Application of Artificial Neural Network Model to Predict Corrosion Rates on Pipeline

This study aims at determining corrosion rates in oil and gas pipelines by application of artificial neural network model to predict corrosion rates on pipeline; and to compare the achieved numerical outcomes with the existing work as special cases. An artificial neural network model capable of predicting the rate of corrosion was developed. The model was able to successfully predict corrosion rate between 0.02mm/yr-0.17mm/yr. The study had a root mean square error of 0.0130; mean absolute error of 0.007, scattered index of 0.1708, and above 91.5% confidence level at training, testing and validation, with coefficient of determination above 95% prediction accuracy, with a relative error of 0.013%-0.047%. Graphs are plotted to show the impact of various physical parameters on pipeline age, environmental pH and temperature. It is detected from the obtained graphical data that multi-factors interactions significantly affect corrosion rates. Furthermore, the contour and surface plots indicate the ascending severity order of the localized attack on the pipes due to factor pairs. The results obtained by ANN predictions are consistent with that of experimental and the validity of the achieved numerical outcomes is ensured by making a comparison with the existing work of special cases. With this concept, the present ANN model reflects the mainstreams understanding of corrosion in acidic environments, and can be easily used to predict the corrosion rates in industrial applications.