Application of artificial neural network for prediction of high temperature oxidation behavior of AlCrFeMnNiWx (X = 0, 0.05, 0.1, 0.5) high entropy alloys

Abstract

The work demonstrates the artificial neural network (ANN) potential applicability to predict the oxidation kinetic of high entropy alloy (HEA). For the establishment of the ANN model, an extensive experiment has been performed. Initially, spark plasma sintered (SPS) AlCrFeMnNiWx (x = 0, 0.05, 0.1, 0.5) HEAs oxidized at 700 °C, 800 °C, and 850 °C isothermally in thermal gravimetric Analyser (TGA) for 50 h and investigated by XRD and SEM. The HEAs exhibited multifarious behavior while adding tungsten and showed various oxides. Admittedly, alloying constituent significantly affects oxidation behavior. Thus alloying composition, exposer time, and oxidation temperature were chosen as input parameters for the modeling. At the same time, the resulting mass gain of the oxidized sample was an output of the ANN model. The ANN model attained outstanding performance during training, testing, and validation (R > 0.999). Several physical models have been compared with the proposed predictive ANN model during the kinetic law interpretation and found accuracy significantly. In comparison, the ANN predictive model provides an excellent result with the experimental data at each studied temperature for the HEAs. Unquestionably, the ANN model is a consistent and precise approach to predicting HEAs' high-temperature oxidation behavior.