地球物理地球化学勘探计算技术的智能化之路OA
The leapfrog development of intelligent computing techniques in geophysical and geochemical exploration:a review from Computing Techniques for Geophysical and Geochemical Exploration
过去十年,大数据与人工智能技术深刻重塑了地球物理与地球化学探查计算技术的研究范式.《物探化探计算技术》作为国内专门领域的学术性刊物,深刻地记载着智能化转型过程,是近 10 年来积极参与、推动地球科学大数据与智能技术应用研究的先锋学术期刊之一.基于该期刊 2016-2026年的文献分析,该领域智能化发展呈现出清晰的阶段性特征.大数据挖掘技术引入自 2016开始变成显著的学术现象,主要聚焦于地震属性融合、多源数据整合、并行计算加速和三维可视化建模.2017-2020年是机器学习算法的快速发展期,支持向量机、随机森林、BP神经网络等算法在储层预测、滑坡评价、地球化学异常识别等领域获得广泛应用,显著提升了数据处理和解释的效率和精度.2020年以来是深度学习算法的全面爆发期,卷积神经网络、U-Net及其变体、Transformer、自编码器、生成对抗网络等深度模型在地震数据去噪、断层识别、速度建模、反演成像、遥感图像处理等任务中取得突破性进展,多项指标超越传统方法.此外,遗传算法、粒子群优化、强化学习与迁移学习等也在各自适用领域发挥了重要作用.当前仍面临若干关键挑战:模型可解释性差、多源异构数据融合困难、标注数据稀缺与类不平衡、跨区域泛化能力不足,以及纯数据驱动结果难以保证物理一致性.未来突破方向包括:多模态大模型与地质基础模型、物理信息神经网络驱动的知识-数据双驱动建模、图神经网络与地质知识图谱、可解释人工智能,以及边缘智能实时处理.物探化探计算技术正迈向更智能、更可靠、更高效的发展阶段.
The rapid development of big data and artificial intelligence technologies has profoundly reshaped the research paradigm of computing techniques in geophysical and geochemical exploration in past decade.Computational Technology in Geophysical and Geochemical Exploration is an academic journal specializing in a domestic field,profoundly documenting the process of intelligent transformation.It is one of the pioneering academic journals that has actively participated in and promoted research on big data and intelligent technology applications in geosciences over the past decade.Based on literature published in the journal Computing Techniques for Geophysical and Geochemical Exploration from 2016 to 2026,this paper systematically reviews the ten-year evolution of intelligent development in geophysical and geochemical exploration computing.First,it elaborates on the introduction and application of big data mining techniques,including seismic attribute fusion,data dimensionality reduction,and association rule mining.Second,it analyzes the introduction and development of machine learning algorithms,focusing on the application progress of algorithms such as support vector machines and random forests in reservoir prediction,landslide susceptibility assessment,and geochemical anomaly identification.Third,it systematically reviews the breakthrough applications of deep learning algorithms,covering convolutional neural networks,U-Net and its variants,Transformers,autoencoders,and generative adversarial networks.Fourth,it introduces the adoption pathways of other cutting-edge algorithms,including genetic algorithms,particle swarm optimization,reinforcement learning,and transfer learning.Finally,it discusses the current challenges,such as data scarcity,poor model interpretability,and insufficient generalization capability,and looks forward to future development directions,including multimodal fusion,physics-informed neural networks,and large models.This paper aims to provide a systematic reference for the intelligent transformation of computing techniques in geophysical and geochemical exploration.
周永章;曹礼刚
中山大学 地球科学与工程学院,珠海 519000||广东省地质过程与矿产资源探查重点实验室,珠海 519000成都理工大学 地球物理学院,成都 610059
天文与地球科学
地球物理探查地球化学探查地球科学智能大数据挖掘机器学习深度学习
geophysical explorationgeochemical explorationgeoscience intelligencebig data miningmachine learningdeep learning
《物探化探计算技术》 2026 (3)
337-349,13
重点研发计划项目(2022YFF0800101)国家自然科学基金项目(U1911202)内蒙古自治区"揭榜挂帅"项目(2025KJTW0020)
评论