The paper proposes an original algorithm which allows a long time scale extrapolation of DEM results at a very low computational cost. This algorithm can be adapted to any periodic processes. In this study, it is applied to the mixing process of powders within a conical screw mixer. The results are then compared with long time DEM simulations. It appears that this method is able to predict the DEM results with a very good accuracy.

Calibration of powder mixing simulation using Discrete-Element-Method is still an issue. Achieving good agreement with experimental results is difficult because time-efficient use of DEM involves strong assumptions. This work presents a methodology to calibrate DEM parameters using Efficient Global Optimization (EGO) algorithm based on Kriging interpolation method. Classical shear test experiments are used as calibration experiments. The calibration is made on two parameters -Young modulus and friction coefficient. The determination of the minimal number of grains that has to be used is a critical step. Simulations of a too small amount of grains would indeed not represent the realistic behavior of powder when using huge amout of grains will be strongly time consuming. The optimization goal is the minimization of the objective function which is the distance between simulated and measured behaviors. The EGO algorithm uses the maximization of the Expected Improvement criterion to find next point that has to be simulated. This stochastic criterion handles with the two interpolations made by the Kriging method : prediction of the objective function and estimation of the error made. It is thus able to quantify the improvement in the minimization that new simulations at specified DEM parameters would lead to.

The macroscopic behaviour of cohesive granular material in the FT4 shear tester is studied using the discrete element method (DEM). The shear test is simulated faithfully to the experimental procedure (filling, compaction, pre-shearing and shearing). The angle of internal friction and the apparent bulk cohesion are the macroscopic properties analysed as a result of the variation of the microscopic parameters: the sliding friction coefficient and the adhesive surface energy. The simplified JKR model was used to account for the cohesive contact between spheres. The results of the shear test show that the adhesive forces influence the dilatancy of the granular bed and the incipient failure point. In general, the shear stress increases with the adhesive energy. The sliding friction coefficient and the adhesive energy affect the Yield locus and therefore the angle of internal friction and the apparent cohesion. Two correlations were established between the angle of internal friction and sliding friction coefficient and between cohesion and adhesive energy. The effect of the initial consolidation on the shear test results is also discussed.