Seismic response and predictive method for corroded buried pipelines under coupled reverse-fault displacement

Conventional finite element methods for large-scale numerical simulations are often constrained by high computational demands and extended runtimes. To enhance efficiency, we developed a predictive model based on a backpropagation (BP) neural network. A three-dimensional finite element model of a buried pipeline with corrosion defects crossing a reverse fault was established using ABAQUS. We systematically analyzed the effects of four key parameters—corrosion depth-to-thickness ratio, diameter-to-thickness ratio, internal pressure, and burial depth—on the seismic response of the pipeline. In this parametric study, fault displacement and the four key parameters served as inputs to the BP neural network, with the pipeline’s axial peak compressive strain as the output. The model was trained and validated using training, validation, and test datasets. Results indicate that increasing the corrosion depth-to-thickness ratio, diameter-to-thickness ratio, internal pressure, or burial depth reduces the fault displacement necessary for the lower section of the pipeline to reach its strain limit. Failure modes differ between unpressurized and pressurized pipelines, exhibiting inward local buckling and outward bulging, respectively, at stress concentration zones. The four parameters are highly correlated with the compressive strain response, with correlations transitioning from linear to nonlinear as fault displacement increases. The trained BP neural network achieves maximum prediction errors of 13.60% on the validation set and 12.84% on the test set, both below 15%, demonstrating robust accuracy and generalization in predicting the seismic response of in-service buried pipelines across reverse faults.