The Proceeding of the 1st International Conference

The aim of this study is to apply artificialintelligence to predict performance of inorganic inhibitors. Theseinhibitors are used to inhibit the localized corrosion of carbonsteel in secondary cooling circuit of a nuclear facility. Theperformance of inhibitors has been evaluated by the pittingpotential (Epit) which was obtained through potentiodynamicpolarization. The relationship between Epit and factorsinfluencing pitting corrosion is very complex. Therefore, DeepNetwork using Auto-Encoders DN-AE was used to predict pittingpotential Epit. Various parameters of DN-AE are optimized todetermine optimal parameters and to obtain the best accuracy.The correlation coefficient (R) is used as evaluation statisticcriteria to determine the performance of DN-AE. Allexperimental results obtained from electrochemical experimentare collected in a table to use in the prediction work.The results obtained after optimization of the parameters ofDN-AE showed the ability of DN-AE to predict Epit with goodperformance. The optimal accuracy was obtained with squaredform of concentration and 24 neurons in each auto-encoderslayers. Moreover, 2/3 represents the optimal fraction betweentraining data.