Multi-fidelity strength monitoring method for dynamic response of deep-sea pipelines based on digital-twin technology

• A multi-fidelity strength monitoring method is proposed using XGBoost-based Stacking-CGAN based on digital twin model. • Full-scale thin-walled pipe test along with numerical simulation is conducted under high-pressure and forced vibration. • Comparison with other methods affirms the validity of the digital twin model in pipeline strength monitoring. Cost-effective structural monitoring of deep-sea pipelines in harsh environments is essential for advancing deep-sea oil and gas development. This paper presents a multi-fidelity monitoring method for deep-sea pipelines based on digital twin (DT) model, applied to dynamic response monitoring under high pressure and forced vibration. This method integrates low-fidelity (LF) data from numerical simulations—clustered using the Hierarchical Geodesic K-means++ (HG-kmeans++) method—with high-fidelity (HF) data from full-scale pipeline tests. A pre-training model is developed using XGBoost based on LF data, and a DT model is established via the XGBoost-based Stacking-CGAN (XS-CGAN) method by integrating multi-fidelity data. A case study involving full-scale experiment demonstrates that the proposed approach significantly enhances both global and local accuracy. Compared to finite element analysis (FEA), the RRMSE and RE max improve by 40 % and 55.61 % respectively, outperforming current advanced method by 5.14 % and 3.79 %. These findings demonstrate the method's practical value for structural monitoring of deep-sea pipelines.