The Interpretation And Prediction Method Of Fault Structure Basedon Deep Learning And Its Application

The study of fault structure has always been the focus in the field of geology.As one of the important forms of geological process,fault plays an irreplaceable role in the process of mineralization.The study of mineral resources is inseparable from the study of fault structure.At present,with the advent of the era of big data,the introduction of it into the field of Geosciences will be conducive to the processing of complex geoscience information,thus contributing to the next step of prediction work.In the field of Geosciences,although deep learning method has been widely used,there are few researches on fault prediction.Based on the deep learning method,this paper discusses the following three aspects.1.In this paper,based on the generative adversarial networks,according to the fault data obtained from the geological maps of the West Qinling Daqiao and Zhaishang area in Gansu Province,the fault data set is obtained through processing,and then input into the model to train the fault prediction model.Then,according to the model,the prediction experiment is carried out.Four predicted fault maps are compared with the original geological images,and the fault structure can be predicted,And the prediction effect is good;On this basis,the model is used to predict the fault image extracted from the geological map in the area of Chonidun-Xixiaokouzi in Gaotai County,Gansu Province,and the fault prediction image of the area is generated.Through the remote sensing interpretation of the remote sensing image of the area,the remote sensing fault interpretation map of the area is also obtained.Comparison with each other: the strike of the fault in the predicted fault map is near EW to NWW,which is consistent with the strike of the fault in the geological map and remote sensing interpretation;The predicted fault map can produce new fault structure,and the position is almost the same as that of remote sensing fault interpretation map.The results show that the GAN model is effective.2.Based on the generative adversarial networks,the fracture interpretation is studied in the UAV aerial orthophoto of Beishan,Gansu Province.The GAN model has a good fault interpretation effect by testing the dislocated dikes and strata in the fault interpretation marks.This method has a good application effect in orthophoto fault interpretation.3.Based on the convolution neural network model,the copper ore prospecting in the area of chonidun-xixiaokouzi,Gaotai County,the western part of Longshou is predicted.The predicted area is 27.3% based on geochemical and aeromagnetic data;The predicted area based on geology,geochemistry and aeromagnetism is 12.1%;Based on the GAN model,the fault prediction map of the area is used as the geological data of the area,and the prediction area of the prediction results obtained by integrating the geochemical and aeromagnetic data is 19.7%.Compared with the prediction results of the geochemical and aeromagnetic data,the prediction area of the Quaternary and Neogene is narrowed or disappeared,and compared with the geological and aeromagnetic data,The prediction results of geochemical and aeromagnetic data are re-delineated in the key prospecting area(Pantoushan),which is more reliable.