基于Bayes判别分析模型的风化基岩富水性预测
Prediction of water enrichment of weathered bedrock based on Bayes discriminant model
-
摘要: 风化基岩含水层是陕北侏罗纪煤田煤炭开采的主要充水含水层,风化基岩含水层富水性的分区预测是矿井防治水的关键。以红柳林井田中西部为研究区,在分析影响风化基岩富水性控制因素的基础上,选取了风化基岩厚度、岩心采取率、风化程度、岩性组合、风化基岩顶面标高以及砂基比6个因素作为判别指标,以研究区内41组有效风化基岩钻孔抽水试验数据以3∶1的随机分配方式作为训练样本及验证样本,构建了富水性Bayes判别分析模型;采用该模型对红柳林井田中西部未进行过抽水试验钻孔的风化基岩富水性类别进行了预测,得到了风化基岩富水性预测图。结果表明:研究区西一盘区内整体富水性相对较强,强富水性区空间分布不均;北二盘区及南二盘区大部分区域富水性相对较弱;极弱富水性区分布于东南部。通过与实际生产中工作面涌水量及出水点位置对比,该富水性分区预测结果与实际吻合。predict the water richness category of weathered bedrock that has not been drilled for pumping test in the central and western part of Hongliulin Minefield, and the water richness prediction diagram of weathered bedrock is obtained. The results show that the overall water-rich water in the western first panel of the study area is relatively strong, and the spatial distribution of strong water-rich area is uneven. Most areas of the north second panel and the south second panel have relatively weak water richness, and the extremely weak water rich areas are distributed in the southeast. By comparing with the actual production of the working face water inflow and the location of the water outlet point, the prediction result of the water-rich zone is consistent with the actual situation.Abstract: Weathered bedrock aquifer is the main water-filled aquifer for coal mining in Jurassic coalfields in northern Shaanxi. The regional prediction of the water richness of weathered bedrock aquifer is the key to mine water control. Taking the central and western part of Hongliulin Minefield as the research area, based on the analysis of the factors affecting the water-richness of the weathered bedrock, the thickness of the weathered bedrock, the core removal rate, the degree of weathering, the combination of lithology, the top elevation of the weathered bedrock and the sand base ratio are selected to use as the discriminant indexes, 41 groups of effective weathered bedrock borehole pumping test data in the study area are used as training samples and verification samples with a 3∶1 random allocation method to construct a water-rich Bayes discriminant analysis model. This model is used to