Prediction of CO2 corrosion rate of X65 steel under intermittent oil/water wetting using GA–BP neural network

Purpose This study aims to develop a corrosion prediction model for pipeline steel under intermittent oil/water wetting – a typical wetting condition in multiphase flow pipelines – providing a mathematical tool for designing effective corrosion control strategies. Design/methodology/approach A genetic algorithm–optimised back-propagation (BP) neural network (GA–BP) was trained on experimental data published online to predict the corrosion mitigation efficiency of X65 steel under intermittent wetting, which can be further used to modify and refine the DW 95 semi-empirical model for the new working phenomenon. Findings A corrosion mitigation efficiency prediction model for X65 steel under intermittent wetting was established based on GA–BP algorithm, achieving a correlation coefficient of 0.99895. Leveraging this model, the relative influence of operating parameters on pipeline steel corrosion under intermittent wetting was ranked as follows: wetting cycle > rotational speed > oil viscosity > temperature > surface roughness. Originality/value This study enables data-driven insights into the corrosion behaviour of pipeline steel under oil–water alternating wetting, providing a valuable tool for corrosion protection in oil–water two-phase flow systems.