Multi-objective optimization of subsea U-shaped metal seals using a surrogate model and NSGA-II

The U-shaped metal seal ring of the subsea tubing hanger is a critical component ensuring wellhead sealing integrity, with its sealing performance determining the safe operation of oil and gas production systems. Addressing the analytical challenges posed by the seal's multi-parameter and highly nonlinear characteristics, this study establishes an optimization method integrating Optimized Latin Hypercube Sampling (Opt-LHD), XGBoost surrogate modeling, and the multi-objective genetic algorithm (NSGA-II). Initially, Finite Element Analysis (FEA) is utilized to evaluate the stress distribution and contact behavior, elucidating the sealing mechanism. Opt-LHD is then employed to construct a high-coverage sample space. Two surrogate models—XGBoost and BP neural network—are established and compared for the U-shaped metal seal. Through cross-validation and error analysis, the XGBoost model, exhibiting superior generalization, is selected to replace time-consuming FEA. Subsequently, a dual-objective global optimization targeting increased maximum contact pressure and reduced maximum equivalent stress is performed using NSGA-II. The Pareto front is validated through FEA to determine the optimized design. Simulation results demonstrate that the optimized seal ring achieves a 10.07% increase in maximum contact pressure and an 8% reduction in maximum equivalent stress, simultaneously enhancing sealing performance and structural safety, thereby validating the effectiveness of this optimization methodology.