人工监督神经网络在煤层裂缝发育区识别中的应用  

Application of artificial supervised neural network in identification of fracture development area in coal seam

作　　者：仇念广 Qiu Nianguang(China Coal Technology & Engineering Group Chongqing Research Institute, Jiulongpo, Chongqing 400039, China)

机构地区：[1]中煤科工集团重庆研究院有限公司,重庆市九龙坡区400039

出　　处：《中国煤炭》2020年第4期32-36,共5页China Coal

基　　金：“十三五”国家科技重大专项(2016ZX05067004-006)。

摘　　要：针对在地震属性应用中难以精细识别煤层裂缝发育区边界的难题,开展了基于人工监督神经网络技术的煤层裂缝发育区应用研究。从地震资料中获取倾角导向体以提取高质量的地震属性,以多属性为指导进行人工拾取样点,并基于多层感知器进行神经网络机器训练学习,建立裂缝的最优属性集,拓展整体数据后获得裂缝概率体,从而识别划分出煤层裂缝发育区。该技术在山西阳泉新元矿区进行了应用,煤层裂缝发育区的识别结果明显优于属性直接识别,但勘探区内暂无钻井资料,预测效果还有待进一步验证。Aiming at the puzzle that it was difficult to identify the boundary of seam fracture development area accurately by seismic attributes,the applied research of artificial supervision neural network technology was carried out.The dip steering from seismic data was help for the high-quality seismic attributes,and the neural network machine training and learning based on multi-layer perceptron was conducted after picking the sample points manually guided by the multiple attributes,and the optimal set of seismic attributes was established,and then the coal fracture development area could be recognized and divided up through developing the overall seismic data and achieving the fracture probabilistic volume.The application in Shanxi Yangquan Xinyuan coal mining area showed that the recoginition result of coal seam fracture development area was obviously better than that of attribute direct recoginition.However,there was no drilling data in the exploration area,so the prediction effect needed to be further verified.

关 键 词：人工监督神经网络 机器训练学习 地震属性 裂缝发育区 

分 类 号：TD163[矿业工程—矿山地质测量] P631.4[天文地球—地质矿产勘探]