Mixing of powders is of great importance in food, chemical and pharmaceutical industries. However, direct online measurement of mixing is impractical due to difficulties in real-time particle sampling. In such systems, soft-sensors placed external to the equipment may be used to indirectly determine the behaviour of the system using relationships between internal and external phenomena. In this paper, a soft-sensor approach is studied for a ribbon powder mixer with 0, 2, 4 and 6 impeller spokes, and 20, 30, 40 and 50% volumetric particle filling, using two types of particles of different densities and a fixed impeller speed of 100 RPM. The particle data are based on DEM simulation results in a previous study. Force sensors along the underside of the mixer identify the number of particle-wall contacts and the force experienced by the sensors during the DEM simulations. This information is used along with mixing data to determine a relationship between external force data and internal mixing behaviour. The dividing rectangles global optimisation technique is used to approximate the mixing rate coefficient from particle data, and principal component analysis is used to develop a fast and practical means to estimate the mixing rate coefficient using only readings from external force sensors. This approach is then extended to allow for real time estimation of the required mixing time. (C) 2018 Elsevier B.V. All rights reserved.