Measurement and numerical simulation of local dust concentrations over time in a vessel can only be done with some uncertainty due to the complexity of the behavior of dust/air mixtures. Dust concentration was measured in a 50 m(3) vessel and compared to simulations with a commercial CFD code. A 50 m(3) silo was used with two different filling methods. In one setup dust was conveyed with pneumatic filling at the top. The other filling was done with pressurized air and a homogenous injection via eight nozzles. Experiments were repeated three to four times with two kinds of dusts and the results were used to evaluate reproducibility of dust concentration measurements over time in a vessel depending on the filling method. Dust concentrations over time varied up to 30% from the average for homogenous injection and even more for pneumatic filling. Numerical investigations were done with maize starch. Measured concentrations were compared to simulated ones with the commercial CFD code ANSYS CFX R14 using an Euler/Lagrange approach. Drag force, turbulent dispersion force, particle size distribution, particle surface area and particle/particle interaction were modeled. A general agreement of measurement and simulation was achieved. Numerical simulations of filling processes were used to predict parts of the vessel where the lower explosion limit is reached and exceeded. This could help to improve dust explosion protection, if it is used to find configurations where the dust concentration exceeds the lower explosion limit only in small parts of the vessel during filling, e.g. using different injection points or injection angles. The volume where LEL is reached or exceeded in a 50 m(3) silo is shown for pneumatic and homogenous filling. Volume of combustible atmosphere in the vessel over time is compared for two pneumatic filling configurations and one worst-case homogenous injection configuration. (C) 2014 Elsevier Ltd. All rights reserved.