基于CNN神经网络的煤层底板突水预测

陈建平; 王春雷; 王雪冬

doi:10.16031/j.cnki.issn.1003-8035.2021.01.07

基于CNN神经网络的煤层底板突水预测

辽宁工程技术大学矿业学院，辽宁阜新　123000

基金项目: 国家自然科学基金项目(51604140)

详细信息

作者简介: 陈建平(1971-)，男，山西保德人，博士，副教授，主要从事矿山水文地质等方面的教学和科研工作。E-mail：chenjianp123000@163.com

中图分类号: TD76

收稿日期: 2020-03-10

修回日期: 2020-04-23

刊出日期: 2021-02-25

Coal mine floor water inrush prediction based on CNN neural network

College of Mining ,Liaoning Technical University, Fuxin, Liaoning　123000, China

Received Date: 10 March 2020

Revised Date: 23 April 2020

Publish Date: 25 February 2021

摘要

摘要:
为了提高煤层底板突水预测的准确性，建立了基于卷积神经网络的煤层底板突水预测模型。通过综合分析，确定了15个影响煤层底板突水的因素，将这些影响因素进行拼接组合，运用建立的深度计算结构模型对影响因素及其相互联系进行特征提取。用已知的115组数据对模型进行学习训练，并进行了预测。为验证模型的准确性，利用相同的数据对BP神经网络模型和LeNet-5模型进行训练，将建立的模型与BP神经网络模型和LeNet-5模型进行对比。结果表明：该模型通过加深模型的计算深度，综合考虑了影响底板突水因素间的相互联系，提高了突水预测准确性。基于卷积神经网络构建的模型可以对煤层底板突水进行预测，并且准确率相对较高。
- 卷积神经网络 /
- 深度学习 /
- 煤层底板 /
- 突水预测 /
- 数据拼接组合
Abstract:
With a view to improving the accuracy of water bursting prediction in coal seam floor, a model based on Convolutional Neural Network (CNN) was established. Through comprehensive analysis of water bursting in coal seam floor, 15 factors affecting water bursting prediction were determined and these factors were combined to stimulate the partial correlation among these factors. These factors and their interrelation on water bursting prediction in coal seam floor were simulated by using the structure model established for depth calculation. Training and prediction were performed by using the known 115 sets of data. To verify the model efficiency, the BP neural network model and the LeNet-5 model were trained by using the same data, and then the established BP neural network model was compared with the LeNet-5 model. The result indicates that the interrelation between factors affecting water bursting prediction is considered comprehensively by deepening the calculation depth of the model, and the accuracy of water bursting prediction is improved. The prediction model of water bursting prediction in coal seam floor based on Convolutional Neural Network (CNN) has high accuracy and small standard error, which improves the accuracy of prediction effectively.
- convolutional neural network /
- deep learning /
- coal seam floor /
- water inrush prediction /
- data stitching combination

HTML全文

图 1 BP神经网络与卷积神经网络连接方式示意图

Figure 1.

下载: 全尺寸图片幻灯片

图 2 卷积神经网络结构示意图

Figure 2.

下载: 全尺寸图片幻灯片

图 3 模型结构示意图

Figure 3.

下载: 全尺寸图片幻灯片

图 4 接触概率设置对比实验结果

Figure 4.

下载: 全尺寸图片幻灯片

图 5 数据拼接组合过程示意图

Figure 5.

下载: 全尺寸图片幻灯片

图 6 CNN模型在训练集和测试集上的正确率

Figure 6.

下载: 全尺寸图片幻灯片

图 7 测试集上的预测值与实际值对比

Figure 7.

下载: 全尺寸图片幻灯片

表 1 影响煤层底板突水的因素

Table 1. Factors affecting water inrush from coal floor

一级指标	二级指标
含水层因素	单位涌水量(x₁)
	水压(x₂)
	含水层厚度(x₃)
隔水层因素	隔水层厚度(x₄)
	隔水层岩石饱和单轴抗压强度(x₅)
	隔水层岩体完整性指数(x₆)
构造因素	单位面积断层条数(x₇)
	大断层条数(x₈)
	裂隙发育程度(x₉)
煤层因素	煤层埋深(x₁₀)
煤层因素	煤层倾角(x₁₁)
开采因素	底板破坏带深度(x₁₂)
	工作面长度(x₁₃)
	采高(x₁₄)
	开采面积(x₁₅)

下载: 导出CSV

表 2 不同卷积层的预测结果

Table 2. Prediction results of different convolution layers

编号	卷积层数量/个	准确率/%	标准误差
1	1	67	0.492
2	2	93	0.268
3	3	80	0.385
4	4	73	0.436

下载: 导出CSV

表 3 部分样本数据

Table 3. Part of the sample data

序号	x₁/(L·s⁻¹·m⁻¹)	x₂/MPa	x₃/m	x₄/m	x₅/MPa	x₆	x₇/(条·km⁻²)	x₈/条	x₉/%	x₁₀/m	x₁₁/(°)	x₁₂/m	x₁₃/m	x₁₄/m	x₁₅/m²	是否突水
1	3.4	2.80	14.35	39.17	34.28	0.64	34.2	4	14.52	347	12	24.33	150	2.75	5 240	是
2	3.5	2.95	14.35	36.11	34.28	0.64	34.2	4	14.52	380	12	24.61	150	2.55	4 260	是
17	0.15	0.73	15	79	42.00	0.80	0.44	1	5	178	13	20.90	200	3.08	8100	否
18	0.15	1.00	15	51	37.12	0.80	0.44	1	5	202	13	18.95	180	1.80	7800	否
53	2.5	1.37	12.54	50.92	38.30	0.53	42.5	5	18.22	332	14	33.17	200	3.50	5 900	是
54	2.6	1.45	13.02	46.23	38.30	0.53	42.5	5	18.22	352	15	33.50	200	3.50	5 400	是
88	1.8	2.25	35.15	30.00	17.38	0.60	21.6	4	16.53	289	13	19.69	100	3.00	1 530	否
89	1.8	2.35	35.15	30.00	17.38	0.60	21.6	4	16.53	323	16	20.47	100	2.75	1 535	否
95	0.16	1.02	70	52	34.60	0.75	0.52	6	11	370	9	12.15	110	2.76	23500	否
111	2.12	2.89	85	67	46.12	0.55	36.11	4	29	351	9	9.84	90	1.56	4180	是
114	0.29	0.84	31	121	38.13	0.80	0.14	1	13	270	4	14.79	150	3.12	8510	否
115	0.29	1.08	31	97	41.89	0.80	0.14	1	13	295	4	15.00	150	1.97	8510	否

下载: 导出CSV

表 4 实验参数

Table 4. Experimental parameters

实验参数	数值
Learning rate	0.00001
Epochs	1800
Batch size	10
Dropout	0.5

下载: 导出CSV

表 5 各个模型的正确率

Table 5. Accuracy of the predicted results

预测模型	训练集准确/%	测试集准确/%	标准误差
BP	74	67	0.450
LeNet-5	83	80	0.430
本文CNN模型	96	100	0.135

下载: 导出CSV

表 6 各个模型的预测结果

Table 6. Testresults of the forecast model

序号	实际情况	BP预测	LeNet-5预测	本文CNN模型
1	不突水	突水	不突水	不突水
2	突水	突水	不突水	突水
3	突水	突水	突水	突水
4	突水	突水	突水	突水
5	不突水	突水	不突水	不突水
6	不突水	突水	不突水	不突水
7	突水	突水	突水	突水
8	突水	突水	突水	突水
9	不突水	不突水	不突水	不突水
10	不突水	不突水	不突水	不突水
11	突水	突水	不突水	突水
12	突水	突水	不突水	突水
13	不突水	突水	不突水	不突水
14	不突水	不突水	不突水	不突水
15	不突水	突水	不突水	不突水

下载: 导出CSV

参考文献(24)

[1]	武强, 涂坤, 曾一凡, 等. 打造我国主体能源(煤炭)升级版面临的主要问题与对策探讨[J]. 煤炭学报,2019,44(6):1625 − 1636. [WU Qiang, TU Kun, ZENG Yifan, et al. Discussion on the main problems and countermeasures for building an upgrade version of main energy(coal)industry in China[J]. Journal of China Coal Society,2019,44(6):1625 − 1636. (in Chinese with English abstract)
[2]	施龙青, 谭希鹏, 王娟, 等. 基于PCA_Fuzzy_PSO_SVC的底板突水危险性评价[J]. 煤炭学报,2015,40(1):167 − 171. [SHI Longqing, TAN Xipeng, WANG Juan, et al. Risk assessment of water inrush based on PCA_Fuzzy_PSO_SVC[J]. Journal of China Coal Society,2015,40(1):167 − 171. (in Chinese with English abstract)
[3]	张文泉, 张广鹏, 李伟, 等. 煤层底板突水危险性的Fisher判别分析模型[J]. 煤炭学报,2013,38(10):1831 − 1836. [ZHANG Wenquan, ZHANG Guangpeng, LI Wei, et al. A model of Fisher's discriminant analysis for evaluating water inrush risk from coal seam floor[J]. Journal of China Coal Society,2013,38(10):1831 − 1836. (in Chinese with English abstract)
[4]	靳德武, 马培智. 华北煤层底板突水的随机—信息模拟及预测[J]. 煤田地质与勘探,1998,26(6):36 − 39. [JIN Dewu, MA Peizhi. Random information simulation and forecast of water inrush through coal seam floor in mining areas of Northern China[J]. Coal Geology & Exploration,1998,26(6):36 − 39. (in Chinese with English abstract)
[5]	刘伟韬, 张文泉. 用层次分析—模糊评判进行底板突水安全性评价[J]. 煤炭学报,2000,25(3):278 − 282. [LIU Weitao, ZHANG Wenquan. An evaluation of the safety of floor water irruption using analytic hierarchy process and fuzzy synthesis methods[J]. Journal of China Coal Society,2000,25(3):278 − 282. (in Chinese with English abstract) doi: 10.3321/j.issn:0253-9993.2000.03.013
[6]	武强, 庞炜, 戴迎春, 等. 煤层底板突水脆弱性评价的GIS与ANN耦合技术[J]. 煤炭学报,2006,31(3):314 − 319. [WU Qiang, PANG Wei, DAI Yingchun, et al. Vulnerability forecasting model based on coupling technique of GIS and ANN in floor groundwater bursting[J]. Journal of China Coal Society,2006,31(3):314 − 319. (in Chinese with English abstract) doi: 10.3321/j.issn:0253-9993.2006.03.010
[7]	刘伟韬, 廖尚辉, 刘士亮, 等. 主成分Logistic回归分析在底板突水预测中的应用[J]. 辽宁工程技术大学学报(自然科学版),2015,34(8):905 − 909. [LIU Weitao, LIAO Shanghui, LIU Shiliang, et al. Principal component Logistic regression analysis in application of water outbursts from coal seam floor[J]. Journal of Liaoning Technical University (Natural Science Edition),2015,34(8):905 − 909. (in Chinese with English abstract)
[8]	高延法, 章延平, 张慧敏, 等. 底板突水危险性评价专家系统及应用研究[J]. 岩石力学与工程学报,2009,28(2):253 − 258. [GAO Yanfa, ZHANG Yanping, ZHANG Huimin, et al. Research on expert system for risk assessment of water inrush from coal floor and its application[J]. Chinese Journal of Rock Mechanics and Engineering,2009,28(2):253 − 258. (in Chinese with English abstract) doi: 10.3321/j.issn:1000-6915.2009.02.005
[9]	张和生, 薛光武, 石秀伟, 等. 基于地学信息复合叠置分析对煤层底板突水的预测[J]. 煤炭学报,2009,34(8):1100 − 1104. [ZHANG Hesheng, XUE Guangwu, SHI Xiuwei, et al. Prediction of water inrush from coal seam floor confined based on geo-information composite overlay analysis[J]. Journal of China Coal Society,2009,34(8):1100 − 1104. (in Chinese with English abstract) doi: 10.3321/j.issn:0253-9993.2009.08.019
[10]	肖建于, 童敏明, 姜春露. 基于模糊证据理论的煤层底板突水量预测[J]. 煤炭学报,2012,37(增刊1):131 − 137. [XIAO Jianyu, TONG Minming, JIANG Chunlu. Prediction of water inrush quantity from coal floor based on fuzzy evidence theory[J]. Journal of China Coal Society,2012,37(Sup1):131 − 137. (in Chinese with English abstract)
[11]	董东林, 孙文洁, 朱兆昌, 等. 基于GIS—BN技术的范各庄矿煤12底板突水态势评价[J]. 煤炭学报,2012,37(6):999 − 1004. [DONG Donglin, SUN Wenjie, ZHU Zhaochang, et al. Water-inrush assessment of coal 12 floor using a GIS-based Bayesian network for Fangezhuang Coal Mine with collapse column[J]. Journal of China Coal Society,2012,37(6):999 − 1004. (in Chinese with English abstract)
[12]	宋国娟. 基于极限学习机的煤矿突水预测及避险路线优化研究[D]. 徐州: 中国矿业大学, 2016. SONG Guojuan. Research on mine water inrush prediction based on extreme learning machine and route optimization[D]. Xuzhou: China University of Mining and Technology, 2016. (in Chinese with English abstract)
[13]	姜谙男, 梁冰. 基于最小二乘支持向量机的煤层底板突水量预测[J]. 煤炭学报,2005,30(5):613 − 617. [JIANG Annan, LIANG Bing. Forecast of water inrush from coal floor based on least square support vector machine[J]. Journal of China Coal Society,2005,30(5):613 − 617. (in Chinese with English abstract) doi: 10.3321/j.issn:0253-9993.2005.05.016
[14]	曹庆奎, 赵斐. 基于模糊-支持向量机的煤层底板突水危险性评价[J]. 煤炭学报,2011,36(4):633 − 637. [CAO Qingkui, ZHAO Fei. Risk evaluation of water inrush from coal floor based on fuzzy-support vector machine[J]. Journal of China Coal Society,2011,36(4):633 − 637. (in Chinese with English abstract)
[15]	乔育锋. 遗传算法和BP神经网络在煤矿突水预测中的应用研究[D]. 西安: 西安建筑科技大学, 2011. QIAO Yufeng. Application research of genetic algorithm and artificial neural networks in the prediction of mine water inrush[D]. Xi’an: Xian University of Architecture and Technology, 2011. (in Chinese with English abstract)
[16]	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. 2014: arXiv: 1409.1556[cs. CV]. https://arxiv.org/abs/1409.1556
[17]	刘小安. 卷积神经网络在自然语言处理中的应用研究综述[C]//中国计算机用户协会网络应用分会. 中国计算机用户协会网络应用分会2017年第二十一届网络新技术与应用年会论文集. 北京联合大学北京市信息服务工程重点实验室, 2017: 5. LIU Xiaoan. Application of convolutional neural network in nature language processing[C]//Network application branch of China Computer Users Association. Papers of network application branch of China Computer Users Association at the 21st Annual Meeting of network new technology and application in 2017. Beijing Key Laboratory of information service engineering, Beijing Union University, 2017: 5. (in Chinese)
[18]	ER M J, ZHANG Y, WANG N, et al. Attention pooling-based convolutional neural network for sentence modelling[J]. Information Sciences,2016,373:388 − 403. doi: 10.1016/j.ins.2016.08.084
[19]	吴素雯, 战荫伟. 基于选择性搜索和卷积神经网络的人脸检测[J]. 计算机应用研究,2017,34(9):2854 − 2857. [WU Suwen, ZHAN Yinwei. Face detection based on selective search and Gabor optimizing convolutional neural network[J]. Application Research of Computers,2017,34(9):2854 − 2857. (in Chinese with English abstract) doi: 10.3969/j.issn.1001-3695.2017.09.064
[20]	秦品乐, 李鹏波, 曾建潮, 等. 基于级联全卷积神经网络的颈部淋巴结自动识别算法[J]. 计算机应用,2019,39(10):2915 − 2922. [QIN Pinle, LI Pengbo, ZENG Jianchao, et al. Automatic recognition algorithm of cervical lymph nodes using cascaded fully convolutional neural networks[J]. Journal of Computer Applications,2019,39(10):2915 − 2922. (in Chinese with English abstract)
[21]	WALLACH I, DZAMBA M, HEIFETSA. AtomNet: A deep convolutional neural network for bioactivity prediction in structure-based drug discovery[J]. Mathematische Zeitschrift, 2015.
[22]	HINTON G E, SRIVASTAVA N, KRIZHEVSKY A, et al. Improving neural networks by preventing co-adaptation of feature detectors[EB]. 2012: arXiv: 1207.0580[cs. NE].
[23]	KINGMA D P, BA J. Adam: a method for stochastic optimization[EB]. 2014: arXiv: 1412.6980[cs. LG].
[24]	LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE,1998,86(11):2278 − 2324. doi: 10.1109/5.726791