Abstract <p>The article considers the problem of rebuilding the surface computational mesh in the problem of ice accretion. In this problem, the surface mesh should be rebuilt in accordance with the volume of ice accumulated in each mesh cell. The main factor in choosing a mesh rebuilding method is the accuracy of the method, characterized by the magnitude of the deviation of the volume swept by the cell from the target value. Another determining factor is the ability of the rebuilding method to smooth out mesh defects—sharp peaks and caverns. Without this smoothing, the computational mesh can accumulate defects, which sooner or later leads to the impossibility of continuing the calculations. The paper considers theoretical estimates of three methods of rebuilding the surface mesh—the rectangles method, the trapeziums method and the neighbourhoods method—from the point of view of the accuracy of rebuilding and the ability to smooth out defects. The results of the analysis showed that the rectangles and trapeziums methods cannot smooth out the defects of the computational mesh during rebuilding, which means that use of them in the problem of ice formation is unsafe. The neighbourhoods method has an acceptable rebuilding accuracy and can smooth out the defects of the computational mesh. All results were obtained in a two-dimensional case, but the conclusions can be transferred to analogous methods of rebuilding the surface mesh in a three-dimensional case.</p>