Modelling high temperature oxidation behaviour of Ni-Cr-W-Mo alloys with Bayesian neural network

High temperature oxidation property of Ni-Cr-W-Mo alloys is modelled as a function of alloy composition. A database has been constructed from cyclic oxidation experiments performed between 1150 °C and 400 °C of 27 alloys having various contents of Cr, W, and Mo. The Bayesian neural network technique was used for the modelling of cyclic oxidation experiments. With a 3-17-1 neural network architecture, the model shows a precise prediction of oxidation property of Ni-Cr-W-Mo alloys (R = 0.999). Automatic relevance determination (ARD) analysis reveals that the influence of alloying elements on the output is in the order of Cr, Mo, and W.