A multi-scale model is presented for predicting the magnitude and rate of powder mixing in a Tote blender. The model combines particle diffusion correlations calibrated from experiments with transient advective flow field information from finite element method simulations. Predictions of the mixing rate from the multi-scale model compare well quantitatively to published experimental data for side-side and top-bottom loading conditions. The multi-scale model, since it does not directly model individual particles, is expected to scale well to systems of industrial interest. (C) 2018 Elsevier B.V. All rights reserved.