3D Modeling Technology For Geological Bodies Based On Knowledge Graph

By creating a three-dimensional geological model,the spatial distribution and correlation between elements such as terrain,topography,lithology,and geological structures on the earth’s surface can be clearly displayed.Especially for constructing geological models,it is a necessary tool for understanding underground spaces,detecting oil and gas resources,and conducting mining.However,the existing modeling methods only consider computer graphics,without fully extracting information from complex and heterogeneous seismic data,resulting in unreasonable model constructions.Therefore,we need a modeling method and process that can effectively extract information from geological data and fully represent geological knowledge to solve this problem.To address the problems of existing tectonic modelling mentioned earlier,this thesis investigates the construction of a knowledge map to characterise the tectonic knowledge in geological models,so as to achieve knowledge-based modelling instead of the traditional method of data-based modelling.By referring to the construction of knowledge maps in traditional fields and combining seismic interpretation data and knowledge in the field of tectonic geology,this thesis proposes a knowledge map construction method for geological bodies and tectonic geology,and proposes a knowledge-driven knowledge map constraint-interpolation technique for 3D modelling of geological bodies,and achieves the following results:Firstly,this thesis uses a fishtail representation model based on faceted entities,combined with profile-based hierarchical lines of interpretation of data features,to construct a knowledge graph in a bottom-up manner,solving the problem of expressing the constructed semantics of the model spatially.The knowledge graph and fishtail representation models are topologically consistent,using triples as data structures to characterise the topological relationships between entities by linking nodes through topological relationships.In this thesis,an initial data-layer knowledge graph is successfully constructed from which information is extracted at the ontology and schema levels.Secondly,this thesis investigates a deep learning-based knowledge extraction method for geological bodies and tectonic geology,taking the mound-bank complex as an example,to build a knowledge map in a top-down manner.Considering the problems of imperfect corpus,poor ability to handle heterogeneous data from multiple sources and difficulty in knowledge sharing in the field of geological and tectonic geology,this thesis firstly constructs a rule-based dataset;considering the problems of large text volume and long text statements in this field,this thesis introduces two types of Bi LSTM and CRF on the basis of traditional BERT structure.Through experimental comparison,this thesis outperforms the comparison model in terms of Precision,Recall and F1 score evaluation indexes,which verifies the effectiveness of the model;considering the problems of different importance of entities in the text and the distance between entities and relations in this domain,this thesis introduces the self Att structure based on Bi LSTM.Through experimental comparison,this thesis’ s method clearly outperforms the comparison model in terms of Marco-P,Marco-R and Marco-F1 value evaluation indexes,and achieves very good results,thus realizing knowledge extraction in the field of geological bodies and tectonic geology with mound-bank complex as an example,and obtaining a pattern layer knowledge map in this field.Finally,this thesis proposes a knowledge graph constraint-interpolation technique.This thesis uses knowledge fusion and knowledge inference to improve the model layer and update the data layer based on the improved model layer,so as to obtain a complete knowledge map of the geological and tectonic geology domain with the mound-bank complex as an example.Based on the completed knowledge map of geological and tectonic geology of the mound-bank complex,this thesis combines the actual work area data,obtains nodal constraints and other relationships through the knowledge map,and then performs kriging interpolation to obtain detailed information of the tectonic body,and finally successfully constructs a smooth,reasonable and geological model under the guidance and constraints of the knowledge map.