A combined support vector regression and genetic algorithm for prediction of burst pressure of corroded pipelines

The burst pressure of oil and gas pipelines containing corrosion defects is an important factor in structural reliability of such structures. An accurate prediction of this parameter can provide sufficient safety levels of corroded pipelines. Usually, the burst pressure of corroded pipelines is estimated based on empirical or probabilistic models that have the narrow limitation for large variety of steel grades. In this paper, an artificial intelligence technique called support vector regression (SVR) is proposed to predict the burst pressure of corroded pipelines. To ensure an optimal selection of SVR hyper parameters, meta-heuristic technique widely known as genetic algorithm (GA) is applied. The SVR-GA model development is developed using real burst pressure experimental databases obtained from distinct grades of oil and gas pipelines gathered from literature. The performances of the proposed technique in terms of accuracy predictions are investigated using comparative statistics such as root mean square error (RMSE), average absolute percent error (AAPD), mean absolute error (MAE) and coefficient of determination (R2).