Evaluation of residual strength of externally-corroded submarine pipelines based on deep learning

Abstract

Corrosion is one of the important causes for submarine pipeline failure. Accurate prediction of the residual strength of corroded submarine pipelines is a key to evaluate the integrity and the subsequent service ability of submarine pipelines. Based on the nonlinear finite element method, a residual strength analysis model was established for the corroded submarine pipelines to predict their residual strength, and the impact of the depth, length and width of the external corrosion on the residual strength of the corroded pipelines was analyzed. Besides, a residual strength prediction model was also established based on the deep learning theory, the residual strength of the corroded submarine pipeline was predicted by the deep learning model that is trained with a dataset comprising 114 groups of calculation results, and the prediction results of the model were compared with the finite element calculation results. The results indicate that the deep learning model has a rapid calculation speed and high prediction precision, which verifies the feasibility and effectiveness of the evaluation method of the residual strength of externally-corroded submarine pipelines based on deep learning.

Keywords

• submarine pipeline
• external corrosion
• residual strength
• deep learning
• finite element simulation