The problem of simulating floating-zone crystal-growth experiments in microgravity, in monoellipsoidal mirror furnaces, is addressed; that is, the problem of making quantitative predictions about the behavior of the system. It is shown that having a validated model available is not enough, since some physical parameters are not known accurately and the problem happens to be ill conditioned, that is, varying all those parameters within physically reasonable ranges leads to quite different results, so that accurate quantitative predictions are difficult to make. This difficulty is solved by "adjusting" the model to the experimental configuration to be simulated, with the help of reference experiments performed with a setup similar to the one used in the mug experiments. This adjustment process defines those parameters not well known by requiring a "best" comparison with the reference experiments. Performing this set of reference experiments (which must be properly designed) turns out to be a necessary task for the simulation, thus showing the complementarity of the experimental and numerical tools. Microgravity experiments performed in the MAXUS programmes are simulated in this way. Important parameters for the design of the experiments, like the Marangoni number or the temperature gradient at the melt-crystal interface, are estimated. The effect in the results of parameters like the radius of the sample or the mirror reflectance is analyzed, and shown to be important. Comparison with available mug experimental data shows very good agreement. (C) 2003 Elsevier B.V. All rights reserved.