Model-based prediction of fluid bed state in full-scale drinking water pellet softening reactors