面向钻孔策略优化的机器学习三维地层建模研究OA

To reduce the adverse impact of sparse geological exploration data on the precision of stratigraphic models,this study proposes an implicit three-dimensional(3D)stratigraphic modeling method integrating ensemble learning and multi-fidelity geological survey data.A genetic algorithm(GA)is utilized to construct an optimized Stacking-based ensemble learning framework,which learns stratigraphic distribution patterns from raw geological exploration data and generates stratigraphic classification datasets corresponding to virtual boreholes.A radial basis function with a multiquadric kernel is adopted for implicit stratigraphic modeling.On this basis,information entropy is introduced to quantify the modeling uncertainty and further optimize the borehole layout scheme.A case study is conducted using geological survey data from an urban rail transit project.The results indicate that the GA-Stacking model achieves an F1-score of 90%and an accuracy of 89%on the test dataset.Compared with standalone machine learning models,the proposed method exhibits superior performance in stratigraphic classification and prediction.The mean absolute error between the modeled stratigraphic distribution and benchmark stratigraphic profiles is 0.23 m,with an average profile consistency of 85%,demonstrating the reliability of the developed model.Moreover,the overall stratigraphic uncertainty is reduced by 24.1%after supplementing five additional boreholes,which validates the effectiveness of the proposed optimization strategy.This research provides a novel insight into refined 3D stratigraphic modeling and offers theoretical guidance for borehole layout optimization in geological exploration.