Application of RBF neural network and ANFIS on the prediction of corrosion rate of pipeline steel in soil

Abstract

Many factors can affect the corrosion rate of pipeline steel in soil. It is difficult to entirely simulate these factors in lab. Based on a few corrosion data by modeling emerging corrosion tests, the corrosion rate of L245NB steel in soil was predicted by using RBF Neural Network and Adaptive Neural-Fuzzy Inference System (ANFIS). The results showed that for the two models there were almost no differences about the prediction precision of test data. While predicting the actual data of coupons in-situ using the models trained by laboratory data, the precision of ANFIS model was better than that of RBF model. And the trend curves about corrosion rate and each factor were correspondent with the truth more. © 2012 Springer-Verlag GmbH.