Efficient calibration of discrete element material model parameters using Latin hypercube sampling and kriging

Material model parameter identification for discrete element models (DEM) is typically done using a trial-and-error approach and its outcome depends largely on the experience of the DEM user. This paper describes a work flow which facilitates the efficient and systematic calibration of discrete element material models against experimental data. The described workflow comprises three steps. In the first step, an approach based on the design and analysis of computer experiments (DACE) is adopted in which data is generated for the parametrisation of Kriging models based on Latin hypercube sampling. In the second step, multi-objective optimisation is applied to the Kriging models. This study introduces an additional cost criterion, which includes the Rayleigh time step, in order to reduce the solution set size to one element. In the third step, the optimisation procedure is repeated with the actual DEM models, using the optimal parameter set from the Kriging models as the start value. This final step with the full DEM models refines the parameter set against experimental data. Since DEM material model calibration is time-consuming, the workflow is implemented into an automated process chain. In this paper, the methodology is described in detail and results are shown which illustrate the usefulness and effectiveness of this approach. Initial verification simulations run using the calibrated parameters give good agreement with experimental results.