Prediction of particle mixing in rotary drums by a DEM data-driven PSO-SVR model

Particle mixing in rotary drums is of significant industrial importance, but very complex. Detailed simulation can be achieved using the discrete element method (DEM), but the computational time cost is enormous. Therefore, we combined machine learning with DEM simulations and established a DEM data-driven particle swarm optimization and support vector regression (PSO-SVR) model to predict mixing time and mixing degree at steady mixing state for binary sphere mixtures in horizontal rotary drums with four independent variables: revolution frequency, particle density ratio, particle size ratio, and drum length. After hyperparameter tuning by PSO on a validation set of 25 simulations, the SVR model was trained on 81 DEM simulations. Testing on another 25 simulations yielded excellent results with R2 = 0.95 for mixing time and R2 = 0.90 for mixing degree. These results indicate that the PSO-SVR model is suitable for rapid predictions of particle mixing in rotary drums and other particle processing equipment.