To date, the efficiency of industrial-size bioreactors has mainly been improved based on empirical knowledge. Computer simulation may help to understand the processes that occur inside the reactor and to develop new reactor designs. Euler-Lagrange simulations of the two-phase flow in large bioreactors, which could not be performed within a timeframe suitable for engineering purposes due to the limited computation resources, were made possible by the calculation power of graphic cards. The lattice Boltzmann method is well suited for parallelization which makes it ideal for calculating the fluid field inside a reactor driven by multiple Rushton turbines on graphic processing units. The bubble movemnts were captured via a Lagrangian approach by solving the Newton's equations of motion. A two-way coupling between the disperse and continuous phases was applied. Break up and coalescence of the bubbles were modeled via stochastic algorithms using the approach rate of small turbulent eddies and the comparison of the contact time and film breakage time, respectively. To gather experimental data, a conductivity sensor was used to measure the local gas holdup. The rate and the duration of current drops were recorded to estimate the bubble size and the void fraction around the sensor's tip position. The sensor was used in a 1501 custom-built acrylic reactor. Several flow regimes with varying gas flow rates and stirrer speeds were investigated. The experimental results were in good agreement with the simulation data, especially at low stirring and low aeration rates. To prove the applicability of the code to large-scale problems, a 40 m(3) reactor was simulated. (C) 2016 Elsevier Ltd. All rights reserved.