A novel framework for calibrating DEM parameters: A case study of sand and soil-rock mixture.

The discrete element method (DEM) has been widely used to simulate geotechnical works, e.g. slopes, embankments, tunnels, pipelines, etc. since the last decade. However, how to choose the appropriate input parameters still remains problematic and greatly affects the accuracy of DEM modelling results. One of the common practice is to calibrate DEM parameters using lab or in-situ tests. Such operation is largely dependent on personal judgement and thus induces errors to the numerical modeling results. Therefore, a novel framework based on particle swarm optimization (PSO) algorithm was proposed in this study, which could calibrate parameters using 3 or more macro-indicators. The framework was written in Python, which performed automatic calibration process and thus minimized involvement of personal judgement. To validate its effectiveness, the framework was applied to the simulation of biaxial tests on sand first. The calibration results showed that similar parameters were yielded when different initial values were fed to the framework. Then, the framework was applied to simulation of soil-rock mixture (S-RM), and the calibration process was effective as well. These two cases proved that the proposed framework was an efficient tool for DEM parameter calibration and largely reduced computation errors resulting from personal judgement. [ABSTRACT FROM AUTHOR]