A combined neural network and mechanistic approach for the prediction of corrosion rate and yield strength of magnesium-rare earth alloys

► This study presents a body of corrosion data for a set of custom alloys and displays this in multivariable space. These alloys represent the next generation of Mg alloys for auto applications. ► The data is processed using an ANN model, which makes it possible to yield a single expression for prediction of corrosion rate (and strength) as a function of any input composition (of Ce, La or Nd between 0 and 6 wt.%). ► The relative influence of the various RE elements on corrosion is assessed, with the outcome that Nd additions can offer comparable strength with minimal rise in corrosion rate. ► The morphology and solute present in the eutectic region itself (as opposed to just the intermetallic presence) was shown – for the first time – to also be a key contributor to corrosion. ► The above approach sets the foundation for rational alloy design of alloys with corrosion performance in mind. Additions of Ce, La and Nd to Mg were made in binary, ternary and quaternary combinations up to ∼6 wt.%. This provided a dataset that was used in developing a neural network model for predicting corrosion rate and yield strength. Whilst yield strength increased with RE additions, corrosion rates also systematically increased, however, this depended on the type of RE element added and the combination of elements added (along with differences in intermetallic morphology). This work is permits an understanding of Mg–RE alloy performance, and can be exploited in Mg alloy design for predictable combinations of strength and corrosion resistance.