E mphasis in recent years has been focused on improving processes which lead to enhanced phosphate recovery. This paper studies the precipitation features of calcium phosphate in a uidized bed reactor in a concentration range between 4 and 50 mg l 1 and establishes the conditions for optimum phosphate removal ef ciency. For this purpose, two models are coupled for predicting the pellet reactor ef ciency. First, a thermodynamical model is used for predicting calcium phosphate precipitation vs. initial conditions (pH, [P], [Ca], temperature). The second one is a reactor network model. Its parameters are identi ed by an optimization procedure based on simulated annealing and quadratic programming. The ef ciency is computed by coupling a simple agglomeration model with a combination of elementary systems representing basic ideal ow patterns (perfect mixed ow, plug ow, etc.). More precisely, the superstructure represents the hydrodynamical conditions in the uidized bed. The observed results show that a simple combination of ideal ow patterns is involved in pellet reactor modelling, which seems interesting for a future control. The experimental prototype used for validation purpose is rst described. Then, the thermochemical model is presented for calcium phosphate precipitation. The third part is devoted to the reactor networkoriented model. The approach presented is nally validated with experimental runs.