Performances of three models in predicting packing densities and optimal mixing fractions of mixtures of micropowders with different sizes

Abstract

Mixing powders with different sizes is a common method to tune the packing density. For the first time, predicted packing density and optimal mixing fraction from different linear packing models (de Larrard's, Kwan's, and Yu's) are compared against the experimental results of mixtures of micropowders (with different particle sizes: 2, 10, and 70 μm). Regarding the predicted packing density, Kwan's model achieved the smallest prediction deviations for three mixing systems of 10 μm and 2 μm powders, 70 μm and 10 μm powders, and 70 μm, 10 μm, and 2 μm powders, while de Larrard's model achieved the smallest prediction deviation for the mixing system of 70 μm and 2 μm powders. Overall for the predicted packing density, Kwan's model achieved the best prediction performance with the lowest average mean absolute error of 2.2%. Regarding the predicted optimal mixing fraction, Kwan's model outperformed the other two models for the mixing system of 10 μm and 2 μm powders and the mixing system of 70 μm and 2 μm powders, while Yu's model outperformed the other two models for the mixing system of 70 μm and 10 μm powders. Possible reasons of the better performances of Kwan's model in both prediction aspects include the consideration of wedging effect.