Predict Soil Corrosion Rate of Pipeline Steel Using ANFIS

In soil environment, due to the consideration of the experimental cost and environmental conditions, samples of material corrosion data are limited and affected by many factors. These features determine that soil corrosion data are typical small samples and high dimensional data, and they are strongly correlated with each other, it is difficult to establish accurate mathematical models through theoretical analysis to predict soil corrosion. In this study, Adaptive Neuro-Fuzzy Interference System (ANFIS) and RBF Neural Network were established based on simulated corrosion experiments to predict corrosion rate. The results showed that the two models' predicting accuracy for nontraining data from experiment were almost the same, but ANFIS was more accurate than RBF Neural Network when predicting actual on-site soil corrosion rate and could better reflect the relationship between corrosion rate and each of the corrosion factors.