>> 自然科学版 >> 当期目录 >> 地质•测绘•环境 >> 正文
基于稀疏地勘数据的岩土地层信息智能解析研究
时间: 2026-11-20 次数:

姜秀峰,庄培芝,钱远顺, 等.基于稀疏地勘数据的岩土地层信息智能解析研究[J].河南理工大学学报(自然科学版),2027,46(1):70-78.

Jiang X F , Zhuang P Z, Qian Y S,et al.Intelligent analysis of stratigraphic information using sparse geological survey data[J].Journal of Henan Polytechnic University(Natural Science) ,2027,46(1):70-78.

基于稀疏地勘数据的岩土地层信息智能解析研究

姜秀峰1, 庄培芝2, 钱远顺2, 厉超吉2, 石光3, 张明晶3

1.山东高速建设管理集团有限公司,山东 济南 250098;2.山东大学 齐鲁交通学院,山东 济南  250002;3.山东省交通规划设计院集团有限公司,山东 济南  250101

摘要: 目的  地质勘查中勘查数据稀疏时传统数据处理手段和深度学习技术难以准确高效地获取岩土地层空间分布特征,基于此,开展基于稀疏地勘数据的岩土地层信息智能解析研究。 方法  使用矩阵方法对已知稀疏地勘数据进行转化,采用循环卷积集成算法对参考图进行地层分布特征提取,并使用提取特征对分类器进行训练,随后基于训练的分类器对待测地层图进行多尺度循环插值计算,最终形成二维待测岩土地层空间分布。进一步地,在2张互相垂直的参考地层图基础上,采用本文方法对三维空间的地层分布进行特征提取,并将提取特征与稀疏地勘数据进行循环插值计算,最终实现三维地层模型的预测构建。 结果 以实际地层分布图为标准计算预测精度,在使用单张地层参考图和6条稀疏数据的二维验证算例中,本文方法预测准确度为96%;在基于2张地层参考图和36条稀疏地勘数据的三维验算算例中,本文方法预测准确度为92.3%,说明本文方法预测精度高于现行地质勘查处理手段和传统机器学习方法。 结论  本文算法可有效提取地层空间分布特征,可仅使用稀疏地勘数据对地层空间分布进行高效准确预测,预测精度和计算效率满足实际勘察精度要求和工程需求,并可进一步扩展到地质灾害防治和矿产开发中。

关键词:稀疏数据;卷积神经网络;集成学习算法;地层重构

doi:10.16186/j.cnki.1673-9787.2024110035

基金项目:国家自然科学基金资助项目(52108374);山东省交通运输厅科技项目(2022B82)

收稿日期:2024/11/19

修回日期:2025/02/24

出版日期:2026/11/20

Intelligent analysis of stratigraphic information using sparse geological survey data

Jiang Xiufeng1, Zhuang Peizhi2, Qian Yuanshun2, Li Chaoji2, Shi Guang3, Zhang Mingjing3

1.Shandong Hi-Speed Construction Management Group Co., Ltd., Jinan  250098,Shandong,China;2.School of  Qilu Transportation, Shandong University, Jinan  250002,Shandong,China;3.Shandong Provincial Communication Planning and Design Institute Group Co., Ltd., Jinan  250101,Shandong,China

Abstract: Objectives  A method for predicting the spatial distribution of strata from sparse geological survey data was developed to address the difficulty of obtaining accurate and efficient stratigraphic distributions using conventional geological survey data processing and deep-learning methods. Methods  Sparse survey data were first transformed into matrix form. A convolutional integration algorithm was then used to extract stratigraphic distribution features from reference maps, and the extracted features were used to train a classifier. The trained classifier was subsequently used to perform multiscale cyclic interpolation on maps for which stratigraphic distributions were to be predicted, generating two-dimensional stratigraphic distributions. In addition, features of the three-dimensional distribution of strata were extracted from two mutually perpendicular reference maps and cyclically interpolated with the sparse survey data to construct a three-dimensional stratigraphic model. Results  When the actual stratigraphic distribution map was used as the reference, the prediction accuracy was 96% in a two-dimensional validation example based on one reference map and six sparse data points. In a three-dimensional validation example based on two reference maps and 36 sparse survey data points, the prediction accuracy was 92.3%. The proposed method achieved higher prediction accuracy than existing geological survey data processing and conventional machine-learning methods. Conclusions  The method can effectively extract stratigraphic distribution features and predict stratigraphic distributions accurately and efficiently from sparse survey data. Its prediction accuracy and computational efficiency meet practical survey and engineering requirements. The method may also be applied to geological hazard prevention and mineral resource development.

Key words:sparse data;convolutional neural network;ensemble learning algorithm;stratigraphic reconstruction

最近更新