Regression Analysis Of High-Temperature Oxidation Of Ni-Based Superalloys Using Artificial Neural Network

The high-temperature oxidation resistances of Ni-based superalloys with the compositions of Ni-(0-15)Co-(8-15)Cr-(0-5)Mo-(0-10)W-(3-8)Al-(0-5)Ti-(0-10)Ta-0.1C-0.01B were analyzed using an artificial neural network (ANN). The oxidation resistances of the alloys were evaluated based on the weight change measured during cyclic oxidation tests. An ANN was constructed with the contents of the alloying elements as input and mass gains as output. The ANN provided highly accurate regression results. The present results were compared with those obtained using response surface methodology (RSM) in a previous study. The regression model of the ANN could effectively detect the effect of an additional alloying element explicitly. The main and interaction effects of the alloying elements were plotted based on the results with random compositions. The optimum composition of superalloys with the highest oxidation resistance at high temperatures was determined using the trained ANN.