Introduction Bow thrusters are of great help for the navigation at quay walls, but the high and turbulent velocities can result in a bed load exceeding the strength of the bed or bed protection. To be able to design a stable bed the velocities at the bed need to be accurately determined. In design practise the velocities generated by a propeller are determined with formulae based on a mix of the momentum theory and measurements. The application of the formulae is often limited to cases for which measurements have been carried out and do not allow a secure design for more complicated structures and the different velocity field of a bow thruster. Scale model measurements To improve the calculation of velocities on a slope, a large number of measurements were done by Van Doorn [TU Delft, 2012] for several scenarios with and without piles and resulted in an amplification of the design formula for some of his scenarios. To also predict the velocities for other scenarios these measurements are used to build and calibrate a numerical model. Numerical bow thruster implementation The open source CFD package OPENFOAM is used for the construction of this numerical model. As the implementation of a rotating propeller in the mesh will result in high computational costs and to allow a fine calibration of the propeller efflux, the propeller is simplified to an actuator disc. At the actuator disc an axial and tangential body force, varying over the radius, are added to the momentum equations in the OPENFOAM solver. Functions for both a ducted and a free propeller are simulated and show comparable results, the free Goldstein propeller functions are further applied. The coefficients are estimated based on the measured thrust and torque and calibrated to achieve a good fit to the measured efflux. A local increase of the turbulence at the hub and the propeller tip is not implemented in the numerical computations. Results Comparing the calibrated model to the measured diffusion in axial direction, shows a very good agreement and the numerical model nearly exactly computes the distribution as derived by Blaauw and van der Kaa. When comparing the velocities at the slope to both theory and the scale model measurements, it shows an underestimation of the velocities at the toe for steeper slopes, which is explained by unexpected velocities in the wall boundary layer, as a result of the wall functions in OPENFOAM. The model is exerted on different geometries (with piles) to get insight in the velocities for quay structures.