Reliability analysis of spherical pressure hull based on active-learning hierarchical Kriging method

This study examines the reliability of a spherical pressure hull using an active-learning hierarchical Kriging method. The proposed approach incorporates an empirical formula to develop a low-fidelity model and employs an active-learning hierarchical strategy in conjunction with finite element results to build a high-fidelity model. Compared to the conventional active-learning Kriging method, the active-learning hierarchical Kriging method reduces the quantity of high-fidelity samples required while preserving accuracy and stability in failure probability estimation. Sensitivity analysis identifies key design parameters of the spherical pressure hull, and corresponding optimization recommendations are presented. The integration of the empirical formula effectively decreases the quantity of high-fidelity samples needed and improves the robustness of high-fidelity model construction process. The results offer an efficient modelling framework for the reliability evaluation of deep-sea equipment.