摘要
煤层火成岩侵入给煤矿生产以及经济效益带来了极大的影响,属于亟待解决的问题.本文充分发挥测井信息的作用,基于测井数据建立正演模型,获得不同侵入模型叠加记录,并提取多种地震属性;利用灰色关联和模糊聚类方法对提取的地震属性进行分类和优化,得到与地质目标相关性较好,且相互独立的4种地震属性;利用井旁道地震记录和井信息作为BP神经网络的学习样本进行训练,在训练好的BP神经网络中输入从地震数据中提取的优化后的地震属性,预测煤层火成岩侵入区的分布情况.从实际测区的预测情况看,该方法准确性和可靠性较高,可对实际生产进行理论指导.
Coal seam igneous intrusion, an urgent issues need to be solved, results in an seriously impact on production and economic benefits of mining. In the paper, to make full using of well logging data, we modeled different intrusion forms based on it and exacted many seismic attributes from the stacked recordings after forward modeling. Four seismic attributes which show good correlation with geological target and independent of each other were selected though classifying and optimizing the exaction seismic attributes with gray correlation and fuzzy clustering methods. Then the four seismic attributes of borehole-side seismic traces, as well as the well logging information were input as the learning samples of BP neural network, the igneous intrusion zones could be division by the BP neural network with the four seismic attributes exacted from 3-D seismic recordings after training. A case from a real testing areas shows that this method can acted as the mining guidance because of high accuracy and reliahility.
出处
《地球物理学进展》
CSCD
北大核心
2015年第3期1376-1381,共6页
Progress in Geophysics
基金
江苏省自然科学基金(BK20130201)
中国博士后科学基金(2014M551703)联合资助
关键词
地震属性
灰色关联
模糊聚类
BP神经网络
测井
seismic attributes
gray correlation
fuzzy clustering
BP neural network
well logging