Milling is a typical intermittent cutting process. As a result, tool wear is generated cyclically due to periodic process variables. However, the traditional tool wear prediction strategy based on continuous cutting model is no longer applicable. In this paper, a novel geometric approach through mesh node rigid moving for the milling cutter tool wear prediction has been developed. Firstly, a unified tool wear predictive model is established through bridging the two wear configurations before and after worn. A coupled abrasive–diffusive model is employed to calculate the tool wear volume of each point on tool face. Further, a novel iterative algorithm for tool wear prediction through mesh node rigid moving layer-by-layer and process variables redistribution is designed in discrete-time domain, which is generally decomposed into two phases according to cutting heat equilibrium state, FEM simulation and offline calculation. Last, a series of numerical and saw-milling experiments for flank wear prediction were implemented to verify the developed approach. The AISI304 and the high vanadium high-speed steel tool without coating were adopted. By comparison, the predicted results were consistent with the experimental overall. It has been proved that the proposed approach is more effective than pure FEM simulation and is suitable for long-term milling tool wear prediction.