An approach for a design of functionally gradated hard metals processing combining thermodynamics and data-driven machine learning

Processing of FGM hard metals is possible starting from two homogenous compositions where the final gradient of properties (hardness, toughness) will result from liquid phase migration. Control and prediction of this process demands a knowledge of thermodynamics of the systems. Here such analysis for the WC-Co-TiC system has been performed to obtain the driving force for the migration of liquid phase during and after sintering of FGM hard metals, with updated data on volumetric fractions of compounds and liquid phase. Parameters of starting compositions (density, grain size, chemistry) have been linked with the liquid phase composition, properties and its migration pressure. Creation of functionally gradated composition of hard metals starting from uniform compositions for two initially homogeneous specimens of the WC-TiC-Co system has been simulated for various masses of the specimens. It is shown that the grain size of the WC is the most critical factor for liquid migration, which in many cases overcomes influence of TiC additions (wetting effect and the phases equilibrium). To develop a proper FGM design process, simple machine learning methods were applied to do feature extraction in order to get links between compositions, masses, WC grain size and properties like hardness and fracture toughness yet without deployment of a specific model. Obtained data can be used of backwards engineering to select starting specimens for a desired hardness and toughness gradients either directly or via metamodels when sufficient amount of experimental data are available and when known engineering correlations for properties could be considered sufficient.