Theory and Machine Learning Modeling for Burst Pressure Estimation of Pipeline with Multipoint Corrosion
Hongfang Lü, Haoyan Peng, Zhao‐Dong Xu, Guojin Qin, Mohammadamin Azimi, John C. Matthews, Li Cao
Abstract
Burst pressure is an essential parameter to measure the residual bearing capacity of pipelines with corrosion defects. Accurate prediction of burst pressure is beneficial to evaluate the remaining life of pipelines. The traditional estimation formulas mainly focus on single-point corrosion but seldom consider the case of multipoint corrosion. Although DNV-RP-F101 considers multipoint corrosion, this method only considers the influence of axial spacing of adjacent corrosion defects on burst pressure but does not consider the influence of circumferential spacing. This work refers to the pipe corrosion criterion (PCORRC) and DNV-RP-F101 methods and modifies the calculation formula of the PCORRC method by introducing (D−t). The average relative error is only 11.46%, which is 27.65% lower than the previous formula. In addition, this paper presents a method for estimating the burst pressure of pipelines with multipoint corrosion defects. This method considers not only the axial spacing between defects but also the circumferential spacing. The results show that the maximum relative error is 6.52%. A machine learning modeling framework for predicting burst pressure in multipoint corroded pipelines is proposed. The case using radial basis function neural network as the prediction tool shows that the average Pearson correlation coefficient of the prediction results in the test set is 0.903, indicating that the prediction performance is good.