When using a Computational Fluid Dynamics (CFD) code for the simulation of a two phase system, special attention must be given to the solution procedure. Fluid dynamics in CFD codes are obtained by iterative computations, and criteria must be established to decide the number of iterations needed to obtain a convergent solution under certain defined conditions. In this work, the predicted liquid and gas flow patterns obtained in the simulation of hydrodynamics in bubble columns are analysed throughout the computational process, and the effects of some parameters affecting the computation are studied (dimension of the grid, working conditions, scale-up, relaxation used in the calculations, free surface performance). It was found that each of these parameters affects considerably the solution procedure (though in a different manner) and therefore no general criteria can be established to decide the number of sweeps necessary for a certain simulation. It is concluded that every simulation should be analysed individually each time the conditions are changed. This analysis clearly helps to understand the reliability of CFD predictions in complex systems.