Exploration Geochemical Data Processing And Inference Of Buried Bedrock In Jining Shallow Covered Area,Inner Mongolia

Prospecting in covered area is the frontier research direction of mineral resources exploration.The Jining city is located in the central part of Inner Mongolia,which is in the northern margin of the North China Craton.It is covered by Cenozoic sediments and basalt.In recent years,three large or giant molybdenum deposits have been discovered in the area,implying great prospecting potential.Taking Jining shallowcoverage area as a study subject,through the statistical analysis of geochemical data of stream sediments and soil,the Singularity theory and Principal component analysis(PCA)method are applied to the identification of geochemical anomalies in the covered area.The geochemical data of the study area are analyzed in detail and pattern recognition is carried out by using Local singularity analysis(LSA)method.Combined with the characteristics of typical deposits,aeromagnetic and geochemical anomaly,and based on partially exposed bedrock as training set,the underlying bedrock in the coverage area is predicted by means of convolution neural network algorithm.It is found that the geochemical element concentrations on the surface can reflect the distribution of underlying bedrock to a certain extent,but the results are not accurate due to the existence of overburden.Further analysis of the singularity of geochemical elements shows that the singularity index can effectively highlight the weak and gentle anomalies under the overburden,and effectively indicate concealed mineralization or element enrichment.Moreover,the element distribution model obtained by the singularity index is stable and will not change accompanying the thickness of the overburden.Therefore,the stable singularity index which can reflect the concealed mineralization could be used for the prediction of concealed ore bodies.Finally,through convolutional neural network processing of image spatial information such as ore-forming geological bodies,structures,aeromagnetic,geochemical singularities,etc.,a prediction map of the concealed bedrock under the shallow-coverage area of Jining is obtained.A comparison of the influence of the convolutional neural network method considering spatial information and that without considering spatial information in bedrock inference shows that spatial information should be considered when making geological bedrock prediction.Otherwise,although data with higher accuracy will be obtained,accurate geological information cannot be obtained in actual conditions.The results could provide a basis for future research to predict mineral resources.Combining deep learning with mineral forecasting,based on the original data of regional geochemical data and its local singularity index,we can utilize the feature discovery capabilities of deep learning and get more accurate results.Convolutional neural networks can better reflect the extraction and retention of spatial information when making mineral predictions.