STRATIGRAPHIC SEQUENCE SIMULATION BASED ON MACHINE LEARNING

The structure and distribution simulation of strata is an urgent demand in geological informatization as well as engineering. Current study methods are mostly based on borehole data,drawing stratigraphic section or building three-dimensional geological model through interpolation. Interpolation is an important factor in accuracy. However,the determination of interpolation method is subjective,lacking of scientific consideration and therefore difficult to apply to other distinct. Therefore,this study proposes a stratigraphic sequence simulation method based on machine learning. The method considers the borehole data as type and thickness sequence,and presents the stratigraphic sequence simulation model based on recurrent neural network and sequence to sequence learning. Comparing the simulation sequence to the actual borehole data,the result indicates that the machine learning-based model are capable of describing the stratigraphic sequence above bedrock using coordinate information. This study provides a new method for stratigraphy study.