摘要
探地雷达地质预报具有高效、高精度及无损探测等优势,是当前桥梁、隧道及边坡等工程勘察领域理论与应用研究的热点。然而,传统探地雷达隧道地质预报解译推断因观测数据的欠定性,常常具有病态性、非线性和多解性;或定性依赖于解译人员的工程经验或简化方法,其强主观性常导致错误推断。随着人工智能的发展,基于机器学习的探地雷达隧道地质预报解译方法等成果不断涌现,但缺乏系统的归纳总结。本文首先回顾和阐述了探地雷达隧道地质预报智能化解译的研究现状及存在的问题;其次,介绍了人工智能探地雷达隧道地质预报解译基本原理;然后分别从人工智能算法框架、复杂环境人工智能训练集获取、跨域泛化能力的提高等方面阐述了当前探地雷达隧道地质预报的最新进展及其具体应用效果,最后,对探地雷达隧道地质预报智能化解译方法发展趋势进行了总结,指出针对不规则灾害预报学习模型构建、区域检测数据集驱动下增量学习及迁移学习等或将是未来极具潜力的发展方向。
Geological prediction using Ground-penetrating radar(GPR)has the advantages of high efficiency,high precision and non-destructive detection.It is a hot topic in theory and application research of bridge,tunnel and slope engineering investigation.However,the interpretation and inference of the traditional GPR tunnel geological prediction are often ill-conditioned,nonlinear and multi-solution due to the lack of qualitative observation data.Or qualitative reliance on the interpreter’s engineering experience or simplified methods,its strong subjectivity leads to false inference.With the development of artificial intelligence,achievements such as geological prediction and interpretation methods of GPR tunnel based on machine learning continue to emerge,which need to be summarized systematically.This paper firstly reviews and expounds the research status and existing problems of intelligent interpretation of GPR tunnel geological prediction.Secondly,it introduces the basic principle of artificial intelligence GPR tunnel geological prediction interpretation.This paper expounds the latest progress and specific application effects of the geological prediction of GPR tunnel from the aspects of artificial intelligence algorithm framework,acquisition of artificial intelligence training set in complex environment,improvement of cross-domain generalization ability,and finally summarizes the development trend of intelligent interpretation methods of GPR tunnel geological prediction.The construction of learning model for irregular disaster prediction,incremental learning and transfer learning driven by regional detection data sets may be the most promising development directions in the future for GPR tunnel geological prediction.
作者
杨庭伟
冉孟坤
李静和
姜洪亮
卢超波
Yang Tingwei;Ran Mengkun;Li Jinghe;Jiang Hongliang;Lu Chaobo(College of Earth Sciences,Guilin University of Technology,Guilin Guangxi 541000,China;Guangxi Transportation Science and Technology Group Co.,Ltd.,Institute of Tunnel and Geotechnical Engineering,Nanning Guangxi 530007,China;Highway Tunnel Safety Early Warning Research Center of Guangxi Zhuang Autonomous Region,Nanning Guangxi 530007,China;Guangxi Key Laboratory of Road Structure and Materials,Nanning Guangxi 530007,China;Research and Development Center of Transportation Industry for High-grade Highway Construction and Maintenance Technology,Materials and Equipment,Nanning Guangxi 530007,China)
出处
《工程地球物理学报》
2023年第6期702-715,共14页
Chinese Journal of Engineering Geophysics
基金
广西自然科学基金(编号:2021GXNSFAA196056)。
关键词
探地雷达隧道地质预报
智能化解译
机器学习
神经网络
研究进展
ground penetrating radar tunnel geological prediction
intelligent resolution translation
machine learning
neural networks
research progress
