煤巷支护参数预测研究

陈攀; 马鑫民; 向俊杰; 陈莉影; 梁厅皓

doi:10.13272/j.issn.1671-251x.2022120047

煤巷支护参数预测研究

doi: 10.13272/j.issn.1671-251x.2022120047

陈攀^{1, 2,},
马鑫民^1, ,,
向俊杰¹,
陈莉影¹,
梁厅皓¹

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，云南昆明　650021

基金项目: 国家自然科学基金资助项目（52074301）。

详细信息

作者简介:
陈攀（1998—），男，云南曲靖人，硕士，主要从事巷道支护和水利勘察工作，E-mail：18811432245@163.com

通讯作者:
马鑫民（1979—），男，山东荷泽人，副教授，主要研究方向为矿山工程爆破和巷道支护智能化技术，E-mail：mxm@cumtb.edu.cn。

中图分类号: TD353
计量
- 文章访问数: 171
- HTML全文浏览量: 20
- PDF下载量: 41
- 被引次数: 0
出版历程
- 收稿日期: 2022-12-15
- 修回日期: 2023-09-20
- 网络出版日期: 2023-10-23

Research on prediction of support parameters for coal roadways

1.
School of Mechanics and Civil Engineering, China University of Mining and Technology-Beijing, Beijing 100083, China
2.
Yunnan Institute Co., Ltd. of Water & Hydropower Engineering Investigation Design, Kunming 650021, China

摘要

摘要: 目前支持向量机（SVM）和随机森林（RF）等算法在煤矿巷道支护领域应用较少。研究了不同的机器学习模型进行支护参数设计的适用性，以建立一个更高性能的模型来实现锚杆支护的合理、科学设计。首先建立煤巷支护智能预测数据库：采用现场调研、问卷调查和文献检索等方式收集煤矿巷道样本；采用缺失值填补、箱形图修改离群点和局部异常因子剔除等方式对数据进行处理，建立煤巷支护数据库。提出一种基于合成少数类过采样（SMOTE）−遗传算法（GA）−SVM的煤巷支护参数预测模型：将数据库中的数据分成训练集与测试集，采用SMOTE技术平衡训练样本，提高模型对少数类样本的拟合能力；训练过程采用GA对SVM的超参数进行全局寻优，进一步提高模型整体性能。测试结果表明，SMOTE−GA−SVM模型的分类精度达到83.8%，比传统的SVM模型提高了21.8%。将SVM、人工神经网络（ANN）、RF、AdaBoost（ADA）和朴素贝叶斯分类器（NBC）等机器学习方法引入到煤巷锚杆支护参数预测中，建立对应的支护参数预测模型，比较结果表明：从最优到最差的预测模型排序分别为SMOTE−GA−SVM、RF、GA−ANN、SVM、NBC和ADA，6种模型的平均分类精度达69.9%，验证了机器学习方法在煤巷锚杆支护参数预测方面的可行性。在山西霍宝干河煤矿有限公司对SMOTE−GA−SVM模型进行了应用，模型预测准确率达87.5%，具有较强的适用性和可靠性。
- 煤矿巷道 /
- 机器学习 /
- 锚杆支护参数 /
- 合成少数类过采样 /
- 遗传算法优化支持向量机
Abstract: Currently, algorithms such as support vector machine (SVM) and random forest (RF) are less applied in the field of coal mine roadway support. The paper studies the applicability of different machine learning models for support parameter design.Thus a higher performance model would be established to achieve reasonable and scientific design of anchor bolt support. Firstly, it is suggested to establish an intelligent prediction database for coal mine roadway support. Through on-site research, questionnaire survey, and literature search, the coal mine roadway samples are collected. The data is processed using methods such as filling in missing values, modifying outliers in box charts, and removing local abnormal factors to establish a coal roadway support database. The paper proposes a coal roadway support parameter prediction model based on synthetic minority oversampling technique (SMOTE) - genetic algorithm (GA) - SVM. The data in the database is divided into training and testing sets. The SMOTE technology is used to balance training samples, and improve the model's fitting capability for minority class samples. The training process uses GA to globally optimize the hyperparameters of SVM, further improving the overall performance of the model. The test results show that the classificaton precision of the SMOTE-GA-SVM model reaches 83.8%, which is 21.8% higher than the traditional SVM model. The machine learning methods such as SVM, artificial neural network (ANN), RF, AdaBoost (ADA), and naive Bayesian classifier (NBC) are introduced into the prediction of coal roadway anchor support parameters. The corresponding support parameter prediction models are established. The comparison results showed that the best to worst prediction models are ranked as SMOTE-GA-SVM, RF, GA-ANN, SVM, NBC, and ADA, with an average classificaton precision of 69.9%. The result verifies the feasibility of machine learning methods in predicting the parameters of coal roadway bolt support. The SMOTE-GA-SVM model is applied in Shanxi Huobaoganhe Coal Mine Co., Ltd., with a precision of 87.5% and strong applicability and reliability.
- coal mine roadway /
- machine learning /
- anchor rod support parameters /
- oversampling of synthesized minority classes /
- genetic algorithm optimizing support vector machine

HTML全文

图 1 GA对SVM超参数寻优流程

Figure 1. GA optimization process for super parameters of SVM

下载: 全尺寸图片幻灯片

图 2 原始数据的箱形图

Figure 2. Box diagram of the original data

下载: 全尺寸图片幻灯片

图 3 训练集输入参数分布统计

Figure 3. Distribution statistics of input parameters of training set

下载: 全尺寸图片幻灯片

图 4 SMOTE平衡样本流程

Figure 4. Sample balancing flow by SMOTE

下载: 全尺寸图片幻灯片

图 5 煤巷支护数据库的整体构建流程

Figure 5. The overall building process of the coal roadway support database

下载: 全尺寸图片幻灯片

图 6 SMOTE−GA−SVM支护参数预测模型建立流程

Figure 6. SMOTE-GA-SVM supporting parameter prediction model establishment process

下载: 全尺寸图片幻灯片

图 7 输入变量在支护参数预测模型上的重要性

Figure 7. Importance of input variables in support parameter prediction model

下载: 全尺寸图片幻灯片

图 8 顶板锚杆间距GA−ANN预测模型的网络拓扑结构

Figure 8. Network topology of GA-ANN prediction model of roof bolt spacing

下载: 全尺寸图片幻灯片

表 1 基于LOF的异常样本检测结果

Table 1. Test results of abnormal samples based on local outlier factor(LOF)

k=4	k=5	k=6	k=7	k=8	k=9	k=10
105	105	105	105	33	14	14
84	33	33	33	105	33	25
129	84	84	84	14	84	49
33	129	154	124	84	25	33
124	124	129	154	154	49	84
154	154	124	14	124	105	105
36	75	90	129	25	154	154
90	36	17	25	49	124	124
75	44	44	109	109	75	75
70	14	109	75	129	109	59

下载: 导出CSV

表 2 顶板锚杆支护参数统计

Table 2. Statistics of roof anchor bolt support parameters

参数名称	参数值	频数	参数名称	参数值	频数
直径	18 mm	11	间距	700 mm	5
	20 mm	46		800 mm	39
	22 mm	60		900 mm	36
长度	2 000 mm	14		1 000 mm	16
	2 200 mm	17		1 100 mm	9
	2 400 mm	57		1 200 mm	12
	2 500 mm	20	排距	700 mm	10
	2 600 mm	9		800 mm	33
数量	4 根	13		900 mm	16
	5 根	28		1 000 mm	38
	6 根	53		1 100 mm	7
	7 根	23		1 200 mm	13

下载: 导出CSV

表 3 顶板锚索支护参数统计

Table 3. Statistics of roof anchor cable support parameters

参数名称	参数值	频数	参数名称	参数值	频数
直径	15.24 mm	14	长度	5300 mm	20
	17.89 mm	36		6300 mm	39
	18.7 mm	18		7300 mm	22
	21.6 mm	17		8300 mm	32
	22 mm	32		9300 mm	4
数量	1 根	4	布置方式	1	24
	2 根	68		2	71
	3 根	39		3	22
	4 根	6

下载: 导出CSV

表 4 帮部支护参数统计

Table 4. Side support parameter statistics

参数名称	参数值	频数	参数名称	参数值	频数
直径	16 mm	7	间距	700 mm	10
	18 mm	20		800 mm	41
	20 mm	41		900 mm	21
	22 mm	49		1 000 mm	27
长度	1 800 mm	16		1 100 mm	4
	2 000 mm	25		1 200 mm	14
	2 200 mm	12	排距	700 mm	10
	2 400 mm	43		800 mm	33
	2 500 mm	21		900 mm	15
数量	2 根	7		1 000 mm	39
	3 根	29		1 100 mm	7
	4 根	52		1 200 mm	13
	5 根	29

下载: 导出CSV

表 5 GA全局寻优结果

Table 5. Global optimization results of GA

支护特征	c_best	g_best	支护特征	c_best	g_best
顶板锚杆直径	85.26	2.32	帮部锚杆间距	33.65	5.30
顶板锚杆长度	6.75	0.63	帮部锚杆排距	54.23	2.65
顶板锚杆间距	70.76	7.44	帮部锚杆数量	30.23	4.64
顶板锚杆排距	81.04	2.07	锚索直径	74.62	7.55
顶板锚杆数量	32.97	3.71	锚索长度	61.50	1.51
帮部锚杆直径	73.86	8.96	锚索数量	87.00	2.65
帮部锚杆长度	10.53	5.07	锚索布置	70.72	5.30

下载: 导出CSV

表 6 机器学习模型在测试集上的分类精度

Table 6. Classification precision of machine learning model on test set %

模型	顶板锚杆精度					帮部锚杆精度					顶板锚索精度				平均精度
模型	直径	长度	间距	排距	数量	直径	长度	间距	排距	数量	直径	长度	数量	布置	平均精度
SVM	66.0	83.6	66.0	79.2	74.8	54.0	70.4	70.4	64.8	57.2	54.0	79.2	74.8	69.2	68.8
SMOTE−GA−SVM	86.0	89.8	82.4	84.3	82.1	75.3	79.2	88.4	77.1	77.2	83.3	93.5	90.6	84.0	83.8
RF	79.2	74.8	52.8	83.6	66.0	70.4	61.6	61.6	74.8	52.8	57.2	92.4	88.0	79.2	71.0
ADA	72.6	64.5	53.8	57.8	57.8	48.4	51.1	37.6	59.2	61.8	48.4	78.7	80.7	75.3	60.5
GA−ANN	93.4	63.4	64.4	55.0	66.0	62.7	66.0	57.2	71.5	55.0	74.8	89.8	79.8	82.5	70.1
NBC	66.0	66.0	44.0	44.0	79.2	70.4	57.2	57.2	57.2	74.8	48.4	81.8	88.0	74.8	64.9

下载: 导出CSV

表 7 霍州矿区干河煤矿的特征参数

Table 7. Characteristic parameters of Ganhe Coal Mine in Huozhou Mining area

序号	巷道名称	煤层厚度/m	煤层强度/MPa	基本顶厚度/m	基本顶强度/MPa	直接顶厚度/m	直接顶强度/MPa	直接底厚度/m	直接底强度/MPa	埋深/m	巷道高度/m	巷道宽度/m
1	2−1161巷	4.20	9.38	1.70	65.06	6.40	51.79	1.10	51.79	450	3.6	5.0
2	2−1261巷	3.75	15.00	4.80	86.09	2.45	65.06	2.90	19.16	420	3.7	5.0
3	三采区辅助运输巷	0.78	9.38	3.12	65.06	4.96	86.09	4.60	45.40	420	3.5	4.8
4	2−1021巷	4.20	14.54	3.10	57.32	3.89	25.00	0.80	37.45	500	3.8	5.0

下载: 导出CSV

表 8 SMOTE−GA−SVM模型应用结果

Table 8. Application result of SMOTE-GA-SVM model

序号	巷道名称		顶板锚杆					帮部锚杆					顶板锚索
序号	巷道名称		直径/ mm	长度/ mm	间距/ mm	排距/ mm	数量/ 根	直径/ mm	长度/ mm	间距/ mm	排距/ mm	数量/ 根	直径/ mm	长度/ mm	布置方式	数量/ 根
1	2−1161巷	真实值	22	2500	900	900	6	22	2500	900	900	4	21.60	8300	2	3
1	2−1161巷	测试值	22	2400	900	900	6	22	2400	900	900	4	21.60	8300	2	3
2	2−1261巷	真实值	22	2500	800	800	7	22	2500	800	800	5	21.60	8300	3	3
2	2−1261巷	测试值	22	2500	800	800	7	22	2500	800	800	5	21.60	8300	3	3
3	三采区辅助运输巷	真实值	20	2000	800	800	7	20	2000	800	800	2	15.24	7300	2	2
3	三采区辅助运输巷	测试值	20	2000	800	1000	7	20	2000	800	1000	2	17.89	7300	2	2
4	2−1021巷	真实值	22	2400	900	1000	6	22	2400	1000	1000	4	21.60	6300	2	2
4	2−1021巷	测试值	22	2400	900	1000	6	20	2400	1000	1000	4	21.60	7300	2	2

下载: 导出CSV

参考文献(21)

[1]	康红普. 我国煤矿巷道锚杆支护技术发展60年及展望[J]. 中国矿业大学学报,2016,45(6):1071-1081. KANG Hongpu. Sixty years development and prospects of rock bolting technology for underground coal mine roadways in China[J]. Journal of China University of Mining & Technology,2016,45(6):1071-1081.
[2]	康红普. 我国煤矿巷道围岩控制技术发展70年及展望[J]. 岩石力学与工程学报,2021,40(1):1-30. KANG Hongpu. Seventy years development and prospects of strata control technologies for coal mine roadways in China[J]. Chinese Journal of Rock Mechanics and Engineering,2021,40(1):1-30.
[3]	单仁亮,彭杨皓,孔祥松,等. 国内外煤巷支护技术研究进展[J]. 岩石力学与工程学报,2019,38(12):2377-2403. SHAN Renliang,PENG Yanghao,KONG Xiangsong,et al. Research progress of coal roadway support technology at home and abroad[J]. Chinese Journal of Rock Mechanics and Engineering,2019,38(12):2377-2403.
[4]	顾清华,江松,李学现,等. 人工智能背景下采矿系统工程发展现状与展望[J]. 金属矿山,2022(5):10-25. GU Qinghua,JIANG Song,LI Xuexian,et al. Development status and prospect of mining system engineering under the background of artificial intelligence[J]. Metal Mine,2022(5):10-25.
[5]	谢广祥,曹伍富,王德润,等. 基于人工神经网络的煤巷锚杆支护设计研究[J]. 煤炭学报,1999(6):599-604. XIE Guangxiang,CAO Wufu,WANG Derun,et al. The study on bolting support design in coal roadway based on artificial neural networks[J]. Journal of China Coal Society,1999(6):599-604.
[6]	王茂源. 煤巷锚杆支护设计混合智能系统研究[D]. 北京:中国矿业大学(北京),2016. WANG Maoyuan. Hybrid intelligent system on coal roadway bolting design[D]. Beijing:China University of Mining and Technology-Beijing,2016.
[7]	王哲哲,许梦国,程爱平,等. 模糊神经网络在巷道支护方案选择中的应用[J]. 化工矿物与加工,2019,48(1):16-19,23. WANG Zhezhe,XU Mengguo,CHENG Aiping,et al. Application of fuzzy neural network in selection of roadway support scheme[J]. Industrial Minerals & Processing,2019,48(1):16-19,23.
[8]	XU Qingyun,LI Yongming,LU Jie,et al. The use of surrounding rock loosening circle theory combined with elastic-plastic mechanics calculation method and depth learning in roadway support[J]. PLoS ONE,2020,15(7). DOI: 10.1371/journal.pone.0234071.
[9]	REN Heng,ZHU Yongjian,WANG Ping,et al. Classification and application of roof stability of bolt supporting coal roadway based on BP neural network[J]. Advances in Civil Engineering,2020. DOI: 10.1155/2020/8838640.
[10]	ZHANG Xiliang,NGUYEN H,BUI X,et al. Evaluating and predicting the stability of roadways in tunnelling and underground space using artificial neural network-based particle swarm optimization[J]. Tunnelling and Underground Space Technology,2020,103. DOI: 10.1016/j.tust.2020.103517.
[11]	PU Yuanyuan,APEL D,HALL R. Using machine learning approach for microseismic events recognition in underground excavations:comparison of ten frequently-used models[J]. Engineering Geology,2020. DOI: 10.1016/j.enggeo.2020.105519.
[12]	马鑫民,范皓宇,林天舒,等. 基于GA−SVM的煤矿岩巷爆破效果智能预测[J]. 煤炭工程,2019,51(5):148-153. MA Xinmin,FAN Haoyu,LIN Tianshu,et al. Intelligent prediction of blasting effect of coal mine roadway based on GA-SVM[J]. Coal Engineering,2019,51(5):148-1533.
[13]	MAHDEVARI S,KHODABAKHSHI M B. A hierarchical local-model tree for predicting roof displacement in longwall tailgates[J]. Neural Computing and Applications,2021,33(21):14909-14928. doi: 10.1007/s00521-021-06127-y
[14]	赵汝星. 基于随机森林的回采巷道围岩稳定性分类[J]. 煤矿安全,2014,45(11):200-202,206. ZHAO Ruxing. Classification of roadway surrounding rock stability based on random forest[J]. Safety in Coal Mines,2014,45(11):200-202,206.
[15]	汪海燕,黎建辉,杨风雷. 支持向量机理论及算法研究综述[J]. 计算机应用研究,2014,31(5):1281-1286. WANG Haiyan,LI Jianhui,YANG Fenglei. Overview of support vector machine analysis and algorithm[J]. Application Research of Computers,2014,31(5):1281-1286.
[16]	JU Xuchan,YAN Zhenghao,WANG Tianhe,et al. Overview of optimization algorithms for large-scale support vector machines[C]. IEEE International Conference on Data Mining Workshops,Ningbo,2021:909-916.
[17]	CHAHAR V,KATOCH S,CHAUHAN S S. A review on genetic algorithm:past,present,and future[J]. Multimedia Tools and Applications,2020,80(5):8091-8126.
[18]	谭文侃,叶义成,胡南燕,等. LOF与改进SMOTE算法组合的强烈岩爆预测[J]. 岩石力学与工程学报,2021,40(6):1186-1194. TAN Wenkan,YE Yicheng,HU Nanyan,et al. Severe rock burst prediction based on the combination of LOF and improved SMOTE algorithm[J]. Chinese Journal of Rock Mechanics and Engineering,2021,40(6):1186-1194.
[19]	FENG Shuo,KEUNG J,YU Xiao,et al. Investigation on the stability of SMOTE-based oversampling techniques in software defect prediction[J]. Information and Software Technology,2021,139(6). DOI: 10.1016/j.infsof.2021.106662.
[20]	CHAO Ying,YIN Kunlong,ZHOU Chao,et al. Establishment of landslide groundwater level prediction model based on GA-SVM and influencing factor analysis[J]. Sensors,2020,20(3):845. doi: 10.3390/s20030845
[21]	吕红燕,冯倩. 随机森林算法研究综述[J]. 河北省科学院学报,2019,36(3):37-41. doi: 10.16191/j.cnki.hbkx.2019.03.005 LYU Hongyan,FENG Qian. A review of random forests algorithm[J]. Journal of the Hebei Academy of Sciences,2019,36(3):37-41. doi: 10.16191/j.cnki.hbkx.2019.03.005