Accelerated design of chromium carbide overlays via design of experiment and machine learning

Chromium carbide overlays with MC-type primary carbides show a significantly lower cracking susceptibility during welding. However, optimizing the composition and improving the wear-resistant properties is difficult by traditional one-variable-at-a-time methods. Here we propose a method using design of experiments and machine learning to accelerate this process. A closed-loop design process was developed by using existing data to predict future experiments and suggest new ones. The influence of Nb, V, and Ti on the overlay hardness was established in an active database. XRD analysis shows that Nb would promote the formation of austenite in the matrix while Ti would foster the formation of martensite. The efficacy of this method is demonstrated by predicting the simultaneous effects of the carbide formers on overlay hardness, which highlights the possibilities of using machine learning in the alloy design of chromium carbide overlays.