Inverse analysis of friction coefficient in discrete element method using powder properties by support vector regression

This study demonstrates the inverse prediction of the discrete-element method (DEM) input friction coefficient from the DEM output powder properties such as the outflow rate, aerated bulk density, and repose angle. Support vector (SV) regression can reasonably reproduce the friction coefficient from the DEM output powder properties. Among the powder properties considered in this study, the outflow rate is found to be the most effective for predicting the friction coefficient. Other powder properties also contribute to improving the prediction accuracy. The accuracy of the SV model depends on the number of cases used for the inverse regression of the friction coefficient, and models with practical accuracy can be obtained for 100–200 cases. These results will lead to further use of machine learning in DEM simulations.