中文 | English
Hefei University of Technology Logo J. Hefei Univ. Tech. (Nat. Sci.)

Navigation

Safety factor prediction of mine roadway bolts based on ensemble learning

Download Full Text

基于集成学习的矿山巷道锚杆安全系数预测

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Aug. 2026, Vol. 49 No. 8: 1089-1097, 1106

DOI: 10.3969/j.issn.1003-5060.2026.08.013

Authors

YUAN Haiping $ ^{1} $, PAN Luyang $ ^{1} $, LI Hengzhe $ ^{1} $, ZHU Chuanqi $ ^{2} $, NI Bin $ ^{1} $

(1. School of Civil and Hydraulic Engineering, Hefei University of Technology, Hefei 230009, China; 2. State Key Laboratory of Mining Response and Disaster Prevention in Deep Coal Mines, Huainan 232001, China)

Abstract and Keywords

Abstract: The stable anchor network support of roadways signifies an important guarantee for safe and efficient exploitation of mine, while the traditional design methods often have the problem of excessive support. In order to reasonably determine the anchor network support parameters, economically and effectively realize the anchor network support design, and quantitatively predict the safety of rock roadway, this paper adopts the geological strength index (GSI) method based on the Hoek-Brown strength criterion to divide the surrounding rock grades of mine roadways, and carries out the numerical analysis of different anchor network support schemes under the condition of each surrounding rock grade. Multiple groups of rock bolt stress data have been obtained to establish the safety factor database of mine roadway rock bolts. In addition, a prediction model based on three ensemble algorithms has been proposed: random forest (RF), eXtreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM). The hyperparameters of each model have been searched and optimized by the Bayesian optimization algorithm (BOA). Several single learners were employed to compare the prediction accuracy. Furthermore, an importance analysis of rock roadway eigenvalues was carried out to study the relative importance of each eigenvalue. The results indicate that the ensemble learning has higher prediction accuracy than single learners, and the BOA-LightGBM model has the shortest training time and the highest prediction accuracy, which can be adopted as the optimal algorithm for intelligent decision-making of rock roadway support system. The importance analysis demonstrates that various parameters have different effects on the safety factor of rock bolts in the mine roadways, and the surrounding rock grade represents the most influential factor. This study offers novel insights into the design optimization of anchor network support and the safety prediction of rock roadways to improve the economic benefits of mine.

Keywords: mine roadway; Hoek-Brown strength criterion; anchor network support; ensemble learning; Bayesian optimization algorithm(BOA); safety factor

作者信息

袁海平 $ ^{1} $，潘璐阳 $ ^{1} $，李恒喆 $ ^{1} $，朱传奇 $ ^{2} $，倪斌 $ ^{1} $

(1. 合肥工业大学土木与水利工程学院，安徽合肥 230009；2. 深部煤矿采动响应与灾害防控国家重点实验室，安徽淮南 232001)

摘要和关键词

摘要: 稳定的巷道锚网支护是矿山安全、高效开采的重要保障,传统设计方法常存在过度支护的问题。为合理地确定锚网支护参数,经济有效地实现锚网支护设计,定量地预测岩巷安全性,文章采用 Hoek-Brown 强度准则中地质强度指标(geological strength index, GSI)法将矿山巷道围岩等级进行精细化划分,对各围岩等级下的不同锚网支护方案开展数值分析,获取多组锚杆应力数据以建立矿山巷道锚杆安全系数数据库。基于数据库利用随机森林(random forest, RF)、极限梯度提升回归树(eXtreme gradient boosting, XGBoost)和轻量级梯度提升机(light gradient boosting machine, LightGBM)这3种集成算法建立预测模型,使用贝叶斯优化算法(Bayesian optimization algorithm, BOA)对各模型进行参数寻优优化,同时采用几种单一学习器进行预测精度对比,最后进行数据特征值重要度分析。结果表明:集成学习比单一学习器预测精度更高,其中 BOA-LightGBM 模型训练时间最短,预测精度最高,可作为岩巷支护体系智能决策的优选算法;各参数对岩巷锚杆安全系数的影响不同,其中围岩等级是影响最大的因素。该研究为矿山巷道锚网支护设计优化和岩巷安全性预测提供新的思路,以期提高矿山的经济效益。

关键词: 矿山巷道；Hoek-Brown强度准则；锚网支护；集成学习；贝叶斯优化算法（BOA）；安全系数

Funding

国家自然科学基金资助项目（51874112）；深部煤矿采动响应与灾害防控国家重点实验室开放基金重点资助项目（SKLMRDPC22KF02）

Previous Next
Back to Contents

Account

Submission Login Reviewer Login Editor Login