A two-phase Eulerian-Eulerian volume-averaged model was used to predict the outcome of the twin-roll casting process for inoculated Al-4 wt.% Cu alloys. The model is able to address the deformation of the mush during hot rolling by treating the solid/liquid mixture as a viscoplastic skeleton saturated with liquid. A parameter measuring the mean absolute deviation around the nominal composition was applied to the normalized macrosegregation distribution and the results were then used to determine an optimal process window where minimal compositional variation is achieved in the final strip. In the case of casting an 8 mm thick Al-4%Cu strip with rolls of 800 mm diameter, casting speeds between 40 and 42 mm/s and heat transfer rates given by HTCs between 5000 and 5250 W/m2/K minimize the macrosegregation deviations in the strip and thus predict an optimal strip quality.