基于深度卷积神经网络与多源信号的煤岩识别研究  

Research on Coal Rock Identification Based on Deep Convolution Neural Network and Multi-source Signals

作　　者：李富强 LI Fuqiang(CHN Energy Wuhai Energy Co.,Ltd.,Wuhai 016000,China)

机构地区：[1]国家能源集团乌海能源有限责任公司,内蒙古乌海016000

出　　处：《煤炭技术》2025年第3期233-238,共6页Coal Technology

摘　　要：煤岩识别是采煤装备自主调高的关键技术,也是实现综采工作面智能化的重要难点,针对低照度噪音大的工作面环境下,给出了一种基于深度卷积神经网络的煤岩识别方法。搭建截割实验台,浇筑不同煤岩硬度的试件,采集截割过程中的三向振动信号、截割电机电流信号、声发射信号波形图,基于深度卷积神经网络对信号识别,分析煤岩特性,并通过现场实验进行了验证。实验结果表明:多源信号的组合作为煤岩识别的特征信号泛化性更好,验证了卷积神经网络模型具有较高的识别精度,极大地提高综采工作面的整体辨识精度,能够实现截割过程中煤岩界面的准确、快速识别,该模型的构建为实现综采工作面自动化、智能化开采提供了理论基础和技术前提。Coal rock identification is a key technology for autonomous height adjustment of coal mining equipment,and it is also an important difficulty in achieving intelligence in fully mechanized mining faces.Proposes a coal rock identification method based on deep convolutional neural networks for low illumination and high noise working face environments.Build a cutting experimental platform,pour specimens with different coal and rock hardness,collect waveforms of three-dimensional vibration signals,cutting motor current signals,and acoustic emission signals during the cutting process,recognize signals based on deep convolutional neural networks,analyze coal and rock characteristics,and verify them through on-site experiments.The experimental results show that the combination of multi-source signals has better generalization as the characteristic signals of coal and rock recognition,which verifies that the convolutional neural network model has higher recognition accuracy,greatly improves the overall recognition accuracy of the fully mechanized working face,and can achieve accurate and rapid recognition of middling coal rock interface in the cutting process.The construction of the model provides the theoretical basis and technical premise for realizing the automation and intelligent mining of the fully mechanized working face.

关 键 词：煤岩识别 多源信息融合 深度卷积神经网络 

分 类 号：TD67[矿业工程—矿山机电]