基于数据驱动的TBM掘进地层岩性识别预测方法  

Prediction method for data-driven lithology identification of TBM tunneling strata

作　　者：丁自伟[1] 高成登 张玲 张旭 侯涛 翟剑平 王家行 董云俊 DING Ziwei;GAO Chengdeng;ZHANG Ling;ZHANG Xu;HOU Tao;ZHAI Jianping;WANG Jiahang;DONG Yunjun(College of Energy Engineering,Xi'an University of Science and Technology,Xi'an,Shaanxi 710054,China;Shandong Energy Group Xibei Mining Co Ltd,Xi'an,Shaanxi 710018,China)

机构地区：[1]西安科技大学能源学院,陕西西安710054 [2]山东能源集团西北矿业有限公司,陕西西安710018

出　　处：《采矿与安全工程学报》2025年第1期147-160,共14页Journal of Mining & Safety Engineering

基　　金：山东能源科技计划重大项目(SNKJ2022A15);国家自然科学基金项目(52074209);陕西省自然科学基础研究计划联合基金项目(2021JLM-06)。

摘　　要：随着全断面硬岩掘进机(TBM)在煤矿巷道掘进施工中的广泛应用,地层信息的准确、实时识别已成为保证掘进效率的关键因素。为了研究掘进参数与地层岩性的相互作用关系,以西北矿业高家堡煤矿西区开拓大巷为工程背景,通过对稳定阶段掘进参数的深入分析,建立地层岩性与关键掘进参数之间的“机-岩”感知关系,提出基于数据驱动的TBM掘进地层岩性识别的Stacking集成预测算法,确定与地层岩性预测相关的主要输入特征参数,包括推进速度v、刀盘转速n、刀盘推力F、刀盘扭矩T和贯入度P。训练结果表明,Stacking预测模型在5个输入参数的平衡精度和训练时间下均获得了最佳性能;预测模型多元评价结果显示,Stacking模型的AUC曲线面积指数为0.97,比单一的XGBoost,ANN和SVM模型(0.94,0.94,0.95)预测精度更高,且在处理不均衡数据预测表现出明显的优势。因此,基于本研究的预测模型可以很好地指导现场TBM掘进参数的调整,可有效减少TBM故障停机和刀头的磨损,提高掘进效率。With the wide application of full-face hard rock tunnel boring machine(TBM)to coal mine roadway excavation,the accurate and real-time identification of stratum information has become a key factor for ensuring the excavation efficiency.In order to study the interaction between tunneling parame-ters and stratum lithology,this study takes the development roadway in the west area of Gaojiapu Coal Mine of Northwest Mining Industry as the engineering background.To be specific,an in-depth analysis was conducted on the tunneling parameters in the stable stage first.Based on the analysis,the“machinerock”perception relationship between stratum lithology and key tunneling parameters was established,and a Stacking integrated prediction algorithm for data-driven lithology identification of TBM tunneling strata was proposed.Moreover,the main input characteristic parameters related to stratum lithology prediction were determined,including the propulsion speed v,the cutterhead speed n,the cutterhead thrust F,the cutterhead torque T,and the penetration degree P.According to the training results,the Stacking integrated prediction model achieves the best performance under the balance accuracy and training time of the five input parameters.The multivariate evaluation results show that the AUC curve area index of the Stacking integrated prediction model is 0.97,which is higher than the prediction accuracy of the single XGBoost,ANN,and SVM models(0.94,0.94,and 0.95),and boasts obvious advantages in unbalanced data prediction.Therefore,the prediction model proposed in this study can well guide the adjustment of TBM tunneling parameters on site,effectively reduce TBM fault shutdown and cutter head wear,and improve tunneling efficiency.

关 键 词：TBM掘进参数 岩性识别 机器学习 Stacking集成学习算法 

分 类 号：TD353[矿业工程—矿井建设]