Novel Graph Convolutional Network for Predicting Geological Profiles

doi:10.16183/j.cnki.jsjtu.2024.264

Abstract

Abstract:

The graph convolutional network (GCN) algorithm is applied to geological profile prediction, and an improved GCN algorithm is proposed through targeted modifications. This improved GCN algorithm can flexibly establish connection graphs based on unequal spacing and sparse boreholes within the site, enabling the prediction of two-dimensional geological profiles. A boundary accuracy metric is introduced to better evaluate the accuracy of the predicted geological profile. Applied to actual geological profile cases and compared with the existing Markov random field and IC-XGBoost methods, the proposed method improves both global accuracy and boundary accuracy, which indicates its accuracy in predicting two-dimensional geological profiles. Finally, the influence of the borehole spacing and unequal spacing on preliction results is also discussed. The results indicate that the reduction of the borehole spacing generally improves the accuracy of the predicted geological profile boundary. However, the relationship between the reduction of borehole spacing and the improvement in prediction accuracy is not simply linear. The improvement in prediction accuracy depends on the soil layer spatial information provided by the additional boreholes. When the number of boreholes is fixed, the influence of unequal spacing depends on the soil layer spatial information provided by the locations of the boreholes. The more soil layer spatial information provided by the boreholes, the more accurate the prediction will be. The improved GCN method applied in Hangzhou to a real case to predict the geological profile, with all the measurement accuracy of validated boreholes above 0.8.

Key words: graph convolutional network (GCN), sparse boreholes, unequal spacing, stratigraphic interface, boundary accuracy

CLC Number:

TU433

QIU Yashi, LIU Hongchi, CAO Huajin, FENG Guohui, YANG Kaifang, XU Changjie. Novel Graph Convolutional Network for Predicting Geological Profiles[J]. Journal of Shanghai Jiao Tong University, 2026, 60(6): 1008-1016.

Figures/Tables 9

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Tab.1

Tab.2

References 17

[1]	SHI C, WANG Y. Nonparametric and data-driven interpolation of subsurface soil stratigraphy from li-mited data using multiple point statistics[J]. Canadian Geotechnical Journal, 2021, 58(2): 261-280. doi: 10.1139/cgj-2019-0843 URL
[2]	QI X H, LIU H X. Estimation of autocorrelation distances for in-situ geotechnical properties using limited data[J]. Structural Safety, 2019, 79: 26-38. doi: 10.1016/j.strusafe.2019.02.003 URL
[3]	GONG W, ZHAO C, JUANG C H, et al. Coupled characterization of stratigraphic and geo-properties uncertainties—A conditional random field approach[J]. Engineering Geology, 2021, 294: 106348. doi: 10.1016/j.enggeo.2021.106348 URL
[4]	ZHAO C, GONG W, LI T, et al. Probabilistic cha-racterization of subsurface stratigraphic configuration with modified random field approach[J]. Engineering Geology, 2021, 288: 106138. doi: 10.1016/j.enggeo.2021.106138 URL
[5]	JUANG C H, ZHANG J, SHEN M, et al. Probabilistic methods for unified treatment of geotechnical and geological uncertainties in a geotechnical analysis[J]. Engineering geology, 2019, 249: 148-161. doi: 10.1016/j.enggeo.2018.12.010 URL
[6]	JIANG S H, HUANG J, QI X H, et al. Efficient probabilistic back analysis of spatially varying soil parameters for slope reliability assessment[J]. Engineering Geology, 2020, 271: 105597. doi: 10.1016/j.enggeo.2020.105597 URL
[7]	LI D Q, QI X H, CAO Z J, et al. Evaluating slope stability uncertainty using coupled Markov chain[J]. Computers and Geotechnics, 2016, 73: 72-82. doi: 10.1016/j.compgeo.2015.11.021 URL
[8]	QI X H, LI D Q, PHOON K K, et al. Simulation of geologic uncertainty using coupled Markov chain[J]. Engineering Geology, 2016, 207: 129-140. doi: 10.1016/j.enggeo.2016.04.017 URL
[9]	DENG Z P, LI D Q, QI X H, et al. Reliability eva-luation of slope considering geological uncertainty and inherent variability of soil parameters[J]. Computers and Geotechnics, 2017, 92: 121-131. doi: 10.1016/j.compgeo.2017.07.020 URL
[10]	WANG Y, ZHAO T. Statistical interpretation of soil property profiles from sparse data using Bayesian compressive sampling[J]. Géotechnique, 2017, 67(6): 523-536. doi: 10.1680/jgeot.16.P.143 URL
[11]	ZHAO T, HU Y, WANG Y. Statistical interpretation of spatially varying 2D geo-data from sparse measurements using Bayesian compressive sampling[J]. Engineering Geology, 2018, 246: 162-175. doi: 10.1016/j.enggeo.2018.09.022 URL
[12]	HU Y, WANG Y, ZHAO T, et al. Bayesian supervised learning of site-specific geotechnical spatial variability from sparse measurements[J]. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 2020, 6(2): 04020019. doi: 10.1061/AJRUA6.0001059 URL
[13]	HUANG Y, BECK J L, WU S, et al. Robust Bayesian compressive sensing for signals in structural health monitoring[J]. Computer-Aided Civil and Infrastructure Engineering, 2014, 29(3): 160-179. doi: 10.1111/mice.12051 URL
[14]	WEI X, WANG H. Stochastic stratigraphic modeling using Bayesian machine learning[J]. Engineering Geology, 2022, 307: 106789. doi: 10.1016/j.enggeo.2022.106789 URL
[15]	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. doi: 10.1061/(ASCE)GT.1943-5606.0002583 URL
[16]	QIU Y S, ZHANG N, YIN Z Y, et al. Novel multi-spatial receptive field (MSRF) XGBoost method for predicting geological cross-section based on sparse borehole data[J]. Engineering Geology, 2024, 338: 107604. doi: 10.1016/j.enggeo.2024.107604 URL
[17]	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[C]// International Conference on Learning Representations. Vancouver, British Columbia, Canada: Curran Associates, Inc.,2017: 231-245.

验证钻孔位置/m	A_M
验证钻孔位置/m	图7(b)	图7 (c)	图7 (d)
28	0.82	—	—
57	—	0.94	—
86	0.86	—	—
115	—	—	1
144	0.82	0.96	—
173	—	—	0.8
231	—	0.88	0.9
平均A_M	0.83	0.93	0.9

不同预测结果	A_G	A_B
图7(b)	0.88	0.512
图7(c)	0.89	0.527
图7(d)	0.89	0.526