Research on classification and identification of mine microseismic signals based on deep learning method
-
摘要: 为了精准识别矿山微震信号，本文提出了一种适用于识别矿山微震信号的VGG4-CNN深度学习网络模型，该模型采用Python语言进行编写，基于PyTorch深度学习网络架构框架进行搭建。根据矿山生产过程中的岩石破裂、爆破作业、背景噪声等9类事件的微震信号的时域特征，VGG4-CNN深度学习网络模型实现了对3 835组矿山微震信号数据进行监督学习训练和分类识别应用。研究结果表明：本文构建的VGG4-CNN神经网络识别精度高达94 %，在采用该模型时不需要对原有波形信号进行去噪且鲁棒性强于现存其他模型，可在中等层次GPU上实现，满足工程需要。Abstract: In order to accurately identify mine microseismic signals, this paper proposes a VGG4-CNN deep learning network model suitable for identifying mine microseismic signals. The model is written in Python language and built based on the PyTorch deep learning network architecture framework. Based on the time-domain characteristics of the microseismic signals of 9 types of events such as rock fracture, blasting operations, and background noise in the mine production process, VGG4-CNN has realized the supervised learning training and classification recognition application of 3 835 sets of mine microseismic signal data. The research results show that the recognition accuracy of the VGG4-CNN neural network constructed in this paper is as high as 94 %. This model does not require denoising of the original waveform signal and is more robust than other models. The implementation can be performed by a medium-level GPU to meet engineering requirements.
-
Key words:
- microseismic signal /
- identification technology /
- deep learning /
- time domain analysis
-
图 4 VGG4-CNN训练过程中准确率和损失变化情况
Figure 4. Accuracy and loss changes in VGG4-CNN training process
图 5 各类信号经t-SNE降维后的平面分布
Figure 5. Plane distribution of various signals after dimensionality reduction by t-SNE
图 6 4种神经网络训练过程损失及准确率对比
Figure 6. Comparison of loss and accuracy of 4 types neural network training process
表 1 神经网络训练各项参数
Table 1. Parameters of neural network training
模型 单次输入量 训练步伐 学习率 输入尺寸 随机失活率 VGG4-CNN 16 50 0.01 100×100 0.1 
下载: 导出CSV
表 2 VGG4-CNN神经网络对于不同种类信号各层输出情况
Table 2. VGG4-CNN neural network output for different types of signals
事件名称 C1-1 P1-1 C2-1 C3-1 P2-1 铲煤 
背景 
爆破 
锄煤 
破裂 
喊话 
矿车 
敲击 
镐煤 
下载: 导出CSV
表 3 三类矿山微震信号P2-2特征层
Table 3. P2-2 layer characteristics of microseismic signals of three types of mines
事件名称 喊话 敲击 镐煤 P2-2层 
下载: 导出CSV
表 4 VGG4-CNN神经网络对于各项数据识别结果及耗时
Table 4. Identification results and time consuming of VGG4-CNN neural network
类别 VGG4-CNN 精确率 召回率 F1系数 识别数目 铲煤 0.73 0.63 0.68 35 背景 1.00 0.99 1.00 130 爆破 0.88 0.80 0.84 35 锄煤 0.96 0.92 0.94 52 破裂 0.99 0.98 0.99 414 喊话 1.00 0.96 0.98 46 矿车 0.97 1.00 0.99 72 敲击 0.84 0.72 0.78 29 镐煤 0.82 0.92 0.87 139 均值 0.91 0.88 0.89 加权均值 0.95 0.94 0.94 耗时/ms 231.89
下载: 导出CSV
表 5 4类神经网络耗时及准确率对比
Table 5. Comparison of time consumption and accuracy of four types neural networks
模型 准确率 耗时/ms 分类器 VGG4-CNN 0.94 231.89 Softmax LSTM-RNN 0.89 469.20 Full Connection BP 0.54 583.21 - DCNN 0.72 1 173.95 SVM
下载: 导出CSV
-
[1] 李铁, 倪建明, 李忠凯. 采动岩体强矿震破裂机制反演及其防治对策[J]. 采矿与安全工程学报, 2016, 33(6): 1110-1115. https://www.cnki.com.cn/Article/CJFDTOTAL-KSYL201606021.htmLi Tie, Ni Jianming, Li Zhongkai. Rupture mechanism inversion of mining-induced strong mine earthquake and its preventive methods[J]. Journal of Mining & Safety Engineering, 2016, 33(6): 1110-1115. https://www.cnki.com.cn/Article/CJFDTOTAL-KSYL201606021.htm [2] 赵洪宝, 刘一洪, 程辉, 等. 回坡底煤矿回采巷道非对称底鼓机理及防治措施[J]. 矿业科学学报, 2020, 5(6): 638-647. doi: 10.19606/j.cnki.jmst.2020.06.006Zhao Hongbao, Liu Yihong, Cheng Hui, et al. Mechanism and prevention measures of asymmetric floor heave in mining roadway of Huipodi Coal Mine[J]. Journal of Mining Science and Technology, 2020, 5(6): 638-647. doi: 10.19606/j.cnki.jmst.2020.06.006 [3] 徐奴文, 唐春安, 沙椿, 等. 锦屏一级水电站左岸边坡微震监测系统及其工程应用[J]. 岩石力学与工程学报, 2010, 29(5): 915-925. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201005010.htmXu Nuwen, Tang Chun'an, Sha Chun, et al. Microseismic monitoring system establishment and its engineering applications to left bank slope of Jinping I hydropower station[J]. Chinese Journal of Rock Mechanics and Engineering, 2010, 29(5): 915-925. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201005010.htm [4] 程辉, 赵洪宝, 徐建峰, 等. 基于滑移线场理论的巷道底鼓机理与防治技术研究[J]. 矿业科学学报, 2021, 6(3): 314-322. doi: 10.19606/j.cnki.jmst.2021.03.008Cheng Hui, Zhao Hongbao, Xu Jianfeng, et al. Study on floor heave mechanism and control technology of roadway based on slip line field theory[J]. Journal of Mining Science and Technology, 2021, 6(3): 314-322. doi: 10.19606/j.cnki.jmst.2021.03.008 [5] 陈宇龙, 曹鹏. 混凝土时间相关断裂模型研究综述[J]. 矿业科学学报, 2021, 6(3): 290-295. doi: 10.19606/j.cnki.jmst.2021.03.005Chen Yulong, Cao Peng. Review of time-dependent fracture model of concrete[J]. Journal of Mining Science and Technology, 2021, 6(3): 290-295. doi: 10.19606/j.cnki.jmst.2021.03.005 [6] 朱梦博, 王李管, 刘晓明, 等. 基于波形参数的微震P波到时拾取值质量控制方法[J]. 岩土力学, 2019, 40(2): 767-776. https://www.cnki.com.cn/Article/CJFDTOTAL-YTLX201902040.htmZhu Mengbo, Wang Liguan, Liu Xiaoming, et al. A quality control method for microseismic P-wave phase pickup value based on waveform parameters[J]. Rock and Soil Mechanics, 2019, 40(2): 767-776. https://www.cnki.com.cn/Article/CJFDTOTAL-YTLX201902040.htm [7] 雷艺繁, 徐奴文, 董林鹭, 等. 开挖卸荷下双江口水电站地下厂房上游拱肩破坏机制研究[J]. 岩土力学, 2020, 41(S2): 1-11.Lei Yifan, Xü Nuwen, Dong Linlu, et al. Failure mechanism of spandrel under the excavation and unloading condition, at the upstream of underground powerhouse, Shuangjiangkou hydropower station[J]. Rock and Soil Mechanics, 2020, 41(S2): 1-11. [8] Jeffrey F T, Henry C B, Robert R S. Stewart classification of microseismic events from bitumen production at Cold Lake, Alberta[J]. CREWES Research Report, 2007, 19(1): 1-24. [9] Le C 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 [10] 赵国彦, 邓青林, 李夕兵, 等. 基于EMD和形态分形维数的微震波形识别[J]. 中南大学学报: 自然科学版, 2017, 48(1): 162-167. https://www.cnki.com.cn/Article/CJFDTOTAL-ZNGD201701023.htmZhao Guoyan, Deng Qinglin, Li Xibing, et al. Recognition of microseismic waveforms based on EMD and morphological fractal dimension[J]. Journal of Central South University: Science and Technology, 2017, 48(1): 162-167. https://www.cnki.com.cn/Article/CJFDTOTAL-ZNGD201701023.htm [11] 赵国彦, 邓青林, 马举. 基于FSWT时频分析的矿山微震信号分析与识别[J]. 岩土工程学报, 2015, 37(2): 306-312. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC201502018.htmZhao Guoyan, Deng Qinglin, Ma Ju. Recognition of mine microseismic signals based on FSWT time-frequency analysis[J]. Chinese Journal of Geotechnical Engineering, 2015, 37(2): 306-312. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC201502018.htm [12] 尚雪义, 李夕兵, 彭康, 等. 基于EMD-SVD的矿山微震与爆破信号特征提取及分类方法[J]. 岩土工程学报, 2016, 38(10): 1849-1858. doi: 10.11779/CJGE201610014Shang Xueyi, Li Xibing, Peng Kang, et al. Feature extraction and classification of mine microseism and blast based on EMD-SVD[J]. Chinese Journal of Geotechnical Engineering, 2016, 38(10): 1849-1858. doi: 10.11779/CJGE201610014 [13] 董陇军, 孙道元, 李夕兵, 等. 微震与爆破事件统计识别方法及工程应用[J]. 岩石力学与工程学报, 2016, 35(7): 1423-1433. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201607013.htmDong Longjun, Sun Daoyuan, Li Xibing, et al. A statistical method to identify blasts and microseismic events and its engineering application[J]. Chinese Journal of Rock Mechanics and Engineering, 2016, 35(7): 1423-1433. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201607013.htm [14] Zhang Xingli, Jia Ruisheng, Lu Xinming, et al. Identifi cation of blasting vibration and coal-rock fracturing microseismic signals[J]. Applied Geophysics, 2018, 15(2): 280-289, 364. doi: 10.1007/s11770-018-0682-9 [15] 朱权洁, 李青松, 李绍泉, 等. 煤与瓦斯突出试验的微震动态响应与特征分析[J]. 岩石力学与工程学报, 2015, 34(S2): 3813-3821. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX2015S2023.htmZhu Quanjie, Li Qingsong, Li Shaoquan, et al. Microseismic dynamic response and characteristic analysis of coal and gas outburst experiment[J]. Chinese Journal of Rock Mechanics and Engineering, 2015, 34(S2): 3813-3821. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX2015S2023.htm [16] Bi L, Xie W, Zhao J J. Automatic recognition and classification of multi-channel microseismic waveform based on DCNN and SVM[J]. Computers & Geosciences, 2019, 123: 111-120. [17] 王国法, 徐亚军, 孟祥军, 等. 智能化采煤工作面分类、分级评价指标体系[J]. 煤炭学报, 2020, 45(9): 3033-3044. https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB202009001.htmWang Guofa, Xu Yajun, Meng Xiangjun, et al. Specification, classification and grading evaluation index for smart longwall mining face[J]. Journal of China Coal Society, 2020, 45(9): 3033-3044. https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB202009001.htm [18] 李彦冬, 郝宗波, 雷航. 卷积神经网络研究综述[J]. 计算机应用, 2016, 36(9): 2508-2515, 2565. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201609029.htmLi Yandong, Hao Zongbo, Lei Hang. Survey of convolutional neural network[J]. Journal of Computer Applications, 2016, 36(9): 2508-2515, 2565. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201609029.htm [19] Bottou L, Curtis F E, Nocedal J. Optimization methods for large-scale machine learning[J]. SIAM Review, 2018, 60(2): 223-311. doi: 10.1137/16M1080173 -
点击查看大图
图(6) / 表(5)
计量
- 文章访问数: 716
- HTML全文浏览量: 616
- PDF下载量: 123
- 被引次数: 0