Field analysis of metallogenic information and its application

Focusing on the issue to deal with inadequate extraction of metallogenic information especially geological information, a new method of extracting metallogenic information based on field model, i.e. the field analysis method of metallogenic information, was proposed. In addition, a case study by using the method of the extraction of metallogenic information from the west Guangxi and southeast Yunnan district as an example was performed. The representation method for the field models of metallogenic information, including the metallogenic influence field model and the metallogenic distance field model, was discussed by introducing the concept of the field theory, based on the characteristic analysis of the distance gradualness and the influence superposition of metallogenic information. According to the field theory superposition principle and the spatial distance analysis method, the mathematical models for the metallogenic influence field and the metallogenic distance field of point, line and area geological bodies were derived out by using parameter equation and calculus. Based on the metallogenic background analysis, the metallogenic information field models of synsedimentary faults and manganese sedimentary basins were built. The relationship between the metallogenic information fields and the manganese mineralization distribution was also investigated by using the method of metallogenic information field analysis. The instance study indicates that the proposed method of metallogenic information field analysis is valid and useful for extracting the ore-controlling information of synsedimentary faults and manganese sedimentary basins in the study area, with which the extraction results are significant both statistically and geologically.     

基于机器学习的地球化学采样下伏基岩类型判别—以青海省察汗乌苏河地区为例

基岩类型的判别是地质调查中十分重要的内容,也是开展油气勘探和矿产勘探的重要基础工作.本文采用决策树、随机森林、XGBoost、LightGBM四种机器学习的方法,实现了基于地球化学采样数据的基岩类型判别.以15种地球化学元素含量及其局部空间自相关莫兰指数和地形因子为特征,训练了不同的分类模型,通过10折交叉验证对模型做...     

基于水系沉积物地球化学采样的地形加权图卷积网络基岩填图方法（英文）

为了探索高效的第四系覆盖及露头较少区域的基岩智能填图方法，应用图卷积网络(GCN)对青海省察汗乌苏河地区水系沉积物地球化学采样的下伏基岩进行分类。基于Delaunay三角化采样点被组织为一个地形加权的有向图来表达水系沉积物地球化学采样点之间的河流上下游关系。实验结果表明：半监督的GCN模型仅使用了20%的采样点标签，分类精度达到68.20%(10类基岩)和78.31%(5类基岩)。该方法能有效利用水系沉积物地球化学采样中的元素含量进行基岩填图，且能提高基岩填图的效率并能进行大面积应用。