The Research on the Construction Method of a Slope Digital Twin Model Driven by Surrogate Models

In view of the shortcomings of traditional slope safety monitoring methods, such as low calculation efficiency and poor real‐time performance, this paper proposes a digital twin modelling method for slopes based on a surrogate model. Through a stacking integrated learning model combined with a multidimensional error analysis mechanism and adaptive parameter optimisation strategy, a high‐precision surrogate model for slope displacement prediction was constructed. Compared with the real displacement data obtained from laboratory experiments and digital image technology (DIC), the prediction error of the optimised surrogate model is reduced from 20% to 5%. In this study, a digital twin system of the Donglingxin landslide was successfully constructed, and the deformation process from 2010 to 2022 was dynamically displayed. The results show that the deformation of the Donglingxin landslide clearly experienced acceleration−stabilisation, and the displacement rate first increased from 12.5 mm/y to 15.5 mm/y and then decreased to 11.8 mm/y, indicating that the landslide entered a relatively stable deformation stage.