The broadband noise prediction of low speed axial fans is investigated, with a focus on coupling Reynolds-Averaged Navier-Stokes (RANS) simulation with analytical acoustic prediction models. The aim is a good compromise between result accuracy and prediction time for early design studies. The analytical Amiet trailing-edge and turbulence-interaction noise generation models, adapted to rotating blade, with no blade to blade interaction, and with free-field propagation, is chosen. The python implementation of these models is successfully validated using prediction and test results on one isolated two-bladed fan and one six-bladed fan installed in a short duct from two previously published cases. These prediction models are then applied to an automotive axial fan installed on a full engine cooling module, including electrical motor, shroud and heat exchangers. The flow parameters and turbulence characteristics, required analytical model inputs, are obtained from the CFD simulation of a simplified periodic geometry where only a single fan blade and an axisymmetric module are modeled. Fan module simulated flow performance are validated at a single fan rotational speed on a large airflow range by the appropriate match with AMCA plenum experimental measurements. Aeroacoustic prediction results at working fan speed compared to experimental data measured in anechoic chamber show the relative importance of each noise mechanism and the performance of the prediction strategy considering the geometry simplifications. Finally, a direct Lattice-Boltzmann Method (LBM) simulation of the noise from the complete module with no simplification is performed. The analysis of the accurate results gives perspectives for improvement of the simplified method.