君豪,等.基于深度学习理论的煤层底板突水危险性预测[J].地下空间与工程学报,2023,19(6):2090-2100.(Zhao Chende, Wang Xinyi, Ren Junhao, et al. Prediction of water inrush risk of coal seam floor based on deep learning theory [J].Underground Space and Engineering Journal, 2023,19 (6): 2090-2100.(in Chinese))
[7] Sun S, Zhang L,Wang H, et al. A new method for predicting the shale distribution of the Wufeng Formation in the Upper Yangtze Region, China[J]. BSGF-EARTH Science Bulletin, 2020,191(1):8.
[8] Han L, Wang L, Zhang W, et al. Rockhead profile simulation using an improved generation method of conditional random field[J]. Journal of Rock Mechanics and Geotechnical Engineering,2022,14(3): 896-908.
[9] Li J, Cassidy M J, Huang J, et al. Probabilistic identification of soil stratification[J]. Géotechnique, 2016,66(1): 16-26.
[10] 郭福钟,郑博文,祁生文,等.三维地质建模技术与方法综述[J].工程地质学报,2024,32(3):1143-1153.(Guo Fuzhong, Zheng Bowen, Qi Shengwen, et al. Review of 3D geological modeling techniques and methods [J]. Journal of Engineering Geology, 2024,32 (3): 1143-1153.(in Chinese))
[11] Zhang Y, Ji W,Saurette D D, et al. Three-dimensional digital soil mapping of multiple soil properties at a field-scale using regression kriging[J]. Geoderma, 2020,366: 114253.
[12] De Benedetto D, Barca E, Castellini M, et al. Prediction of soil organic carbon at field scale by regression kriging and multivariate adaptive regression splines using geophysical covariates[J]. Land, 2022,11(3): 381.
[13] 温嘉琦,汤雷.岩石平面应变状态辨识特征与精度控制[J].地下空间与工程学报,2024,20:471-479.(Wen Jiaqi, Tang Lei. Identification characteristics and precision control of rock plane strain state[J]. Chinese Journal of Underground Space and Engineering,2024,20:471-479.(in Chinese))
[14] 仉文岗,王琦,刘汉龙,等.岩体空间变异性对隧道拱顶失效概率的影响[J].岩土力学,2021,42:1462-1472.(Zhang Wengang, Wang Qi, Liu Hanlong, et al. Effect of spatial variability of rock mass on failure probability of tunnel vault[J]. Rock and Soil Mechanics,2021,42:1462-1472.(in Chinese))
[15] Gao H, Zhang J, Liu C, et al. Efficientbayesian inverse modeling of water infiltration in layered soils[J]. Vadose Zone Journal, 2019, 18(1): 1-13.
[16] 王宇超,谢雄耀,黄昌富,等.基于混合深度学习算法的盾构运动轨迹实时预测[J].地下空间与工程学报,2024,20(增1):59-69.(Wang Yuchao, Xie Xiongyao, Huang Changfu, et al. Real-time prediction of shield motion trajectory based on hybrid deep learning algorithm[J]. Chinese Journal of Underground Space and Engineering, 2019,20(Supp.1):59-69. (in Chinese))
[17] 朱合华,潘柄屹,武威,等.岩体结构面信息采集及识别方法研究进展[J].应用基础与工程科学学报,2023,31(6):1339-1360.(Zhu Hehua, Pan Pingyi, Wu Wei, et al. Research progress of rock mass structural plane information acquisition and identification methods [J]. Chinese Journal of Basic and Engineering Sciences,2023,31(6):1339-1360.(in Chinese))
[18] Jordan M I, Mitchell T M. Machine learning: Trends, perspectives, and prospects[J]. Science,2015,349(6245): 255-260.
[19] Ali S, Wani M A. Gradient-Based Neural Architecture Search: A Comprehensive Evaluation[J]. Machine Learning and KnowledgeExtraction, 2003,5(3): 1176-1194.
[20] 刘汉龙, 肖杨, 崔允亮. 结构性软土三维弹塑性损伤本构模型研究[J]. 岩土工程学报, 2011, 33(4): 637-642.(Liu Hanlong,Xiao Yang,Cui Yunliang. Elasto-plastic damage constitutive model in three-dimensional stress space for structured soft soils[J]. Chinese Journal of Geotechnical Engineering,2011,33(4): 637-642. (in Chinese))
[21] Elkatatny S, Ahmed A, Abughaban M, et al. Deep illustration for loss of circulation while drilling[J]. Arabian Journal for Science and Engineering, 2020,45: 483-499.
[22] Wu J, Boucher A, Zhang T. ASGeMS code for pattern simulation of continuous and categorical variables: FILTERSIM[J]. Computers & Geosciences, 2008,34(12): 1863-1876.
[23] Pérez C, Mariethoz G, Ortiz J M. Verifying the high-order consistency of training images with data for multiple-point geostatistics[J]. Computers & Geosciences, 2014,70: 190-205.
[24] Mariethoz G, Caers J. Multiple-point geostatistics: stochastic modeling with training images[M]. John Wiley & Sons,2014.
[25] Shi C, Wang Y. Non-parametric machine learning methods for interpolation of spatially varying non-stationary and non-Gaussian geotechnical properties[J]. Geoscience Frontiers, 2021,12(1): 339-350.
[26] Shi C, Wang Y. Development of subsurface geological cross-section from limited site-specific boreholes and prior geological knowledge using iterative convolution XGBoost[J]. Journal of Geotechnical and Geoenvironmental Engineering, 2021,147(9): 04021082.
[27] Shi C, Wang Y. Training image selection for development of subsurface geological cross-section by conditional simulations[J].Engineering Geology, 2021,295: 106415.
[28] 程德强,程志威,马祥,等.基于渐近式感受野的轻量级图像超分辨率重建方法[J].控制与决策,2025,40(4): 1237-1246. (Cheng Deqiang, Cheng Zhiwei, Ma Xiang, et al. Lightweight image super-resolution reconstruction method based on progressive receptive field[J]. Control and Decision, 2025,40(4): 1237-1246. (in Chinese))
[29] Valsesia D, Fracastoro G, Magli E. Image denoising with graph-convolutional neural networks[A]//2019 IEEE International Conference on Image Processing (ICIP)[C]. IEEE, 2019: 2399-2403.
[30] 伍浩,邓丽钰,刘文源,等.考虑空间信息的地连墙变形智能动态预测方法[J].地下空间与工程学报,2025,21(增1):95-101.([J].Chinese Journal of Underground Space and Engineering,2025,21(Supp.1):95-101.(in Chinese))
[31] Guardiano F B, Srivastava R M. Multivariate geostatistics: beyond bivariate moments[J].Quantitative Geology and Geostatistics, 1993: 133-144.
文章导航

/