High-precision fault prediction technologies and applications
Using fault-sensitive attributes to identify faults is limited in accuracy and applicability,mainly able to identify major large-scale faults.As medium and small-scale faults lack sufficient resolution and continuity,the predicted number of faults is inaccurate.The credibility of predicting faults solely relying on a single attribute is relatively lower.In addition,for a long time,fault interpretation has been mainly done manually,seriously affect-ing working efficiency.To overcome the limitations of a single method,thus identifying more fault layers,im-proving the resolution and continuity of fault prediction,and eliminating the"double boundary"effect,this pa-per creates a high-precision artificial intelligence fault prediction process to gradually improve the accuracy and efficiency of fault interpretation.Firstly,post-stack seismic data preprocessing is carried out to improve the quality of original seismic data and establish a foundation for generating high-precision fault attributes.Sec-ondly,artificial intelligence fault prediction is carried out to further improve the prediction accuracy and process the artificial intelligence prediction results in the"fault enhancement-skeletonization-ant body"way.Again,multi-attribute fusion is utilized to enrich multi-scale fault information.Finally,with high-precision fault predic-tion results and intelligent technology,the efficiency of fault interpretation is improved.The high-precision arti-ficial intelligence fault prediction process was applied to the actual data of Zone X in the Bohai Bay Basin,and the results showed that the profile normal faults are mainly composed of stepped and Y-shaped combinations.Most of them are SE orientation and form simultaneously with the fault depression.The dominant orientation of the fault on the plane is NE,and forms a comb-like combination with secondary faults in the near SN direc-tion,indicating that the area may have undergone sinistral strike slip while deforming extensionally.
国家科技期刊平台
NSTL
万方数据
