Tool wear remains of high interest for industry, as it influences process costs and part’s surface integrity. Although experimental and analytical investigations have been the main ways to investigate wear, the growing development of computational power enables predicting tool wear based on chip formation simulations. If this has been quite successful in turning, developments in milling are still limited due to the specific nature of this machining operation characterized by an interrupted cutting process leading to mechanical and thermal cyclic loadings onto the cutting tool. Wear modes are often not well characterized and become even more difficult to model as far as hard to machine material such as martensitic stainless steels are concerned. The present work propose to investigate wear in orthogonal milling of a 15-5PH martensitic stainless steel. An experimental campaign is first performed to identify the wear modes when cutting this material with uncoated and coated carbide tools. Milling forces, tool wear and material transfer are especially studied. A multi-scale numerical procedure is then developed by combining an Arbitrary-Lagrangian-Eulerian (ALE) thermomechanical model to a pure thermal sub-model in order to predict the thermomechanical loadings withstood by the tool. The thermal sub-model is applied at the scale of the coating in order to extract the thermal gradients generated by the interrupted cutting. These loadings are finally compared to the reported wear modes to identify a correlation and improve their understanding.