基于模糊C均值聚类的空—地—井垂直重力梯度数据反演方法  

Inversion of airborne⁃ground⁃borehole vertical gravity gradient data based on fuzzy C⁃means clustering

作　　者：张显 侯振隆 赵福权 秦朋波 赵信阳 王家辉 ZHANG Xian;HOU Zhenlong;ZHAO Fuquan;QIN Pengbo;ZHAO Xinyang;WANG Jiahui(School of Resources and Civil Engineering,Northeastern University,Shenyang,Liaoning 110819,China;Qingdao Geological Exploration Institute of Metallurgical Geology of China,Qingdao,Shandong 266109,China;Guangzhou Marine Geological Survey,Guangzhou,Guangdong 511458,China)

机构地区：[1]东北大学资源与土木工程学院,辽宁沈阳110819 [2]中国冶金地质总局青岛地质勘查院,山东青岛266109 [3]广州海洋地质调查局,广东广州511458

出　　处：《石油地球物理勘探》2024年第3期629-639,共11页Oil Geophysical Prospecting

基　　金：国家自然科学基金项目“大规模全张量重力梯度数据源体生长快速三维反演方法研究”(42204140)和“重力和重力梯度联合精细化三维密度模型构建方法研究”(42104136);广州市基础与应用基础项目“基于重力和重力梯度数据的先验信息的提取及其在正则化反演中的应用”(202201011216);中国地质调查项目“南海中北部重点海域精细测量”(DD20230405)和“海洋地质精细测量”(DD20230070)联合资助。

摘　　要：通过重力梯度数据三维反演能够获得地下密度结构模型,用于地质资源勘探等领域。航空、地面和井中观测的重力梯度数据含有不同频率的信息,通过数据联合可以降低反演多解性,提高成像分辨率。对于具有复杂形态的地下异常体,目前这种多尺度数据联合反演的纵向空间分辨率,尤其是异常体底部的成像分辨率有待提升。针对该问题,开展了航空—地面—井中垂直重力梯度数据的联合反演方法研究。首先,在正则化反演中引入模糊C均值聚类算法,通过在迭代过程中加入聚类约束降低多解性;其次,联合航空、地面和井中垂直重力梯度数据,提出一种联合反演方法,并使用GPU加速计算;然后,将反演应用于理论模型数据与美国文顿盐丘地区重力梯度数据,验证方法的效果,并讨论了井位置对结果的影响;最后,对基于GPU加速的并行反演方法进行性能分析。数据试验证明了模糊C均值聚类算法能够降低反演的多解性,通过联合反演能够获得准确的密度分布,该方法具有一定的抗噪能力;使用异常旁井和穿异常井数据的成像分辨率更高。计算的文顿盐丘地区密度分布与其他学者的结论相近,证明了方法是有效且可行的。试验还表明,GPU并行方法具有较高的加速比,提出的方法能够为地质找矿等研究提供技术支撑。Three-dimensional inversion of gravity gradient data can obtain a subsurface density structure model,which can be used for geological resource exploration and other fields.Gravity gradient data from airborne,ground,and borehole observations contain information at different frequencies.Joint inversion of these data can reduce the non-uniqueness of inversion and improve imaging resolution.However,for underground anomalous bodies with complex morphology,the vertical resolution of current joint inversion of multi-scale data,especially the imaging resolution of the anomalous body bottom still needs improvement.To address this issue,the joint in-version of airborne-ground-borehole vertical gravity gradient data was studied in this paper.Firstly,a fuzzy C-means clustering algorithm was introduced into the regularization inversion,and the clustering constraint was added during the iteration process to reduce non-uniqueness.Secondly,a joint inversion method was proposed by combining airborne,ground,and borehole vertical gravity gradient data,and GPU acceleration was utilized for computation.Thirdly,the inversion was applied to theoretical model data and gravity gradient data from the Vinton Dome,USA,so as to verify the effectiveness of the method and discuss the influence of borehole loca-tions on the results.Finally,the performance of GPU acceleration-based parallel inversion was analyzed.The data test demonstrates that the fuzzy C-means clustering can reduce the non-uniqueness of inversion,and the joint inversion can obtain accurate density distributions with certain anti-noise ability.Using the data from anomaly-side and anomaly-through boreholes can improve imaging resolution.The computed density distribu-tion of the Vinton Dome area is close to the conclusions of other researchers,demonstrating the effective-ness and feasibility of the proposed method.The test results also show that the GPU parallel method has a high acceleration ratio,and the proposed method can provide technical support for geological explorations

关 键 词：空—地—井垂直重力梯度 密度反演 模糊C均值聚类 文顿盐丘 GPU加速 

分 类 号：P631[天文地球—地质矿产勘探]