Intelligent calibration method for microscopic parameters of soil-rock mixtures based on measured landslide accumulation morphology

The discrete element method (DEM) has been widely used in landslide research. However, selecting microscopic parameters is complicated due to the diversity and complexity of landslide material composition, induced mechanism and geological conditions. This study proposed an intelligent calibration method of microscopic parameters based on the measured landslide accumulation morphology. Post-sliding accumulation thickness, sliding height difference, sliding distance and accumulation dip were proposed to describe the morphological characteristics of landslide accumulation. Latin hypercube sampling (LHS) was used to establish the training sample set. The surrogate model of the discrete element numerical simulation method was established by combining the relevance vector mechanism (RVM) and the third-generation nondominated sorting genetic algorithm (NSGA-III) to find the nonlinear relationship between the accumulation characteristics after landslide and microscopic parameters. Based on the calibration results, the area that may be affected by landslides was predicted. The results showed that the similarity between the calculated value of the intelligent model and the measured value was 97.49%. The intelligent calibration method had high accuracy and applicability.