Research On Large-scale Fault Reconstruction Method Based On Manifold Learning

Faults,diverse and complex geological phenomenon,are the migration channel of oil,gas,coal-bed methane and other hydrocarbon resources,which have an extremely critical impact on the distribution of oil and gas resources.Therefore,it plays a vital role in the exploration of oil and gas reservoirs to explore the spatial distribution and the morphological characteristics of faults.Three-dimensional fault interpretation is of great significance to the exploration of oil and gas reservoirs.At present,most fault modeling methods reconstruct fault surfaces from seismic attributes,which tend to result in broken faults and incomplete faults.In addition,as the scale of seismic data acquisition expanding with the deepening of exploration work,it is imminent to develop a fault modeling method suitable for large-scale seismic data.To address the problems mentioned above,the contributions of this thesis are as follows:1.Aiming at broken faults and incomplete faults,this thesis proposes a fault surface reconstruction algorithm based on manifold learning.Using locality preserving projections manifold learning method reduces the dimension of 3D fault scattered points while maintaining the internal local neighborhood structure of these points,which can reduce the difficulty of reconstructing 3D fault surface,and tend to yield less broken and more complete fault surfaces.This method can also deal with the situation of intersecting faults and prevent the existence of holes near the intersecting positions.2.Aiming at fault modeling for large-scale seismic data,this thesis proposes a fault surface reconstruction algorithm for large-scale seismic data based on the local geometric features of fault point clouds on the basis of the fault surface reconstruction algorithm based on manifold learning.Through data partitioning,the single memory load is reduced.The local geometric features of fault point clouds are used to merge the surface elements in adjacent seismic data blocks.In this way,the large-scale fault surface reconstruction is realized on the computing platform with limited memory resources.3.In this thesis,a fault surfaces reconstruction and visualization system is initially constructed based on a system with simple basic functions.Based on the MVC framework,the system integrates two fault surface reconstruction algorithms proposed in this thesis and related support algorithms,provides basic human-computer interaction functions,and realizes the end-to-end visualization of 3D seismic data to 3D fault surfaces.In this thesis,the synthetic seismic data and the field seismic data are used to test the proposed methods,whose results are compared with those of professional commercial software.The result shows that the fault surfaces reconstruction effect of the methods proposed is good and the proposed methods can reconstruct fault surfaces effectively and reliably.