3D Quantitative Prediction Of Mineral Resources At Depth Based On Geology And Geochemistry

The research on three-dimensional(3D)quantitative prediction methods of mineral resources at depth provides important technical support for deep mineral resources exploration.The three core scientific problems to be solved in three-dimensional quantitative mineral resources prediction are: 3D meticulous modeling of mineralization space,deep geological and geochemical information extraction and inference,and quantitative metallogenic prediction at large depth.In view of this,this paper mainly discusses from the following aspects: based on the combination of explicit and implicit modeling method,3D visualization of geology and geochemistry are carried out;based on the multifractal content volume(C-V)model,the spatial distribution of primary halo geochemical elements is studied;based on data-and knowledge-driven compositional data analysis framework,the characteristics of element associations are extracted and inferred;On the basis of machine learning and deep learning methods,quantitative mineral prediction at large depth is carried out.The details are as follows:(1)Aiming at the scientific problem of 3D meticulous modeling of metallogenic space,the 3D geological and geochemical model is constructed by combining 3D explicit and implicit modeling.In the superficial area,the explicit modeling method is used to control the modeling accuracy as much as possible.In areas with insufficient data at large depths,the spatial distribution of deep ore-controlling factors is deduced by using the implicit modeling method under geological constraints.3D geological models,including topography,ore bodies,rocks and faults are constructed.The primary geochemical data volume models based on explicit and implicit modeling are constructed.The 3D model of deep quantitative mineral prediction is also constructed,which provide visual support for deep quantitative prediction.(2)In view of the scientific problems of deep extraction and inference of favorable geological and geochemical information for deep mineralization,the following three parts are analyzed.First,with the guidance of nonlinear theory,the3 D multifractal method for geochemical anomaly extraction is discussed,and the spatial anomaly characteristics of 12 primary geochemical elements related to mineralization are extracted,which lays the elements distribution,zoning and association characteristics of metallogenic spatial elements;Second,based on the compositional data analysis approach,the element association extraction method is studied.Meanwhile,follows the framework of data-driven compositional data analysis,the geochemical element association(Sb-Hg)corresponding to ore-controlling structure are quantitatively extracted,which provide data support for inferring the deep ore-controlling structure;based on the knowledge-driven compositional data analysis framework,the element associations of front halo indicators(As-Sb-Hg),recent halo indicators(Au-Ag-Cu-Pb-Zn)and the tail halo indicators(W-Mo-Co-Bi)are quantitatively extracted,which provide quantitative indicators for deep exploration.Through the above analysis,a 3D primary halo model is constructed.Compared with the conventional profile primary halo method,it not only delineates the deep targets of mineral resources,but also provides a qualitative reference for deep quantitative mineral prediction.(3)Aiming at the scientific problem of quantitative metallogenic prediction at large depth,based on the metallogenic model of Zaozigou deposit,the deep geological and geochemical indicators are quantitatively extracted,and the deep geological and geochemical exploration model is constructed.Three quantitative mineral prediction models of machine learning and deep learning,including maximum entropy model,Gaussian mixture model and convolutional neural network model,are designed.These three models are used for deep quantitative mineral prediction.Taking the five prospecting indicators of structural buffer zone,ore-controlling structure element association Hg-Sb,ore-forming element Au,recent halo element association Au-Ag-Cu-Pb-Zn,the ratio of front halo and tail halo association(As-Sb-Hg)/(W-Mo-Bi)as input variables,quantitative,positioning and uncertainty mineral resources prediction are carried out for the occurrence area of large-depth ore bodies.The three-dimensional quantitative mineral resources prediction methods based on deep geological and geochemical information are formed in this paper,which relates to the analysis of geological metallogenic model,quantitatively extracting the geological and geochemical prospecting indicators,constructing the deep geological and geochemical prediction model at large depth.It is worth mentioning that in 2021,Zaozigou deep scientific drill has achieved remarkable results in the deep prediction.On the one hand,the deep drill has verified the reliability of the deep prediction in this paper.On the other hand,on the basis of adding the deep drilling data,the deep quantitative prediction provides a new prospecting direction.