Machine Learning-Enabled Prediction of High-Temperature Oxidation Resistance for Ni-based Alloys

A machine learning (ML) model was developed to predict the oxidation resistance, with the natural logarithm of the parabolic rate constant (lnkp) as the output. Five algorithms were used to training the model, revealing that backpropagation neural network and gradient boosting exhibited superior performance with an R2 of 0.908 and 0.907. The trained models were employed to predict lnkp for changes in temperature, time, and elemental compositions. At last, promising areas with slower oxidation kinetics in typical commercial alloys were predicted under the guidance of the present ML methods.