| With the increasing demand for mineral resources in China,the shallow and easy to identify deposits are difficult to meet the needs of China's development.Nowadays,the research and exploration of deep deposits have attracted more and more attention of geological exploration industry.However,the research of deep mineral resources in China is still slow,and many deposits are also in the state of secondary development.With the development of Industrial Science and technology,our country's demand for mineral resources increases.However,the resources of the deep Jinchuan deposit have not been further studied,which is a very meaningful work.The deep prospecting mainly considers the deposit with a depth of 3000 meters.Therefore,with the support of National Key R&D Program of China "geophysical response and comprehensive index system of deep mineral resources",this paper studies the deep mineral resources.The research content is mainly divided into two parts: one is to establish the geophysical response index system of deep mineral resources on the medium and small scale to evaluate the ability of geophysical methods to detect deep deposits;the other is to use the machine learning method to extract the anomaly index of geophysical data in a large scale.In the first part,taking Jinchuan copper nickel sulfide deposit as an example,this paper mainly uses geophysical methods such as gravity,magnetic method,magnetotelluric(MT)and seismic method to carry out forward calculation of Jinchuan copper nickel sulfide deposit,and analyzes the calculation results accordingly.Then the gravity,magnetic method and CSAMT data of Jinchuan deposit are inverted,and the inversion results are analyzed and evaluated in combination with the geological background of the deposit.Then,the analytic hierarchy process(AHP)is used to establish the index system of the deep mineral resources.25 kinds of typical deposits of 6 different genetic types are used for the numerical simulation of gravity,magnetism,electricity and earthquake,and the corresponding relationship between the simulated geophysical response signal and the deposit is analyzed.The characteristic signal of geophysical response is scored,and then the four aspects of gravity,magnetism,electricity and earthquake are analyzed the response indexes of four geophysical methods are analyzed and the weight coefficients are obtained.Finally,the evaluation index system of geophysical response of deep mineral resources is constructed by using the weight coefficients and scores of the four methods.In the second part,we use the method of machine learning to extract the indexes of geophysical anomalies in a large range,and use the method of GMM to test the gravity and aeromagnetic data of Lake Superior in Canada,and extract the indexes of geophysical anomalies to draw the mineral prospect map of the region.Firstly,this paper collects the geological background data,geological genesis and other information of Jinchuan copper nickel sulfide deposit,which can effectively determine the deep deposit,effectively analyze the geophysical response results according to the geological information,and ensure the accuracy of the subsequent index system.Next,the geological and geophysical modeling of Jinchuan sulfide Cu Ni deposit is carried out.In order to more accurately describe the morphological characteristics of the ore body,on the basis of rectangular grid,the adaptive Delaunay grid modeling method is added,and the quality of the grid generated by this method is analyzed.The Jinchuan sulfide Cu Ni deposit model based on adaptive Delaunay grid modeling has the characteristics of grid refinement in the boundary and complex areas,and can realize grid sparseness in areas with small physical differences,thus saving the use of computer memory.The mesh generation algorithm can control the quality of mesh and ensure the accuracy of numerical simulation.Then,based on the theory of Geology and mineralization,the data related to geology and geophysics are collected comprehensively.The space and plane geophysical model of Jinchuan sulfide Cu-Ni deposit is established.Based on the geophysical model,numerical simulation methods such as integral equation method,barycenter volume method,finite difference algorithm and finite element method are applied to the numerical calculation of geophysical method.According to the simulation results,geological background and genesis,the occurrence of the deposit and the signal response of deep deposit are discussed.According to the response information obtained by four geophysical methods,four methods are determined and the weight coefficients are given.Through the given coefficient,it is used in the index system of geophysical exploration of deep minerals.According to the geological background of Jinchuan copper nickel sulfide deposit,the typical structure and distribution structure of physical properties of Jinchuan copper nickel sulfide deposit are qualitatively analyzed.According to the geophysical response signals and inversion results,four geophysical methods of gravity,magnetism,electricity and earthquake are evaluated and analyzed,and the analysis results are taken as the important basis of the weight coefficient selected in the index system.Then,in order to construct the geophysical index system of deep mineral resources,this paper uses analytic hierarchy process to select six types of deposits with different genesis as the secondary index of deep mineral resources,and uses 25 typical deposits as the secondary index,namely the third index.The geophysical responses of 25 typical ore deposits are taken as the interpretation indexes of three-level indexes.The geophysical response signal characteristics of the deposit are analyzed and scored,and the weights of the four geophysical methods are determined at the same time.Then the feasibility of the obtained weights is verified.After the verification results are passed,the weights are added according to the weight coefficients of the four geophysical methods.Finally,the weight coefficient and the geophysical response score are calculated synthetically and the geophysical index scores of each deposit are obtained.Then the scores of 25 deposits are evaluated synthetically and the comprehensive index system scores of deep mineral are obtained.In view of these simulation and processing results,the relationship between different indexes is established by using AHP,and the geophysical index system of deep mineral resources is established at the same time.This index system not only has a evaluation system for the selected deposits in this paper,but also has a comprehensive evaluation of the geophysical response of 25 deposits,and can effectively guide the exploration of deep minerals by geophysical methods.Finally,this paper has carried on the extraction work of geophysical anomaly index,using the method of Gauss mixture model in machine learning to test the geophysical data of Lake Superior area.When the Gaussian mixture function has enough Gaussian functions,it can simulate any continuous complex probability distribution.The contour map of gravity and magnetic method can be regarded as a continuous probability density distribution function,so GMM method can be used to model it,and the maximum likelihood estimation method can be used to effectively delineate the Gaussian mixture function in the region.All geophysical data in this area that do not conform to the distribution of Gaussian mixture model are considered as abnormal areas.All the different anomaly points in the plane can be extracted to form the geophysical anomaly index plan of the area,which can also be called the mineral prospect map.According to the adjustment of the number of Gauss functions,five Gauss functions are finally determined to realize the geophysical data modeling of the Lake Superior area,and the index distribution map of the area is obtained.Through comparison,83% of the deposits are located in the range of the abnormal index.At the same time,the receiver operating characteristic curve(ROC)and the percentage of anomaly area(PAA)are used to evaluate the simulation results,which proves that this paper has a good performance Effectiveness of the work. |
PDF Full Text Request