Linking model tribological tests and milling wear via a specific wear rate descriptor for PVD TiN and TiAlN model coatings

This work presents a laboratory-scale tribological framework for predicting the flank wear of coated milling tools, developed and validated for monolithic PVD TiN and Ti₁₋ₓAlₓN coatings on WC–10 wt% Co inserts in dry milling of normalised C45 steel, using a compact tribological descriptor derived from a ring-on-cylinder model test. For this purpose, a set of DC magnetron-sputtered TiN, TiAlN (Ti:Al 75/25), TiAlN (Ti:Al 50/50), and a HiPIMS TiAlN (Ti:Al 50/50) model coatings was used for both the tribological model and real cutting tests. A commercial TiAlN reference coating was also included in these investigations. The model coatings were simple monolithic coatings with variation of elemental composition, microstructure and morphology, deposited onto cemented carbide model samples and tool inserts. These coatings were characterized by electron probe microanalysis (EPMA), X-ray diffraction (XRD), and scanning electron microscopy (SEM). All materials were evaluated in both disc-on-ring tests against C45 steel and dry milling of C45 with uncoated and coated WC–Co inserts with regard of the effective wear mechanisms and resulting wear rates. Model tribological tests were designed to reproduce line contact conditions while remaining short in duration (2000 s) and moderate in temperature (up to 100 °C). The rationale behind these model tribological tests was to replicate the wear mechanism occurring in real-world milling, without wearing through the coating completely, so that substrate effects would not influence the prediction of wear in milling experiments. Combined SEM/FIB analyses of worn tribological samples confirmed the dominant wear mechanisms in the model tribotests: diffusion for uncoated, adhesion for TiN and abrasion-dominated wear for Al-rich TiAlN coated samples. These findings very closely matched the wear mechanisms observed on worn milling tools. Consequently, the specific wear rate, defined as wear volume normalised by load and sliding distance, was then used as a coating-specific descriptor ( W s ), derived from the model tribological tests. This descriptor was then utilised in a grey-box tool-wear model that combines milling data with thermo-mechanical load scenarios from chip-formation simulations. The resulting framework correctly reproduced the coatings' ranking observed in terms of wear progression during milling and enabled extrapolative prediction of an unseen TiAlN coating using only tribological data obtained from the model tribological test. Compared with a purely data-driven black-box model, the grey-box model achieved a higher R 2 in all coating–parameter combinations and a lower RMSE, with the largest gains at the highest cutting speed and feed, demonstrating that physics- and tribology-informed descriptors can enhance predictive robustness within this material window while reducing experimental effort. Validation is restricted to monolithic TiN/TiAlN coatings against C45 steel; transferability to multilayer, graded or quaternary coating architectures remains to be established.