为提升施工阶段围岩级别判识的智能化水平，提升围岩级别智能判识的准确性，提出一种基于多元地质信息和信息融合的围岩级别智能判识方法，并开发基于多元地质信息的钻爆法隧道围岩级别智能判识软件系统，实现隧道工程多元信息的自动采集和围岩级别自动化判识。依托我国西部山区隧道工程，采集随钻参数、掌子面高清数码图像、超前地质预报信息、地勘信息等4项多元地质信息，开展数据标准化、结构化处理及特征提取，构建基于多元地质信息的围岩级别智能判识模型，共判定844个断面的围岩级别，模型平均准确率达到95.45%，平均精确率为95.05%，平均召回率为93.25%，平均F1分数为94.14%，对软硬不均及局部破碎围岩具有较好的识别效果。
王明年
,
童建军
,
易文豪
,
彭鑫
. Intelligent Classification and Software System for Surrounding Rock Based on Multivariate Geological Information（基于多元地质信息的钻爆法隧道围岩级别智能判识及软件系统）[J]. 隧道建设, 2024
, 44(12)
: 2451
-2468
.
DOI: 10.3973/j.issn.2096-4498.2024.12.
In order to improve the
intelligentization level of surrounding rock classification in the construction
stage and enhance the accuracy of intelligent rock classification, an
intelligent rock classification method based on multivariate geological
information and information fusion is proposed. An intelligent classification
software system based on multivariate geological information is developed to
realize automatic acquisition of multivariate geological information and
automatic classification of surrounding rock for drill-and-blast tunnels. Based on tunnel projects
in the western mountainous area of China, four types of multivariate geological
information, including drilling parameters, highdefinition
digital images of tunnel face, advance geological prediction information, and
geological exploration information, are collected. Standardized and structured
data processing and feature extraction are carried out. An intelligent
classification model of surrounding rock based on multivariate geological
information is constructed, and a web-end intelligent
classification software system of surrounding rock based on multivariate
geological information is developed for drill-and-blast tunnels. The accuracy of the intelligent classification model
proposed in this study reached 95.45%, with an average accuracy of 95.05%, an
average recall rate of 93.25%,
and an average F1 score of 94.14%. Field applications of the software system have been
carried out for rock classification at 844 sections, which had a good
performance on identifying unevenly distributed soft and hard rock or locally
fractured rock.
