Hydro-classification is one of the most important operation to separate particles according with their size and specific gravity. This operation involves many phenomena and its modeling is still not clear. Complex models demand high computational resources and are used to design the dimension of devices or predict the classification but off-line. However, for control proposes, it is necessary to get a suitable model with acceptable prediction level to be implemented online. This work presents a phenomenological based semi-physical model (PBSM) that describes the phenomena of the hydro-classification and forecasts its performance according to the operative conditions, summarizing all phenomena in the main physical properties of production.