基于重磁三维建模与机器学习的成矿预测方法——以朱溪外围为例

三维地质建模是实现深部矿产勘查突破的重要途径,通过对控矿地质体建模,可以直观刻画控矿要素之间关系,实现矿集区“透明化”,进行深部找矿及预测.本文利用赣东北矿集区及外围1：5万重力数据和1：5万航磁数据,开展位场分离获取用于重磁反演的异常数据,再分析物性和岩性之间的关系,完成人机交互重磁反演,最后通过剖面建立朱溪矿区及邻区的三维地质模型.在三维模型建立的基础上,利用BP神经网络,选用随机单元作为样本训练,对朱溪矿区进行成矿预测并获得以下主要认识：(1)获得了赣东北矿集区朱溪外围的三维地质-地球物理模型,获取了地下5 km深度的岩体、地层的三维空间形态;(2)利用BP神经网络对矿集区进行成矿预测,得到了矿集区不同位置的成矿有利度分布图,成矿有利单元与前人划分的成矿远景区大致相符,为赣东北矿区继续找矿提供了依据,也为类似矿集区寻找深部金属资源提供了思路和技术．     

基于剩余异常相关成像的重磁物性反演方法

将场源区剖分成长方体单元,通过采集的重磁数据反演出这些单元的密度或者磁化率变化,勾画出场源的分布图像,这种方式是重磁三维反演的重要方向.重磁相关成像通过计算测量的重磁异常与地下各点在测区上的重磁异常的归一化相关,显示出异常地质体的空间赋存状态和等效剩余重磁物性.该方法计算速度快,方法简单、稳定,但是反演的结果只是在-1到+1之间的等效物性,不能够直接反演剩余密度或者磁化率,并且无法引入已知的地质约束.本文通过对物性模型的正演和实测结果的残差进行相关成像,迭代更新物性模型实现对物性参数的反演过程.模型实验证明该方法相对相关成像不仅能提高分辨率,还能够得到真正的物性参数.     

基于概率成像技术的低纬度磁异常化极方法

化极转换是磁异常解释的重要基础,为了克服在地磁纬度较低的地区尤其是磁赤道附近化极不稳定的问题,出现了多种化极方法．本文基于概率成像技术提出了一种等效物性的反演方法,实现对地下等效场源的反演成像,取得了对低纬度磁异常稳定化极的效果．化极反演中逐次对剩余异常进行反演成像,实现由概率模型到物性模型的复杂映射,避免了类似反演中需要大型方程组求解等问题,并将概率模型的构制、物性参数的反演和反演评价有机地集成到一起,加速了反演成像的进程,使反演成像速度与目前概率成像的计算速度达到可相比拟的程度．对理论模型和实际资料的计算表明,该方法对低纬度磁异常化极处理是稳定有效的,而且可以较好地压制噪声干扰,能够在噪声干扰条件下进行反演化极．     

重力异常的BP神经网络三维物性反演

三维物性反演参数多,计算量巨大,传统的方法难以实现.本文使用BP神经网络实现重力三维物性反演,介绍了BP神经网络的基本原理及特性,并构造一个适用于重力位场反演的BP神经网络.并用其对模型进行反演计算,结果表明:BP网络具有较好的泛化能力和容错能力,反演速度快、准确,并且较好的反应了场源的分布情况.     

南海共轭大陆边缘构造属性的综合地球物理研究

利用南海最新的重磁资料,在岩石物性分析和全海域分带变倾角化极磁异常反演磁性基底分布的基础上,选择6条典型剖面拟合反演其密度与磁性结构,在此基础上进行了深部结构的对比分析.反演中尽可能以海底地震仪探测数据(OBS)、多道地震等结果作为约束,其中FF'剖面层速度分析是利用“南海大陆边缘动力学及油气资源潜力”973项目200...     

基于交叉梯度约束的重力、磁法和大地电磁三维联合反演

为了降低单一地球物理方法反演的多解性及受噪声的影响程度,本文围绕重力、磁法和大地电磁法开展了三维联合反演的研究.重、磁采用基于对数障碍法的正则化反演算法,大地电磁使用limited-memory BroydenFletcher-Goldfarb-Shanno(L-BFGS)反演算法,引入交叉梯度函数实现了三种物性结构的相互耦合,最终开发出一套重磁电三维联合反演算法,并实现MPI并行加速计算.通过理论模型算例验证了算法的准确性,结果表明:不论是单棱柱体模型还是组合棱柱体模型,联合反演结果相较单独反演对于异常体的空间形态刻画以及物性数值恢复具有较好的提升;单棱柱体模型算例使得异常体的物性参数(密度、磁化率和电阻率)更加接近于真实的物性参数;组合棱柱体模型的联合反演结果不仅仅消除了围岩物性参数的假异常,而且还增强了异常体边界结构的恢复程度.     

九江-瑞昌矿集区的3D结构及对区域找矿的启示

本文分析了九瑞地区重、磁场的分布特征.运用最新的方法技术对已有的重磁资料重新进行处理:采用重磁多尺度边缘检测方法,对九瑞矿集区区域重力和航磁数据进行了边缘检测,并根据检测结果重新厘定了断裂系统的展布位置.在整理、分析九瑞地区地质、地层物性资料的基础上,对实测的1∶5万重磁数据进行较细致的准三维反演.同时将重磁三维物性反演应用到岩浆岩空间结构研究中,获得了矿集区地层结构及岩浆岩三维空间形态特征.根据反演所得磁化率强弱,分析了岩体的基性程度,为寻找与火山岩、侵入岩体有关的金属矿产提供了指示信息.最终建立的模型给出了地下地层的分布特征、控矿构造的展布规律、与成矿相关岩体的三维形态以及已知矿点的空间分布特征,为在九瑞矿集区的深部寻找隐伏矿体提供了新的信息.     

对应分析约束下的重磁联合反演及其在塔里木盆地的应用

综合应用重磁对应分析和重磁联合反演的基础上提出对重磁联合反演的一种改进方法：约束联合反演，即利用重磁异常相关系数对联合反演中目标函数进行加权。实验结果表明，该方法有利于迭代反演的收敛性。把该方法应用于塔里木盆地东部库尔勒～若羌剖面，结果说明，以对应分析的结果作为先验认识，有助于合理地建立最优化反演的初始模型，能提高解释结果的质量     

基于全卷积神经网络的磁异常及磁梯度异常反演

地球物理反演问题具有病态性、不适定性,传统的线性反演方法面临着次优逼近和初始模型选择等挑战,为了提高磁场数据反演的精度,受深度学习卓越的非线性映射能力的启发,本文提出了一种基于全卷积神经网络的磁异常及磁梯度异常反演方法.文中首先提出了一种基于网格点几何格架的磁异常及磁梯度异常的空间域快速正演算法,这为本文全卷积神经网络...     

基于拟大地电磁二维反演技术的磁源瞬变电磁实测资料处理

介绍了磁源瞬变电磁法(TEM)拟大地电磁法(MT)二维反演技术,总结出一套基于拟MT二维反演技术的磁源TEM实测资料处理流程,并对其关键流程进行了较为详细的介绍.对陕西某矿区磁源TEM实测资料进行拟MT二维反演,结果表明:反演结果对异常刻画的更加细致,反映出了近地表的低阻异常,深部的低阻异常范围也更加集中,优于视电阻率拟断面图结果;根据15、17和19线TE和TM联合反演的结果,推断了两条断裂F和F_1~2,F_1~2断裂与已知地层岩性分界面处的断裂相吻合,据此划分了勘探区的地层分布;结合矿区岩矿石电性特征、成矿构造特征和化探异常范围,推断反演结果中的低阻异常可能与成矿相关,为下一步地质下钻提供了物探依据.文章利用拟MT二维反演方法实现了磁源TEM实测数据二维反演解释,并取得了较好的地质效果,得出磁源TEM拟MT二维反演技术可在实际资料处理中予以推广使用的结论.     

Using constrained inversion of gravity and magnetic field to produce a 3D litho‐prediction model

Geologically constrained inversion of gravity and magnetic field data of the Victoria property (located in Sudbury, Canada) was undertaken in order to update the present three‐dimensional geological model. The initial and reference model was constructed based on geological information from over 950 drillholes to constrain the inversion. In addition, downhole density and magnetic susceptibility measured in six holes were statistically analysed to derive lower and upper bounds on the physical properties attributed to the lithological units in the reference model. Constrained inversion of the ground gravity and the airborne magnetic data collected at the Victoria property were performed using GRAV3D and MAG3D, respectively. A neural network was trained to predict lithological units from the physical properties measured in six holes. Then, the trained network was applied on the three‐dimensional distribution of physical properties derived from the inversion models to produce a three‐dimensional litho‐prediction model. Some of the features evident in the lithological model are remnants of the constraints, where the data did not demand a significant change in the model from the initial constraining model (e.g., the thin pair of diabase dykes). However, some important changes away from the initial model are evident; for example, a larger body was predicted for quartz diorite, which may be related to the prospective offset dykes; a new zone was predicted as sulfide, which may represent potential mineralisation; and a geophysical subcategory of metabasalt was identified with high magnetic susceptibility and high density. The litho‐prediction model agrees with the geological expectation for the three‐dimensional structure at Victoria and is consistent with the geophysical data, which results in a more holistic understanding of the subsurface lithology.     

The application of rapid potential field methods for the targeting of IOCG mineralisation based on physical property data,Great Bear magmatic zone,Canada

New techniques for the integration of gravity and aeromagnetic data through rock physical property relationships are tested for the rapid location of potential iron oxide–copper–gold (IOCG) exploration targets across the Great Bear magmatic zone (GBMZ). These techniques are based on the recognition of coincident or near-offset magnetic and gravity anomalies associated with IOCG deposits, resulting from their intense and combined iron-oxide and alkali alteration. In the GBMZ, the NICO (Au–Co–Bi–Cu) deposit is an atypical magnetite-group IOCG hosted within an intensely hydrothermally-altered metasedimentary sequence. Prospectivity maps for the NICO area are derived from the integration of high-resolution aeromagnetic and surface gravity data with physical property measurements of magnetic susceptibility and density. Method 1 combines the pseudo-gravity (derived from the aeromagnetic data) and Bouguer gravity data to determine the locations of superimposed high anomalies. Method 2 estimates the apparent susceptibility and density subsurface distribution based on vertical prism models of the magnetic and gravity data. The apparent susceptibility data are transformed to apparent density using physical property relationships. The two apparent density datasets are then combined to determine the location of coincident magnetic and gravity derived anomalies. The results of these methods are supported by, and compared with, similar prospectivity maps generated through weights of evidence techniques. At NICO, the prospectivity maps accurately locate the deposit and nearby occurrences. Despite the lower resolution of the regional gravity data, application of the techniques across the GBMZ is successful at locating mineralised IOCG systems including the magnetite-group systems at Grouard, Fab, DeVries and Cole Lake, the magnetite to hematite-group systems of the Port Radium-Echo Bay district, and has highlighted possible targets for future exploration.     

Structure and State of Stress of the Chilean Subduction Zone from Terrestrial and Satellite-Derived Gravity and Gravity Gradient Data

It is well known that the quality of gravity modelling of the Earth’s lithosphere is heavily dependent on the limited number of available terrestrial gravity data. More recently, however, interest has grown within the geoscientific community to utilise the homogeneously measured satellite gravity and gravity gradient data for lithospheric scale modelling. Here, we present an interdisciplinary approach to determine the state of stress and rate of deformation in the Central Andean subduction system. We employed gravity data from terrestrial, satellite-based and combined sources using multiple methods to constrain stress, strain and gravitational potential energy (GPE). Well-constrained 3D density models, which were partly optimised using the combined regional gravity model IMOSAGA01C (Hosse et al. in Surv Geophys, 2014, this issue), were used as bases for the computation of stress anomalies on the top of the subducting oceanic Nazca plate and GPE relative to the base of the lithosphere. The geometries and physical parameters of the 3D density models were used for the computation of stresses and uplift rates in the dynamic modelling. The stress distributions, as derived from the static and dynamic modelling, reveal distinct positive anomalies of up to 80 MPa along the coastal Jurassic batholith belt. The anomalies correlate well with major seismicity in the shallow parts of the subduction system. Moreover, the pattern of stress distributions in the Andean convergent zone varies both along the north–south and west–east directions, suggesting that the continental fore-arc is highly segmented. Estimates of GPE show that the high Central Andes might be in a state of horizontal deviatoric tension. Models of gravity gradients from the Gravity field and steady-state Ocean Circulation Explorer (GOCE) satellite mission were used to compute Bouguer-like gradient anomalies at 8 km above sea level. The analysis suggests that data from GOCE add significant value to the interpretation of lithospheric structures, given that the appropriate topographic correction is applied.     

基于柯西分布约束和快速近端目标函数优化的三维重力反演方法

优化算法的选取在很大程度上影响着三维重力反演的计算效率,从而制约着三维重力反演的实用性.在复杂地质构造背景下,不同岩性单元之间可能会发生物性突变,产生尖锐边界.为此,本文提出了一种新的基于柯西分布约束和快速近端目标函数（Fast Proximal Objective Function,FPOF）优化的三维重力反演方法.FPOF优化方法的一个突出特点是在每一步迭代过程中逐一计算剖分网格内的未知密度参数,因此,有较低的计算复杂度和较高的计算效率.此外,目标函数中柯西范数（Cauchy norm）的引入会对反演结果施加稀疏性,有助于产生块状效果.理论模型测试表明,本文方法不仅能产生更加聚焦的反演效果,而且反演所需的时间也比传统的共轭梯度优化方法少.最后将本文方法应用于我国西部某地区实际重力数据,反演结果与已知的地质信息有较好的一致性.     

3D AND 2½ D MODELLING OF GRAVITY ANOMALIES WITH VARIABLE DENSITY CONTRAST1

The decrease of density contrast in sedimentary basins may be approximated by a quadratic function. A sedimentary basin may be viewed as a number of prisms placed in juxtaposition. Equations in closed form for the gravity anomalies of 3D and 2½ D prismatic models are derived. Approximate equations for these models are also derived for rapid calculations. Efficient methods are developed for anomaly calculation by an appropriate use of the exact and approximate equations, and hence, for 3D and 2½ D modelling. The depths to the basement are adjusted iteratively by comparing the calculated anomalies with the observed anomalies. These methods are applied for analysis of the residual anomaly map of the Los Angeles Basin, California.     

Crustal Structure of Turkey from Aeromagnetic, Gravity and Deep Seismic Reflection Data

In this paper, aeromagnetic and gravity anomalies obtained from the General Directorate of Mineral Research and Exploration were subjected to upward continuation to 3?km from the ground surface to suppress shallow effects and to expose only regional, deep sources. Then, a reduction to pole (RTP) map of aeromagnetic anomalies was produced from the 3?km upward continued data. A sinuous boundary to the south of Turkey is observed in the RTP map that may indicate the suture zone between the Anatolides and African/Arabian Plates in the closure time of the Tethys Ocean. The sinuous boundary can be correlated with the recent palaeo-tectonic maps. The southern part of the sinuous boundary is quite different and less magnetic in comparison with the northern block. In addition, maxspots maps of the aeromagnetic and gravity anomalies were produced to find out and enhance the boundaries of tectonic units. Crustal thickness, recently calculated and mapped for the western Turkey, is also extended to the whole of Turkey, and the crustal thicknesses are correlated with the previous seismological findings and deep seismic sections. The average crustal thickness calculations using the gravity data are about 28?km along the coastal regions and increase up to 42?km through the Iranian border in the east of Turkey. Density and susceptibility values used as parameters for construction of two-dimensional (2D) gravity and magnetic models were compiled in a table from different localities of Turkey. 2D models indicate that all of the anomalous masses are located in the upper crust, and this could be well correlated with the earthquakes which occurred at shallow depths.     

基于3DDelaunay剖分算法的重力建模与分析

地质体的重力建模是正确解释和应用重力资料的关键问题之一.针对非规则形状变密度的三度体,本文提出了基于3D Delaunay剖分算法的重力建模方法.采用3D Delaunay剖分算法将三维目标地质体分解为若干变密度四面体体元,推导了基于四面体体元的重力正演公式,建立了剩余密度值与重力异常值的线性方程组;以变密度的长方体和倾斜台阶组合体为例,比较分析了常规块体算法和3D Delaunay剖分算法应用于重力正演的有效性,并采用共轭梯度法加密度约束条件对非规则形状变密度的倾斜台阶组合体进行了密度反演.计算结果验证了本文方法的正确性和有效性.基于3D Delaunay剖分算法的重力建模可应用于存在褶皱、断层、裂缝等复杂地质体的重力正反演计算.     

AUTOMATIC INTERPRETATION OF GRAVITY GRADIOMETRIC DATA IN TWO DIMENSIONS: VERTICAL GRADIENT1

The magnetic and gravity field produced by a given homogeneous source are related through Poisson's equation. Starting from this consideration, it is shown that some 2D interpretation tools, widely applied in the analysis of aeromagnetic data, can also be used for the interpretation of gravity gradiometric data (vertical gradient). This paper deals specifically with the Werner deconvolution, analytic signal and Euler's equation methods. After a short outline of the mathematical development, synthesized examples have been used to discuss the efficiency and limits of these interpretation methods. These tools could be applied directly to airborne gravity gradiometric data as well as ground gravity surveys after transformation of the Bouguer anomalies into vertical gradient anomalies. An example is given of the application of the Werner deconvolution and Euler's equation methods to a microgravity survey.     

Three-dimensional gravimetric modelling to detect the deep structure of the region Vogtland/NW-Bohemia

The occurrence of swarm earthquakes in the Vogtland/NW-Bohemia area results probably from the physical interactions of fluids, the stress field and the geometry of the geological units. Therefore the present study aims at the development of a 3-D density model of the region with a vertical range of 35 km. A new Bouguer anomaly map is presented containing about 17 000 gravity data points. Prominent Bouguer anomalies are produced by the granites of Eibenstock and Karlovy Vary (low with −75 mGal), the metabasites near Mariánzké Lázně (high with 5 mGal) and the Münchberg Gneiss Massif (gravity high of Hof with 10 mGal). The geometry of the internal model structures correspond to geological units and, thus, the modelled gravity fits well the observed Bouguer anomaly. The 3-D gravimetric modellings indicate detailed geometries of the geological settings. With regard to the periodic occurrence of swarm earthquakes in the Vogtland region the existence of an upwelling mantle or a magmatic body is investigated. Precise information only can be given, if the vertical extension of the near surface bodies is known.     

Gravity and magnetic investigation on the distribution of volcanic rocks in the Qinggelidi area,north‐eastern Junggar Basin (north‐west China)

A detailed investigation on the location of magmatic intrusions in the Carboniferous strata of the Qinggelidi area, north‐eastern Junggar Basin, is presented based on the interpretation of gravity and magnetic data constrained by petrophysical data, seismics and surface geology. The wavelet multi‐resolution analysis based on the discrete wavelet transform is adopted to the regional‐residual separation of gravity and magnetic anomalies. A power spectrum analysis is applied to estimate the source depths corresponding to different scales. A comparative analysis on the characteristics of local gravity and magnetic anomalies improved our understanding of volcanic rock distribution in the Carboniferous strata. Generally speaking, in total 75 anomalies are recognized, among which 23 are inferred to be the responses of basalts, diabases and andesites with high density and strong magnetization. Twelve anomalies are assumed to be caused by andesites, rhyolites and volcanic breccias with medium‐low density and high magnetization. There are still five anomalies that are believed to be generated by volcanic tuffs with low density and weak magnetization. Lastly, four cross‐sections in 3D gravity and magnetic modelling are displayed to provide a more thorough image of volcanic rocks in our study area.