Application of Box-Behnken Method Based ANN-GA to Prediction of wt.% of Doping Elements for Incoloy 800H Coated by Aluminizing-Chromizing

Abstract

In this work , an effective procedure of Box-Behnken based-ANN (Artificial Neural Network) and GA (Genetic Algorithm) has been utilized for finding the optimum conditions of wt.% of doping elements (Ce,Y, and Ge) doped-aluminizing-chromizing of Incoloy 800H . ANN and Box-Behnken design method have been implanted for minimizing hot corrosion rate k_p (10^-12g².cm^-4.s^-1) in Incoloy 800H at 900^oC . ANN was used for estimating the predicted values of hot corrosion rate k_p (10^-12g².cm^-4.s^-1) . The optimal wt.% of doping elements combination to obtain minimum hot corrosion rate was calculated using genetic algorithm approach . The predicted optimal values for minimizing  hot corrosion rate for  Incoloy 800H coated by (Ce-Y-Ge) doped-aluminizing-chromizing are (3wt.%Ce, 3wt.%Y, and  3wt.%Ge) ,  the hot corrosion rate k_p (10^-12g².cm^-4.s^-1) value at these conditions was found to be 71.701 . The results have been verified by confirmation experiment , results obtained by GA method match closely with experimental values (R²=98.30) . EDS and XRD results show that the formation of protective layers Al₂O₃ and Cr₂O₃ during hot corrosion tests.