煤层中异常体的存在对煤炭资源的安全开采具有重要影响，开展煤层异常体的探测 识别对于煤层异常体的处置至关重要。如何快速的定位煤层异常体，获取煤层异常体的 种类、形状以及位置信息，是煤层异常体探测的重点研究内容。探地雷达（Ground Penetrating Radar，GPR）以应用灵活方便、高分辨率、高效性、以及高抗干扰性和无损 检测等优势，逐渐成为煤矿领域的研究热点。但多数研究集中在煤层厚度的探测以及煤 矿上保护层的厚度探测等工作，在探测煤层异常体的应用还未全面普及，钻探打孔仍是 主要的煤层超前探测手段；针对上述问题，本文通过煤层异常体的数值正演模拟结果图 像，以研究煤层异常体的雷达反射波图像信号特征，分析电磁波在不同煤层异常体中的 传播规律；结合实际探测煤层工程的雷达剖面图，分析它们结果之间的异同，并研究产 生异同的原因，通过数值正演模拟为探地雷达在煤层异常体的实际工程探测中提供更好 的应用；最后，利用基于深度学习的三种目标检测算法对煤层异常体数据集进行训练和 测试，实现对煤层异常体的智能检测，为实际的煤层探测数据解译提供了一种较为高效 准确的方法。 本文主要研究结果如下： （1）当煤层无异常体时，正常煤层的数值正演模拟结果图只显示一条能量较强的直 达波，电磁波在单一介质中传播时只会缓慢衰减，不会产生反射波；实际的正常煤层由 于受掘进工作的影响，煤层表面出现破碎现象，所以实际正常煤层的雷达剖面反射波表 现为连续、信号强度不均匀的缓慢衰减。由于空气的相对介电常数小于煤层的相对介电 常数，电磁波从空气层进入煤层时，分界面的反射振幅为负值。 （2）建立煤层异常体探测模型，研究煤层空洞、煤层夹石、煤层富水区的不同形状、 不同规模、不同埋深以及复杂煤层异常体的探地雷达反射回波信号剖面图特征，建立对 应的病害样本，提高对煤层异常体解译的效率和精度。 （3）结合实际探测工程中的雷达剖面图，验证了煤层异常体数值正演模拟结果的有 效性，研究表明煤层异常体的数值正演模拟对于实际工程探测图像解译有很好的指导意 义。 （4）通过基于深度学习的三种目标检测算法对煤层异常体数据集进行训练和测试，实现了对煤层异常体的智能检测识别，为实际的煤层探测数据解译提供了一种较为高效 准确的方法；检测结果表明，YOLO-v5 相比于 Faster R-CNN 和 YOLO-v3 作为煤层异常 体的目标检测算法更具有优势，能够更加准确的识别目标，煤层异常体的识别结果与实 际情况符合度高，整体煤层异常体识别率为 90%，能够满足实际的工程应用。

The existence of abnormal bodies in coal seams has a significant impact on the safe mining of coal resources. The detection and identification of abnormal bodies in coal seams are very important for the treatment of abnormal bodies in coal seams. How to quickly locate coal seam abnormal bodies and obtain information on their types, shapes, and positions is the key research content of coal seam abnormal body detection. Ground penetrating radar has gradually became a research hotspot in the field of coal mines due to its advantages of flexible, convenient application, high resolution, high efficiency, as well as high anti-interference and nondestructive testing. However, most research has focused on detecting the thickness of coal seams and the thickness of protective layers in coal mines. The application of detecting abnormal bodies in coal seams has not yet been fully popularized, and drilling are still the main advanced detection methods for coal seams. In response to the above issues, this article uses numerical forward simulation results of coal seam abnormal bodies to study the radar reflection wave signal characteristics of coal seam abnormal bodies and analyze the propagation law of electromagnetic waves in different coal seam abnormal bodies. Based on the radar profiles of actual coal seam exploration projects, analyze the similarities and differences between their results, and study the reasons for the differences. Through numerical forward simulation, providing better applications for ground penetrating radar in the actual engineering detection of coal seam anomalies. Finally, three object detection algorithms based on deep learning were used to train and test coal seam abnormal bodies, achieving intelligent detection of coal seam abnormal bodies, providing a more efficient and accurate method for interpreting actual coal seam detection data.

The main research results of this article are as follows:

(1)When there is no abnormal body in the coal seam, the numerical forward simulation results of the normal coal seam only show a direct wave with strong energy. When electromagnetic waves propagate in a single medium, they only slowly decay and do not produce reflected waves; Due to the influence of excavation work, the surface of the actual normal coal seam experiences fragmentation, so the radar profile reflection waves of the actual normal coal seam exhibit continuous and uneven signal strength with slow attenuation. Due to the fact that the relative dielectric constant of air is smaller than that of coal seam, the reflection amplitude of electromagnetic waves entering the coal seams from the air layer is negative. (2) Establish a detection model for coal seam abnormal bodies, study the characteristics of ground penetrating radar reflection echo signal profiles of different shapes, scales, burial depths, and complex coal seam abnormal bodies in coal seam cavities, coal seam rocks, and water rich areas, establish corresponding disease samples, and improve the efficiency and accuracy of interpreting coal seam abnormal bodies. (3) The effectiveness of the numerical forward simulation results of coal seam abnormal bodies was verified by combining radar profiles in actual detection engineering. Research has shown that the numerical forward simulation of coal seam abnormal bodies has good guiding significance for the interpretation of actual engineering detection images. (4) Through the training and testing of three object detection algorithms based on deep learning, intelligent detection and recognition of coal seam anomalies have been achieved, providing a more efficient and accurate method for the interpretation of actual coal seam detection data. The detection results show that YOLO-v5 has more advantages compared to Faster R-CNN and YOLO-v3 as target detection algorithms for coal seam abnormal bodies, and can more accurately identify targets. The recognition results of coal seam abnormal bodies are highly consistent with the actual situation, and the overall recognition rate of coal seam abnormal bodies is 90%, which can apply to practical engineering applications.

[1]解海军.煤矿积水采空区瞬变电磁法探测技术研究[D].北京:中国地质大学(北京),2009.

[2]罗彩祯.空间步进探地雷达煤层异常体电磁特性正演研究[D].西安:西安科技大学,2021.

[3]王莹,王勇强,邱德生,等.探地雷达在煤巷超前探测中的应用[J].中国科技信息,2011,No.437(24):79+81.

[4]杨明.煤矿高效掘进技术质量研究[J].中国石油和化工标准与质量,2019,39(15):231-232.

[5]许忠.煤矿巷道快速掘进及高效支护技术标准的探讨[J].中国石油和化工标准与质量,2019,39(15):9-10.

[6]吴群英,胡雄武,王宏科.陕北矿区地下水资源地面瞬变电磁法探查实践[J].煤炭科学技术,2022,50(05):208-215.

[7]张淼淼.矿井直流电法超前探测三维数值与物理模拟研究[D].北京:煤炭科学研究总院,2022.

[8]仇念广.瞬变电磁与地震反射波技术在巷道掘进中的联合应用[J].中州煤炭,2015,No.239(11):106-110.

[9]王树奇,宁箭,王振.基于FDTD的煤层工作面上保护层检测正演分析[J].采矿技术,2020,20(01):145-149.

[10]曾昭发.探地雷达方法原理及应用[M].科学出版社,2006.

[11]Cook J C. Radar exploration through rock in advance of mining[J]. Society Mining Engineers, 1973, 254: 140-146.

[12]丁春雨,封剑青,郑磊,等.雷达探测技术在探月中的应用[J].天文研究与技术,2015,12(02):228-242.

[13]罗回国,张清利.不同中心频率天线探地雷达在古城墙病害体探测中的应用研究[J].岩土工程技术,2018,32(05):226-230+265.

[14]张爱江,吕祥锋,周宏源,等.地质雷达快速检测城市道路路基病害应用研究[J].公路,2017,62(12):270-274.

[15]谢勇勇,廖红建,昝月稳.探地雷达检测铁路路基病害的二维正演模拟[J].浙江大学学报(工学版),2010,44(10):1907-1911.

[16]程琦,张世文,罗明,等.基于探地雷达粉煤灰充填复垦土壤含水率反演[J].地球物理学进展,2021,36(05):2159-2167.

[17]朱兆荣,赵守全,秦欣,等.探地雷达在铁路隧道衬砌质量无损检测中的应用研究[J].工程地球物理学报,2021,18(05):703-708.

[18]朱兆荣,赵守全,秦欣,等.探地雷达隧道衬砌空洞探测效果模型和现场实体试验研究[J].岩土工程学报,2022,44(S1):132-137.

[19]刘宗辉,刘毛毛,周东,等.基于探地雷达属性分析的典型岩溶不良地质识别方法[J].岩土力学,2019,40(08):3282-3290.

[20]杨道学,赵奎,曾鹏,等.基于露天开采地下盲空区探地雷达正演模拟及应用[J].岩石力学与工程学报,2019,38(11):2288-2298.

[21]刘帅,赵文生,高思伟.超宽带探地雷达煤层厚度探测试验研究[J].煤炭科学技术,2019,47(08):207-212.

[22]张璐云,崔喜红,全振先,等.野外自然条件下探地雷达识别植物根系的有效性研究[J].地球物理学进展,2021,36(06):2764-2774.

[23]Lai J L, Xu Y, Bugiolacchi R, et al. First look by the Yutu-2 rover at the deep subsurface structure at the lunar farside[J]. Nature Communications, 2020, 11(1): 3426.

[24]Sengani F. The use of ground Penetrating Radar to distinguish between seismic and non-seismic hazards in hard rock mining[J]. Tunnelling and Underground Space Technology, 2020, 103: 103470.

[25]Chufo, R L, Johnson W J. A radar coal thickness sensor[J]. IEEE Transactions on Industry Applications, 1993, 29(5): 834-840.

[26]Peng S, Yang F. Fine geological radar processing and interpretation[J]. Applied geophysics, 2004, 1(2): 89-94.

[27]王连成,高克德,朱敏川,等.用地质雷达控探测采区煤厚[J].煤炭工程师,1997,24(05):46-47.

[28]王太元,董远浪,屈旭辉.地质雷达技术在阳泉矿区井下超前探测中的应用[J].中州煤炭,2015,No.230(02):99-101+105.

[29]王振.探地雷达在采煤工作面上保护层检测中的应用[D].西安:西安科技大学,2016.

[30]李亮,王昕,胡克想,等.探地雷达探测煤岩界面的方法与试验[J].工矿自动化,2015,41(09):8-11.

[31]程久龙,潘冬明,李伟,等.强电磁干扰区灾害性采空区探地雷达精细探测研究[J].煤炭学报,2010,35(02):227-231.

[32]文虎,张铎,郑学召,等.基于FDTD的电磁波在煤中传播特性[J].煤炭学报,2017,42(11):2959-2967.

[33]齐承霞.煤层超前探测中的探地雷达信号处理[D].西安:西安科技大学,2014.

[34]习建军,曾昭发,黄玲,等.阵列式探地雷达信号极化场特征[J].吉林大学学报(地球科学版),2017,47(02):633-644.

[35]伊世学.放顶煤开采的顶煤厚度雷达探测技术研究[D].徐州:中国矿业大学,2021.

[36]刘炳岐.复杂电磁问题的高效数值模拟技术研究[D].成都:电子科技大学,2022.

[37]Warren, Craig, Giannopoulos, et al. gprMax: Open source software to simulate electromagnetic wave propagation for Ground Penetrating Radar[J]. Computer Physics Communications, 2016, 209: 163-170.

[38]Teixeira, F.L, Chew, et al. General closed-form PML constitutive tensors to match arbitrary bianisotropic and dispersive linear media[J]. IEEE Microwave & Guided Wave Letters, 1998, 8(6): 223-225.

[39]Loewer M, Igel J. FDTD simulation of GPR with a realistic multi-pole Debye description of lossy and dispersive media[C]. //2016 16th International Conference on Ground Penetrating Radar (GPR). IEEE, 2016: 1-5.

[40]Luo T X H, Lai W W L, Giannopoulos A. Forward modelling on GPR responses of subsurface air voids[J]. Tunnelling and Underground Space Technology, 2020, 103: 103521.

[41]Namiki T. 3-D ADI-FDTD method-unconditionally stable time-domain algorithm for solving full vector Maxwell's equations[J]. IEEE Transactions on Microwave Theory and Techniques, 2000, 48(10): 1743-1748.

[42]Heh D Y, Tan E L. Unconditionally stable multiple one-dimensional ADI-FDTD method for coupled transmission lines[J]. IEEE Transactions on Antennas and Propagation, 2018, 66(12): 7488-7492.

[43]Gaffar M, Jiao D. An explicit and unconditionally stable FDTD method for electromagnetic analysis[J]. IEEE Transactions on Microwave Theory and Techniques, 2014, 62(11): 2538-2550.

[44]Gaffar M, Jiao D. Alternative method for making explicit FDTD unconditionally stable[J]. IEEE Transactions on Microwave Theory and Techniques, 2015, 63(12): 4215-4224.

[45]Gaffar M, Jiao D. An explicit and unconditionally stable FDTD method for the analysis of general 3-D lossy problems[J]. IEEE Transactions on Antennas and Propagation, 2015, 63(9): 4003-4015.

[46]Cooke S J, Botton M, Antonsen T M, et al. A leapfrog formulation of the 3‐D ADI‐FDTD algorithm[J]. International Journal of Numerical Modelling: Electronic Networks, Devices and Fields, 2009, 22(2): 187-200.

[47]Wang X H, Yin W Y, Chen Z Z D. One-step leapfrog ADI-FDTD method for simulating electromagnetic wave propagation in general dispersive media[J]. Optics express, 2013, 21(18): 20565-20576.

[48]Chang C, Sarris C D. A spatially filtered finite-difference time-domain scheme with controllable stability beyond the CFL limit: Theory and applications[J]. IEEE Transactions on Microwave Theory and Techniques, 2012, 61(1): 351-359.

[49]Sarris C D. Extending the stability limit of the FDTD method with spatial filtering[J]. IEEE microwave and wireless components letters, 2011, 21(4): 176-178.

[50]冯德山,陈佳维,吴奇.混合ADI-FDTD亚网格技术在探地雷达频散媒质中的高效正演[J].地球物理学报,2014,57(04):1322-1334.

[51]仇念广.煤矿掘进巷道地质雷达超前探测数值模拟及应用[J].矿业安全与环保,2019,46(05):42-46.

[52]崔凡,耿晓航,俞慧婷,等.基于探地雷达的煤层层状超前探测[J].煤矿安全,2019,50(05):153-157.

[53]张铎.超宽带雷达波在煤中传播规律与定位基础研究[D].西安:西安科技大学,2018.

[54]Tong Z, Gao J, Zhang H. Innovative method for recognizing subgrade defects based on a convolutional neural network[J]. Construction and Building Materials, 2018, 169: 69-82.

[55]滕坤阳.基于探地雷达的地下管线病害正演模拟及回波解译研究[D].西安:西安科技大学,2021.

[56]Lei W, Man M, Shi R, et al. Target detection based on automatic threshold edge detection and template matching algorithm in GPR[C]. //2018 IEEE 3rd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC). IEEE, 2018: 1406-1410.

[57]Zhou H, Feng X, Zhang Y, et al. Combination of support vector machine and H-alpha decomposition for subsurface target classification of GPR[C]. //2018 17th International Conference on Ground Penetrating Radar (GPR). IEEE, 2018: 1-4.

[58]Zhang B, Liu J, Zhong Y, et al. A BP neural network method for grade classification of loose damage in semirigid pavement bases[J]. Advances in Civil Engineering, 2021, 2021: 1-11.

[59]Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90.

[60]Lei W, Luo J, Hou F, et al. Underground cylindrical objects detection and diameter identification in GPR B-scans via the CNN-LSTM framework[J]. Electronics, 2020, 9(11): 1804.

[61]Tong Z, Gao J, Han Z, et al. Recognition of asphalt pavement crack length using deep convolutional neural networks[J]. Road Materials and Pavement Design, 2018, 19(6): 1334-1349.

[62]赵迪,叶盛波,周斌.基于Grad-CAM的探地雷达公路地下目标检测算法[J].电子测量技术,2020,43(10):113-118.

[63]Zhang J, Yang X, Li W, et al. Automatic detection of moisture damages in asphalt pavements from GPR data with deep CNN and IRS method[J]. Automation in Construction, 2020, 113: 103119.

[64]游日,董茂干.探地雷达的发展与应用[J].筑路机械与施工机械化,2010,27(05):20-23+37.

[65]张劲松,丛鑫,杨伯钢,等.地下管线探测雷达图特征分析[J].地球物理学进展,2019,34(03):1244-1248.

[66]王中林,邵佳佳.面向工程电磁学的动生麦克斯韦方程组及其求解方法[J].中国科学:技术科学,2022,52(09):1416-1433.

[67]索明康.探地雷达信号处理及反演方法研究[D].哈尔滨:哈尔滨工业大学,2021.

[68]柴正昆.地质雷达技术在兰渝铁路隧道施工中的应用[D].成都:西南交通大学,2015.

[69]胡群芳,郑泽昊,刘海,等.三维探地雷达在城市市政管线渗漏探测中的应用[J].同济大学学报(自然科学版),2020,48(07):972-981.

[70]Yee K S.Numerical solution of initial boundary value problems involving Maxwell's equations in isotropic media[J]. IEEE Trans Antennas Propag, 1966, 14: 49-45.

[71]卞邶.基于gprMax的探地雷达在公路路基检测中的正演模拟[D].荆州：长江大学,2018.

[72]李世念,王秀荣,林恬,等.基于GprMax的道路空洞三维探地雷达正演数值模拟[J].中国地质灾害与防治学报,2020,31(03):132-138.

[73]戴前伟,冯德山,王启龙,等.时域有限差分法在地质雷达二维正演模拟中的应用[J].地球物理学进展,2004,19(04):898-902.

[74]杨光照.基于探地雷达的煤岩界面识别技术研究[D].徐州:中国矿业大学,2019.

[75]蒋昊林.时域有限差分法中高效算法的实现[D].南京:东南大学,2020.

[76]李少杰.基于探地雷达系统地下管线探测及正演模拟计算[D].西安:西安科技大学,2020.

[77]吕恒.基于探地雷达地面波技术探测土壤含水率及有效深度研究[D].贵阳:贵州大学,2021.

[78]徐世国.断层对煤矿生产的影响与处理探讨[J].山西焦煤科技,2011,35(01):22-23+26.

[79]李智,郑士田,石志远,等.极软中厚煤层越界开采区老空水害防治技术[J].煤田地质与勘探,2021,49(06):167-174.