Evaluation of the Ability of ANFIS and SVMR Models to Predict the Corrosion Inhibition of Cerium Conversion Coating

This work investigates the inhibition effect of cerium conversion coating on various aluminium alloys. Two different computational methods, ANFIS (adaptive-network-based fuzzy inference system) and SVMR (support vector machine regression) were developed to predict the inhibitory power of coated Al-alloys. The corrosion behaviour of the coated samples was also examined using potentiodynamic polarization testing. Results showed that the cerium oxide layer decreased the corrosion rate of Al-alloys, but the efficiency of the protective layer depended on the alloying composition. From the experimental results, the alloying elements of aluminium were used as input parameters to train the models. The optimum ANFIS model was developed by varying the clustering parameters manually. The optimum structure is achieved when the squash factor, the range of influence, and the reject ratio and accept ratio are taken as 15, 0.3, and 0.5 respectively. The root-mean-square error (RMSE) of the optimized fuzzy model was 3.67 × 10 –5 . Various statistical criteria showed that the ANFIS model ( R 2 = 0.99) could predict inhibitory power more accurately than SVMR with R 2 = 0.86. A predictive model was thus created to classify and predict coated Al-alloys AA1xxx to AA8xxx, which could solve the shortage of data sets. Furthermore, investigation of the effect of the inputs and a sensitivity analysis for the ANFIS model showed the remarkable impact of Mg, Mn, and Zn alloying elements on the inhibitory power of coated Al-alloys. The results also indicated that higher inhibition efficiencies were obtained for coated 5xxx and 6xxx aluminium series.