Evolution of ultimate pressure of randomly corroded pipes based on stochastic numerical analysis and CNN

Abstract

Research on corroded pressure pipelines is of great importance, given their critical role in industries including oil and gas, chemical processing, water distribution, and power generation. Corrosion is among the most detrimental factors impacting pressure pipeline safety. Via stochastic finite element analysis, this study explores the degradation law of external ultimate pressure for randomly corroded pipelines concerning the mass loss rate χ. It also examines the effect of axial loading on the radial pressure of such pipelines and clarifies how geometric dimensions (characterized by diameter-to-thickness ratio) and material yield strength impact their ultimate radial compressive bearing capacity. Furthermore, a reduction coefficient is proposed to quantify the degradation of corroded pipelines' ultimate compressive bearing capacity. Results show that the study's conclusions apply to pressure pipelines of various strength grades. Meanwhile, a theoretical simplified formula for the reduction coefficient—covering the range of diameter-to-thickness ratios commonly used in engineering practice—is established. Moreover, an optimized convolutional neural network (CNN) is introduced to effectively predict the reduction coefficient for corroded pipelines with varying diameter-to-thickness ratios. Overall, this research lays a foundation for the safety assessment of corroded pressure pipelines.