Research on coal and rock recognition model based on improved 1DCNN
-
摘要: 随着煤矿智能化建设的加速推进，煤岩高效识别已成为煤炭智能化开采亟待解决的技术难题。针对复杂煤矿地质条件下现有煤岩识别方法精度低、通用性差且难以工程应用等问题，提出了一种基于改进一维卷积神经网络（1DCNN）的煤岩识别模型。以1DCNN为基础，使用多个连续卷积层提取一维振动信号特征，通过全局均值池化（GAP）层代替全连接层，以减少模型训练参数，节省计算资源，同时采用带有线性热启动的余弦退火衰减方法优化学习率，以避免模型训练陷入局部极小值区域，提升训练质量。为直观描述改进1DCNN模型对煤岩截割振动数据的特征提取过程和分类能力，采用t−分布随机近邻嵌入（t−SNE）流形学习算法对模型的特征学习过程进行可视化分析，结果表明，改进1DCNN模型通过逐层特征学习，很好地实现了对煤岩截割状态的识别。以陕西某矿MG650/1590−WD型采煤机截割煤岩时的实测振动数据为样本进行模型训练，结果表明，改进1DCNN模型在训练集上的准确率为99.91%，在测试集上的准确率为99.32%，可直接用于采煤机截割煤岩时的原始振动信号分类，并能够有效识别煤岩截割状态。与传统机器学习、集成学习及未改进的1DCNN模型相比，改进1DCNN模型具有明显优势，平均识别准确率达99.56%，同时大大节约了计算成本，提高了模型识别速度。Abstract: With the acceleration of intelligent construction of coal mines, efficient recognition of coal and rock has become a technical problem to be solved urgently in intelligent coal mining. The existing coal and rock recognition methods under complex coal mine geological conditions have problems of low precision, poor universality and are difficult to apply in engineering. In order to solve the above problems, a coal and rock recognition model based on improved 1-dimensional convolutional neural network (1DCNN) is proposed. Based on the 1DCNN, a plurality of continuous convolution layers are used for extracting one-dimensional vibration signal features. The global average pool (GAP) layer is used for replacing the full connection layer. The model training parameters are reduced, and computing resources are saved. At the same time, a cosine annealing attenuation method with a linear hot start is adopted for optimizing the learning rate. Therefore, the model training is prevented from falling into a local minimum region, and the training quality is improved. In order to intuitively describe the feature extraction process and classification capability of the improved 1DCNN model for coal and rock cutting vibration data, the t-distributed stochastic neighbor embedding (t-SNE) manifold learning algorithm is used to visually analyze the feature learning process of the model. The results show that the improved 1DCNN model can realize the recognition of coal and rock cutting states well through feature learning layer by layer. Based on the measured vibration data obtained in the process of coal and rock cutting of the MG 650/1590-WD shearer in a mine in Shaanxi province, the model is trained and the result shows that the accuracy of the improved 1DCNN model is 99.91% on the training set and 99.32% on the test set. The model can be directly used to classify the original vibration signals of the shearer in coal and rock cutting, and can effectively identify the cutting state of coal and rock. Compared with traditional machine learning, ensemble learning and the unmodified 1DCNN model, the improved 1DCNN model has obvious advantages. The average recognition accuracy rate reaches 99.56%. The calculation cost is greatly saved, and the model recognition speed is improved.
-
图 1 改进1DCNN模型结构
Figure 1. Structure of improved 1-dimensional convolutional neural network model
图 3 滑窗法构造的煤岩信号样本
Figure 3. Coal and rock signal samples constructed by sliding window method
图 4 训练集在各网络层的特征学习效果
Figure 4. Feature learning effect of training set at each network layer
图 5 改进1DCNN模型在训练集与测试集上的准确率曲线
Figure 5. Accuracy curves of improved 1-dimensional convolutional neural network model on training set and test set
图 6 改进1DCNN模型在测试集上的混淆矩阵
Figure 6. The confusion matrix of improved 1-dimensional convolutional neural network model on test set
表 1 样本数据集信息
Table 1. Sample dataset information
截割状态 样本长度 训练样本数 测试样本数 标签 割煤 400 6 392 660 0 割岩 400 6 392 660 1 
下载: 导出CSV
表 2 改进1DCNN模型结构
Table 2. Structure of improved 1-dimensional convolutional neural network model
网络层 输出形状 参数量/个 Input(Input Layer) (400，1) 0 Conv1(Conv1D) (400， 64) 256 BN2 (Batch Normalization) (400， 64) 256 Map3 (MaxPooling1D) (200， 64) 0 Conv4 (Conv1D) (100， 128) 24 704 Conv5 (Conv1D) (50， 128) 49 280 Conv6 (Conv1D) (25， 256) 98 560 BN7 (Batch Normalization) (25， 256) 1024 Map8 (MaxPooling1D) (12，256) 0 Conv9 (Conv1D) (6，256) 196 864 GAP10 (GlobalAveragePooling1D) (256，1) 0 Dropout (Dropout) (256，1) 0 Output (Dense) (2，1) 514
下载: 导出CSV
表 3 模型对比实验结果
Table 3. Model comparison experiment results
% 模型 准确率 实验1 实验2 实验3 实验4 实验5 平均值 KNN 93.66 94.50 94.00 93.67 94.50 94.07 RF 76.50 75.50 76.16 75.66 75.83 75.93 SVM 92.83 94.00 94.50 93.00 94.83 93.83 XGBoost 92.66 93.00 93.83 89.50 94.50 92.70 R1DCNN 98.82 99.19 99.03 98.85 98.94 98.97 改进1DCNN 99.32 99.69 99.70 99.55 99.54 99.56
下载: 导出CSV
-
[1] 张强,张润鑫,刘峻铭,等. 煤矿智能化开采煤岩识别技术综述[J]. 煤炭科学技术,2022,50(2):1-26.ZHANG Qiang,ZHANG Runxin,LIU Junming,et al. Review on coal and rock identification technology for intelligent mining in coal mines[J]. Coal Science and Technology,2022,50(2):1-26. [2] 王增才,王汝琳,徐建华,等. 自然 $ \text{γ} $ 射线法在采煤机摇臂自动调高中检测煤层厚度的研究[J]. 煤炭学报,2002,27(4):425-429. doi: 10.3321/j.issn:0253-9993.2002.04.020WANG Zengcai,WANG Rulin,XU Jianhua,et al. Research on coal seam thickness detection by natural Gamma ray in shearer horizon control[J]. Journal of China Coal Society,2002,27(4):425-429. doi: 10.3321/j.issn:0253-9993.2002.04.020[3] 王昕,丁恩杰,胡克想,等. 煤岩散射特性对探地雷达探测煤岩界面的影响[J]. 中国矿业大学学报,2016,45(1):34-41. doi: 10.13247/j.cnki.jcumt.000455WANG Xin,DING Enjie,HU Kexiang,et al. Effects of coal-rock scattering characteristics on the GPR detection of coal-rock interface[J]. Journal of China University of Mining & Technology,2016,45(1):34-41. doi: 10.13247/j.cnki.jcumt.000455 [4] 张强,孙绍安,张坤,等. 基于主动红外激励的煤岩界面识别[J]. 煤炭学报,2020,45(9):3363-3370.ZHANG Qiang,SUN Shao'an,ZHANG Kun,et al. Coal and rock interface identification based on active infrared excitation[J]. Journal of China Coal Society,2020,45(9):3363-3370. [5] 苗曙光,邵丹,刘忠育,等. 基于太赫兹时域光谱技术的煤岩识别方法研究[J]. 光谱学与光谱分析,2022,42(6):1755-1760. doi: 10.3964/j.issn.1000-0593(2022)06-1755-06MIAO Shuguang,SHAO Dan,LIU Zhongyu,et al. Study on coal-rock identification method based on terahertz time-domain spectroscopy[J]. Spectroscopy and Spectral Analysis,2022,42(6):1755-1760. doi: 10.3964/j.issn.1000-0593(2022)06-1755-06 [6] 杨恩,王世博,宣统. 融合近红外光谱的煤岩界面分布感知研究[J]. 工矿自动化,2022,48(7):22-31,42. doi: 10.13272/j.issn.1671-251x.17950YANG En,WANG Shibo,XUAN Tong. Research on coal-rock interface distribution perception based on near-infrared spectra[J]. Journal of Mine Automation,2022,48(7):22-31,42. doi: 10.13272/j.issn.1671-251x.17950 [7] 孙传猛,王燕平,王冲,等. 融合改进YOLOv3与三次样条插值的煤岩界面识别方法[J]. 采矿与岩层控制工程学报,2022,4(1):81-90.SUN Chuanmeng,WANG Yanping,WANG Chong,et al. Coal-rock interface identification method based on improved YOLOv3 and cubic spline interpolation[J]. Journal of Mining and Strata Control Engineering,2022,4(1):81-90. [8] ZHANG Qiang, GU Jieying, LIU Junming. Research on coal and rock type recognition based on mechanical vision[J]. Shock and Vibration, 2021. DOI: 10.1155/2021/6617717. [9] 刘春生,刘延婷,刘若涵,等. 采煤机截割状态与煤岩识别的关联载荷特征模型[J]. 煤炭学报,2022,47(1):527-540.LIU Chunsheng,LIU Yanting,LIU Ruohan,et al. Correlation load characteristic model between shearer cutting state and coal-rock recognition[J]. Journal of China Coal Society,2022,47(1):527-540. [10] 赵丽娟,王雅东,张美晨,等. 复杂煤层条件下采煤机自适应截割控制策略[J]. 煤炭学报,2022,47(1):541-563. doi: 10.13225/j.cnki.jccs.yg21.1862ZHAO Lijuan,WANG Yadong,ZHANG Meichen,et al. Research on self-adaptive cutting control strategy of shearer in complex coal seam[J]. Journal of China Coal Society,2022,47(1):541-563. doi: 10.13225/j.cnki.jccs.yg21.1862 [11] 张启志,邱锦波,庄德玉. 基于倒谱距离的采煤机煤岩截割振动信号识别[J]. 工矿自动化,2017,43(1):9-12. doi: 10.13272/j.issn.1671-251x.2017.01.003ZHANG Qizhi,QIU Jinbo,ZHUANG Deyu. Vibration signal identification of coal-rock cutting of shearer based on cepstral distance[J]. Industry and Mine Automation,2017,43(1):9-12. doi: 10.13272/j.issn.1671-251x.2017.01.003 [12] 周飞燕,金林鹏,董军. 卷积神经网络研究综述[J]. 计算机学报,2017,40(6):1229-1251. doi: 10.11897/SP.J.1016.2017.01229ZHOU Feiyan,JIN Linpeng,DONG Jun. Review of convolutional neural network[J]. Chinese Journal of Computers,2017,40(6):1229-1251. doi: 10.11897/SP.J.1016.2017.01229 [13] 陈鹏,赵小强. 一种改进1DCNN的滚动轴承变工况故障诊断方法[J]. 轴承,2022(5):65-69. doi: 10.19533/j.issn1000-3762.2022.05.014CHEN Peng,ZHAO Xiaoqiang. An improved 1DCNN fault diagnosis method for rolling bearings under variable conditions[J]. Bearing,2022(5):65-69. doi: 10.19533/j.issn1000-3762.2022.05.014 [14] 刘云飞, 张俊然. 深度神经网络学习率策略研究进展[J/OL]. 控制与决策: 1-15[2022-08-18]. DOI: 10.13195/j.kzyjc.2022.0147.LIU Yunfei, ZHANG Junran. Research advances in deep neural networks learning rate strategies[J/OL]. Control and Decision: 1-15[2022-08-18]. DOI: 10.13195/j.kzyjc.2022.0147. [15] 陈琼,谢家亮. 基于自适应采样的不平衡分类方法[J]. 华南理工大学学报(自然科学版),2022,50(4):26-34,45.CHEN Qiong,XIE Jialiang. An imbalanced classification method based on adaptive sampling[J]. Journal of South China University of Technology (Natural Science Edition),2022,50(4):26-34,45. [16] 徐卫鹏, 徐冰. 基于卷积神经网络的轴承故障诊断研究[J]. 山东科技大学学报（自然科学版）, 2021, 40（6）: 121-128.XU Weipeng, XU Bing. Study on bearing fault diagnosis based on convolution neural network[J]. Journal of Shandong University of Science and Technology（Natural Science）, 2021, 40（6）: 121-128. [17] 陈剑,程明. 基于tSNE−ASC特征选择和DSmT融合决策的滚动轴承声振信号故障诊断[J]. 电子测量与仪器学报,2022,36(4):195-204. doi: 10.13382/j.jemi.B2104696CHEN Jian,CHENG Ming. Fault diagnosis of rolling bearing acoustic vibration signal based on tSNE-ASC feature selection and DSmT evidence fusion[J]. Journal of Electronic Measurement and Instrumentation,2022,36(4):195-204. doi: 10.13382/j.jemi.B2104696 -
点击查看大图
图(6) / 表(3)
计量
- 文章访问数: 876
- HTML全文浏览量: 69
- PDF下载量: 47
- 被引次数: 0
