To improve surface roughness of components without altering geometrical dimensions, polishing operations are commonly used. These operations may be timeconsuming and expensive. Thus, the prediction of polished surface quality is a key issue to reduce the cost of these operations. Resulting surface quality of polishing operations is the outcome of a large number of local cutting phenomena (grains-material). The control of the multi-scale physical phenomena is a challenge when it comes to simulate these operations. In this paper, an Analytical-Method for Polishing-Surface Prediction (AMPSP) that considers the tool flexibility and each grain-material interaction is proposed. The objective is to predict accurately the polished surface topography and the material removal rate, and to keep the history of all the local cutting phenomena, in order to define a digital twin of polishing operation. Experimental validation demonstrates that this AMPSP predicts the material removal rate (less than 35% of error) and the surface topology (less than a few percent). AMPSP will enable engineers to quickly and accurately predict the polished topology obtained with 5-axis toolpaths.