Blog
How to calibrate a DEM (Discrete Element Method) simulation
This article explains how to use GranuTools instruments to calibrate DEM simulations with a practical approach.
How to calibrate a DEM (Discrete Element Method) simulation
The powder we wish to simulate in a process is first measured in a simple geometry with a well-controlled stress field. Afterward, a set of simulations are performed with an optimisation process to obtain the same behaviour than in the measurement cell. It is a kind of learning stage for the DEM algorithm. Afterward the trained (or calibrated) algorithm is used to simulate a complex process.
DEM simulations are useful to predict the behaviour of powders and granular materials inside industrial processes
The principle consists basically in solving the Newton equations to compute the motion of the grains. Two main difficulties are encountered when dealing with DEM simulations. First, the number of grains in a process is large and therefore the number of contacts (with the corresponding forces) is huge. Then, the computing power needed to solve the wide set of equations is significant. This first difficulty is overcome by using super-computers and by optimizing the algorithm. The second difficulty is more fundamental and is related to powder characterization. The powder properties must be precisely measured to obtain final numerical results corresponding to reality. Indeed, according to the value of the input parameters, a numerical simulation could give a wide range of results including totally unrealistic ones.
Unfortunately, tuning the microscopic parameters (coefficients of friction and restitution, grain stiffness, cohesive force coefficients, ...) is a difficult task when dealing with real powders made of grains having complex characteristics.
Conceptually, two methods can be used for the calibration of DEM simulation parameters. The first method is a bottom-up approach based on the direct measurement of the microscopic parameters at the scale of the contacts between the grains. These measurements can be performed with AFM (Atomic Force Microscopy) techniques. Even if this bottom-up method makes sense conceptually, it is often practically impossible to use this approach based on complex measurement. The distribution of grain size and shape in real powders leads to strong fluctuations of the results. In addition, in many simulation models, the input parameters are not rigorously linked with these physical parameters.
GranuTools proposes an alternative top-bottom method based on a set of macroscopic measurements that can be used practically without difficulties to calibrate the simulation
The principle consists in measuring the powder properties in a laboratory and afterward to simulate the same process to tune the simulation parameters following an optimisation process. These measurements are performed in well-known geometries (tubes or rotating drum) and can be easily simulated. Finally, the best set of parameters will be used to simulate a complex industrial process. For example, this method has been applied in the framework of powder bed based additive manufacturing processes [1] and to simulate feeders in pharmaceutical continuous manufacturing processes.
As for any powder characterisation task, the measurement geometry and the stress applied on the sample must be selected in accordance with the application. For a process involving high speed powder flow with a low confinement, the rotating drum methods (GranuDrum) will extract parameters that make sense for the application. On the other side, the automatic packing dynamics analysis (GranuPack) will be more pertinent for a powder submitted to higher pressure in the process. In between, the optimised GranuHeap angle of repose measurement is a good compromise.
LEARN MORE ABOUT THE GRANUHEAP
[1] N. Preud’homme, A. Neveu, F. Francqui, E. Opsomer, N. Vandewalle and G. Lumay, Simulating Powder Bed Based Additive Manufacturing Processes: From DEM Calibration to Experimental Validation, WCCM-ECCOMAS2020