基于随机森林算法的煤巷顶板位移预测与应用  

Roof displacement prediction of coal roadways based on random forest algorithm and the applica-tion

作　　者：陈攀 马鑫民[1] 向俊杰 陈莉影 翟中华 杨雯清 CHEN Pan;MA Xinmin;XIANG Junjie;CHEN Liying;ZHAI Zhonghua;YANG Wenqing(School of Mechanics and Civil Engineering,China University of Mining and Technology-Beijing,Beijing 100083,China;Yunnan Institute of Water and Hydropower Engineering Investigation,Design and Research,Kunming 650021,China)

机构地区：[1]中国矿业大学(北京)力学与土木工程学院,北京100083 [2]云南省水利水电勘测设计院有限公司,云南650021

出　　处：《煤炭工程》2024年第7期127-135,共9页Coal Engineering

基　　金：国家自然科学基金项目(52074301)。

摘　　要：煤巷围岩稳定控制是保障煤矿安全高效开采的关键,煤巷顶板位移量是反应煤巷围岩稳定性的关键指标,本研究提出机器学习方法对煤巷顶板位移进行超前预测研究。确定了煤巷顶板位移的8个重要影响指标,建立了煤巷顶板位移预测数据库并对数据进行了指标相关性和重要性分析。基于RF、GA-SVM和GA-ANN分别建立了三种煤巷顶板位移预测模型,选用RMSE、MAE和R2三个指标来评价模型的性能。结果显示,RF模型测试性能最佳,R2=0.909,RMSE=20.475,MAE=16.790,GA-ANN模型的性能最差。采用十折交叉方法对RF模型和GA-ANN模型进行可靠性验证,结果显示RF模型的稳定性更高,平均R2为0.891。将RF模型应用到干河煤矿2-1121巷,预测值与实际值的绝对误差为19 mm,相对误差为11.18%,说明了RF模型对煤巷顶板位移预测的准确性与可靠性,研究结果对煤巷顶板位移预测提供了一种新途径。The stability control of surrounding rocks in coal roadways is the key to ensure the safe and efficient mining of coal mine.Roof displacement is a crucial indicator to present wall rock stability.In this paper,machine learning method was introduced to predict the roof displacement of coal roadways in advance.Eight important influence indexes of roof displacement of coal roadways were defined,and the prediction database of roof dis-placement of coal roadway was established,from which the data were analyzed for correlation and importance.Based on RF,GA-SVM and GA-ANN,three roof displacement prediction models of coal roadways were estab-lished respectively,and RMSE,MAE and R2 were selected to evaluate the performance of the models.The results show that the RF model has the best performance,R2=0.909,RMSE=20.475,MAE=16.790,while the GA-ANN model has the worst performance.The ten-fold cross-validation method was used to verify the reliability of RF and GA-ANN models,and it is found that RF model has higher stability,with an average R2 of 0.891.When RF model is applied to No.2-1121 roadway of Ganhe Coal Mine,the absolute error between the predicted and the actual value is 19 mm,and the relative error is 11.18%.This study shows RF model can accurately and stably predict roadway roof displacement,which provide a new reference for roadway roof displacement.

关 键 词：随机森林 煤矿巷道 顶板位移 机器学习 预测 

分 类 号：TD353[矿业工程—矿井建设]