The wear of refractories deserves investigation owing to the potential in cost reduction and process optimization in the steel industry. One of the main lacking points of refractory erosion investigations is the absence of an established erosion law and appropriate methods for its quantification. The method presented here aims to resolve this issue. The experimental set-up is given by finger test experiments with the capability of laser scanning of the eroded surface after the experiment. The erosion model is a computational fluid dynamics model for which the erosion law is obtained from the field of soil erosion. This law is a three-parameter function of the wall shear stress exerted by the melt flow. The refractory, which merely represents a boundary of the model, is altered by the erosion via a user defined function programmed to displace the nodes on this boundary. Dynamic meshing methods, including boundary layer smoothing and remeshing are employed to maintain mesh quality throughout the simulation. The output of the model is the erosion profile. Coupling of this model with an optimization software permits parameter identification by minimization of the sum of square differences between simulated erosion profile and that obtained experimentally by laser scanning of the surface. The mathematical behavior of the method was tested via an artificially designed test-problem. Consequently, the method has been applied to real coarse grain refractories and here the corresponding erosion parameters have been successfully inversely determined.